Citation
American secrets, American transparency

Material Information

Title:
American secrets, American transparency analysis of the Freedom of Information Act as implemented by the U.S. intelligence community
Creator:
Ward-Hunt, Robert Cameron ( author )
Place of Publication:
Denver, Colo.
Publisher:
University of Colorado Denver
Publication Date:
Language:
English
Physical Description:
1 electronic file (194 pages). : ;

Thesis/Dissertation Information

Degree:
Doctorate ( Doctor of Philosophy)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
School of Public Affairs, CU Denver
Degree Disciplines:
Public Affairs
Committee Chair:
Heikkila, Tanya
Committee Co-Chair:
Gerber, Brian
Committee Members:
Jacob, Benoy
Piotrowski, Suzanne

Subjects

Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Review:
The US Freedom of Information Act (FOIA) places citizens' rights to request and discover information in competition with the right of the government to conceal official secrets (Bathory & McWilliams, 1977; Rourke 1957). Simply put, FOI in practice pits secrecy versus transparency, with administrative discretion in the middle. Though FOIA applies to all federal government departments and agencies in the United States, the tension becomes most acute when applied to the most secretive participants in the national security enterprise, the 16 member intelligence community. Secret-keeping can exert psychological and social pressure on organizations (Simmel, 1906; Weber, 1920/2009) that have the potential to impact individual decision making and shape collective norms (Freidman, Landes & Posner, 1991; Keane, 2008; Sandfort, 2000). This effort examines how secrecy impacts transparency initiatives by researching how the US Intelligence Community FOIA programs perform compared to other federal agencies using multivariate analysis of FOIA annual report data in conjunction with an experimental methodology. The research results show that while intelligence agencies release similar information to control agencies, they differ significantly in the how and when that information is provided. In particular, intelligence agencies take longer to respond than peer agencies and offer less information with the response. These findings contribute to theory by leading to a refined model of transparency and contribute to practice by supporting recommendations for policy makers and FOIA program administrators.
Thesis:
Thesis (Ph.D.)--University of Colorado Denver. Public affairs
Bibliography:
Includes bibliographic references.
General Note:
School of Public Affairs
Statement of Responsibility:
by Robert Cameron Ward-Hunt.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
902812160 ( OCLC )
ocn902812160

Downloads

This item has the following downloads:


Full Text
AMERICAN SECRETS, AMERICAN TRANSPARENCY: ANALYSIS OF THE FREEDOM OF
INFORMATION ACT AS IMPLEMENTED BYTHE US INTELLIGENCE COMMUNITY
by
ROBERT CAMERON WARD-HUNT
B.AV Boston University, 1997
University of Colorado Denver, 2006
M.SV University of Colorado Denver, 2006
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Public Affairs Program
2014


2014
ROBERT CAMERON WARD-HUNT
ALL RIGHTS RESERVED


II
This thesis for the Doctor of Philosophy degree by
Robert Cameron Ward-Hunt
has been approved for the
Public Affairs Program
By
Tanya Heikkila, Dissertation Chair
Brian Gerber, Examination Chair
Benoy Jacob
Suzanne Piotrowski
8 July 2014


iii
Ward-Hunt, Robert, Cameron (Ph.Dv Public Affairs)
American Secrets, American Transparency: Analysis of the Freedom Of Information Act as
Implemented by the US Intelligence Community
Thesis directed by Associate Professor Tanya Heikkila.
ABSTRACT
The US Freedom of Information Act (FOIA) places citizens' rights to request and discover
information in competition with the right of the government to conceal official secrets (Bathory
& McWilliams, 1977; Rourke 1957). Simply put, FOI in practice pits secrecy versus transparency,
with administrative discretion in the middle. Though FOIA applies to all federal government
departments and agencies in the United States, the tension becomes most acute when applied
to the most secretive participants in the national security enterprise, the 16 member
intelligence community. Secret-keeping can exert psychological and social pressure on
organizations (Simmel, 190b; Weber, 1920/2009) that have the potential to impact individual
decision making and shape collective norms (Freidman, Landes & Posner, 1991; Keane, 2008;
Sandfort, 2000). This effort examines how secrecy impacts transparency initiatives by
researching how the US Intelligence Community FOIA programs perform compared to other
federal agencies using multivariate analysis of FOIA annual report data in conjunction with an
experimental methodology. The research results show that while intelligence agencies release
similar information to control agencies, they differ significantly in the how and when that
information is provided. In particular, intelligence agencies take longer to respond than peer
agencies and offer less information with the response. These findings contribute to theory by
leading to a refined model of transparency and contribute to practice by supporting
recommendations for policy makers and FOIA program administrators.
The form and content of this abstract are approved. I recommend its publication.
Approved: Tanya Heikkila


iv
DEDICATION
This work is dedicated to my loving wife, Cassi, who inspired, encouraged and supported
me throughout this process. It is also dedicated to my son, Cavan, with hope for a future of
more open government.


ACKNOWLEDGEMENTS
v
I give my sincere thanks to Drs. Brian Gerber, Benoy Jacob, and Suzanne Piotrowski for
their support and guidance. I owe special acknowledgment and deep gratitude to Dr. Tanya
Heikkila for agreeing to serve as my committee chair, reviewing innumerable drafts and
generously giving her time to support me.
This effort would have not been possible without the financial support given by the Post
9/11 Gl Bill and the flexibility provided by my employer. Thank you all.


vi
TABLE OF CONTENTS
CHAPTER
I. INTRODUCTION.........................................................................1
Problem of Interest and its Importance..........................................1
Piecing together the Secrecy and Transparency Puzzle............................3
Freedom of Information Program Outcomes..........................................4
Discretion in Request Dispositions...............................................6
Institutional Transparency Rules.................................................7
Dissertation Outline.............................................................8
II. LITERATURE REVIEW....................................................................9
Introduction.....................................................................9
Secrecy as a Phenomenon.........................................................10
Transparency as both Intrinsic and Instrumental Value...........................15
Secrecy and Transparency Decisions: Costs and Benefits..........................22
Discretion and its Application to Secrecy and Transparency Decisions............26
Conclusions from the Literature.................................................31
III. FREEDOM OF INFORMATION: RIGHTS AND RESEARCH........................................34
Introduction....................................................................34
Freedom of Information as a Redress for Government Secrecy......................34
FOIA Establishes Public and Government Rights to Information....................36
The Institutional Grammar of FOIA........................................37
A Closer Look at the 20-Day Rule Needed..................................39
Executive Order 13392 Supplements Legislation............................40
Shared Rules, Similar Syntax.............................................42
Institutional Grammar Conclusion.........................................43


VII
Contemporary FOIA Research.....................................................43
Problems with FOIA: The National Security Archive at George Washington
University..............................................................44
Effects of Administration Change on FOIA outcomes: Kim (2007)...........45
The Effects of New Public Management on FOIA Programs: Piotrowski and
Roberts.................................................................46
Implications for FOIA research.................................................47
IV. CONCEPT DEVELOPMENT AND HYPOTHESES.................................................49
Introduction...................................................................49
Transparency Model.............................................................49
Refining Key Concepts into Operational Definitions.............................51
Operationalizing Secrecy................................................51
Testing Transparency....................................................52
Dealing with Discretion.................................................53
Hypotheses.....................................................................53
Evaluating Program-Level Outcome Differences............................54
Evaluating Individual Request Disposition Variances.....................57
Exploring Institutional Rules...........................................59
V. RESEARCH DESIGN....................................................................62
Overview.......................................................................62
Annual Report Data used to Evaluate Program Outcomes...........................64
Data Source FOIA Annual Reports.......................................64
Variables Used..........................................................66
Analytical Techniques
71


viii
Experimental Requests used to Evaluate Disposition Variances......................72
Experiment Design.........................................................72
Experimental Analysis.....................................................78
Agency Records and Artifacts Support Exploratory Questions........................80
Threats to Design.................................................................84
Secrecy Operationalization................................................84
Measuring Transparency....................................................84
Bias and the use of Mixed Methods.........................................85
Data and Information Accuracy.............................................85
Small-n Evaluation........................................................86
VI. TESTING HYPOTHESES USING ANNUAL REPORT DATA...........................................87
Introduction......................................................................87
Descriptive Statistics and Preliminary Analyses...................................88
Analyses for Hypothesis 1-Intelligence Agencies Grant Fewer Requests..............92
Intelligence Agencies Exhibit Lower Released in Full Rates................93
Intelligence Agencies Exhibit Higher Denied in Full Rates.................95
Discussion and Assessment.................................................97
Analyses for Hypothesis 2 Intelligence Agencies use Partial Grants and Administrative
Denials...........................................................................98
Analysis Released in Part as a Dependent Variable.......................98
Analysis Administrative Denials as a Dependent Variable.................99
Discussion and Assessment................................................100
Analyses for Hypothesis 3 Intelligence Agencies have Higher Backlog and Longer
Response Times...................................................................101


Intelligence Agencies exhibit Higher Backlog Ratios.......................101
Intelligence Agencies show Longer Mean Durations..........................104
Discussion and Assessment.................................................106
Chapter Summary...................................................................107
VII. EXPERIMENTAL RESULTS.................................................................109
Overview..........................................................................109
Round 1 Results...................................................................110
Objective.................................................................110
Response Summary..........................................................Ill
Service Summary...........................................................114
Material Summary..........................................................116
Round 2 Results...................................................................119
Objective.................................................................119
Response Summary..........................................................119
Service Summary...........................................................122
Materials Summary.........................................................124
Round 3 Results...................................................................125
Objective.................................................................125
Response Summary..........................................................125
Service Summary...........................................................128
Materials Summary.........................................................129
Discussion and Key Findings.......................................................130
Hypothesis 5 Little Difference in Disposition Outcomes..................130
Hypothesis 6 Delays by Intelligence Agencies............................132


X
Hypothesis 7 Intelligence Agencies lag on some Service Indicators.134
Records Shed Some Light on Exploratory Questions....................138
Summary of Experimental Round..............................................139
VIII. SUMMARY OF FINDINGS AND RECOMMENDATIONS......................................141
Introduction...............................................................141
Summary of Findings........................................................142
Analysis of Annual Report Data......................................142
Comparing Transparency about Transparency...........................144
Comparing Experimental Results......................................146
Implications for Theory....................................................148
Recommendations for Practice...............................................152
Policy Recommendations..............................................153
Recommendations for FOIA Administrators.............................155
Summary....................................................................157
IX. CONCLUSION...................................................................159
Introduction...............................................................159
Contribution to Theory.....................................................159
Contribution to Methodology................................................161
Contribution to Practice...................................................162
Limitations................................................................163
Recommendations for Future Research........................................165
Deepening Research in Secrecy and Transparency......................lbb
An Expanded Agenda for Secrecy and Transparency Research............168
Conclusion.................................................................170


REFERENCES................................................................171
APPENDIX A................................................................178
APPENDIX B................................................................181


CHAPTER I
1
INTRODUCTION
Problem of Interest and its Importance
The application of the Freedom of Information Act (FOIA) in the United States presents
an interesting and critical public administration problem. The US FOIA was designed to promote
transparency by providing citizens with the right to request records from Federal Government
agencies. A citizen's right to information about the government is considered an essential value
of democratic societies and successful implementation of freedom of information is a hallmark
of open governments (Galnoor, 1977; Openness, 2007; Piotrowski & Rosenbloom, 2003).
However, the right of citizens to request and discover is in potential competition with the right
of the government to conceal official secrets (Bathory & McWilliams, 1977; Rourke, 1957) and
may be impaired by bureaucratic patterns of administrative secrecy (Weber, 1920/2009). In
between these competing rights is the implied right of discretion exercised by executive
departments and agencies (Bathory & McWilliams, 1977; Bay, 1977; O'Brien, 1979; Rourke,
1957). Put simply, FOI in practice pits secrecy against transparency, with administrative
discretion in the middle.
Though FOIA applies to all federal government departments and agencies in the United
States, the tension becomes most acute when applied to the most secretive participants in the
national security enterprise, the 17 member intelligence community, consisting of six primary
intelligence agencies and eleven sub-agency intelligence components (Roberts, 2006, p. 48).
Compelled to both protect national security information and comply with the letter (if not
intent) of the FOIA, these agencies are presented with a dilemma that cannot be resolved
through policy alone. In short, they face a problem of discretion (Galnoor, 1975; O'Brien, 1979),
impacted by the institutional, organizational and administrative effects of secrecy. This


2
condition leads to the principal research question: How does secrecy in government
organizations impact execution of transparency initiatives? JW\s dissertation presents research
questions, hypotheses and research results that address that question, adding to the knowledge
on transparency in and by the US intelligence community.
This topic examines a key conflict between conceptually opposing themes, open
government and transparency on the one hand and government's right to secrecy on the other
hand. Freedom of information, in general and the intelligence community's application of FOIA
in particular, sits at the fissure between these two themes and serves as an excellent
opportunity to examine the dynamic in detail. Research on the US FOIA is both timely and
relevant. The FOI problem has been amplified in recent decades by two movements, the E-
Government movement which seeks to shift an increasing set of information and task execution
online and the reinventing government movement spawning from New Public Management
thinking. In the case of E-Government, federal agencies face increasing expectations of
openness and voluntary disclosure in online environments that may conflict with organizational
culture and evolved agency rule sets. In contrast, the NPM movement has the potential to treat
freedom of information as a non-mission essential task, outsourced and devalued by the
bureaucracy (Newbold, 2011; Piotrowski, 2007; Roberts, 2000; Roberts, 2002). Both examples
demonstrate the need for continual research on the implementation of FOIA by the US
government.
Additionally, events like Bradley Manning's mass divulgence of information to Wikileaks
and Edward Snowden's authorized dissemination of key NSA documents to US press show that
transparency of and in the intelligence community is a perennial topic of interest (Bannister &
Connolly, 2011). This research effort and dissertation builds on scant literature on the effects of


3
FOIA implementation in general and fills a gap in the literature with respect to FOIA programs in
the intelligence community in particular. In addressing the principal research question, this
research will provide analyses and context to address the question if America's most secretive
agencies deliver on the right of transparency through the Freedom of Information process.
Piecing together the Secrecy and Transparency Puzzle
Secrecy itself presents a bit of puzzle as there are secrets about secrets. By their very
nature, secret organizations are hard to study and secrecy is difficult to observe directly.
Instead, researchers must often look for the influences of secrecy indirectly. The literature on
secrecy reveals complex interpersonal and institutional dynamics, opening up the possibility of
using a variety of lenses to examine the impact of secrecy on organizations and systems. Use of
an organization-level lens, such as administrative discretion, allows for the possibility of
incorporating both institutional and interpersonal effects into a single viewpoint.
Scholarship on administrative discretion indicates that a number of factors affect
administrative discretion, including collective processes and socialized norms (Sandfort, 2000),
management (Riccucci, 2005), organizational structure (Meier & Bohte, 2001) clients (Scott,
1997), and the self-perception of administrators (Sowa & Selden, 2003). Drawing from
administrative and sociological concepts of secrecy (i.e. Simmel, 1906; Weber, 1920/2009), it is
reasonable to suspect that administrative and cultural aspects of secrecy might also influence
the exercise of discretion and the creation of institutional rules, particularly with respect to the
administration of FOIA by the intelligence community. Similarly, transparency can be viewed as
a set of instrumental rules and directions affecting program outcomes, discrete request
responses and internal processes (Meijer, 2013). These rules and the resultant outcomes can be
evaluated comparatively, using intelligence agencies as the subject set for secret organizations.


4
Collectively, looking at aggregate outcomes, discrete request responses and institutional rule
sets provides a multi-dimensional portrait of how intelligence agencies differ from their non-
secret counterparts on transparency.
Freedom of Information Program Outcomes
Although the First Amendment guarantees the right of citizens to petition the
government for a redress of grievances, no other constitutional rights are explicitly granted to
either the executive or the citizen for the withhold or discovery of information1. When FOIA was
enacted in 1966, the Act not only established a legal right for citizens to petition for information
from government agencies, but also provided the mechanism for them to do so, equipping
'citizens with spades' (as paraphrased from Seymour-Ure, 1977) to uncover information of
interest by compelling executive agencies to receive and disposition FOIA requests. In doing so,
the US Congress formalized the long-recognized value and made freedom of information an
essential task for federal executive agencies. The Act served as progenitor for a wave of first-
generation FOI legislation in other countries during the 1960s and 1970s. In more recent
decades, a second wave of legislation adoption has occurred, embracing second-generation FOI
principles. Modifications to FOIA in the last decade added emphasis on agencies providing FOIA
as a service (OPEN, 2007).
Since 1966, citizens and press have used FOIA requests as a method to monitor
government performance. In Fiscal Year (FY) 2011 alone, federal agencies received over 644,000
requests for information or records (Justice, 2012). Unsurprisingly, not all of these requests
were granted. Though the Act established citizens' rights to information, it also established the
1 The constitution of Sweden, often cited as an exemplar for freedom of information, does explicitly
contain FOI rights (Galnoor,1975)


5
legal right of the government to withhold information in nine enumerated categories such as
trade secrets, sensitive law enforcement information and classified national defense
information. Tens of thousands of requests per year are denied in full or in part on the basis of
these enumerated exemptions (Justice, 2012). In addition to these denials, agencies have the
ability to deny requests for administrative reasons, a class of denial that includes the inability to
find records or unwillingness to accrue excessive costs to retrieve the requested information.
Both the exemptions and administrative denials are used frequently. For FY11, of the requests
dispositioned by the federal government, only 53.91% were fully granted. However, this coarse
measure only tells a portion of the story, and doesn't account for the administrative purgatory
between granted and denied. Although federal agencies processed (granted or denied) over
631,000 requests in FY11, they also carried forward a substantial backlog of 108,763 requests
into the next fiscal year, an almost 2% increase over the previous year. Though the Act specifies
that agencies are to return a response within 20 business days of receiving a FOIA request, many
agencies comply with the letter of the law by returning a response that indicates that the
request has been added to the backlog.
Treated as a system, agency administrators have considerable latitude in implementing
the FOIA, deciding how requests should be received (thus potentially imposing high transaction
costs on requestors), the amount of resources to dedicate to FOIA request processing, how
requests should be dispositioned and if and how reports should be made available to the
general public. As a result, despite legislated transparency in the form of the FOIA, the amount
of effective transparency provided by the government remains largely at the discretion of
individual agencies. Agencies have the discretionary latitude to set policies and allocate
resources that have the potential to greatly affect the overall outcomes of their FOI programs
and hence their effective transparency. Yet, little is known about intelligence agency


6
performance compared to non-intelligence agencies. This raises the question of how intelligence
agencies might exercise this differently when it comes to FOI. If these variances exist, they merit
additional investigation to discover proximate causes and patterns. Research Question 1:How
do Intelligence Agency FOIA program outcomes differ from their less-secretive counterparts?
Discretion in Request Dispositions
A number of discretionary decisions exist for any potential FOIA request. Agencies can
select to invoke an exemption, respond with an administrative denial, segment the request for
partial grant or delay processing the request. In exercising discretion, the agencies must
navigate between several different and sometimes opposing policy objectives, presenting a
difficult administrative discretion scenario. A prime example exists with respect to national
security. Though the agencies are compelled by executive order to protect classified
information, non-classified national security information is subject to FOIA requests, leaving a
significant amount of sensitive-but-unclassified information in a 'gray' area potentially
releasable upon request. During the Clinton Administration, the White House issued guidance
establishing presumption of openness as a policy objective (Kim, 2007). Following the terrorist
attacks of 2001, the George W. Bush administration issued guidance designed to protect
national security information to the maximum extent possible (Kim, 2007). In both cases, the
general executive intent was to shape execution of FOIA programs within the law and the
specific intent was to influence the implied exercise of discretion, particularly with respect to
information in gray areas.
The main discretionary decision made for any single request is to grant or deny the
request, with a companion decision on which records (if any) to release. Several secondary
decisions exist as well, such as which track to assign the request to (in the case of multi-track


processing), whether or not to grant a fee waiver and whether or not to seek remedial fees for
7
expenses. In addition, agencies also make undocumented decisions regarding how much search
to accomplish to look for responsive records and what constitutes a record, in the case of
ambiguous documentation. Documented decisions have the potential to shed insight on
systematic or individual variation between agencies, raising the question of if secrecy results in
different disposition results. In other words, given a request for unclassified information, do
secret agency responses differ significantly from their counterparts? While research question 1
examines program outcomes as a whole, this research question is concerned with individual
request decisions. Research Question 2: How do Intelligence Agency FOIA disposition decisions
differ from their less-secretive counterparts?
Institutional Transparency Rules
The first two research questions will provide data about the effects of secrecy on
programs as a whole and variances in individual request responses, but offer little insight about
direct or proximate causes. For this, a deeper look is needed at internal mechanisms, such as
institutional rule sets. Transparency is affected by a number of statues, orders, policy
statements and organizationalevel practices. Specifically, rules regarding the definition and
retention of records and information classification have a direct impact on transparency in
practice. Understanding how the intelligence community implements these key rules will
improve our understanding of how secrecy and transparency impact discretion in FOIA
programs. Research Question 3: How do FOIA-related rules differ between intelligence agencies
and their less-secretive counterpartsP


8
Dissertation Outline
This dissertation begins with a review of several sets of literature relevant to the FOI
problem, starting with secrecy as a phenomenon, informed by work in sociology and philosophy.
Definitions and concepts for transparency follow and set the framework for transparency
research. Cost and benefit considerations, found in literature on the economics of secrecy help
provide insight into secrecy decision-making made by individuals. Additionally, the literature on
administrative discretion provides a basis for considering aspects influencing the exercise of
discretion in federal agencies. Following the literature review, Chapter III contains an
introduction to the US Freedom of Information Act, an analysis of the institutional grammar of
the Act to draw out specific elements of the law that can be observed and a review of
contemporary research on FOIA. In broad strokes, this chapter helps place the specific elements
of FOIA in the overarching context of the secrecy and transparency literature.
Chapter IV builds on the literature review and analysis of FOIA to present a model of
transparency and convert key concepts to operational definitions. These operational definitions
are in turn used to generate testable hypotheses and exploratory questions. The dissertation
continues with a research design chapter (Chapter V) containing the approach to data collection
and analytical methodology. Results are captured in three chapters, beginning with analysis of
quantitative data (Chapter VI) and followed by a summary of experimental results (Chapter VII)
and concluding with an integrated summary of findings and recommendations for policy makers
and agency program managers (Chapter VIII). The dissertation concludes with an assessment of
the impact of this research and recommendations for future research.


9
CHAPTER II
LITERATURE REVIEW
Introduction
The research questions for this dissertation compare intelligence agencies and their
less-secretive counterparts in three areas:1)FOIA Program Outcomes, 2) FOIA request
disposition decisions and 3) FOIA-related rules. There are a number of key concepts at work that
are needed to understand the essential dynamics behind these three questions. The first is the
concept of secrecy which distinguishes intelligence agencies from other federal civil agencies. As
the principal causal relationship of interest, a nuanced understanding of the social, psychological
and economic impacts of secrecy on individuals and organizations helps inform interpretation of
observable outcomes. Transparency is the driving value underpinning the FOIA. Understanding
the intrinsic and instrumental nature of transparency helps illuminate a duality within this
concept that influences both the expectations of the citizenry and the actions of federal
agencies.
Hypothesized differences between outcomes in the first research question and
disposition results in the second research question are likely partially attributable to individual
and organizational decision-making processes. The literature on the economics of secrecy and
transparency helps to make sense of this explanatory relationship. Similarly, the literature on
administrative discretion helps point towards areas for investigation in examining how FOIA
programs and rule sets are implemented by agencies and groups. Collectively, these four
streams of literature help provide the conceptual framework for this dissertation and inform
relationships of interest that merit additional exploration and research.


10
Secrecy as a Phenomenon
This dissertation draws on classical and more recent work on secrecy to understand the
phenomenon of secrecy and its impact on organizations, particularly useful for appreciating how
intelligence agencies differ from their federal counterparts. The literature on secrecy is as
fascinating as it is varied, with contributions in the fields of public administration, psychology,
sociology and economics. Four scholars, Georg Simmel, Max Weber, Edward Shils and Sissela
Bok, established the central writings on secrecy as a phenomenon, incorporating individual and
institutional perspectives of the topic that serve to outline the essential features of secrecy
(Blank, 2008/2009). Their collective understanding illuminates potential aspects of secrecy
needing examination in the context of FOI research.
As a sociologist, Simmel (1906) focused on the interpersonal aspects of secrecy with
society and secret societies in particular, with insights on the reciprocal nature of secrets and
the role of trust in social dynamics. Although also a German sociologist like Simmel, Weber
(1920/2009) takes a different approach, examining the dynamics of power and secrecy in
organizations, particularly the bureaucratic form. Edward Shils, an American sociologist writing
in the mid-1950s on the heels of the McCarthy-led Second Red Scare era, adopts an institutional
approach to secrecy. In many ways, Shils embraces Weberian concepts of secrecy as an
organizational phenomenon and applies an institutional construct, examining the mechanisms
of secrecy and looking at broad social effects. Bok continues Shils line of reasoning by separating
privacy from secrecy, finding the latter to hide more and be a qualitatively different
phenomenon than the former. Like Shils, Bok rejects the normatively neutral frame of Weber
and Simmel (190b), finding secrecy inherently problematic and ethically dangerous (Blank,
2008/2009, p. 64).


11
Despite differences in viewpoint, when taken as a whole, these four scholars present a
consistent approach to secrecy as a phenomenon and social construct. The essential
characteristic of secrecy common to all four scholars is that it involves concealment of
information described as "compulsory withholding" by Shils (1956, p.26), "consciously willed
concealment" by Simmel (Blank, 2008/2009, p. 61)and "intentional concealment" by Bok (1982,
p.11). The element of purposeful concealment distinguishes secrecy from deception, the act of
presenting false representations, and privacy, which is thought of as roughly passive and
complemented by indifference (Blank, 2008/2009). The aspect of intentionality has implications
for social dynamics present in organizations and institutions.
Simmel, Weber, Shils and Bok find secrecy to be a powerful social construct. Purposeful
concealment implies concealed by as well as concealed from group structures or, as Keane
describes it, ''boundaries of social inclusion and exclusion" (2008; p.108). Simmel found it
remarkable that information could be concealed effectively by a group. "Secrecy in this sense
i.ev which is effective through negative or positive means of concealment is one of the
greatest accomplishments of humanity" (Simmel,1906, p.462). Although they express the
argument differently, Simmel and Bok in terms of secret societies, Shils and Weber in terms of
institutions and organizations, all four present secrecy as a potentially divisive social construct
because it separates one group from another by means of purposeful concealment. Simmel
argues that secrets become a source of individual power in interpersonal relationships, at once
strengthening the bond between individuals who share secrets and creating barriers between
those who hold the secret and those that do not. ''Secrecy secures, so to speak, the possibility of
a second world alongside of the obvious world, and the latter is most strenuously affected by
the former" (Simmel, 1906, p.462). This possibility of dual realities, one presumably enlightened
with secret knowledge and information and the other presumably left to make do with only


12
ignorance, forms the basis of fascination with secret societies and conspiracy theories (Shils,
1956). 'The mere existence of secrecy is bound to set nervous minds on edge" (Shils, 1956, p.
12). Extending this argument to contemporary bureaucratic structures implies the existence of a
'secret' government operating within an open one. It is precisely this condition that the framers
of the FOIA found following the military build-up during the Second World War. The growth of
dedicated intelligence agencies, operating behind a curtain of official secrecy, exemplifies this
second, secret government. As Bok suggests, this presents alarming ethical implications for
citizen participation in democratic government and increases the chances that deviant behavior
and legal transgressions will go undetected in secret agencies.
Simmel also lays a portion of the foundation for discussions of disclosure, articulating a
dynamic tension between the act of concealment and the forces of discovery. The power of
secrecy, Simmel argues, is magnified when concealment is successful against discovery efforts
and quickly reduced when the secret is a secret no more. As a result, groups and societies of
secret keepers can exert significant psychology pressure on the individual to maintain secrets,
reinforcing the social boundaries of secrecy (1906/2009). Weber also recognized that the
division between secret keepers and the uninformed was a significant source of power and
applied it specifically to the bureaucratic form of organizations. Weber's basic premise is that
the information or knowledge differential that the bureaucracy creates is a source of power for
the bureaucrats who will attempt to further or maintain that differential, preserving their job
and protecting the bureau. Weber credits the concept of an official secret to the bureaucracy
and links the two together (1920/2009). In this, Weber does not mince words, "every
bureaucracy seeks to increase the superiority of the professionally informed by keeping their
knowledge and intentions secret" (1920/2009, p 47). For Weber, the designation of an official
secret is only one of several means for officials in the bureaucracy to maintain secret knowledge


13
in order to further their power. In practice, information or knowledge need not be designated
'secret' in order to effectively be so; the ordinary process barriers a bureaucracy might apply
offer an effective deterrent against disclosing information, even if the information itself was not
intended to be protected. Applied on a large scale, as in the case of many contemporary federal
governments, there can exist a significant cadre of secret-keepers concealing information at
various levels, some classified and some unclassified and some intentionally and some
unintentionally. By extension, Niskanen's (1968) work on budget-maximizing bureaus might also
be applied here, substituting budget for power. In this light, bureaus might attempt to maximize
power through the retention of secrets, creating an environment significantly adverse to sharing
information with citizens, other bureaus and legislative authorities (Weber, 1920/2009). This is
part of the problem that FOI laws were designed to correct.
Shils and Bok address the question of the morality of secrecy in a way that Simmel and
Weber do not. Whereas Simmel states that ''Secrecy is not in immediate interdependence with
evil, but evil with secrecy" (p.21); Shils and Bok are less than convinced. While acknowledging
the right or duty of government to withhold some information, Shils (1956) warns against both
of the perils of secrecy and of an obsession with conspiracies and secrets, no doubt influenced
by the American political events of his time. In fact, the title of his work, The Torment of Secrecy,
succinctly reveals his frame of reference for the problem. Shils' emphasis is on the broad
institutional effects, linking secrecy to conspiracies to social and political anxiety. Bok (1982)
frames the problem differently, with a focus on individual and social effects. For Bok, the
essence of the danger of secrecy lies precisely in its definition, the intentional concealment that
impacts individual and societal decision-making by removing information and knowledge from
consideration. In particular, Bok finds collective secrecy dangerous, asking the question "Is there
something about collective secrecy that renders it more problematic than individual secrecy?"


14
(Bok, 1982, p.107). Bok answers in the affirmative, reasoning that collective secrecy removes
accountability and diminishes the sense of personal responsibility from collective decision
making. This leads to the reasoning that secrecy can have 'spillover' effects with impacts
broader than just the individuals party to the secret. Even with this normative framing, both
Shils and Bok admit that some secrecy is necessary, though they advocate for limited secrecy
and strong publicity as a form of societal and institutional checks and balance.
The work of Simmel, Weber, Shils and Bok points to the certainty of the bureaucracy
attempting to conceal information from both the citizens it serves and the officials elected to
govern it. This phenomenon is most acutely experienced in a nation's national security
apparatus, which has a legitimate need for security, but yet may be most in need of
transparency so that citizens may keep a watchful eye on it (Roberts, 2006). This concern is
clearly evident in the work of Shils and Bok. Writing about foreign and military policy, Shils
states, ''In these fields, secrecy has been accepted in the liberal democracies as a necessary evil"
(1956, p. 26). Bok, writing after a decline in trust in the executive branch in general and the
national security apparatus in particular following a decade of scandal and impropriety, states
that ''we have created, in the last quarter century, a new culture, a national security culture,
protected from the influences of American life by the shield of secrecy" (1982, p. 206). In
democratic societies this is an especially troubling and significant problem because it effectively
restricts the participative element from the business of democratic government (Piotrowski &
Rosenbloom, 2003). Deprived of information, the electorate is a disadvantaged participant in
decision-making, a condition which would seem to violate a core tenet of any true democratic
society (Newbold, 2011). Viewed through this lens, the rights and obligations of the government
to conceal are placed in opposition to the rights of the citizenry to discover and know (Galnoor,
19//; Roberts, 2006). The bureaucracy, in the form of Executive branch agencies and


15
departments, is placed in the awkward, but unavoidable position of simultaneously creating and
enforcing the processes of secrecy and information protection while executing the policies of
disclosure and transparency (Piotrowksi & Rosenbloom, 2003).
The literature on secrecy as a phenomenon raises questions that require additional
research. At a basic level it raises the questions of if and how America's secret agencies live up
to the promise of FOIA and if the presence of official secrecy produces broader impacts. More
specifically, it leads to the assumption that agencies with more official secrecy would exhibit
fewer tendencies to disclose information than their less secretive peers, even after accounting
for exemptions used to protect ofncial secrets. Put succinctly, while agencies dealing with
classified information would obviously protect those secrets, the influence of secrecy as
phenomenon might also lead them to over-protect unclassified information that would and
should ordinarily be disclosed. The literature strongly suggests that research should reveal
meaningful differences between secret agencies and their less-secretive counterparts.
Transparency as both Intrinsic and Instrumental Value
Secrecy is only one part of the larger research puzzle. Absent other compelling forces,
the secrecy literature makes it clear that organizations and agencies would likely prefer
concealment over disclosure. However, secrecy is not the only force at work on individuals and
organizations. American citizens have a history of expecting transparency from public and, to a
lesser degree, private organizations. Just as the literature on secrecy informs the essence of
secrecy as a phenomenon, the public administration literature on transparency helps to unravel
this complex concept and points to areas needing further examination and research. Of
particularly interest given the main research question How does secrecy in government


16
organizations impact execution of transparency initiatives? is how the literature deals with
transparency as both a value and as an outcome.
The word itself is compound of trans- and parene, literally meaning to see through. But
this seemingly simple and straightforward definition belies a more complex set of relationships
between an observer and the observed. At the conceptual level, transparency in public
administration literature is represented by two different streams; the first viewing transparency
normatively as an intrinsic value, the second viewing transparency descriptively as an
instrument.
The normative stream of literature characterizes transparency as an intrinsic social and
individual value (Etzioni, 2010; Hood, 2006; Piotrowski, 2007, p.3). This is a classical view of
transparency summarized by the often quoted paraphrase from Justice Brandeis: ''Sunlight is ...
the best of disinfectants". The normative frame owes much to the work of enlightenment
scholars like Rousseau and Bentham who articulated both a personal and public need for
disclosure and openness (Hood, 2006). This is transparency as an unalloyed good, described by
Bentham as an "indisputable truth"' (p. 9), and by Rousseau as a "lost state of nature'Hood,
2006, p.7). To Bentham and Rousseau, whether the observed is the government, public officials
or fellow citizens, unfettered visibility provides the mechanism for reaching a more utopic state.
To scholars in this stream, transparency is key condition of democratic society; reducing the
information asymmetries needed for the citizen and press to maintain a watchful eye on
government (Bathory & McWilliams, 1977; Fung, Graham, & Weil, 2007). Framed normatively,
secrecy and privacy are inescapably oppositional in nature as barriers to transparency. In this
context, transparency and disclosure are fulfilling the ideals of openness and are vehicles by


17
which government accountability can be realized, all with the objective of increasing trust
between people and their government (Meijer, 2009; O'Neill, 2006).
However, this normative framing is not without challengers, particularly in
contemporary scholarship which follows more of post-modern bent. Heald (200b) argues that a
degree of ignorance is socially useful and posits that the optimal level of transparency is
somewhere below maximum, at a level allowing for some privacy and intentional concealment.
Bannister and Connolly (2011) argue that there are transparency costs and risks which create
the potential for disutility. Bannister and Connolly cite the potential for misinterpretation and
difficulties in citizen comprehension as potential risks, along with the possibility that agencies
may shift towards increased failure to acknowledge problems or record key opinions or
processes for fear of mass publication. Etizoni (2010) advances along bounded rationality lines,
arguing that more transparency doesn't necessarily result in better outcomes and that an
unlimited amount of transparency can be counter-productive as it has the potential to
overwhelm the cognitive processing abilities of the supposed beneficiary. Similarly, Fung,
Graham and Weil (2007) find that cognitive biases and limitations limit the effectiveness of
disclosure regulations and practices.
The descriptive perspective of transparency focuses on the instrumental nature of
transparency, represented by Meijer's (2013) definition of transparency as ''the availability of
information about an actor that allows other actors to monitor the workings or performance of
the first actor" (p.430). Meijes definition has a number of key aspects worth highlighting.
First, Meijer's definition revolves around principal-agent dynamics, making transparency
not only instrumental but also directional. In other words, the question to be asked is
'transparency o/whom by whom?' If Meijer's definition is accepted, then there can be many


18
directions of transparency in public affairs. For any given agency or agent, there is the potential
for transparency to a number of differing principals including other agencies, the executive, the
legislature, the press and the public. This directional nature is derived from Heald (2006) who
provides a four direction framework for consideration: Upwards, Downwards, Inwards and
Outwards. The implication of directional transparency is that it must also be viewed as relative:
transparency to one observer might be perceived as opaqueness by another observer.
Regardless of direction, the implicit assumption is that the availability of information allows the
principal actor to more closely ensure that the agent's actions are in alignment with the
principaTs goals. However, this assumption is challenged in separate arguments by Bannister
and Connolly (2011), Prat (2006) and Etzioni (2010).
Second, Meijer's definition is institutional in nature, involving structured rules and
relationships. Support can be found in the work of Dawes (2010), who posited that information
can be both an object and an instrument of policy and in the work of Hood (2006) who offered a
definition of transparency as fixed and predictable rules. Viewed through this institutional lens,
information and power are directly related. Though not specifically referenced by Meijer, the
possibility of socially constructed transparency is clear, some groups of actors will be more
deserving of monitoring or surveillance (and hence less privacy) than others and some groups of
actors will be more deserving of information than others. By extension, one might observe
agencies exercising discrimination in transparency; choosing some classes of individuals as more
deserving of transparency than others, particularly if the agencies perceive those classes as
being more powerful relative to other individuals. For example, agencies might treat requesters
identified as 'press' differently from requesters identified with educational institutions.


19
Third, by using an instrumental definition, Meijer raises the question of 'transparency
about what?' Meijer uses the terms 'workings and performance' in his definition, referencing
Heald's (2006) distinction between event and process transparency. Heald (2006) describes the
former as observables inputs, outputs and outcomes while the latter consists of less
measurable, but not unobservable, procedural sequences connecting events. Event
transparency is similar to Pasquier and Villeneuve's (2007) conception of document
transparency: transparency about the artifacts and records of government. The US FOIA
provides an example of document transparency, though the operational definition of a 'record'
by agencies means that many documents fall outside of its scope. Process transparency can be
found in requirements to post certain procedural instructions to the federal register and in some
'Sunshine Laws' regarding meeting openness and public hearings.
Meijer's definition takes for granted the awareness of the first actor about the existence
of the second, for monitoring the 'workings and performance' of an agency first requires that
one knows that the agency exists. This has particular relevance with secret agencies. As a case in
point, the US Government did not openly acknowledge the existence of the National
Reconnaissance Office until 1992, more than three decades after it was formed (Berkowitz,
2011). While an extreme case, the precept stands; there must be sufficient information about
the existence and activities of an organization for any transparency to exist. A more recent
example can be found in the 2013 scandal involving NSA collection of information on US
persons. Without awareness of the 'fact of such a program existing, transparency about the
program was impossible. This line of argument suggests that Meijer's definition is missing the
element of awareness or 'fact of, but does not otherwise detract from his definition.


20
Fourth, Meijer's definition raises the question of medium. While some transparency is
direct between the observers and observed, most transparency is mediated, most often through
electronic means (Meijer, 2009). This complicates the relationship between observers and
observed and raises the issue of the trust in the medium as well as trust in the transparency of
the original actor. Quickly, the problem resembles Russian nesting dolls with the question of
transparency through, not only to a given actor. In a sense, US FOIA is a mediated rather than
direct transparency law, offering citizens information only through the structured medium of
official records (Piotrowski, 2007; Roberts, 2006). Roberts (2001) places the issue of medium in
the context of the structural pluralism of NPM, questioning the government's responsibility to
ensure transparency of its contracted actors or partners. Interestingly enough, this places FOIA
offices in government agencies in the position of being both the object of transparency and the
mediators of transparency, raising the question of how transparent agencies are about
transparency.
Though different, the two streams of transparency literature are not entirely exclusive
and have areas of commonality. Both the normative and descriptive frames are based on the
principal-agent model and the ability of the principal to monitor the agent. The discretion
exercised by the agent and the limits upon the principal's monitoring abilities are important
factors of the model. Discretion is an inevitable element of the principal-agent theory and offers
the opportunity for the agent to select transparency outcomes that differ from the principaTs
preferences. Bounded rationality concepts indicate that the principal experiences cognitive
limitations and biases which may limit the ability to process an unbounded amount of
information and impact decision-making.


21
In some aspects, the normative and descriptive transparency streams merge
unexpectedly. The normative framing for transparency still has an instrumental aspect, as
transparency is an enabler to obtaining trust between the people and government. Similarly, the
descriptive framing cannot entirely escape a normative implication, as the core assumption is
that monitoring of the agent yields more positive outcomes. Transparency is not costless, but
rather incurs costs on the part of both the principal and the agent. The literature overwhelming
assumes that the net benefit of transparency outweighs its costs, with only a few voices (i.e.
Bannister & Connolly, 2011; Etizoni, 2010; Heald, 2006) warning of the potential for
diseconomy.
Secrecy has a different and more complex relationship with instrumental transparency
than it does with normative transparency. Secrecy has a simple, albeit oppositional, relationship
to normative transparency. However, this relationship looks different when using an
instrumental definition of transparency. Through this lens, secrecy and transparency can be
viewed as co-existing institutional influences, with the potential to be competing or
complementary influences. This institutional view points to the potential of transparency, like
secrecy, being observable not just through outputs but also through organizational rules and
structures. Viewing transparency and secrecy as co-resident institutional influences also
highlights the role of administrative discretion in secrecy and transparency decision-making.
Of the two frames, the descriptive frame, based on Meijer's (2013) definition holds
more promise. The supporting literature leads to questions of how the co-resident influences of
secrecy and transparency are resolved by individual administrators and agencies as a whole. For
the second research question, meaningful differences attributable to these institutional
influences might be found by comparing individual request dispositions and responses from


22
agencies. The literature also suggests a deeper look at secrecy and transparency costs and
benefits is needed. Finally, using a definition of instrumental transparency raises the question of
medium and to what degree agencies are 'transparent about transparency/ In terms of the
principal research question, this transparency literature suggests the likely outcome that
intelligence agencies are less likely to be transparency about their transparency initiatives than
other agencies. For the third research question involving FOIA-related rules, it also suggests
directionality, importing that intelligence agencies are more likely to have institutional rules that
favor secrecy at the expense of transparency.
Secrecy and Transparency Decisions: Costs and Benefits
While the classic literature on secrecy and transparency describes well the macro-level
influences of secrecy and transparency, it falls short in illuminating how individual concealment
and disclosure decisions are made. The literature on the economics of secrecy and disclosure
helps ameliorate this gap with micro- and meso-level models of individual and organizational
decision-making, drawn first from rational choice economics and then tempered by behavioral
economics assumptions. This stream of literature is key to understanding discrete outcomes of
the FOIA process like formal and informal disclosure decisions involved in answering the second
research question, How do Intelligence Agency FOIA disposition decisions differ from their less-
secretive counterparts?.
Scholarship on trade secrets and information economics indicates that organizations
experience both costs and benefits associated with secrecy and transparency (Aftergood,
1999/2009; Kultti, Takalo & Toikka, 2006; Stigler, 1980). Secrecy and transparency can be
viewed as aspects of the information marketplace, with the government, press and public as
participants in the exchange producing and demanding information as a resource (Galnoor,


23
1975; Lowi, 1977). The protection of information conveys certain benefits to organizations
(Dufresne & Offstein, 2008; Hermalin & Katz, 2006). Through secrecy, companies may achieve
strategic surprise in the marketplace, obtaining a first-mover advantage and deterring or
delaying new market entrants (Keane, 2008). Designation of a trade secret may allow companies
to protect intellectual property without disclosure, obtaining a competitive advantage over
potential imitators. Similarly, state secrets allow governments to maintain advantages over rival
states, obtain differential information advantages in diplomatic negotiations and conduct covert
operations to conceal action. However, the act of concealment also generates costs, both direct
and indirect (Stigler, 1980). The direct costs of implementing a secrecy program are non-trivial,
including personnel costs such as vetting and clearance, materiel costs for storage and security
and, in the aigital age, significant information technology costs. The indirect costs of secrecy,
while difficult to quantify with precision, can also be significant and include intra-firm effects
such as impaired organizational learning (Jensen & Webster, 2009), weakened knowledge
management and other intra-firm information asymmetries that may impair exploration and
exploitation (March, 1991) as well as extra-firm effects like suppressed diffusion of innovation
(Aftergood, 1999/2009; Kultti, Takalo & Toikka, 2006; Rourke, 1960). On balance, this set of
literature views secrecy negatively, as an impedance to the free exchange of information
necessary for ideal market operations (Hermalin & Katz, 2006).
However, full transparency, in the form of unfettered information disclosure, is not
costless either (Prat, 2006). Organizations also experience benefits and costs to disclosing
information, whether voluntary or compulsory. The work of Fung, Weil, and Graham (2007)
indicates that there are some benefits to voluntary disclosure as it reduces information
asymmetry on the part of consumers/clients. In some cases, compulsory disclosure as a result of
transparency laws improves consumer choice, though in other cases, the disclosure appears to


24
have little to no effect. A key conclusion from their research is that transparency is as much
about how, where and when as the what, in other words, the mechanisms and timing of
disclosure are as significant as the content disclosed (Fung, Weil, & Graham, 2007). This
conclusion has direct implications for FOIA, which can be considered a procedural law, intended
to provide the mechanism for request based disclosure, rather than a transparency law,
specifying content and information to be provided by the government. Additional, it suggests
that research examining FOIA mechanisms and procedures might be fertile ground for
uncovering differences between agencies.
Additionally, secrecy has spillover effects, resulting in social costs and potential value
conflicts (Galnoor, 1975). External parties, including private companies and public organizations,
might bear the costs of deviant behavior by a single organization made easier through concealed
action and impaired discovery (i.e. social losses caused by fraud or theft). Moreover, the gains
and costs of secrecy are distributed unequally. The gains from secrecy belong primarily to the
organization keeping the secret and begin immediately (the gain being prevention of loss or
retention of competitive advantage) (Galnoor, 1975). The benefiting organization bears some
direct and indirect costs of secret-keeping, but other organizations or institutions may also bear
costs due to impeded diffusion of innovation (Rourke, 1960), impaired decision-making and
increased risk of deviant behavior. These costs are not likely to accrue immediately, but rather
over time and are experienced largely by extra-organization (i.e. second & third) parties.
Conversely, the benefits of transparency/disclosure are largely realized by external parties
whereas the costs are due to the disclosing party or observed agent (Prat, 2006). The patent
system is the exemplar of this effect, with the addition of limited-duration protection for the
patented party to realize competitive advantage (Friedman, Landes & Posner, 1991). Prat (2006)
states that more observations (or disclosures) will not harm the principal(s), but the same


25
cannot be said for the agent. The presence of social costs for secrecy and social benefits for
transparency and disclosure drives the collective action response that has resulted in a role for
government to regulate secrecy and mandate transparency and disclosure in certain industries
(Roberts, 2006). Viewed through this lens, government agencies present a distinctive case as
government organizations in general and US federal agencies in particular are charged with a
special self-regulation problem, effectively enforcing two potentially competing rule regimes.
This suggests that research on FOIA should examine the possibility of rule conflict as a factor in
decision-making.
Principal-Agent Theory provides a model of the individual to frame and formalize
assumptions about secrecy costs and benefits. Agency theory is essentially a relational construct
between individuals placed in the role of principal and agent (Eisenhardt, 1989). At the simplest
level, both the principal and the agent have different utility functions that create the
opportunity for variance between the two. The principal seeks to align the agent's functions to
his/her own by presenting incentives and inducements and must monitor the agent to identify
deviation. Conversely, the agent seeks to maximize his/her own utility and desires wide
discretionary authority to do so. As a resultnformation is a critical aspect of principaagent
relationships (Eisenhardt, 1989). The principal seeks information to monitor the activities of the
agent, but this information is not costless, nor does the principal have limitless capacity for
processing information (Simon, 1957). The agent has rational cause to restrict availability of
information to the principal and enjoy the benefits of a deliberate information asymmetry. Both
principal and agent maintain a subjective and boundedly rational perceptive of the costs and
benefits of information disclosure (Stigler, 1980).


26
The literature on the costs and benefits of secrecy and transparency suggests some
general suppositions can be made using the principal-agent model. In general, principals will
likely overvalue information disclosure and agents will likely overvalue information secrecy. In
organizations with a high degree of cultural secrecy or an institutional bias toward non-
disclosure, the perceived costs of information disclosure might be quite high with relatively low
benefits. Conversely, because of the relative lack of information about agent activities for secret
organizations, principals might place a premium on information disclosure from these agencies.
Regardless, both principal and agent are likely to base decisions on the perceived, rather than
objective, costs and benefits. This leads to Assumption 1:Principals and agents will select
desired information disclosure functions so that the perceived benefits exceed the perceived
costs.
Though not directly testable, this assumption suggests that individual decisions likely
play a significant role in determining the quantity and quality of transparency provided by an
agency. Moreover, these decisions are likely affected by potentially competing goals and
institutional influences. In short, administrative discretion is a key element in secrecy and
transparency programs and bears examination.
Discretion and its Application to Secrecy and Transparency Decisions
The competing conditions of secrecy and transparency in organizations suggest that
FOIA administrators are confronted with potentially conflicting policy objectives which
necessitate the exercise of considerable discretion. Understanding discretion within the specific
context of secrecy and transparency decision-making and rule sets directly informs
methodological approaches for addressing the research questions by suggesting observable
indicators of decision-making.


27
At the heart of the administrative discretion condition is a potential for variance
between policy objectives and program outcomes caused by boundedly rational humans
operating in a complex social context. Work on discretion by Lipsky (1980), Riccucci (2005) and
Sandfort (2000) shows that policy inputs do not necessarily translate into outcomes as intended,
a condition expected to be present in FOIA programs as well. In the case of FOI policy for
national security organizations, the problem is compounded by the competing guidance to both
conceal and disclose and the potential for value conflict experienced by the individual
administrator.
The condition of administrative discretion is rooted in the organization as a social
construct consisting of individuals. Viewed through the lens of principal-agent theory, discretion
may be seen as a bundle of principal-agent situations. At the macro-level, government
organizations are agents to multiple principals, at once beholden to the chief executive as chief
administrator, to the legislature as representatives of the people and originators of policy and to
the people themselves as both ultimate shareholders and ultimate beneficiaries of services
(Finer, 1941; Mosher, 1968). In the event that these principals conflict, the organization faces an
ordering problem to determine which set of responsibilities take precedence. At the micro-level,
principal-agent dynamics manifest themselves between individuals as managers and workers. In
both macro and micro cases, the principal faces monitoring and enforcement costs which
prohibit the principal from obtaining perfect information as to the agent's behaviors and restrict
the principal's ability to ensure perfect compliance with directives. The theory suggests that
agents (both in individual and organizational forms) will have different preferences from their
principals and can be expected to make decisions which deviate from the principaTs optimal
decisions in favor of obtaining some advantage on the part of the agent (Maynard-Moody &
Musheno, 2000). For organizations, this might mean seeking more control of resources or


28
information as Niskanen (1968) suggests, or expressing different values, as Riccucci's (2005)
work indicates. Prat's (2006) argument that additional disclosure cannot harm the principal, yet
results in potential costs for the agent leads to the conclusion that agents will attempt to
minimize costs by disclosing less or resisting transparency initiatives. In the case of FOIA, this
literature suggests that agencies will adopt imperfect realizations of policy, opting for more
secrecy than desired by their multiple principals. However, principal-agent theory is limited in
perspective and does not account for the social influences on the agent or the possibility of
socially beneficial discretion (Maynard-Moody & Musheno, 2000). In practice, both principal and
agent are embedded in a social construct (i.e. teams, groups and organizations) that produces
informal rules, establishes collective norms and ultimately influences administration discretion.
FOIA administrators in the modern federal bureaucracy operate in a complex
environment with the potential for confusing and conflicting rule sets. Bozeman's (1993; 2000)
work on Red Tape indicates that administrators respond to the complexity by adding
endogenously created rule sets on top of externally imposed requirements and guidance. This
creates further complexity for individuals in the workplace. These rule sets constrain the
individual exercise of discretion, particularly by workers and street-level bureaucrats who may
be forced to adhere to sequences or protocols that serve contrary purposes (Maynard-Moody &
Musheno, 2000; Riccucci, 2005; Scott, 1997). The effect is felt by organizations as well as
individuals. Red Tape Theory indicates that organizations may become overwhelmed by the
density of rules and the compliance burden they impose (Bozeman, 1993). Dealing with these
rule sets often requires mandatory training, compliance monitoring and other costs, diverting
resources from the organizations' primary mission (Bozeman, 2000; Riccucci, 2005; Sandfort,
2000). Similar to having multiple principals, a sufficient density of rules can result in conflicting
rule sets, presenting managers and workers with a discretionary choice of which rules to adhere


29
to and which rules merit disregard or deferred compliance (Maynard-Moody & Musheno, 2000;
Sandfort, 2000). In addition to the formal rule sets, workers develop short-cut practices and
tools to assist in dealing with day-to-day tasks without referencing statute or formal procedures.
These informal constructs form a second phalanx of rules and procedures that offer the
opportunity for discretion to be exercised. In the case of FOI, agency and department level rule
sets have the potential to influence FOI outcomes in ways unintended by policy as codified in
law.
Organization norms may also influence discretion. Organizations are social constructs,
and as such, work is a social occasion. Work by Riccucci (2005), Sandfort (2000), Sowa and
Selden (2003) shows that elements of organizational culture, such as shared values, collective
norms and perception of goals and priorities influence individual worker behavior. While
organizational rules may indicate the range of actions that are technically permitted or
restricted, organizational norms influence what actions are socially acceptable, what behaviors
are informally rewarded and what attitudes are expected. These norms help workers make
sense of complex cases and scenarios in a way that statute or formal direction cannot (Sandfort,
2000). This social process helps workers establish collective perceptions of their work role and
'real' goals.
Simmel and Weber suggest that secrecy can have a significant impact on social
processes in organizations, leading to more secrecy. Building off of Simmel's model of secrets as
a social construct, Anand and Rosen (2008) and Keane (2008) find that secrets can be embedded
in and reinforce organizational culture. Keane (2008) suggests that secrets can become
symbolically important for organizations and increase the social cohesion of the organization by
providing a clear boundary. Anand and Rosen (2008) make a distinction between who has


30
sanctioned a secret, arguing that secrets sanctioned by organizational insiders are more likely to
transgress ethical boundaries that secrets sanctioned by societal norms or law. Additionally, this
process may lead to acculturation of unethical behavior or excessive secret-keeping (Anand &
Rosen, 2008). The literature on Britain's Official Secrets Act leads to the conclusion that Britain's
government developed an unreasonable amount of secrecy for a liberal democracy due the
acculturation of secrecy as a value (Bennett, 1985; Christoph, 1975; Galnoor, 1975; Griffith,
1989).
Expanding the principal-agent dyad to the perspective of relations between
organizations rather than individuals provides a frame for viewing the relationships between
federal agencies and their stakeholders through the lens of agency theory. Both principal and
agent will create institutional relationships and mechanisms to achieve their goals, the former to
obtain relevant low-cost information about the agent, and the latter to restrict information
disclosure. In organizations, both principal and agent are affected by institutional and social
forces that shape the desired selection of disclosure functions. The secrecy literature indicates
that secrecy can exert a powerful psychological force and creates social pressure to maintain
collective secrets. Informal norms (Sandfort, 2000) and endogenously created rules (Bozeman,
1993; 2000) influence the exercise of individual administrator discretion. For example, federal
agencies may selectively use a combination of ofncial and administrative secrecy rules to justify
disclosure decisions. This leads to Assumption 2: In organizations, the disclosure function of the
principal-agent dyad is affected by institutional forces. These institutional forces are likely to
shape both program-level and discrete outcomes and offer the potential for creating
unintended effects and variance from policy goals.


31
While social norms do help to reduce individual variation within a given organization
through an acculturation process which inculcates shared norms and hence priorities, the
nature of norm formation indicates that significant organization to organization variation may
still be present, suggesting additional analysis is needed to understand variations between
organizations (Sandfort, 2000). The implication from the literature is that secrecy can become
inculcated in organizational norms and some organizations will be more secretive than others,
opening up the possibility to developing a scale of secrecy or transparency which might inform
the principal research question.
In sum, the literature on discretion suggests that understanding of FOI in practice
requires a deeper look at specific agency rule sets and cultural factors that might influence
individual and organizational discretion in the execution of agency FOI programs. Taken with the
literature on secrecy as a phenomenon and instrumental transparency, this set of literature
suggests that the effects of secrecy and transparency institutional influences should be
observable in individual decisions as well as program outcomes, as individual administrators
make disposition choices as influenced by organizational norms and internal rule-sets.
Additionally, this literature leads to the conclusion that variance might be observable not just in
disposition decisions, but also in more nuanced instrumental choices such as the choice of
phrasing or form of response, suggesting a qualitative component to transparency research is
warranted.
Conclusions from the Literature
Collectively, the literature helps illuminate the topic of FOI in practice, pointing to
several areas for examination and research. The literature on secrecy indicates that withholding
information is linked to organizational power dynamics and can be a significant force in


32
organizational culture. This directly informs the principal research question How does secrecy
in government organizations impact execution of transparency initiatives? and suggests that
research is likely to discover meaningful differences between secret agencies and their less-
secretive counterparts. The transparency literature highlights the role of institutional rules and
individual discretion on the part of agents as influences on transparency outcomes. Viewing
transparency as an instrument leads to the need to examine not just the coarse outputs of
transparency programs, but also some of the more nuanced implementation aspects of
transparency programs. Literature on secrecy in practice indicates that the costs and benefits of
secrecy and transparency are important factors to consider, as heuristics are used in day-to-day
decision making. Scholarship on administrative discretion leads to the conclusion that
examination of agency internal rules sets is warranted, as the possibility for conflicting guidance
is substantial. Additionally, the discretion literature suggests that social norms play a significant
role in individual decision-making and offer a possible explanation for systematic variances
between organizations.
A major conclusion from the literature is that there is a strong potential for over-
production of secrecy, particularly in government organizations (Vaughn, 1997/2009; Weber,
1920/2009). This overproduction is caused in part to administrative discretion decision, such as
over-classification (Roberts, 2006; Rourke, 1960; Rowat, 1965) but also in part due to the
compartmentalization that the bureaucracy uses as an organizing technique which creates intra-
firm organizational secrecy (Dufresne & Offstein, 2008; Weber, 1920/2009). In some cases this
allocative inefficiency can be significant, increasing program costs by 25% (Aftergood,
1999/2009). In the case of FOI programs, over-production of secrecy can have important
consequences for democratic values and citizen participation by diminishing transparency
precisely on those agencies that need it most. This conclusion provides both impetus for


33
additional public administration research and directionality for hypotheses regarding secrecy
and its impacts on FOIA programs.
In sum, the literature provides the broad strokes for understanding the transparency
dynamic. Stakeholders acting as principals desire information about organizations as agents
through a medium. The literature suggests that secret-keeping can impact both the principal
and agent, chiefly through social norms and constructs. However, the literature makes it clear
that our collective understanding of secrecy is limited, particularly in how secrecy regimes and
practices impact transparency outcomes and mechanisms. This research addresses that gap by
exploring how the US intelligence community as a specific instance of secrecy impacts the
transparency outcomes outlined in the FOIA.


34
CHAPTER III
FREEDOM OF INFORMATION: RIGHTS AND RESEARCH
Introduction
The literature summarized in the previous chapter provides a conceptual basis for
understanding secrecy and transparency. To understand how these concepts are applied in
practice in the case of FOIA, a deeper look is needed. This chapter begins with a brief glance at
the historical origins of the Act to establish the intent of the Act. An analysis of the institutional
grammar of the contemporary version of the Act and current executive orders helps shed light
on how policy intent aligns to specific actions. Additionally, this analysis reveals actors and
measures for potential observation and comparison. The chapter concludes with a review of
contemporary research and implications for this research effort.
Freedom of Information as a Redress for Government Secrecy
The Freedom of Information Act of 1966 emerged as a result of a decades-long struggle
by the press and sympathetic legislators to peel back the veil of government secrecy described
as a '"paper curtaiiV" (Archibald, 1979; Rourke, 1960). Executive secrecy had grown considerably
during and following World War II, causing concern amongst legislators, journalists and
scientists (Rourke, 1960). Archibald describes a portion of the letter to Representative John
Moss creating a Special Sub-committee on Government Information as a "battle call to tear
down the walls of government secrecy" (1979, p. 313). The text of this paragraph succinctly
frames the problem that the FOIA would eventually be enacted to address:
An informed public makes the difference between mob rule and democratic
government. If the pertinent and necessary information on governmental activities is


35
denied [to] the public, the result is weakening of the democratic process and the
ultimate atrophy of our form of government. (Archibald, 1979, p. 313)
The Moss commission began a ten-year effort to examine the extent of secrecy in the
federal government and propose changes. Unsurprisingly, the commission found an over-
abundance of secrecy, both official and administrative. The root of executive branch secrecy laid
not in any explicit secrecy law, as in Britain's Official Secrets Act, but in housekeeping laws and
the 1946 Administrative Procedures Act which afforded agencies and departments sufficient
discretion to conceal if "good cause" warranted (Archibald, 1979; Rourke, 1960). Given this
leeway, administrators and bureaucrats performed as the Weber (1920/2009) anticipated, over-
producing secrecy and concealing far more than one would anticipate in the public interest. The
commission highlighted a few egregious examples for effect. "The Moss subcommittee disclosed
that the Pentagon had stamped 'secret' on the bow and arrow and had also classified the fact
that water runs downhilT (Archibald, 1979, p. 314).
After more than a decade of what Archibald describes as a public relations campaign,
the Moss subcommittee secured the passage of a deliberately weakened FOIA in 19bb, with an
effective date of July 4,1967. Executive departments, despite testifying on the benefits of full
disclosure, resisted the spirit of the act with vigor, developing internal rule sets and mechanisms
to defeat the purpose of the law (Archibald, 1979; Relyea, 1975; Roberts, 1979). In some cases,
agencies began to delay responding to the requests, creating secrecy by delay, in other cases
they charged fees intended to make completion of the request cost prohibitive by the requester
(Archibald, 1979). In response, The Congress overrode President Ford's veto to pass
amendments to the FOIA in 1974, correcting these deficiencies and making the law more
enforceable by both the executive and judicial branches (Relyea, 1979). Though the act would


36
see additional modifications in subsequent decades, the 1974 version established the essential
rule set in force today.
Substantial updates were made again in 1996 to bring the law in line with the digital age
with the passage of the Electronic Freedom of Information Act (E-FOIA) which added
requirements pertaining to the agencies posting and receiving information via electronic means.
Additional updates were made in 2007 to incorporate practices established by executive order
and add regulatory language to further define agency FOIA activities. In the 2007 update,
subtitled "Openness promotes effectiveness in our national (OPEN) government act of 2007'
Congress reaffirmed the intent of FOIA, linking constitutional democracy to informed consent
and citing Supreme Court rulings (OPEN, 2007). The finding portion of the act reiterated "the
presumption in favor of disclosure" as a preamble to modifying code ''based not upon the 'need
to know' but upon the fundamental 'right to know"' (OPEN007).
FOIA Establishes Public and Government Rights to Information
The Freedom of Information Act is instantiated in law as Section 552 of Title 5,
Government Organization and Employees, of the United States Code (5 U.S.C. §552). Containing
both constitutive and regulatory clauses, the Act establishes the mechanisms for citizens to
request information and enumerates the responsibilities of agencies in responding to those
requests. Broadly speaking, the Act has four complementary aims. First, in order to increase
transparency, it contains some requirements for agencies to make a minimal set of information
available to the public. Second, the Act details administrative procedures and required
conditions for agencies to appropriately service information requests. Third, the Act establishes
the government's right to withhold information in nine enumerated areas (5 U.S.C. §552 (b))
and in the case of the intelligence community, compels those agencies to withhold records from


37
foreign government entities (5 U.S.C. §552 (a)(3)(E)). Fourth, the Act contains performance and
reporting requirements for agencies to ensure appropriate congressional and executive
oversight of agency FOIA programs.
The Institutional Grammar of FOIA
To examine how the intelligence agencies implement the Act, a thorough examination
of the text of the legislation is needed, particularly in comparison or contrast to executive
instructions that also have a potential to impact agency discretion. The Institutional Grammar
Tool provides a methodology for analyzing institutional statements in the law and producing
insights to help frame research efforts. In particular, use of the institutional grammar approach
allows for the identification of clauses requiring action or compliance from specific actors. Of
particular interest for this research effort are institutional statements involving federal agencies
that provide standards for performance or compliance that can be used to create testable
hypotheses and observations for research.
The Act contains both definitive clauses, to establish the meaning of particular
elements, and institutional statements, providing instructions for execution. The coding method
proposed by Basurto et al. (2010) provides a baseline for evaluating the syntax of regulatory
statements using Crawford and Ostrom's (1995) five components: Attribute (A), Deontic (D),
Aim (I), Condition (C) and Or Else (O). These five components indicate what actions are required,
permitted or forbidden by which actors for what purpose under what conditions and with what
penalties for non-compliance/non-performance. A modification proposed by Siddiki, Weible,
Basurto and Calanni (2012) includes an object (B) code to identify animate and inanimate
targets of legislation and is roughly analogous to the direct object of sentences. This
modification reduces ambiguity in attribute coding and makes it easier to distinguish between


38
two statements that have identical attribute, deontic and aim elements but differing objects.
The entirety of Title 5, Section 552 U.S.C. was coded using Basurto et al/s (2010) methodology
as modified by Siddiki, Weible, Basurto and Calanni (2012).
Coding of the section yielded 98 observable units of institutional syntax. Statements
containing the five ADICO elements are categorized as rules, statements with the first four
(ADIC) are norms and those without a Deontic are categorized as strategies. All 98 units of
observation within the FOIA section were categorized as norms, meaning that all institutional
statements including a deontic, but not an "Or else" statement.
Most statements pertained to the agency or agency officials (The/Each Agency, Head
of Agency, and Chief FOIA Officer). Seventy of the 98 institutional statements are directed at the
agency or agency official level, reflecting an administrative law emphasis (See Table 3.1). Other
actors include the US District Courts (n=10), the Attorney General (n=7) and Special Counsel
(n=4). The 2007 amendments established the Office of Government Information Services (OGIS)
within the National Archives and Records Administration and created obligatory statements for
OGIS (n=3) and the General Accounting Office (GAO) (n=2).
Table 3.1
Institutional Grammar Analysis Results 5 U.S.C. 552
Deontic
Actor Obligatory Permissive Restrictive Total
Agency 47 8 8 63
The Courts 3 6 1 10
Attorney General 6 1 0 7
Chief FOIA Officer 6 0 0 6
Special Counsel 4 0 0 4
OGIS 3 0 0 3
GAO 2 0 0 2
Perscrn 1 0 1 2
Head of Agency 1 0 0 1
Total 73 15 10 98


39
Over seventy-four percent of the statements were obligatory, using the language
shall' in most cases indicating actions that agencies were required to take. The syntax
combination of "Each agency shall" or variations thereof, account for over half (47 of 98)
statements. A minor share of statements were permissive (15 of 98), offering actors the latitude
to take an optional course of action. Of these permissive statements, 6 of 15 were directed at
the courts, offering latitude for the courts to enforce rulings. Only 10 of 98 statements were
restrictive, and 8 of these 10 were aimed at limiting the actions of agencies.
A Closer Look at the 20-Day Rule Needed
Despite the law's overall intent to provide citizens with the right to request information,
the individual appears in only two statements, and in neither case is the syntax permissive. The
first statement is in awkwardly worded restrictive syntax which indicates that:
Except to the extent that a person has actual and timely notice of the terms thereof, a
person may not in any manner be required to resort to, or be adversely affected by, a
matter required to be published in the Federal Register and not so published. (FOIA,
2009)
The second statement involving an individual person requires that individuals requesting
expedited processing show compelling need. As written, the law does not expressly permit
individuals to submit requests, but rather obligates agencies to receive them. The presumption
of disclosure, while stated clearly in congressional findings, is not strongly represented in
regulatory statements. Instead, the bulk of the burden for transparency falls on the citizenry to
request rather than the agencies to provide.


40
Though the law establishes a number of temporal conditions designed to prevent
secrecy by delay and compel agencies to perfect requests within 20 working days, the section
contains no explicit "Or Else" clauses indicating a penalty for non-compliance. Additionally,
although agencies are obliged to report the number of requests determined within 20 days,
there is no syntax compelling agencies to report their overall compliance with this standard. The
information specified in the annual reports provides only an incomplete benchmark of
compliance, making detection of non-compliance challenging. Moreover, the law does not
require agencies to report how often they invoked a delay or for what purposes, leaving open
the possibility of abuse of the "unusual circumstances" clause contained in 5 U.S.C. §552
(a)(6)(B)(i) which allows agencies to delay perfecting the request if unusual circumstances are
present. This gap in the legislation suggests that a more detailed look at agencies performance
under the 20-day standard is warranted and may yield useful insights.
Executive Order 13392 Supplements Legislation
The FOIA is not only the direction that agencies have received for execution of freedom
of information. Periodically, Presidents issue executive orders aimed at improving FOIA
programs in federal agencies. Executive Order 13392 ''Improving Agency Disclosure of
Information" was issued by President Bush in 2005 as a directive to federal agencies to improve
information disclosure consistent with NPM concepts. The order establishes FOIA as an essential
service of the government and directs agencies to establish service centers and public liaisons
for handling requests. Additionally, the order required agencies to submit and post plans
containing milestones for reducing the backlog of unanswered requests and improving the
quality of FOIA service. The order further directed agencies to conduct follow-up analysis of
those milestones and issue reports on milestone progress and deficiency correction.


41
The Institutional Grammar of EO 13392
The text of this order was also coded using the same method described above to
identify institutional statements of interest. In contrast to the FOIA, EO 13392 contains both
norms (n=60) and strategies (n=7). Five of the seven strategies can be found in the first few
paragraphs of the order and express a desire for a well-informed citizenry and articulate how
agencies should view FOIA requesters as customers of the agencies and respond accordingly.
Like the FOIA, the norms in the Executive Order are primarily obligatory statements
directed at the agencies or agency officials. Forty-nine of 60 norms fit this description (see Table
3.2). The order emphasizes the role of the Chief FOIA Officer, FOIA Public Liaisons, and FOIA
Request Service Centers; directing 30 statements at these three actors. This emphasis on this
structure removes some discretion from agencies and directs a degree of uniformity in
administrative practices. The order also establishes new reporting criteria, requiring the Chief
FOIA Officer to review agency practices and issue a report to the Attorney General. This report
may be a valuable source of information on how the agencies perceive their FOIA programs and
their ability to service the public.


42
Table 3.2
Institutional Grammar Analysis Results E013392
Actor Obligatory Deontic Permissive Restrictive Strategy Total
Chief FOIA Officer 21 0 0 0 21
Agency 12 0 0 0 12
Head of Agency 9 0 0 0 9
FOIA Public Liaisons 5 2 0 0 7
Public 1 0 0 5 6
Attorney General 4 1 0 0 5
All 1 0 1 0 2
President 0 0 0 2 2
FOIA Requester Service Center 2 0 0 0 2
OMB Director 0 1 0 0 1
Total 55 4 1 7 67
Shared Rules, Similar Syntax
The syntax between the FOIA and EO 13392 has a number of similarities. Neither source
contains 'Or Else' clauses, leaving the penalties for non-compliance unstated. Six statements
match exactly between the two texts, as phrasing from the 2005 Executive Order was
incorporated into the 2007 update for FOIA. Neither document contains much in the way of
permissive statements for FOIA requesters, instead burdening the agencies with compliance. In
fact, as many executive orders do, E013392 closes with a statement that the order does not
create a right or benefit enforceable in law against the government. As a consequence, though
the order is clearly intended to improve the quality of FOIA servicing by federal agencies, access
to information falls short of being enshrined as a right.
The order specifies some additional reporting criteria for the agencies, including
narrative reports and plans from Chief FOIA Officers that address agency performance. These
plans offer a potential source of insight into agency issues with providing transparency and
servicing FOIA requests. The strategies included in the order are also significant. In the first few


43
paragraphs, the order specifies that FOIA is to be treated as a service. This framing highlights the
need to examine FOIA outcomes through a service lens, looking not just at requests granted or
denied and records released or withheld, but also examining service performance issues like
timeliness and quality of response. This suggests a qualitative coding approach to agency
responses and interactions might yield useful insights.
Institutional Grammar Conclusion
Analysis of 5 U.S.C. §552 and EO 13392 using the Institutional Grammar Tool proved
useful to this research effort. The coded syntax of institutional statements provides a baseline
for determining agency compliance with codified norms. Additionally, analysis of the FOIA text
leads to the conclusion that agency performance against the 20-day rule bears additional
scrutiny and may be a key measure of performance. Analysis of the executive order suggests
that service quality issues may also be worthy of examination. The literature suggests that the
secret agencies will exhibit meaningful differences in their FOIA programs compared to their
counterparts. Analysis using the institutional grammar coding suggests that examining
differences in service timeliness and quality may also yield fruitful results.
Contemporary FOIA Research
Contemporary research on the US FOIA has explored several hypotheses related to the
effects of the administration and the effects of the NPM movement on FOIA execution by
federal agencies. This dissertation builds upon this contemporary research base to explore new
territory.


44
Problems with FOIA: The National Security Archive at George Washington University
George Washington University's (GWU) The National Security Archive (TNSA) performs
monitoring of and selected research on FOIA results from across the federal government. Since
2003, the archive has periodically conducted the Knight Open Government Survey to query
selected agencies on their oldest pending FOIA requests, and, since 2006, compares those
results to the information contained in agency annual reports (TNSA, 2003; 2010; 2011). The
survey information indicates that backlog of FOIA requests is a sizeable problem, particularly in
agencies with a national security mission, and some requests remain open after more than a
decade of 'work' by the agency (TNSA, 2003; 2010; 2011). TNSA deduces that the referral
process is partially to blame for delayed requests, as the receiving agency refers some or all of
the request for records to another agency, adding processing delays and significantly
complicating the records retrieval and declassification process (TNSA, 2003; 2010). Although
useful as a potential indicator of problem areas, the methodology of the survey does not appear
sufficiently stable or rigorous enough to draw firm empirical conclusions about agency
performance nor were the conclusions subject to peer review.
In addition to the Knight survey, TNSA conducted target research to examine the effects
of changes in administration on FOIA implementation. Within the Federal Government, the
Department of Justice, headed by the Attorney General, is responsible for the oversight of the
FOIA program by executive agencies and departments. However, FOIA execution is primarily
decentralized, leaving considerable opportunity for discretion in interpreting the letter of the
FOIA as written. In order to assist agencies and establish clear policy, the Attorney General
issues guidance to the executive branch on how to apply FOIA, consistent with the policy goals
of the President. During the Clinton administration, Attorney General Janet Reno issued


45
guidance aimed at improving transparency to the effect that agencies should presume
disclosure to the maximum extent possible (Kim, 2007; TNSA, 2003). In 2002, under the George
W. Bush administration, John Ashcroft issued a memorandum reversing this presumption and
encouraging agencies to protect unclassified, but potentially sensitive information from
disclosure by using the exemptions established by FOIA (Kim, 2007; TNSA, 2003). In January of
2009, President Obama rescinded this guidance in place of direction similar to the Clinton
administration, encouraging disclosure (WhiteHouse, 2009). TNSA conducted some analysis of
annual reports and interviewed some agency FOIA personnel to assess whether or not these
memoranda had substantive effects on federal government FOIA programs (TNSA, 2003; 2010).
Additionally, TNSA work revealed inconsistencies in how agencies promulgated executive
guidance. Consistent with the administrative discretion literature, this conclusion suggests the
possibility that agency FOIA processes and rules might vary significantly from established policy.
The work done by TNSA points to the unmistakable conclusion that FOIA in the federal
government is not functioning as intended, with exceptional delays in processing and
incomplete documentation approaching pathology. TNSA work reveals that 'secrecy by delay'
was not eliminated by the 1974 amendments, but is, instead, a persistent problem found in
contemporary agency FOIA programs, although the extent and impact of agency backlogs is not
immediately apparent. This dissertation will examine backlog data as a possible point of
disparity between intelligence agencies and their less secretive peers.
Effects of Administration Change on FOIA outcomes: Kim (2007)
Like TNSA, Kim (2007) also selected the effect of a change in administration as a
research area, though with more rigorous methodology based on data from annual FOIA reports
rather than documentation released under request. Kim evaluates the claim that the Ashcroft


46
memo produced changes in agency disclosure patterns using data available in FOIA annual
reports. Kim selected 25 agencies, including the 16 departments and 9 major independent
agencies, including the CIA, using data from annual reports from 1998 to 2005. Kim found that
agencies were less effective in processing requests under the Bush administration and backlogs
increased. Additionally, Kim found that full grant rates had indeed decreased during the Bush
administration and that the use of exemptions had increased, indicating the Bush executive
guidance had indeed made the government more secretive compared to the Clinton
Administration.
Kim's methodology is sound, but is not readily repeatable due to reporting changes in
2008. Indeed, Kim's work indicates that comparing agency outcomes across administration
boundaries (i.e. 2004 vs 2010) may include unwanted variation due to administration change.
This suggests that using data from only a single administration (i.e. the Obama Presidency, 2008-
present) may help focus only on variables of interest and avoid problems with comparing rule
sets under multiple administrations. However, Kim's core approach, using annual report data
provides an exemplar for using this source of data to draw empirically sound conclusions about
FOIA programs.
The Effects of New Public Management on FOIA Programs: Piotrowski and Roberts
As a reform philosophy, New Public Management (NPM) has had the potential to impact
a wide range of government programs and functions. Piotrowski (2003; 2007) and Roberts
(2000; 2002; 2006) independently examined the effects of New Public Management (NPM) on
FOIA programs in the federal government, finding that FOI programs were adversely affected by
the reform movement. Roberts concluded that because the reforms included paring non-
essential functions and spending, FOI programs were a potential target of cost-savings


47
measures. Roberts found evidence of reduced effectiveness in Canadian FOI programs (2002)
and that NPM had weakened the ability of citizens to monitor public institutions (2002),
particularly weakly organized sectors (2000). Additionally, Roberts found national security
networks to be of particular concern, stating that 'Transparency within the network is matched
opacity without [the network]" (2006, p.139).
Piotrowski adopted a case study approach, examining the effects of the National
Performance Review (NPR) on FOIA. While NPR did not focus specifically on FOIA, Piotrowski
(2007) found unintended consequences of the reform activity on FOIA outcomes. Piotrowski
found that FOI was often left out of performance management objectives (2003) and emphasis
by agencies was on cost-effectiveness rather than customer effectiveness (2007). Additionally,
Piotrowski (2007) found variance in how FOIA programs perceived the need for customer
service, and whether or not FOIA officers viewed requesters as agency customers. Given the
NPM emphasis on contracting, Piotrowski raises the question of how contractor records
produced on behalf of the government, but not under government control are potentially out of
reach for FOIA (2007). Additionally, Piotrowski found agencies applying administrative
procedures to subvert the intent of FOIA by designating records as 'pre-decisionaT (2007, p.90).
Piotrowksi's and Roberts' work suggests that FOIA researchers should examine agency
administrative procedures and look at contract resources in the FOIA offices as variables of
interest.
Implications for FOIA research
Analysis of the institutional grammar of the FOIA and recent research provides a point
of departure for examining FOI programs in the intelligence community. The syntax of the law
indicates that compliance with the 20-day response rule bears examination. A careful look at


48
E013392 suggests that researchers should examine not just response outcomes, but also service
issues, like timeliness and quality. Georgetown's TNSA work shows that progress against
backlogs is an important measure of the effectiveness of FOIA programs and should be
considered in evaluating program health. Additionally, TNSA work reveals that the way agencies
respond (or don't respond) to requests is a source of information as potentially valuable as the
records returned, leading to the conclusion that an experimental methodology might yield
valuable information. Kim's (2007) work lays the basis for quantitative assessment of
disposition rates across time periods and between agencies. Piotrowski (2003; 2007) uses
qualitative methods, particularly in examining performance measures, which might also be
applied to analysis of responses and agency planning materials to generate insights on
organizational norms. Roberts' (2002) conclusion that agencies consider FOI programs to be
non-mission-essential might be evaluated by examining the effects of alterations in program
staffing mix on FOIA program performance.
Absent from this contemporary research is any detailed analysis of how the intelligence
community performs with respect to FOIA. Additionally, although discretion is clearly present as
a factor in FOIA decision-making, no extant research examines which or how factors affect the
exercise of discretion in FOIA programs. Although the literature clearly frames FOI as a secrecy
problem and some scholars identify the national security environment as an exemplar of
secretive culture (Roberts, 2006; Shils, 1956), no research can be found that examines how FOI
execution differs by national security agencies versus other agencies. These gaps point to the
need for research on FOI in practice, particularly with respect to America's secret agencies and
the exercise of administrative discretion.


49
CHAPTER IV
CONCEPT DEVELOPMENT AND HYPOTHESES
Introduction
This chapter connects the literature and research summarized in the Chapters II and III
to the research methodology in Chapter V by presenting a model of transparency to be used for
developing operational definitions. The literature referenced in Chapter II on secrecy,
transparency and discretion provides the conceptual basis needed to convert principal-agent
assumptions into guiding propositions for research. The extant research in Chapter III on FOIA
helps identify operational indicators that can be used to develop testable hypotheses. This
chapter offers a description of selected hypotheses and a rationale for selection.
Transparency Model
The literature provides the essential elements needed to construct a micro-level model
of transparency. Transparency can be described as a mediated interchange between secret-
keeper/agent and stakeholder/principal (see Figure 4.1). One or more actors, acting as principal,
desires information about the availability and workings of another actor in order to monitor the
latter's performance (Eisenhardt, 1989; Meijer, 2013). The organization, acting as agent, has
cause to moderate and meter the availability of information about its activities to its principals.
Individuals in secret-keeping organizations are subject to institutional influences, including
formal and informal rules (Bozeman, 1993; 2000), social norms (Sandfort, 2000) and collective
biases (Dufresne & Offstein, 2008). Similarly, stakeholders, either collectively or individually,
may hold a different appreciation of the value of the information than the secret-keeper, even
without a specific knowledge of the information (Prat, 2006; Stigler, 1980).


50
The observation of the agent by the principal is rarely direct, but rather occurs through a
medium (Meijer, 2009). In the case of the federal government, this medium is largely through
official records (Piotrowski, 2007; Roberts, 2006). The interchange is mediated in terms of both
content (the what) (Pasquier & Villeneuve, 2007; Piotrowski, 2007; Roberts, 2006) and
mechanism/method (the how) (Fung, Weil & Graham, 2007; Meijer, 2009). The literature
suggests that secrecy restricts the availability of information, acting as a barrier between
principal and agent (Weber, 1920/2009). These elements can be placed together in simplified
model of transparency (see Figure 4.1) that illustrates the relationship between observed and
observig actors.
Secrecy
~^ Medium
Agent
Figure 4.1
Conceptual Model of Transparency
Rules Norms
Observed Actor
Observing Actor
Principal
This simple model belies a set of complex interactions and potentially observable
indicators. This research applies this model as a basis for understanding US intelligence agency
FOIA outcomes and contributes towards renning this model. In particular, this research helps
refine understanding of how secrecy affects transparency outcomes. The following section
describes how the key conceptual elements of the model will be operationalized for study.


51
Refining Key Concepts into Operational Definitions
Operationalizing Secrecy
The work of Weber, Simmel, Shils and Bok provides the conceptual basis for secrecy. As
the phenomenon of intentionally concealing information, secrecy has several direct and indirect
influences on information release. Secrecy is institutionalized by formal policies and rule sets to
classify and restrict dissemination of categories and classes of information, but it can also create
social influences on shared norms with the broad potential to impact information release
beyond formally classified information (Simmel, 1906; Weber, 1920/2009). These social
influences, in turn, directly influence the exercise of discretion by individual administrators.
Detecting the presence of secrecy and distinguishing between degrees of secrecy can be
difficult, however.
The concept of official secrecy, i.e. state condoned or required secrecy, is difficult to
operationalize, as there are secrets about secrecy. As an example, NARA's Information Security
Oversight Office compiles an annual report to the President containing an estimate of the cost
of the federal government's security classification system. However, the data provided by the
principal intelligence agencies is classified and excluded from the public report. Similarly, the
budget (including the total size) for individual intelligence agencies is considered classified and
rarely released and only then as an aggregate number representing multiple agencies. The
exclusion of this type of data makes the task of operationalizing secrecy difficult at the
organizational level. Nonetheless, an alternative way to measure the concept of secrecy is by
selecting the core membership of the US intelligence community as the test population for
being presumably more secretive than other domestic civil agencies. This assumption is critical


52
to the definition of hypotheses as well as selection of research methods as it calls for a
comparative approach to both.
Testing Transparency
Using Meijer's (2013) instrumental definition, transparency has three operational
elements: the mechanism or means for the principal to compel or request information, the
compulsory or voluntary disclosure of information by the agent and the medium through which
information is disclosed. FOIA provides a portion of the means for requesting information and
nominal guidelines for disclosure decisions, but the agencies are substantially left to their
discretion to select the medium and make a host of decisions influencing disclosure. These
decisions result in indicators observable at the request level by examining individual request
responses and dispositions and at the program level by examining aggregate program results.
The contemporary work on FOIA and analysis of the institutional grammar of the law
helps identify potential operational indicators which can be used to test hypotheses. Comparing
these indicators, using the intelligence agencies as the subject group and a sample of non-
intelligence agencies as the control group, provides a relative method of comparing agency
transparency. Program grant and denial rates provide one measure of relative transparency,
although these are likely affected by additional variables. Examining quality aspects such as
timeliness of response and backlog rates provides another measure of transparency as service.
Many of these indicators derived from analysis of 5 U.S.C. §552 and EO 13392 can be combined
to form a relative scale of FOIA responsiveness.
In sum, for federal agencies both the 'what' of transparency, e.g. records released, and
the 'how' of transparency, e.g. timeliness, medium, method, can be observed and used for


53
evaluation. Some observations can be captured by examining individual request responses,
while others can be gathered by looking at program-level aggregate data. Both data sources are
useful in painting a fulsome portrait of agency transparency.
Dealing with Discretion
Discretion manifests itself in disposition decisions (e.g. fully grant, partially grant) and
timeliness of response, but can also be observed in secondary decisions such as the assignment
of fee category, determination of what artifacts constitute a record and the emphasis of rights
individual response letters. As the literature indicates, individual discretionary decisions are
affected by the institutional context of the decision-maker. Agency internal rule-sets, which are
typically not widely disseminated, can be obtained via FOIA request and provide a partial
glimpse into the complex system of internal rules affecting discretion. While FOIA processing
rules and directions (such as guidelines for multi-track processing or use of exemptions) can be
expected to have a direct effect on FOIA decisions and outcomes, other agency-level rules such
as record management schedules may also exert influence, perhaps unintentionally.
Collectively, these individual decisions have the ability to affect the agency's overall
transparency posture. This research aims to examine some of the aspects affecting discretion
and individual decision-makers by examining selected agency policies.
Hypotheses
The assumptions identified in Chapter II help influence development of guiding
propositions that connect the research questions to operationalevel hypotheses. The
conclusion from the literature that intelligence agencies are likely to over-protect unclassified
information provides directionality for most hypotheses. If over-production of secrecy exists as
suggested; it should be observable in both program outcomes and individual decisions,


54
suggesting a multi-part research design to address research questions. The hypotheses below
are grouped by research question and foreshadow selection of the supporting research design.
As indicated above, secrecy is operationalized by selecting a group of intelligence
agencies as the subject group and non-intelligence agencies as the control group.
Evaluating Program-Level Outcome Differences
The first research question How do Intelligence Agency FOIA program processes and
outcomes differ from their less-secretive counterparts? explores the assumption that the
institutional influence of secrecy will result in observables differences in program-level
outcomes. The core presumption is that the intelligence agencies, acting as agent in a principal-
agent relationship, will select disclosure practices that are more restrictive than their less-
secretive counterparts. This function is not directly observable, however, but may be inferred by
observing program outcomes. From the literature and extant FOIA research, it can be presumed
that intelligence agencies would have lower information disclosure rates; delay disclosure
decisions more and be more secretive about information disclosure. These presumptions are
expressed as guiding propositions. The review of contemporary research in Chapter III provides
operational indicators to convert these guiding propositions into testable hypotheses as shown
in Table 4.1.


55
Table 4.1
Research Questiortl Concept Summary
Research Question Guiding Propositions
How do Intelligence
Agency FOIA program
processes and outcomes
differ from their less-
secretive counterparts?
Intelligence agencies will
restrict information disclosure
more than other agencies
Intelligence agencies will delay
disclosure more than other
agencies
Intelligence agencies will be
more secretive about
information disclosure
Operational-level Hypotheses
HI: Intelligence agencies will
have a lower percentage of fully
granted requests
H2: Intelligence agencies will
exhibit higher partial grant and
admin denial rates
H3: Intelligence agencies will
have longer mean processing
times and higher backlogs
H4: Intelligence agencies will be
less transparent about their FOIA
program
All four hypotheses are comparative in nature, with expected variation between
intelligence agencies and other non-intelligence agencies. These four hypotheses evaluate the
variation at the program outcome level, i.e. the aggregate effects of multiple requests processed
by the agencies' FOIA programs. A detailed description and justification of each hypothesis
follows.
HI: Intelligence agencies' FOIA programs will have a lower percentage of fully granted
requests than other agencies, even factoring for exemption 1 and exemption 3 denials.
Simmel and Weber suggest that secrecy exerts a powerful sociological force on
organizations, effectively creating a barrier between keepers of secrets and those wanting to
discover information. If this is correct, then agencies in which secrecy is a normalized
organizational practice would be expected to use a variety of means to deter information
seekers, resulting in lower than average full grant rates. At least by this one measure, these
agencies would be considered less transparent than their non-secretive peers. To account for
the possibility that requesters have asked for classified information, Exemption 1(national


56
security information) and Exemption 3 (other statutorily exempted information) denials can be
accounted for as control variables, presumably leaving only requests for unclassified information
for comparison.
H2: Intelligence agencies will use partial grants and administrative denials to protect
more information than other agencies.
Among the means used by agencies is the ability to partially grant or use administrative
denials to effectively deny information requests without appearing to rely on authorized
exemptions. This hypothesis builds on HI by exploring the means of denial other than denial by
exemption. This type of denial may be preferable to agencies over a denial by exemption, which
can be easier to challenge through the appeals process or judicial review.
H3: Intelligence agencies' FOIA Programs will have longer mean processing times and
higher backlog rates than other agencies.
During the early history of FOIA, many agencies used 'secrecy by delay' as a means of
complying with the letter of the law, while resisting the spirit (Archibald, 1979). Work by TNSA
(2003; 2010) suggests that this is still a problem area. Extending the assumption that secret
agencies would use a variety of means to deter information seekers, this hypothesis tests the
statement that secret agencies would use 'secrecy by delay' as a tactic. Evaluating this
hypothesis will require accounting for program resources and request volumes to isolate secrecy
the variable of interest.
H4: Intelligence agencies will be less transparent about their FOIA program than peer
agencies.


57
One of the potential variations between intelligence agencies and their peers is how
transparent they are about transparency. The principal-agent logic assumes that intelligence
agencies will perceive the costs of transparency to be high and the benefits to be low. Although
this maybe evident in a number of areas, examining disclosure of information about
transparency programs provides a useful comparison between agencies.
Evaluating Individual Request Disposition Variances
The second research question How do Intelligence Agency disposition decisions differ
from their less-secretive counterparts explores the secrecy-transparency problem at the
individual request level. Comparing program grant and denial rates (as proposed for hypotheses
1 and 2) relies on the assumption that the requests submitted to different agencies are similar
enough to produce different outcomes based on variations between the programs. This
assumption is suspect, at best, and difficult to verify. An experimental research design, where
the requests are identical to different agencies, might more closely reveal variations in
outcomes due to agency internal dynamics including norms and rule sets. The same set of
principal-agent assumptions used for the first research question apply here and lead to similar
guiding propositions as shown in Table 4.2 with the addition of a service-related guiding
proposition. Recalling Meijer's instrumental definition of transparency which led to the
conclusion that the 'how' of transparency might be as significant as the 'what' and the direction
contained in E013392 for agencies to consider FOIA a key service provided by the government,
exploring differences in how agencies provide FOIA as a service is appropriate. The literature
supports the presumption that the FOIA services offered by intelligence agencies will differ
markedly from the FOIA services provided by non-intelligence agencies. Table 4.2 below
presents the three hypotheses related to evaluating individual requests.


58
Table 4.2
Research Question 2 Concept Summary
Research Question Guiding Propositions
Intelligence agencies will
restrict information disclosure
How do Intelligence
Agency disposition
decisions differ from their Intelligence agencies will delay
less-secretive disclosure
counterparts?
Intelligence agencies will not
offer high-quality FOIA services
Operational-level Hypotheses
H5: Intelligence agencies will
issue more denials and less full
disclosures in response to
requests than the control agency
set
H6: Intelligence agencies will be
more likely to issue delays and
take longer to make release
decisions than the control group
H7: Intelligence agencies will
perform worse on service quality
indicators than the control
group.____________________________
H5: Intelligence agencies will issue more denials and less full disclosures in response to
requests than the control agency set
Comparable to hypotheses HI and H2, this hypothesis tests the premise than secret
agencies will restrict the release of unclassified information more than non-secret agencies. By
controlling for request variations, the experimental results should help isolate the effects of
secrecy in practice in these agencies. The primary indicator is the disposition status and records
released.
H6: Intelligence agencies will be more likely to issue delays and take longer to make
information release decisions than the control agency set.
This hypothesis tests the 'secrecy by delay' premise again. Intelligence agencies are
expected to take longer to respond with a final decision to the experimental requests. This
hypothesis supports hypothesis H3, which tests the effect at a program outcome level, by
controlling for variation in the substance of the requests themselves.


59
H7: Intelligence agencies will perform worse on service quality indicators than the
control group.
E013392 raises the interesting question of how well agencies are providing FOIA as a
service. Extending the principal-agent logic to service provisioning, it is expected that
intelligence agencies would be less likely to prioritize FOIA as service as a key function. Some
service aspects, such as time-to-respond can be easily evaluated. Other aspects, such as
response quality, are more complicated, but can be evaluated by coding responses received.
Service quality has the potential to vary request-by-request and should be evaluated using
individual requests rather than program-level indicators.
Exploring Institutional Rules
By testing Hypotheses 1-7 the effects of secrecy can be evaluated, but not necessarily
the causal mechanisms. The third research question seeks to understand the intervening
influences of agency rule-sets and norms on FOIA programs. Red Tape theory (Bozeman, 1993;
2000) and the discretion literature (Maynard-Moody & Musheno, 2000; Sandfort, 2000) indicate
that organizations will develop internal rule sets that may result in different effects than
intended by policy. Cultural conditions and rules that deviate from established FOIA policy are
expected to be present, but have the potential to influence disposition either positively (more
information released) or negatively (less information).
For the first two research questions, the literature and existing FOIA research supported
a clear connection between guiding principles and operational indicators that allowed for
hypotheses generation. This connection is more tenuous for the third research question,
however, as a connection between specific agency policies and FOIA programs is not well


60
represented either in the literature or in previous research. Some logical inferences can be
made, however, based on the dependency on FOI execution on official records and the
exclusion of classified information. In the case of the former, what an agency designates as a
record directly impacts what the FOIA offices can release to requesters. More indirectly, an
agency's policy for handling unclassified information has the potential to impact records release.
These inferences merit additional exploration. As Table 4.3 indicates, this leads to exploratory
questions to further develop our understanding of how these specific agency policies relate to
FOIA programs.
Table 4.3
Research Question 3 Concept Summary
Research Question Guiding Propositions Exploratory Questions
How do FOIA related rules differ between intelligence agencies and their less-secretive counterparts? FOIA programs are influenced by Records Management policies FOIA programs are influenced by CUI policies EQ1: How do Intelligence agency records management policies differ from non-intelligence agencies? EQ2: How do intelligence agency CUI policies differ from non- intelligence agencies?
EQ1: How do intelligence agency records management policies differ from non-
intelligence agencies?
This question examines the possibility of influence by a specific rule set namely,
records management rules. Agencies maintain record management instructions, directives or
policies to define what constitutes a record and direct employees on storage and management
of records and records disposition schedules that indicate how long certain records should be
kept. Not all of these policies or directions are commonly available via agency web-sites, limiting
the amount of a priori information available. Exploration is needed to uncover potential areas of


61
difference in polices and records management schedules between intelligence agencies and
non-intelligence agencies.
EQ2: How do intelligence agency Controlled Unclassified Information policies differ from
non-intelligence agencies?
Information in the federal government may be broadly binned into three major
categories: Classified National Security Information, governed by E013526, unclassified
information and Controlled Unclassified Information (CUI) governed by E013556. This former
category is a clear representation of official secrecy and explicitly exempt from disclosure under
FOIA (although information may be declassified under certain circumstances). The middle
category may be broadly construed as either information that has yet to be classified (literally
un-classified) or information designated for public release. The latter category is a gray area
between the two and has a special relationship with FOIA. According to Department of Justice
and NARA guidance, designation of a record as CUI does not automatically exempt it from
disclosure under FOIA. However, it is unclear whether individual agencies and administrators
adhere to this guidance. The expectation from the literature is that intelligence agencies will use
a CUI policy to restrict release of information.
Even more than records management policies, agency implementation of E013556 and
CUI rules is something of a black box. In fact, it cannot be readily determined whether or not
agencies even have a CUI policy or use CUI designations. NARA's report (NARA, 2011) to the
President on CUI leads to the impression that CUI categories are widely used throughout the
federal government, but efforts to create a comprehensive catalog of markings and the agencies
that use them is incomplete. Additional exploration is needed on this topic to identify areas of
interest for potential research.


62
CHAPTER V
RESEARCH DESIGN
Overview
This chapter presents a research design for addressing the seven hypotheses and two
exploratory questions. The principal design for this research consists of three parts, a
quantitative portion drawn from annual report data designed to evaluate FOIA program
outcomes, an experimental portion to evaluate differences in agency responses to discrete
requests and a qualitative portion to code results from the experiment and information from
agencies websites. The three part approach addresses different levels of questions, investigating
program-outcome level and individual request level questions. In both cases, the unit of analysis
is the agency, with the five core intelligence agencies (described below) and the Office of the
Director of National Intelligence (ODNI) as the subject population representing secretiveness as
concept.
Hypotheses 1-3, which focus on FOIA program outcomes, will be addressed by
examining data self-reported by the agencies in their annual reports. This data set contains
program-level fields such as the number of fully granted or partially granted requests that can
be normalized into rates for between-agency comparison. Hypothesis 4, which examines agency
transparency about FOIA programs, requires between-agency comparisons of a different sort,
examining information available on agency web sites.
Evaluating hypotheses 5-7, which explore disposition decisions, requires a methodology
that targets the individual request level. An experimental approach in which the researcher
sends identically worded requests to both subject and control agencies will generate
information not available in the annual reports. The first-order data gathered includes easily


63
measurable information, such as disposition result (i.e. granted, denied) and time to return a
disposition. The experiment will also yield second-order qualitative information such as the
phrasing of request responses and content of informal exchanges between the experimenter
and FOIA liaisons. Additionally, the experimental requests are expected to yield records that can
be used for evaluating exploratory questions 1 and 2 related to FOIA-related rules. Table 5.1
below provides a summary of the principal evaluation method in relationship to hypotheses.
Table 5.1
Research Design Summary
Research Question Area
RQ1: How do Intelligence
Agency FOIA program
processes and outcomes differ
from their less-secretive
counterparts?
RQ2: How do Intelligence
Agency disposition decisions
differ from their less-secretive
counterparts?
RQ3: How do FOIA related rules
differ between intelligence
agencies and their less-
secretive counterparts?
Hypotheses
HI: Intelligence agencies will have a
lower percentage of fully granted
requests
H2: Intelligence agencies will exhibit
higher partial grant and admin denial
rates
H3: Intelligence agencies will have longer
mean processing times and higher
backlogs
H4: Intelligence agencies will be less
transparent about their FOIA program
H5: Intelligence agencies will issue more
denials and less full disclosures in
response to requests than the control
agency set
H6: Intelligence agencies will be more
likely to issue delays and take longer to
make release decisions than the control
group
H7: Intelligence agencies will perform
worse on service quality indicators than
the control group.
EQ1: How do Intelligence agency records
management policies differ from non-
intelligence agencies?
EQ2: How do intelligence agency CUI
policies differ from non-intelligence
agencies?
Principal Method
Quantitative
Quantitative
Quantitative
Qualitative
Experiment
Experiment
Experiment
Qualitative
Qualitative


64
Annual Report Data used to Evaluate Program Outcomes
Data Source FOIA Annual Reports
The data for quantitative analysis of hypotheses 1-3, evaluating program outcomes,
come from annual reports submitted by federal agencies. The FOIA requires agencies to provide
to Congress each year a report containing basic, aggregate counts on the disposition of FOIA
requests and backlog information. The reports contain data from independent agencies, some
quasi-governmental agencies, and components or subordinate agencies for all 16 federal
departments. In the latter case, Department-by-Department variation exists in which elements
submit reports. In some cases, such as the Department of Defense, the reporting components
are subordinate operating agencies while other departments use administrative structures (such
as regional designations or a central office) to report data.
For each fiscal year, each agency or reporting component with a FOIA program reports
the number of FOIA requests pending at the beginning of the year, the number received, the
number processed, the number fully granted, the number partially granted, the number denied
due to exemption, by exemption category (1 through 9) and the number denied due to
administrative reasons, by cause. The reports are a rich source of raw information, providing
request and disposition volumes, exemption and administrative denial counts, and
quality/efficiency measures such as backlog volume, median processing delays, and FOIA
program resources (see http://www.iustice.gov/oip/reports.html for examples). The reports
provide information used to assess program outcomes, but do not shed light on individual
requests.


65
Since 2008, agency and component annual report data have been compiled on
www.foia.gov, a website administered by the Department of Justice to offer key FOIA
information in a centralized location. This database provides organization-year observations
from 423 reporting components, which can be used to construct a profile of federal government
programs in aggregate for comparative purposes. This source offers a complete data set, with all
reported fields available.
Data prior to 2008 cannot be easily compared to post-2008 data (M. Kim, personal
communication, March, 29, 2013). Prior to 2008, some agencies reported aggregate values for
FOIA requests and Privacy Act requests together. Some agencies separated the values, but this
was not consistently done. In FY2008 and beyond, the reports pertain to the FOIA program only.
Additionally, individual agency disclosure and posting of FOIA annual reports prior to 2008 is
inconsistent and incomplete. For this reason, FY2008 was selected as the starting period for
annual report data.
The foia.gov record set consists of 1788 organization-year dyads from 423 different
reporting components for fiscal years FY2008 through FY2012. The data are partially
unbalanced, as some entities did not report FY2008 data (the first year of data) and some
departments reorganized during the data period, eliminating agencies or components. To avoid
double-counting, aggregate department values have been excluded (i.e. department-level
totals). The data have been cleaned to remove null observation rows (N=9) and eliminate
department totals. The resultant data set consists of 1715 observations (organization-year
dyads) from over 400 reporting components for most of a five year period.
The primary observations of interest consist of organization-year dyads from the Central
Intelligence Agency (CIA), Defense Intelligence Agency (DIA), the National Geospatial-


66
Intelligence Agency (NGA), the National Security Agency (NSA), and the National Reconnaissance
Office (NRO) plus the Office of the Director of National Intelligence (ODNI). The first five
agencies described are typically considered the 'core' of the intelligence community and have
statutorily specified intelligence gathering and analysis functions. Although nominally a
headquarters organization, the ODNI acts as the center of the intelligence community and
operates some multi-agency operational functions, such as the National Counter-Proliferation
Center and the National Counter-Terrorism Center. As suchhese six agencies are presumed to
have internalized secrecy more than peer organizations, owing to their unique intelligence
mission and are the operationalization of the 'secret agency' set.
The other 417 reporting components in the data set serve as the comparison group.
Though some of these components have national defense or homeland security functions, the
set should sufficiently represent the 'non-secret' agency pool.
Variables Used
The annual report data collected from www.foia.^ov contains a number continuous and
discrete variables for each agency-year pair. Dummy categorical variables were created to
represent Intelligence Community components. The key explanatory variable is a dummy
variable representing the six intelligence agencies described above.
Six dependent variables were created to test the hypotheses. Table 5.2 summarizes the
dependent variables selected. To test the premise of Hypothesis 1 that intelligence agencies
grant fewer and deny more requests than their non-intelligence peers, two ratio variables were
created to serve as dependent variables. The first, Released-in-Full Ratio, is the percentage of
requests processed that were returned as fully granted or released-in-full, indicating that the


67
agency determined that they had satisfied this request. Higher values represent more
transparency, at least in the form of requests categorized as granted. The second ratio variable,
the Denied-in-Full Ratio, is the percentage of requests processed that were denied by asserting
the rights of the government to protect the record of interest using a FOIA or other statutory
exemption. Higher denied-in-full ratios means more information protected by the agency and
less information released to requesters. These two measures are conceptually diametrically
opposed, the first a measure of transparency, the second an assertion of official secrecy.
Converting raw request counts into ratios allows for normalization between programs of
different sizes and simplified reporting of coefficients.
The dependent variables for Hypothesis 2 test the premise that intelligence agencies
may make use of partial grants and administrative denials as a way of restricting information
release. Partial grants represent selective release of information by partially fulfilling the
request. For example, if several records are requested in a single request, the agency might
consider the request partially granted if one or more records were released (e.g. documents
from a single year) or if a portion of a record were released (even if the rest of the document
was redacted). The annual reports reveal very little about this disposition category and leave
much open to interpretation. As such, the partial grants may also be viewed as partial denials
and represent a decision by the agencies to restrict disclosure. A Partial Grant Rate ratio variable
was defined as the percentage of requests processed that were returned as partially granted.
Administrative denials consist of discretionary rejections that are not based on
exemptions. In this case, the government is not asserting its right to protect information (as it is
when invoking an exemption), but the agency has selected another reason for not granting the
request. Typical categories of administrative denials include "No Records'"Request


68
Withdrawn" and "Improper FOIA Request' among others. Viewed as a discretionary disposition
choice, the secrecy literature suggests that administrative denials would be used by intelligence
agencies as a tool to limit information disclosure. Administrative denials are counted differently
than other denials and may be perceived differently by program administrators than a denial
using an exemption. An Administrative Denial Rate ratio variable was defined as the percentage
of requests processed that were returned as partially granted.
Hypothesis 3 is temporal in nature, testing the assumption that intelligence agencies
may practice secrecy-by-delay. Two potential indicators of this assumption can be constructed
from the annual report data. The first indicator is the size of the agency's FOIA request backlog,
defined as a ratio of the number of requests not completed to the number of requests received.
Backlog, the amount of requests agencies carry over from year to year, has been a source of
concern for administrations seeking to improve transparency (TNSA, 2003; 2010). Ideally, the
ratio is low, reflecting only requests exceptionally difficult and/or time-consuming to perfect
and requests received late in the fiscal year. In a few cases (n=32), the ratio is greater than one -
indicating that the agency or component carries forward into the next fiscal year a queue of
unperfected requests greater than the number of requests received that year by the agency.
Higher ratios may be indicative of poor program performance or secrecy-by-delay.
The second indicator is a continuous variable representing the average time taken by
agencies to complete requests. Agencies do not directly report this value. Instead, the data
contain mean values for completed simple requests and complex requests separately. However,
a ratio of simple to complex requests can be calculated allowing an approximate mean to be
constructed. The Mean Request Duration dependent variable is this approximate mean duration
in business days taken to disposition requests. The work by George Washington University


69
suggests a relationship between the time to perfect requests and their disposition, with the
implication that longer processing times were likely correlated to increased denials (TNSA, 2003;
2010).
Table 5.2
Dependent Variable Definitions
Concept Variable Name
Relative
transparency
Non-disclosure
Non-disclosure
Released in Full
Ratio
Denied in Full
Ratio
Partial Grant Ratio
Non-disclosure
Admin Denial
Ratio
Secrecy by delay Backlog Ratio
Secrecy by delay
Mean Request
Duration
Definition Hypothesis
Ratio of requests fully
granted (disclosed) to ^
total requests
dispositioned
Ratio of requests withheld
using exemption to total 1
requests dispositioned
Ratio of requests partially
granted to total requests 2
dispositioned
Ratio of requests not
fulfilled for administrative ^
reasons to total requests
dispositioned
Ratio of requests
backlogged at the end of ^
the FY to requests
received that FY
Mean duration in business
days taken to disposition 3
requests
A number of control variables were used to attempt to isolate the effect of the
explanatory variable (See Table 5.3). To account for the possible influence of labor on the
dependent variables, the Number of Full-Time Employees and Equivalent number of Full-Time
Employees were added as independent variables. The first value captures the number of full-
time employees dedicated to processing FOIA requests. The second value captures non-full time
employees involved in FOIA processing, including part-time employees or full-time personnel


70
who have FOIA responsibilities as a fraction of their job (Justice, 2009). Conceptually, more
employees on the program should expedite processing, resulting in lower Backlog ratios and
shorter mean durations.
Table 5.3
Independent Variable Definitions
Variable Name Definition Use
Intelligence Group Dummy variable for test group Explanatory variable
No. FTE Number of full-time employees employed by FOIA program Control variable
Number of full-time equivalent
No. Equivalent FTE employees employed by FOIA Control variable
program
Ratio of Complex Requests Ratio of complex requests to total requests Control variable
Processing Costs In dollars, agency-reported costs of their FOIA program Control variable
Ratio of agency-reported FOIA-
Litigation Costs related litigation costs to total Control variable
costs
Mean Request Duration Mean duration in business days taken to disposition Control variable
requests
Previous research by George Washington University (2003) suggested that the
complexity of a request could significantly impact both its disposition (i.e. granted, partially
granted, denied) and the amount of time required to perfect the request. The annual report
data reveals little about the nature of requests made to the agencies. One proxy measure
available is the categorization of requests into simple and complex requests by agencies as part
of their multi-track processing rule set, though it appears that there are no standard criteria for
this categorization. Though agencies do not report this figure directly, it may be derived from
other fields in the annual report data. The Complex Request Ratio is defined as the percentage


71
of non-expedited requests designated by agencies as 'Complex'2 to total requests received. This
variable partially captures the concept of differential complexity, that some agencies routinely
receive requests that are more difficult to satisfy than other agencies.
The data contain values for processing and litigation costs incurred by agency FOIA
programs. The value for processing costs is the sum of all labor, technical and administrative
costs incurred by the agency to process requests and administrative appeals (Justice, 2009). This
variable tested high for multicollinearity (likely due to the interaction with the labor variables)
and was dropped from regression models after diagnostics. The value for litigation costs is a
ratio of the costs related to litigating FOIA requests to the total program costs (Justice-
Handbook). This variable captures an aspect of complexity in that higher litigation cost
percentages likely represents increased contestability of agency decisions and may influence
disposition results.
These control variables should help isolate the effects of the explanatory variable by
accounting for the influence of resources and complexity on the dependent variables.
Analytical Techniques
The initial analysis consists of interpretation of descriptive statistics, difference of means
tests and correlations between the explanatory variable and dependent variables. This analysis
provides first-order indicates of the validity of the hypotheses by estimating the size and
significance of the effect of the explanatory variable (intelligence group membership). Stata
software was used to perform all statistical analysis.
2 No discoverable guidance on what constitutes 'complex' may vary from agency to agency.


72
To isolate the effects of the explanatory variable on the dependent variables, the
researcher constructed Ordinary Least Squares (OLS) regression models containing the control
variables. Each model was tested for multicollinearity, heteroskedasticity and influential
outliers. Independent variables with Variance Influence Factor (VIF) scores higher than 4 were
eliminated. Models were run with clustered standard errors to mitigate the effects of
heteroskedascity. After these adjustments and the removal of influential outliers with a dfbeta
score above | .251, each model was re-run. Models with bounded ratios as dependent variables
were subject to an additional treatment and re-run as fractional logit models.
Each hypothesis was evaluated by examining the initial analysis and the regression
model outputs. While high R-squared values for the regression models will provide confidence
that the model adequately captures the variation; the direction, strength and significance of
coefficients, particularly for the explanatory variable, will provide the best indication of the
validity of the hypotheses.
Experimental Requests used to Evaluate Disposition Variances
Experiment Design
While the annual reports provide program outcome data, they do not shed light on how
individual requests are handled, which more directly illuminates the role of administrative
discretion in FOI decision-making. Some agencies produce FOIA logs, detailing individual
requests, but this is not a consistent practice. Within the intelligence community, only the ODNI
and DIA voluntarily disclose logs on their website, and DIA logs do not contain disposition


73
information.3 As a result, there is no source available that provides stable data on individual
requests that can be used for either case study or quantitative analysis. Additionally, comparing
program outcome data as reported in the agency annual reports forces the researcher to make
the assumption that the requests between agencies are comparable. This assumption is not
likely to be true and creates a challenge to the validity of conclusions drawn solely from the
annual report data. Using an experimental approach that holds the requests between agencies
constant can help supplement findings from the quantitative analysis portion of the design. The
experimental approach also directly tests the agencies' FOIA performance in a way that the
annual report data does not, providing insight on individual agency idiosyncrasies and process
variances.
To collect information on the request process, three rounds of experimental requests
were sent to the five core 1C agencies and the ODNI as the subject group and 10 comparable
agencies without an intelligence function and little to no national security function as the
control group (See Table 5.4). The ten control agencies were selected using stratified random
selection from a set of agencies with comparable FOIA programs to produce an equivalent 'peer
group'. Comparable programs were selected using the following parameters: FY2012 FOIA
request volume between 80 and 4119 (+/-10% from subject agencies), FOIA full time staff
greater than or equal to 1,designation as a key component or element of a department on
www.usa.gov and a status other than quasi-official agency or government corporation.
3 ODNI logs are available from 2006 to 2012 and contain request ID, requester name, subject, disposition
and disposition date. DIA logs are available from 2001 to 2010 and contain request ID, requester name
and subject only.


74
Table 5.4
Agency Selection Experimental Portion
Test Group Abbr. Control Group Abbr.
Central Intelligence Agency CIA Transportation Security Administration DHSTSA
Defense Intelligence Agency DIA National Oceanic and Atmospheric Administration DOC NOAA
National Geospatial-Intelligence NGA Defense Contract Management DOD
Agency Agency DCMA
National Reconnaissance Office NRO United States Marshals Service DOJ USMS
National Security Agency NSA Mine Safety and Health Administration DOL MSHA
Office of the Director of National Intelligence ODNI Administration for Children and Families HHS ACF
National Aeronautic and Space Administration NASA
Office of the Comptroller of Currency Treas OCC
Rural Development USDA RD
United States Geological Survey USGS
A three round design provides an opportunity for multiple observation points without
being overly cumbersome to the agencies. These rounds use records returned, disposition
decisions and time-to-return responses as variables for evaluating hypotheses H5: Intelligence
agencies will issue more denials and less full disclosures in response to requests than the control
agency set, H6: Intelligence agencies will be more likely to issue delays and take longer to make
information release decisions than the control agency set and H7: Intelligence agencies will
perform worse on service quality indicators than the control group. Additionally, the records
returned will be used to address exploratory questions EQ1: How do intelligence agency records
management policies differ from non-intelligence agencies? and EQ2: How do intelligence
agency Controlled Unclassified Information policies differ from non-intelligence agencies?
Evaluation of these hypotheses and questions will help formulate an assessment of how


75
transparent agencies are about transparency and how agency rule sets influence disclosure in
practice.
The request rounds were staggered by approximately 45 days to space out the requests
and avoid the appearance of 'shotgunning' requests to agencies (see Table 5.5). Because
responding to requests is a non-trivial action on the part of agencies (the average cost per
request is $637 (Justice, 2012)), the requests were for records that are of inherent value in
addressing the exploratory questions. The first round of requests was released on 25 Sept 2013,
with rounds two and three approximately 45 and 90 days later, respectively. Each experimental
round concluded 100 business days from the request date, allowing roughly five months for
agencies to render a response.
Table 5.5
Experimental Portion Summary
Round Number Request Request Date
1 Records Management Policies and Schedule 25 Sept 2013
2 CUI Policy 12 Nov 2013
3 FOIA Logs 8 Jan 2014
Round 1-Records management policy and schedule. This round requests that
agencies provide their records management program policy documents and their records
disposition schedule. See Appendix A for the full text of the request.
When FOIA was originally enacted, the transactions and decisions of American
government were largely captured in the form of paper records. The access that FOIA provided
to these records offered citizens and stakeholders significant insight into the results, if not the
working processes of government. In the four decades since the original FOIA enactment, the
way in which government business is conducted and documented has changed significantly. The


76
digital revolution facilitated a proliferation of electronic communications that may fail to meet
the threshold of record as originally envisioned by FOIA (Roberts, 2006). Though storage of
electronic documents and files is considerably cheaper than storage of an equivalent amount of
hardcopy documents, storage costs are non-trivial and may be considerable for large agencies.
As a result, not all information is designated as a record, and few records are designated for
permanent storage by the National Archives and Records Administration (NARA).
Informal interviews revealed that conflicting guidance regarding storage/retention of
FOIA records could lead to materials being destroyed. For example, NARA publishes the General
Records Schedule (GRS), a prescriptive guide indicating which records are designated for
permanent storage and how long to retain temporary records. Per the GRS, FOIA annual reports
are designated as permanent records scheduled to be transferred to NARA for safekeeping.
However, both DIA and NGA FOIA officers indicated that their respective agency policies
designated these records as temporary and allowed them to destroy them after a few years.
This indicates that there is the potential for variance in agency-specific records management
policies which may have direct impacts on transparency, as agencies cannot release records
which do not exist or have been destroyed.
Agencies typically publish instructions or directives to their workforce to implement
their records management policy and specify disposition instructions and retention durations
for specific records. Directive documents, if provided, provided insight into directions to
employees for the maintenance and retention of records and were reviewed for their
connection to agency transparency efforts. The records management schedules for each agency,
if provided, were used to identify agency-specific policy variations. Though the records
management schedules and policy documents are presumably unclassified and subject to full


77
release, intelligence agencies are expected to delay and restrict the release of these documents
more than the control group.
Round 2 Controlled Unclassified Information (CUI) policy. This round requests that
agencies provide documents that describe the agency policy or on use of Controlled Unclassified
Information markings such as For Official Use Only (FOUO), Sensitive But Unclassified (SBU) or
Law Enforcement Sensitive (LES). See Appendix A for the full text of the request.
Though not every agency maintains classified information, many agencies potentially
make use of CUI markings to restrict information dissemination. These markings represent a
gray area when it comes to FOIA. In 2011, NARA's Information Security Oversight Office (2011)
identified that there were over 100 agency-specific rule regimes governing CUI in the Federal
government. The 2009 report by the inter-agency Task Force on Controlled Unclassified
Information (CUITF, 2009) indicated that several agencies had used CUI markings incorrectly to
deny release under FOIA. Though Executive Order 13556, issued in 2010, directs agencies not
use CUI markings as a justification for denial of FOIA requests, some agencies may still do so and
may invoke an exemption to withhold documents marked with CUI. This round tests the premise
that intelligence agencies are more likely to make use of internally generated rule sets, like CUI,
to restrict the release of information than agencies in the control set.
Round 3 FOIA logs. This trial round requests that agencies disclose FOIA logs from
FY2008-FY2012. See Appendix A for the full text of the request.
Agency logs will possibly help identify patterns in FOIA requests for a given agency and
the prevalence of requests from different sectors (i.e. academic, commercial, press, etc.).
Additionally, the records requested in this round all have the potential to be exempted from
release using exemption 2, which allows agencies to withhold records they think are solely


78
related to internal practices. This exemption is highly discretionary, however, and variation is
expected between the intelligence agencies and control groups, with the former using
exemption 2 more frequently to deny release of requested records in this round.
Anticipated experiment outcomes. The experiment yielded dozens of individual
responses. The responses to the request are just as significant as any records released. Of
primary importance is whether or not the request was granted, but when and how the agency
responds is also useful data. The time an agency took to respond provided one indicator of how
well the agency is providing FOIA as a service and determined whether or not the agency is
compliant with the requirement to respond to requests within 20 business days. The medium of
the response and any records returned were also viewed as indicators of service. Additionally,
statements in the response were coded to measure service quality dimensions.
In addition to the responses, the experiment is expected to yield some records for
qualitative evaluation. Not all agencies are expected to return records, however, resulting in an
incomplete set.
Experimental Analysis
All responses and correspondence from agencies were logged. For each trial round,
results from the test group, including response times and request disposition, were compared
against results from the control group. The primary analytical technique is the comparison of
results from the test group against results from the control group. Some aspects, like time to
return a response, lend themselves to easy direct comparison, while other aspects, such as
record responses, require a more nuanced interpretation.


79
Hypothesis 5 Intelligence agencies will issue more denials and less full disclosures in
response to requests than the control agency set, will be evaluated by comparing disposition
results between the intelligence agencies and the control group. The intelligence agencies are
expected to issue more denials and partial grants than the control group. Hypothesis 6 -
Intelligence agencies will be more likely to issue delays and take longer to make release decisions
than the control group; will be evaluated by comparing response times between the intelligence
agencies and the control group. Intelligence agencies are expected to have a higher mean
duration and a more frequent use of delay notices than the control group.
The written responses from the agencies contain valuable information, regardless of any
records supplied. Agencies were expected to issue at least one response, but in many cases
released two or three responses to each request. Agencies have broad latitude in determining
how to respond to individual requests, particularly in selecting the medium for communication
and language/phrasing that communicates the agency's FOIA perspective. To address
hypothesis 7 Intelligence agencies will perform worse on service quality indicators than the
control group, responses were coded for a number of quality indicators derived from rules
found in the FOIA text, in EO 13392 and OGIS best practices (See Table 5.6). These indicators
measure a number of aspects of service quality. The first indicator assesses whether or not the
agency provided either an initial or final response within the 20-business-day period required by
the Act. The second indicator assesses the relatively simple requirement to provide requesters
with a tracking number to reference the request. The tracking number is used internal to the
agency for tracking and record keeping purposes. The third and fourth indicators assess the
agency's responsiveness using electronic means. If the agency provided any human-generated
correspondence via e-mail, they received a 'yes' on the third indicator. For the fourth indicator,
agencies received a 'yes' response if they provided to the requester a URL or other electronic


80
means of tracking the request status. The final two indicators assess if the agencies provided
information other than the request disposition in their responses. Agencies received a 'yes' on
the fifth indicator if they provided information related to the request in the response that
supplemented their statement of disposition (granted, denied, released in part). Agencies
received a 'yes' on the sixth and final indicator when they provided a point of contact (other
than an agency mailing address) for requester questions or issues.
Table 5.6
FOIA Service Scale
Indicator No. Question Source
1 Did the agency respond with 20 days? 5 U.S.C552 (a)(6)(i)
2 Did the agency offer information along with the disposition? E013392(l)(b), OGIS Best Practices
3 Did the agency respond electronically? 5 U.S.C552 (a)(3)
4 Did the agency provide a tracking number? 5 U.S.C552 (a)(7)(A)
5 Did the agency provide electronic means to obtain status? 5 U.S.C552 (a)(7)(B)
6 Did the agency provide a point of contact for questions? OGIS Best Practices
The researcher coded responses to requests received in each round against the
questions in the service scale and compared the intelligence agency performance to the control
group. Results of the experiment are detailed in Chapter VII. Each round was analyzed as a
discrete event. The analysis concludes with a summary of key findings from the experiment as
whole.
Agency Records and Artifacts Support Exploratory Questions
Measuring the quantity and quality of information that federal agencies share with the
public is a challenging affair that must be addressed piecemeal. Annual report data provides an


81
aggregate measure of how often requested records are being released, but not necessarily the
quantity of material released or the quality of service provided. The experiment provides a
partial glimpse into both quantity and quality of disclosure, but only in response to requests.
Qualitatively examining other aspects of transparency can help supplement the quantitative and
experimental findings by providing more insight into the instrumental nature of transparency.
Transparency about Transparency
The first research question asks "How do Intelligence Agency FOIA program processes
and outcomes differ from their less-secretive counterparts?" The design for hypotheses HI,H2
and H3 uses a quantitative approach to address this question. Hypothesis H4 "Intelligence
agencies will be less transparent about their FOIA programs than other agencies" requires a
qualitative coding approach. The literature referenced in Chapter II leads to the conclusion that
FOIA programs in government agencies are placed in the special position of being both the
object and instrument of transparency as records might be held by contractors or organizations
within the agency. This role offers an opportunity to compare how transparency agencies are
about their transparency programs, adding another perspective towards answering the first
research question.
The norms in the FOIA text, EO 13392 and best practices documented by the Office of
Government Information Services (OGIS) leads to a set of questions that can help provide a scale
of transparency. This scale (shown in Table 5.7) uses seven indicators to assess the relative
transparency of individual agencies on the topic of transparency. The first two indicators,
derived from the Act and EO 13392 assess whether or not the agency posts the most recent
copy of their annual report and their Chief FOIA Officer report on their public-facing website.
These reports provide the public, press and oversight stakeholders with information about the


82
agency's performance and plans for improvement. Indicators 3 and 4 assess whether or not the
agency maintains an electronic reading room accessible to the public and whether or not that
reading room provides frequently requested records. In an effort to promote proactive
transparency, the Act requires agencies to post certain information and frequently requested
records in an electronic reading room. Indicators 5, 6 and 7 assess whether or not the agencies
provide information that might help requesters submit properly formatted requests for valid
records.
Table 5.7
Transparency about Transparency Scale
Indicator No. Question Source
1 Does the agency make the most recent annual reports available online? 5 U.S.C552 (e)(3)
2 Does the agency make the most recent Chief FOIA Officer reports available online? E013392 (3)(c)(i)
3 Does the agency have an Electronic Reading Room accessible to the public? 5 U.S.C552 (a)(2)
4 Are frequently requested records displayed? 5 U.S.C552 (a)(2)(D)
5 Is the agency policy/regulation for FOIA available online? OGIS best practices
6 Does the agency makes records management policies available online? OGIS best practices
7 Does the agency provide a handbook for requesters? 5 U.S.C552 (g)(3)
The researcher used the same sets of agencies selected for the experimental rounds to
complete the scale using information from the agencies' public web-sites. Agencies were ranked
by the total number of positive answers. The expected direction for the hypothesis is that
intelligence agencies will rank lower on the scale than agencies in the control group.
Records Analysis
Answering EQ1How do Intelligence agency records management policies differ from
non-intelligence agencies? and EQ2 How do intelligence agency CUI policies differ from


83
non-intelligence agencies? requires records returned during the experimental portion.
However, it cannot be assumed that all agencies will return records. Although the records
requested are presumed to be unclassified and eligible for release, this may not be true of every
agency (which is itself, a valuable data point). In fact, the literature suggests that agencies in the
test group are less likely to release information or have a higher probability of delaying release
outside of the experiment period. As a result, analysis of the records is likely to be inconsistent.
Records returned from trial round #1(Records Management Policies and Schedules)
offer several points for evaluation. The researcher compared purpose statements in the policies,
looking for demonstrable differences in policy scope or framing, particularly in terms of how
records management relates to information disclosure. Items of interest in the records
management schedule include retention durations of FOIA related materials, such as annual
reports and FOIA logs. The goal of this analysis is to identify potential areas of examination for
future research.
A similar method was used to evaluate records returned from trial round #2 (CUI Policy).
The relationship between CUI policies and information disclosure in general and FOIA in
particular is unclear. Analyzing records returned in this round helps shed light on potential
aspects of the relationship that merit deeper investigation.
Records returned from trial round #3 (FOIA logs) offer great promise, but also have the
possibility for significant between-agency variation. Agency logs were compared against each
other for relative transparency, examining how much information is revealed by the logs.


84
Threats to Design
Secrecy Operationalization
The research design rests on the assumption that intelligence agencies are indeed more
secretive than non-intelligence agencies. This is a challenging assumption to test. Because of the
organizational structure of the intelligence community, it is impossible to construct a complete
community aggregate set (see Appendix B) as only 7 of the 17 members produce discrete FOIA
data or reports. As such, the set of the five core agencies plus the Office of Director of National
Intelligence provide the best opportunity to observe intelligence agencies as a subject
population.
Measuring Transparency
The research design attempts to assess the concept of transparency in a number of
ways. The design rests heavily on the presumption that responses to FOIA requests can be used
as a partial measure of transparency. Several facets of these responses are captured, examining
not only the 'what' but also the 'how' and 'when' as indicated by Meijer's (2013) instrumental
definition. The FOIA responses capture well elements of what Heald (2006) described as event
transparency. Process transparency is much more difficult to evaluate, but is partially captured
through the indicators on the FOIA service scale. The use of a transparency-about-transparency
scale provides another possible measure of transparency, partially capturing the concept of
proactive transparency. Taken together, these measures provide a defensible representation of
transparency as a concept.


85
Bias and the use of Mixed Methods
FOIA program outcomes, as measured by annual reports from the agencies, likely
include bias from the requesters themselves. That is, some agencies might attract more
frivolous or difficult requests than other agencies, potentially skewing results at the aggregate
level. While this bias is not correctable on a large scale, the use of an experimental design, which
holds the requests constant across agencies, provides some small scale triangulation.
Additionally, accounting for Exemption 1 and Exemption 3 denials offers the possibility of
removing the influence properly classified information from the results, offering a more apt
comparison between agencies.
Each experimental request bears the possibility of idiosyncratic responses from
individual FOIA officers, though the use of multiple rounds provides multiple observation points
and lessens the possibility that a single outlier response will influence the results. The use of
qualitative coding will assist in filling in parts of the picture not provided by quantitative analysis
alone. Taken as a whole, the research design will evaluate program outcomes, individual
disposition responses and a measure of transparency in organizational norms, providing a
fulsome portrait of the agencies' FOIA performance.
Data and Information Accuracy
The research design is heavily dependent on source data self-reported by the agencies.
In the case of the annual reports, the reported values are assumed to be accurate. Responses
returned from experimental requests are assumed to be valid, regardless of their accuracy. For
instance, if an agency responses that no records are available, this is a considered a valid and
useful response, regardless of its factual accuracy. Experimental responses will not be appealed.


86
The experimental nature of this portion involves some unknowns and may result in unexpected
outcomes.
Small-n Evaluation
Selecting only six agencies/offices as representative of the intelligence community
reduces the statistical power of the multivariate analysis. However, the selection is justified
given the impossibility of constructing an intelligence-community wide set. By using multiple
years of data, the small-n problem is ameliorated, though not completely eliminated. The use of
mixed methods provides additional confidence in the results and will help strengthen analytical
co elusions.


87
CHAPTER VI
TESTING HYPOTHESES USING ANNUAL REPORT DATA
Introduction
A portion of the research design calls for comparing intelligence agency FOIA program
performance against other federal agencies (including 'non-agency' reporting components).
Data from annual federal agency reports provided the basis for this comparison. The data
consist of 1715 organization-year dyads from agencies and components across the federal
government from the fiscal years 2008 to 2012. The analysis contained in this chapter uses this
data set to evaluate relationships of interest and draw preliminary conclusions. Specifically, the
data highlight the relationship between intelligence agency membership and aggregate
disposition outcomes (e.g. percentage of requests fully granted, denied or partially granted) as
well as aggregate processing results (e.g. backlog percentages and average time to disposition).


88
Table 6.1
Dependent Variable Overview
Hypotheses Variable Name Definition
HI: Intelligence agencies will have a lower percentage of fully granted requests Released in Full Ratio Denied in Full Ratio Ratio of requests fully granted (disclosed) to total requests dispositioned Ratio of requests withheld using exemption to total
requests dispositioned
Ratio of requests partially
Partial Grant Ratio granted to total requests
H2: Intelligence agencies will dispositioned
exhibit higher partial grant and admin denial rates Admin Denial Ratio Ratio of requests not fulfilled for administrative reasons to total requests dispositioned
Ratio of requests backlogged
H3: Intelligence agencies will have longer mean processing times and higher backlogs Backlog Ratio Mean Request Duration at the end of the FY to requests received that FY Mean duration in business days taken to disposition requests
This chapter begins with a discussion of the descriptive statistics from the data and a
preliminary analysis of means differences and correlations of interest. The chapter continues
with analyses of the multivariate regression models performed for each of the dependent
variables listed in Table 6.1. The chapter concludes with a discussion of the limitations of the
data and a summary of findings.
Descriptive Statistics and Preliminary Analyses
The descriptive statistics from the annual report data provide insight into the federal
government's FOIA performance at a program level. Within the data, there is wide variance
between agencies. Some agencies report receiving no requests, while others receive hundreds
of thousands each year. As a panel set containing organization-year dyads for a five-year period,


Full Text

PAGE 1

AMERICAN SECRETS, AMERICAN TRANSPARENCY: ANALYSIS OF THE FREEDOM OF INFORMATION ACT AS IMPLEMENTED BY THE US INTELLIGENCE COMMUNITY by ROBERT CAMERON WARD HUNT B.A., Boston University, 1997 M.B.A., University of Colorado Denver, 2006 M.S., University of Colorado Denver, 2006 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Doct or of Philosophy Public Affairs Program 2014

PAGE 2

2014 ROBERT CAMERON WARD HUNT ALL RIGHTS RESERVED

PAGE 3

ii This thesis for the Doctor of Philosophy degree by Robert Cameron Ward Hunt has been approved for the Public Affairs Program By Tanya Heikkila Dissertation Chair Brian Gerber Examination Chair Benoy Jacob Suzanne Piotrowski 8 July 2014

PAGE 4

iii Ward Hunt, Robert, Cameron (Ph.D., Public Affairs) American Secrets, American Transparency: Analysis o f the Freedom Of Information Act a s Implemented b y t he U S Intelligence Community Thesis directed by Associate Professor Tanya Heikkila. ABSTRACT The US Freedom of Information Act (FOIA) places ghts to request and discover information in competition with the right of the government to conceal official secrets (Bathory & McWilliams, 1977; Rourke 1957). Simply p ut, FOI in practice pits secrecy versus transparency, with administrative discretion in the middle. Though FOIA applies to all federal government departments and agencies in the United States, the tension becomes most acute when applied to the most secretive participants in the national security enterprise, the 16 member intelligence communit y Secret keeping can exert psychological and social pressure on organizations ( Simmel, 1906; Weber, 1920/2009) that have the potential to impact individual decision making and shape collective norms ( Freidman, Landes & Posner, 1991 ; Keane, 2008; Sandfort, 2000). This effort examines how secrecy impacts transparency initiatives by researching how the US Intelligence Community FOIA programs perform compared to other federal agencies using multivariate analysis of FOIA annual report data in conjunction with a n experimental methodology T he research results show that while intelligence agencies release similar information to control agencies, they differ significantly in the how and when that information is provided. In particular, intelligence agencies take longer to respond than peer agencies and offer less information with the response. These findings contribute to theory by leading to a refined model of transparency and contribute to practice by supporting recomme ndations for policy makers and FOIA program administrators. The form and content of this abstract are approved. I recommend its publication. Approved: Tanya Heikkila

PAGE 5

iv DEDICATION This work is dedicated to my loving wife, Cassi, who inspired, encouraged and supported me throughout this process. It is also dedicated to my son, Cavan, with hope for a future of more open government.

PAGE 6

v ACKNOWLEDGEMENTS I give m y sincere thanks to Drs. Brian Gerber, Benoy Jacob, and Suzanne Piotrowski for their support and guidan ce. I owe special acknowledgment and deep gratitude to Dr. Tanya Heikkila for agreeing to serve as my committee chair, reviewing innumerable drafts and generously giving her time to support me. This effort would have not been possible without the financia l support given by the Post 9/11 GI Bill and the flexibility provided by my employer. Thank you all.

PAGE 7

vi TABLE OF CONTENTS CHAPTER I. INTRODUCTION ................................ ................................ ................................ ................................ .......... 1 Problem of Interest and its Importance ................................ ................................ ............. 1 Piecing together the Secrecy and Transparency Puzzle ................................ ..................... 3 Freedom of Information Program Outcomes ................................ ................................ ..... 4 Discretion in Request Dispositions ................................ ................................ ..................... 6 Institutional Transparency Rules ................................ ................................ ........................ 7 Dissertation Outline ................................ ................................ ................................ ............ 8 II. LITERATURE REVIE W ................................ ................................ ................................ .................... 9 Introduction ................................ ................................ ................................ ........................ 9 Secrecy as a Phenomenon ................................ ................................ ................................ 10 Transparency as both Intrinsic and Instrumental Value ................................ ................... 15 Secrecy and Transparency Decisions: Costs and Benefits ................................ ................ 22 Discretion and its Applic ation to Secrecy and Transparency Decisions ........................... 26 Conclusions from the Literature ................................ ................................ ....................... 31 III. FREEDOM OF INFO RMATION: RIGHTS AND RESEARCH ................................ ............................ 34 Introduction ................................ ................................ ................................ ...................... 34 Freedom of Information as a Redress for Government Secrecy ................................ ...... 34 FOIA Establishes Public and Governm ent Rights to Information ................................ ..... 36 The Institutional Grammar of FOIA ................................ ................................ ..... 37 A C loser Look at the 20 Day Rule Needed ................................ ........................... 39 Executive Order 13392 Supplements Legislation ................................ ................ 40 Shared Rules, Similar Syntax ................................ ................................ ................ 42 Institutional Grammar Conclusion ................................ ................................ ....... 43

PAGE 8

vii Contemporary FOIA Research ................................ ................................ .......................... 43 Problems with FOIA: The National Security Archive at George Washington University ................................ ................................ ................................ ............. 44 Effects of Ad ministration Change on FOIA outcomes: Kim (2007) ...................... 45 The Effects of New Public Management on FOIA Programs: Piotrowski and Roberts ................................ ................................ ................................ ................. 46 Implications for FOIA research ................................ ................................ ......................... 47 IV. CONCEPT DEVELOPM ENT AND HYPOTHESES ................................ ................................ ........... 49 Introduction ................................ ................................ ................................ ...................... 49 Transparency Model ................................ ................................ ................................ ......... 49 Refining Key Concepts into Operational Definitions ................................ ........................ 51 Operationalizing Secrecy ................................ ................................ ..................... 51 Testing Transparency ................................ ................................ ........................... 52 Dealing with Discretion ................................ ................................ ........................ 53 Hypotheses ................................ ................................ ................................ ....................... 53 Evaluating Program Level Outcome Differences ................................ ................. 54 Evaluating Individual Request Disposition Variances ................................ .......... 57 Exploring Institutional Rules ................................ ................................ ................ 59 V. RESEARCH DESIGN ................................ ................................ ................................ ..................... 62 Overview ................................ ................................ ................................ ........................... 62 Annual Report Data used to Evaluate Program Outcomes ................................ .............. 64 Data Source FOIA Annual Reports ................................ ................................ .... 64 Variables Used ................................ ................................ ................................ ..... 66 Analytical Techniques ................................ ................................ .......................... 71

PAGE 9

viii Experimental Requests used to Evaluate Disposition Variances ................................ ...... 72 Experiment Design ................................ ................................ ............................... 72 Experimental Analysis ................................ ................................ .......................... 78 Agency Records and Artifacts Support Exploratory Questions ................................ ........ 80 Threats to Design ................................ ................................ ................................ .............. 84 Secrecy Operationalization ................................ ................................ .................. 84 Measuring Transparency ................................ ................................ ..................... 84 Bias and the use of Mixed Methods ................................ ................................ .... 85 Data and Information Accuracy ................................ ................................ ........... 85 Small n Evaluation ................................ ................................ ............................... 86 VI. TESTING HYPOTHES ES USING ANNUAL REPO RT DATA ................................ ............................ 87 Introduction ................................ ................................ ................................ ...................... 87 Descriptive Statistics and Preliminary Analyses ................................ ............................... 88 Analyses for Hypothesis 1 Intelligence Agencies Grant Fewer Requests ...................... 92 Intelligence Agencies Exhibit Lower Released in Full Rates ................................ 93 Intelligence Agencies Exhibit Highe r Denied in Full Rates ................................ ... 95 Discussion and Assessment ................................ ................................ ................. 97 Analyses for Hy pothesis 2 Intelligence Agencies use Partial Grants and Administrative Denials ................................ ................................ ................................ ............................... 98 Analysis Released in Part as a Dependent Variable ................................ .......... 98 Analysis Administrative Denials as a Dependent Variable ............................... 99 Discussion and Assessment ................................ ................................ ............... 100 Analyses for Hypothesis 3 Intelligence Agencies have Higher Backlog and Longer Response Times ................................ ................................ ................................ .............. 101

PAGE 10

ix Intelligence Agencies exhibit Higher Backlog Ratios ................................ ......... 101 Intelligence Agencies show Longer Mean Du rations ................................ ......... 104 Discussion and Assessment ................................ ................................ ............... 106 Chapter Summary ................................ ................................ ................................ ........... 107 VII. EXPERIMENTAL RE SULTS ................................ ................................ ................................ ....... 109 Overview ................................ ................................ ................................ ......................... 109 Round 1 Results ................................ ................................ ................................ .............. 110 Objective ................................ ................................ ................................ ............ 110 Response Summary ................................ ................................ ........................... 111 Service Summary ................................ ................................ ............................... 114 Material Summary ................................ ................................ ............................. 116 Round 2 Results ................................ ................................ ................................ .............. 119 Objective ................................ ................................ ................................ ............ 119 Response Summary ................................ ................................ ........................... 119 Service Summary ................................ ................................ ............................... 122 Materials Summary ................................ ................................ ............................ 124 Round 3 Results ................................ ................................ ................................ .............. 125 Objective ................................ ................................ ................................ ............ 125 Response Summary ................................ ................................ ........................... 125 Service Summary ................................ ................................ ............................... 128 Materials Summary ................................ ................................ ............................ 129 Discussion and Key Findings ................................ ................................ ........................... 130 Hypothesis 5 Little Difference in Disposition Outcomes ................................ 130 Hypothesis 6 Delays by Intelligence Agencies ................................ ................ 132

PAGE 11

x Hypothesis 7 Intelligence Agencies lag on some Service Indicators ............... 134 Records Shed Some Light o n Exploratory Questions ................................ ......... 138 Summary of Experimental Round ................................ ................................ ................... 139 VIII. S UMMARY OF FINDINGS A ND RECOMMENDATIONS ................................ .......................... 141 Introduction ................................ ................................ ................................ .................... 141 Summary of Findings ................................ ................................ ................................ ...... 142 Analysis of Annual Report Data ................................ ................................ ......... 142 Comparing Transpare ncy about Transparency ................................ .................. 144 Comparing Experimental Results ................................ ................................ ....... 146 Implica tions for Theory ................................ ................................ ................................ ... 148 Recommendations for Practice ................................ ................................ ...................... 152 Policy Recommen dations ................................ ................................ ................... 153 Recommendations for FOIA Administrators ................................ ...................... 155 Summary ................................ ................................ ................................ ......................... 157 IX. CONCLUSION ................................ ................................ ................................ ........................... 159 Introduction ................................ ................................ ................................ .................... 159 Contribution to Theory ................................ ................................ ................................ ... 159 Contribution to Methodology ................................ ................................ ......................... 161 Contribution to Practice ................................ ................................ ................................ .. 162 Limitations ................................ ................................ ................................ ...................... 163 Rec ommendations for Future Research ................................ ................................ ......... 165 Deepening Research in Secrecy and Transparency ................................ ........... 166 An Expanded Agenda for Secrecy and Transparency Research ......................... 168 Conclusion ................................ ................................ ................................ ....................... 170

PAGE 12

xi REFERENCES ................................ ................................ ................................ ................................ 171 APPENDIX A ................................ ................................ ................................ ................................ .. 178 APPENDIX B ................................ ................................ ................................ ................................ .. 181

PAGE 13

1 CHAPTER I INTRODUCTION Problem of Interest and its Importance The application of the Freedom of Information Act (FOIA) in the United States presents an interesting and critical public administration probl em. The US FOIA was designed to promote transparency by pro viding citizens with the right to request records f rom Federal Government agencies. of democratic societies an d successful implementation of freedom of information is a hallmark of open governments (Galnoor, 1977; Openness, 2007; Piotrowski & Ro senbloom, 2003). However, the right of citizens to request and discover is in potential competition with the right of the government to conceal official secrets (Bathory & McWilliams, 1977; Rourke 1957) and may be impaired by bureaucratic patterns of administrative secrecy (Weber, 1920/2009) In between these competing rights is the implied right of discretion exercised by executive departments and agencies (Bathory & 1957). Put simply, FOI in practice pits secrecy against transparency, with administrative discretion in the middle. Though FOIA applies to all federal government departments and agencies in the United States, the tension becomes most acute when applied to the most secretive participants in the national security enterprise, the 1 7 member intelligence community, consisting of six primary intelligence agencies and eleven sub agency intelligence components (Roberts, 2006, p. 48). Compelled to both protect national security information and comply with the letter (if not intent) of the FOIA, these agencies are presented with a dilemma that cannot be resolved through policy alone. In short, they face a problem of discretion impacted by the institutional organizational and administrative effects of secrecy This

PAGE 14

2 condition leads to the principal research question: How does secrecy in government organizations impact execution of transparency initiatives ? This dissertation presents research questions, hypotheses and research results that address that question add ing to the knowledge on transparency in and by the US intelligence community. This topic examines a key conflict between conceptually opposing themes, open hand F reedom of information, in general in particular, sits at the fissure between these two themes and serves as an excellent opportunity to examine the dynamic in detail. Research on the US FOIA is both timely and relevant. The FOI problem has been amplified in recent decades by two movements, the E Government movement which seeks to shift an increasing set of information and task execution online and the reinventing gover nment movement spawning from New Public Management thinking. In the case of E Government, federal agencies face increasing expectations of openness and voluntary disclosure in online environments that may conflict with organizational culture and evolved ag ency rule sets. In contrast, the NPM movement has the potential to treat freedom of information as a non mission essential task, outsourced and devalued by the bureaucracy (Newbold, 2011; Piotrowski, 2007; Roberts, 2000; Roberts, 2002). Both examples demon strate the need for continual research on the implementation of FOIA by the US government. ss show that transparency of and in the intelligence community i s a perennial topic of interest (Bannister & Connolly, 2011) This research effort and dissertation builds on scant literature on the effects of

PAGE 15

3 FOIA implementation in general and fills a gap in the literature with respect to FOIA programs in the intelligence community in particular. In addressing the principal research question, this research will provide analyses and context to address the question if agencies deliver on the right of transparency through the Freedom of Information process Piecing together the Secrecy and Transparency Puzzle Secrecy itself presents a bit of puzzle as there are secrets about secrets. By their very nature, secret organizations are hard to study and secrecy is difficult to observe directly. Instead, researchers must often look for the influences of secrecy indirectly. The literature on secrecy reveals complex interpersonal and institutional dynamics, opening up the possibility of using a variety of lenses to examine the impact of secrecy on organizations and systems. Use of an organization level lens, such as administrative discretion, allows for the possibility of incorporating both institutional and interpersonal effects into a single vi ewpoint. Scholarship on administrative discretion indicates that a number of factors affect administrative discretion, including collective processes and socialized norms (Sandfort, 2000), management (Riccucci, 2005), organizational structure (Meier & Boht e, 2001) clients (Scott, 1997), and the self perception of administrators (Sowa & Selden, 2003). Drawing from administrative and sociological concepts of secrecy (i.e. Simmel, 1906; Weber, 1920/2009), it is reasonable to suspect that administrative and cul tural aspects of secrecy might also influence the exercise of discretion and the creation of institutional rules, particularly with respect to the administration of FOIA by the intelligence community. Similarly, transparency can be viewed as a set of instr umental r ules and directions affecting program outcomes discrete request responses and internal processes (Meijer, 2013). These rules and the resultant outcomes can be evaluated comparatively using intelligence agencies as the subject set for secret orga nizations.

PAGE 16

4 Collectively, looking at aggregate outcomes, discrete request responses and institutional rule sets provides a multi dimensional portrait of how intelligence agencies differ from their non secret counterparts on transparency. Freedom of Informat ion Program Outcomes Although the First Amendment guarantees the right of citizens to petition the government for a redress of grievances, no other constitutional rights are explicitly granted to either the executive or the citizen for the withhold or discovery of information 1 When FOIA was enacted in 1966, the Act not only established a legal right for citizens to petition for information from government agencies, but also provided the mechanism for them to do so, equipping paraphrased from Seymour Ure, 1977) to uncover information of interest by compelling executive agencies to receive and disposition FOIA requests. In doing so, the US Congress formalized the long recognized value and made freedom of information an essential task for federal executive agencies. The Act served as progenitor for a wave of first generation FOI legislation in other countries during the 1960s and 1970s In more recent decades, a second wave of legislation adoption has occurred, embracing second ge neration FOI principles. Modifications to FOIA in the last decade added emphasis on agencies providing FOIA as a service (OPEN, 2007). Since 1966, citizens and press have used FOIA requests as a method to monitor government performance. In Fiscal Year (FY) 2011 alone, federal agencies received over 644,000 requests for information or records (Justice, 2012). Unsurprisingly, not all of these requests were granted. Though the Act also established the 1 The constitution of Sweden, often cited as an exemplar for freedom of information, does explicitly contain FOI rights (Galnoor, 1975).

PAGE 17

5 legal right of the government to withhold information in nine enumerated categories such as trade secret s, sensitive law enforcement information and classified national defense information T ens of thousands of requests per year are denied in full or in part on the b asis of these enumerated exemptions (Justice, 2012). In addition to these denials, agencies have the ability to deny requests for administrative reasons, a class of denial that includes the inability to find records or unwillingness to accrue excessive co sts to retrieve the requested information. Both the exemptions and administrative denials are used frequently. For FY11, of the requests dispositioned by the federal government, only 53.91% were fully granted. However, this coarse measure only tells a port between granted and denied. Although federal agencies processed (granted or denied) over 631,000 requests in FY11, they also carried forward a substantial backlog of 108,763 requests in to the next fiscal year, an almost 2% increase over the previous year. Though the Act specifies that agencies are to return a response within 20 business days of receiving a FOIA request, many agencies comply with the letter of the law by returning a respo nse that indicates that the request has been added to the backlog. Treated as a system, agency administrators have considerable latitude in implementing the FOIA, deciding how requests should be received (thus potentially imposing high transaction costs o n requestors), the amount of resources to dedicate to FOIA request processing, how requests should be dispositioned and if and how reports should be made available to the general public. As a result, despite legislated transparency in the form of the FOIA, the amount of effective transparency provided by the government remains largely at the discretion of individual agencies. Agencies have the discretionary latitude to set policies and allocate resources that have the potential to greatly affect the overall outcomes of their FOI programs and hence their effective transparency Yet, little is known about intelligence agency

PAGE 18

6 performance compared to non intelligence agencies. This raises the question of how intelligence agencies might exercise this differently when it comes to FOI. If these variances exist, they merit additional investigation to discover proximate causes and patterns. Research Question 1: How do Intelligence Agency FOIA program outcomes differ from their less secretive counterparts? Discretion i n Request Dispositions A number of discretionary decisions exist for any potential FOIA request Agencies can select to invoke an exemption, respond with an administrative denial, segment the request for partial grant or delay processing the request. In ex ercising discretion, the agencies must navigate between several different and sometimes opposing policy objectives, presenting a difficult administrative discretion scenario. A prime example exists with respect to national security. Though the agencies are compelled by executive order to protect classified information, non classified national security information is subject to FOIA requests, leaving a significant amount of sensitive but releasable upon r equest. During the Clinton Administration, the White House issued guidance establishing presumption of openness as a policy objective (Kim, 2007). Following the terrorist attacks of 2001, the George W. Bush administration issued guidance designed to protec t national security information to the maximum extent possible (Kim, 2007). In both cases, the general executive intent was to shape execution of FOIA programs within the law and the specific intent was to influence the implied exercise of discretion, part icularly with respect to information in gray areas. The main discretionary decision made for any single request is to grant or deny the request, with a companion decision on which records (if any) to release. Several secondary decisions exist as well, suc h as which track to assign the request to (in the case of multi track

PAGE 19

7 processing), whether or not to grant a fee waiver and whether or not to seek remedial fees for expenses. In addition, agencies also make undocumented decisions regarding how much search to accomplish to look for responsive records and what constitutes a record, in the case of ambiguous documentation. Documented decisions have the potential to shed insight on systematic or individual variation between agencies, raising the question of if s ecrecy results in different disposition results. In other words given a request for unclassified information, do secret agency responses differ significantly from their counterparts? While research question 1 examines program outcomes as a whole, this res earch question is concerned with individual request decisions. Research Question 2: How do Intelligence Agency FOIA disposition decisions differ from their less secretive counterparts? Institutional Transparency Rules The first two research questions wil l provide data about the effects of secrecy on programs as a whole and variances in individual request responses, but offer little insight about direct or proximate causes. For this, a deeper look is needed at internal mechanisms, such as institutional rule sets Transparency is affected by a number of statues, orders, policy statements and organizational level practices. Specifically, rules regarding the definition and retention of records and information classification have a direct impact on transpare ncy in practice. Understanding how the intelligence community implements these key rules will improve our understanding of how secrecy and transparency impact discretion in FOIA programs. Research Question 3: How do FOIA related rules differ between intell igence agencies and their less secretive counterparts?

PAGE 20

8 Dissertation Outline This dissertation begins with a review of several sets of literature relevant to the FOI problem, starting with secrecy as a phenomen on informed by work in sociology and philosoph y Definitions and concepts for transparency follow and set the framework for transparency research Cost and benefit considerations, found in literature on the economics of secrecy help provide insight into secrecy decision making made by individuals. Ad ditionally, the literature on administrative discretion provides a basis for considering aspects influencing the exercise of discretion in federal agencies Following the literature review Chapter III contains an introduction to the US Freedom of Informat ion Act an analysis of the institutional grammar of the Act to draw out specific elements of the law that can be observed and a review of contemporary research on FOIA In broad strokes, this chapter helps place the specific elements of FOIA in the overar ching context of the secrecy and transparency literature. Chapter IV builds on the literature review and analysis of FOIA to present a model of transparency and convert key concepts to operational definitions. These operational definitions are in turn us ed to generate testable hypotheses and exploratory questions. The dissertation continues with a rese arch design chapter (Chapter V) containing the approach to data collection and analytical methodology. Results are captured in t hree chapters, beginning with analysis of quantitative data (Chapter VI) and followed by a summary of experimental results (Chapter VII) and concluding with an integrated summary of findings and recommendations for policy makers and agency program managers (Chapter VIII) The dissertation concludes with an assessment of the impact of this research and recommendations for future research.

PAGE 21

9 CHAPTER II LITERATURE REVIEW Introduction The research questions for this dissertation compare intelligence agencies an d their less secretive counterparts in three areas: 1) FOIA Program Outcomes, 2) FOIA request disposition decisions and 3) FOIA related rules. There are a number of key concepts at work that are needed to understand the essential dynamics behind these thre e questions. The first is the concept of secrecy which distinguishes intelligence agencies from other federal civil agencies. As the principal causal relationship of interest, a nuanced understanding of the social, psychological and economic impacts of sec recy on individuals and organizations helps inform interpretation of observable outcomes. Transparency is the driving value underpinning the FOIA. Understanding the intrinsic and instrumental nature of transparency helps illuminate a duality within this co ncept that influences both the expectations of the citizenry and the actions of federal agencies. Hypothesized differences between outcomes in the first research question and disposition results in the second research question are likely partially attribu table to individual and organizational decision making processes. The literature on the economics of secrecy and transparency helps to make sense of this explanatory relationship. Similarly, the literature on administrative discretion helps point towards a reas for investigation in examining how FOIA programs and rule sets are implemented by agencies and groups. Collectively, these four streams of literature help provide the conceptual framework for this dissertation and inform relationships of interest that merit additional exploration and research.

PAGE 22

10 Secrecy as a Phenomeno n This dissertation draws on classical and more recent work on secrecy to understand the phenomenon of secrecy and its impact on organizations, particularly useful for appreciating how inte lligence agencies differ from their federal counterparts. The literature on secrecy is as fascinating as it is varied, with contributions in the fields of public administration, psychology, sociology and economics. F our scholars, Georg Simmel, Max Weber, E dward Shils and Sissela Bok, established the central writings on secrecy as a phenomenon incorporating individual a nd institutional perspectives of the topic that serve to outline the essential features of secrecy (Blank, 2008/2009). Their collective unde rstanding illuminates potential aspects of secrecy needing examination in the context of FOI research. As a sociologist, Simmel (1906) focused on the interpersonal aspects of secrecy with society and secret societies in particular, with insights on the rec iprocal nature of secrets and the role of trust in social dynamics. Although also a German sociologist like Simmel, Weber (1920/2009) takes a different approach, examining the dynamics of power and secrecy in organizations, particularly the bureaucratic fo rm. Edward Shils, an American sociologist writing in the mid 1950s on the heels of the McCarthy led Second Red Scare era, adopts an institutional approach to secrecy. In many ways, Shils embraces Weberian concepts of secrecy as an organizational phenomenon and applies an institutional construct, examining the mechanisms of secrecy and looking at broad social effects. Bok continues Shils line of reasoning by separating privacy from secrecy, finding the latter to hide more and be a qualitatively different phe nomenon than the former. Like Shils, Bok rejects the normatively neutral frame of Weber and Simmel (1906), finding secrecy inherently problematic and ethically dangerous (Blank, 2008/2009, p. 64).

PAGE 23

11 Despite differences in viewpoint, when taken as a whole, th ese four scholars present a consistent approach to secrecy as a phenomenon and social construct. The essential characteristic of secrecy common to all four scholars is that it involves concealment of Blank, 2008/2009, p. 61) p. 11). The element of purposeful concealment distinguishes secrecy from deception, the act of presenting false representatio ns, and privacy, which is thought of as roughly passive and complemented by indifference (Blank, 2008/2009). The aspect of intentionality has implications for social dynamics present in organizations and institutions. Simmel, Weber, Shils and Bok find secr ecy to be a powerful social construct. Purposeful concealment implies concealed by as well as concealed from group structures or, as Keane remarkable that informat i.e., which is effective through negative or positive means of concealment is one of the Simmel, 1906, p.462). Although they express the argume nt differently, Simmel and Bok in terms of secret societies, Shils and Weber in terms of institutions and organizations, all four present secrecy as a potentially divisive social construct because it separates one group from another by means of purposeful concealment. Simmel argues that secrets become a source of individual power in interpersonal relationships, at once strengthening the bond between individuals who share secrets and creating barriers between those who hold the secret and those that do not. a second world alongside of the obvious world, and the latter is most strenuously affected by the This possibility of dual realities, one presumably enlightened with secret kn owledge and information and the other presumably left to make do with only

PAGE 24

12 ignorance, forms the basis of fascination with secret societies and conspiracy theories (Shils, 1956, p. 12). Extending this argument to contemporary bureaucratic structures implies the existence of a of the FOIA found following the military build up dur ing the Second World War. The growth of dedicated intelligence agencies, operating behind a curtain of official secrecy, exemplifies this second, secret government. As Bok suggests, this presents alarming ethical implications for citizen participation in d emocratic government and increases the chances that deviant behavior and legal transgressions will go undetected in secret agencies. Simmel also lays a portion of the foundation for discussions of disclosure, articulating a dynamic tension between the act of concealment and the forces of discovery The power of secrecy, Simmel argues, is magnified when concealment is successful against discovery efforts and quickly reduced when the secret is a secret no more. As a result, groups and societies of secret keep ers can exert significant psychology pressure on the individual to maintain secrets, reinforcing the social boundaries of secrecy (1906/2009). Weber also recognized that the division between secret keepers and the uninformed was a significant source of pow er and the information or knowledge differential that the bureaucracy creates is a source of power for the bureaucrats who will attempt to further or maintain that differential, preserving their job and protecting the bureau. Weber credits the concept of an official secret to the bureaucracy bureaucracy seeks to increase the sup eriority of the professionally informed by keeping their For Weber, the designation of an official secret is only one of several means for officials in the bureaucracy to maintain secret knowledge

PAGE 25

13 in ord er to further their power. In practice, information or knowledge need not be designated offer an effective deterrent against disclosing information, even if the information itself was not intended to be protected. Applied on a large scale, as in the case of many contemporary federal governments, there can exist a significant cadre of secret keepers concealing information at various levels, some classified and som e unclassified and some intentionally and some unintentionally. maximizing bureaus might also be applied here, substituting budget for power. In this light, bureaus might attempt to maximize power through the retention of secrets, creating an environment significantly adverse to sharing information with citizens, other bureaus and legislative authorities (Weber, 1920/2009). This is part of the problem that FOI laws were designed to correct. Shils and Bok addr ess the question of the morality of secrecy in a way that Simmel and While acknowledging the right or duty of government to withhold some information, Shils (1956) warns against both of the perils of secrecy and of an obsession with conspiracies and secrets, no doubt influenced by the American political events of his time. In fact, the title o f his work, The Torment of Secrecy institutional effects, linking secrecy to conspiracies to social and political anxiety. Bok (1982) frames the problem differently with a focus on individual and social effects. For Bok, the essence of the danger of secrecy lies precisely in its definition, the intentional concealment that impacts individual and societal decision making by removing information and knowledge from con sideration.

PAGE 26

14 (Bok, 1982, p. 107). Bok answers in the affirmative, reasoning th at collective secrecy removes accountability and diminishes the sense of personal responsibility from collective decision impacts broader than just the individuals party to the secret. Even with this normative framing, both Shils and Bok admit that some secrecy is necessary, though they advocate for limited secrecy and strong publicity as a form of societal and institutional checks and balance. The work of Simmel, Weber, Sh ils and Bok points to the certainty of the bureaucracy attempting to conceal information from both the citizens it serves and the officials elected to apparatus, which h as a legitimate need for security, but yet may be most in need of transparency so that citizens may keep a watchful eye on it (Roberts, 2006). This concern is clearly evident in the work of Shils and Bok. Writing about foreign and military policy, Shils st (1956, p. 26). Bok, writing after a decline in trust in the executive branch in general and the national security apparatus in particular following a decade o f scandal and impropriety, states democratic societies this is an es pecially troubling and significant problem because it effectively restricts the participative element from the business of democratic government (Piotrowski & Rosenbloom, 2003). Deprived of information, the electorate is a disadvantaged participant in deci sion making, a condition which would seem to violate a core tenet of any true democratic society (Newbold, 2011). Viewed through this lens, the rights and obligations of the government to conceal are placed in opposition to the rights of the citizenry to d iscover and know (Galnoor, 1977; Roberts, 2006). The bureaucracy, in the form of Executive branch agencies and

PAGE 27

15 departments, is placed in the awkward, but unavoidable position of simultaneously creating and enforcing the processes of secrecy and information protection while executing the policies of disclosure and transparency (Piotrowksi & Rosenbloom, 2003). The literature on secrecy as a phenomenon raises questions that require additional research. At a basic level it raises the questions of if and how Am to the promise of FOIA and if the presence of official secrecy produces broader impacts. More specifically, it leads to the assumption that agencies with more official secrecy would exhibit fewer tendencies to disclose infor mation than their less secretive peers even after accounting for exemptions used to protect official secrets Put succinctly, while agencies dealing with classified information would obviously protect those secrets, the influence of secrecy as phenomenon might also lead them to over protect unclassified information that would and should ordinarily be disclosed. The literature strongly suggests that research should reveal meaningful differences between secret agencies and their less secretive counterparts. Transparency as both Intrinsic and Instrumental Value Secrecy is only one part of the larger research puzzle. Absent other compelling forces, the secrecy literature makes it clear that organizations and agencies would likely prefer concealment over disclos ure. However, secrecy is not the only force at work on individuals and organizations. American citizens have a history of expecting transparency from public and, to a lesser degree, private organizations. Just as the literature on secrecy informs the essen ce of secrecy as a phenomenon, the public administration literature on transparency helps to unravel this complex concept and points to areas needing further examination and research. Of particularly interest given the main research question How does sec recy in government

PAGE 28

16 organizations impact execution of transparency initiatives ? is how the literature deals with transparency as both a value and as an outcome. The word itself is compound of trans and parene literally meaning to see through. But this seemingly simple and straightforward definition belies a more complex set of relationships between an observer and the observed. At the conceptual level, transparency in public administration literature is repre sented by two different streams; the first viewing transparency normatively as an intrinsic value, the second viewing transparency descriptively as an instrument. The normative stream of literature characterizes transparency as an intrinsic social and ind ividual value (Etzioni, 2010; Hood, 2006; Piotrowski, 2007, p.3). This is a classical view of transparency summarized by the often quoted paraphrase from Justice the best of disinfectant s The normative frame owes much to the work of enlightenment scholars like Rousseau and Bentham who articulated both a personal and public need for disclosure and openness (Hood, 2006). This is transparency as an unalloyed good, described by 2006, p.7). To Bentham and Rousseau, whether the observed is the government, public officials or fellow citizens, unfettered visibility provides the mechanism for reaching a more utopic state. To scholars in this st ream, transparency is key condition of democratic society; reducing the information asymmetries needed for the citizen and press to maintain a watchful eye on government (Bathory & McWilliams, 1977; Fung, Graham, & Weil, 2007). Framed normatively, secrecy and privacy are inescapably oppositional in nature as barriers to transparency. In this context, transparency and disclosure are fulfilling the ideals of openness and are vehicles by

PAGE 29

17 which government accountability can be realized, all with the objective o f increasing trust However, this normative framing is not without challengers, particularly in contemporary scholarship which follows more of post modern bent. Heald (2006) argues that a degree of ignorance is socially useful and posits that the optimal level of transparency is somewhere below maximum, at a level allowing for some privacy and intentional concealment. Bannister and Connolly (2011) argue that there are transparency costs an d risks which create the potential for disutility. Bannister and Connolly cite the potential for misinterpretation and difficulties in citizen comprehension as potential risks, along with the possibility that agencies may shift towards increased failure t o acknowledge problems or record key opinions or processes for fear of mass publication. Etizoni (2010) advances along bounded rationality lines, unlimited amount of t ransparency can be counter productive as it has the potential to overwhelm the cognitive processing abilities of the supposed beneficiary. Similarly, Fung, Graham and Weil (2007) find that cognitive biases and limitations limit the effectiveness of disclos ure regulations and practices. The descriptive perspective of transparency focuses on the instrumental nature of information about an actor that allows other actors to monitor the workings or performance of agent dynamics, making transparency not only instrumental b ut also directional. In other words, the question to be asked is of whom by

PAGE 30

18 directions of transparency in public affairs. For any given agency or agent, there is the potential for transparency to a number of differing principals including other agencies, the executive, the legislature, the press and the public This directional nature is derived from Heald (2006) who provides a four direction framework for consideration: Upward s, Downwards, Inwards and Outwards. The implication of directional transparency is that it must also be viewed as relative: transparency to one observer might be perceived as opaqueness by another observer. Regardless of direction, the implicit assumption is that the availability of information allows the and Connolly (2011), P rat (2006) and Etzioni (2010). relationships. Support can be found in the work of Dawes (2010), who posited that information can be both an object and an instrument of p olicy and in the work of Hood (2006) who offered a definition of transparency as fixed and predictable rules. Viewed through this institutional lens, information and power are directly related. Though not specifically referenced by Meijer, the possibility of socially constructed transparency is clear, some groups of actors will be more deserving of monitoring or surveillance (and hence less privacy) than others and some groups of actors will be more deserving of information than others. By extension, one mi ght observe agencies exercising discrimination in transparency; choosing some classes of individuals as more deserving of transparency than others, particularly if the agencies perceive those classes as being more powerful relative to other individuals. Fo r example, agencies might treat requesters

PAGE 31

19 about wha Meijer uses the term former as observables inputs, outputs and outcomes while the latter consists of less measurable, but n ot unobservable, procedural sequences connecting events. Event transparency: transparency about the artifacts and records of government. The US FOIA provides an example of d by agencies means that many documents fall outside of its scope. Process transparency can be found in requirements to post certain procedural instructions to the federal register and in some o ne knows that the agency exists. This has particular relevance with secret agencies. As a case in point, the US Government did not openly acknowledge the existence of the National Reconnaissance Office until 1992, more than three decades after it was forme d (Berkowitz, 2011). While an extreme case, the precept stands; there must be sufficient information about the existence and activities of an organization for any transparency to exist. A more recent example can be found in the 2013 scandal involving NSA c ollection of information on US does not otherwise detract from his definition.

PAGE 32

20 direct between the observers and observed, most transparency is mediated, most often through electronic means (Meijer, 20 09). This complicates the relationship between observers and observed and raises the issue of the trust in the medium as well as trust in the transparency of the original actor. Quickly, the problem resembles Russian nesting dolls with the question of tran sparency through not only to a given actor. In a sense, US FOIA is a mediated rather than direct transparency law, offering citizens information only through the structured medium of official records (Piotrowski, 2007; Roberts, 2006). Roberts (2001) place s the issue of medium in ensure transparency of its contracted actors or partners. Interestingly enough, this places FOIA offices in government agencies in the p osition of being both the object of transparency and the mediators of transparency, raising the question of how transparen t agencies are about transparency. Though different, the two streams of transparency literature are not entirely exclusive and have areas of commonality. Both the normative and descriptive frames are based on the principal agent model and the ability of the principal to monitor the agent. T he discretion factors of the model. Discretion is an inevitable element of the principal agent theory and offers the opportunity for the agent to select transpare preferences. Bounded rationality concepts indicate that the principal experiences cognitive limitations and biases which may limit the ability to process an unbounded amount of information and impact decision m aking.

PAGE 33

21 In some aspects, the normative and descriptive transparency streams merge unexpectedly. The normative framing for transparency still has an instrumental aspect, as transparency is an enabler to obtaining trust between the people and government. Sim ilarly, the descriptive framing cannot entirely escape a normative implication, as the core assumption is that monitoring of the agent yields more positive outcomes. Transparency is not costless, but rather incurs costs on the part of both the principal an d the agent. The literature overwhelming assumes that the net benefit of transparency outweighs its costs, with only a few voices (i.e. Bannister & Connolly, 2011; Etizoni, 2010; Heald, 2006) warning of the potential for diseconomy. Secrecy has a differen t and more complex relationship with instrumental transparency than it does with normative transparency. Secrecy has a simple, albeit oppositional, relationship to normative transparency. However, this relationship looks different when using an instrumenta l definition of transparency. Through this lens, secrecy and transparency can be viewed as co existing institutional influences, with the potential to be competing or complementary influences. This institutional view points to the potential of transparency like secrecy, being observable not just through outputs but also through organizational rules and structures. Viewing transparency and secrecy as co resident institutional influences also highlights the role of administrative discretion in secrecy and tr ansparency decision making. more promise. The supporting literature leads to questions of how the co resident influences of secrecy and transparency are resolved by indi vidual administrators and agencies as a whole. For the second research question, meaningful differences attributable to these institutional influences might be found by comparing individual request dispositions and responses from

PAGE 34

22 agencies. The literature a lso suggests a deeper look at secrecy and transparency costs and benefits is needed. Finally, using a definition of instrumental transparency raises the question of pr incipal research question, this transparency literature suggests the likely outcome that intelligence agencies are less likely to be transparency about their transparency initiatives than other agencies. For the third research question involving FOIA relat ed rules, it also suggests directionality, importing that intelligence agencies are more likely to have institutional rules that favor secrecy at the expense of transparency. Secrecy and Transparency Decisions: Costs and Benefits While the classic literat ure on secrecy and transparency describes well the macro level influences of secrecy and transparency, it falls short in illuminating how individual concealment and disclosure decisions are made. The literature on the economics of secrecy and disclosure he lps ameliorate this gap with micro and meso level models of individual and organizational decision making, drawn first from rational choice economics and then tempered by behavioral economics assumptions. This stream of literature is key to understanding discrete outcomes of the FOIA process like formal and informal disclosure decisions involved in answering the second research question, How do Intelligence Agency FOIA disposition decisions differ from their less secretive counterparts? Scholarship on tra de secrets and information economics indicates that organizations experience both costs and benefits associated with secrecy and transparency (Aftergood, 1999/2009; Kultti, Takalo & Toikka, 2006; Stigler, 1980). Secrecy and transparency can be viewed as as pects of the information marketplace, with the government, press and public as participants in the exchange producing and demanding information as a resource (Galnoor,

PAGE 35

23 1975; Lowi, 1977). The protection of information conveys certain benefits to organizatio ns (Dufresne & Offstein, 2008; Hermalin & Katz, 2006). Through secrecy, companies may achieve strategic surprise in the marketplace, obtaining a first mover advantage and deterring or delaying new market entrants (Keane, 2008). Designation of a trade secre t may allow companies to protect intellectual property without disclosure, obtaining a competitive advantage over potential imitators. Similarly, state secrets allow governments to maintain advantages over rival states, obtain differential information adva ntages in diplomatic negotiations and conduct covert operations to conceal action. However, the act of concealment also generates costs, both direct and indirect (Stigler, 1980). The direct costs of implementing a secrecy program are non trivial, including personnel costs such as vetting and clearance, matriel costs for storage and security and, in the digital age, significant information technology costs. The indirect costs of secrecy, while difficult to quantify with precision, can also be significant an d include intra firm effects such as impaired organizational learning (Jensen & Webster, 2009), weakened knowledge management and other intra firm information asymmetries that may impair exploration and exploitation (March, 1991) as well as extra firm effe cts like suppressed diffusion of innovation (Aftergood, 1999/2009; Kultti, Takalo & Toikka, 2006; Rourke, 1960). On balance, this set of literature views secrecy negatively, as an impedance to the free exchange of information necessary for ideal market ope rations (Hermalin & Katz, 2006). However, full transparency, in the form of unfettered information disclosure is not costless either (Prat, 2006). Organizations also experience benefits and costs to disclosing information, whether voluntary or compulsory. The work of Fung, Weil, and Graham (2007) indicates that there are some benefits to voluntary disclosure as it reduces information asymmetry on the part of consumers/clients. In some cases, compulsory disclosure as a result of transparency laws improves c onsumer choice, though in other cases, the disclosure appears to

PAGE 36

24 have little to no effect. A key conclusion from their research is that transparency is as much about how, where and when as the what, in other words, the mechanisms and timing of disclosure a re as significant as the content disclosed (Fung, Weil, & Graham, 2007). This conclusion has direct implications for FOIA, which can be considered a procedural law, intended to provide the mechanism for request based disclosure, rather than a transparency law, specifying content and information to be provided by the government. Additional, it suggests that research examining FOIA mechanisms and procedures might be fertile ground for uncovering differences between agencies. Additionally, secrecy has spillove r effects, resulting in social costs and potential value conflicts (Galnoor, 1975). External parties, including private companies and public organizations, might bear the costs of deviant behavior by a single organization made easier through concealed acti on and impaired discovery (i.e. social losses caused by fraud or theft). Moreover, the gains and costs of secrecy are distributed unequally. The gains from secrecy belong primarily to the organization keeping the secret and begin immediately (the gain bein g prevention of loss or retention of competitive advantage) (Galnoor, 1975). The benefiting organization bears some direct and indirect costs of secret keeping, but other organizations or institutions may also bear costs due to impeded diffusion of innovat ion (Rourke, 1960), impaired decision making and increased risk of deviant behavior. These costs are not likely to accrue immediately, but rather over time and are experienced largely by extra organization (i.e. second & third) parties. Conversely, the ben efits of transparency/disclosure are largely realized by external parties whereas the costs are due to the disclosing party or observed agent (Prat, 2006). The patent system is the exemplar of this effect, with the addition of limited duration protection f or the patented party to realize competitive advantage (Friedman, Landes & Posner, 1991). Prat (2006) states that more observations (or disclosures) will not harm the principal(s), but the same

PAGE 37

25 cannot be said for the agent. The presence of social costs for secrecy and social benefits for transparency and disclosure drives the collective action response that has resulted in a role for government to regulate secrecy and mandate transparency and disclosure in certain industries (Roberts, 2006) Viewed through this lens, government agencies present a distinctive case as government organizations in general and US federal agencies in particular are charged with a special self regulation problem, effectively enforcing two potentially competing rule regimes. This su ggests that research on FOIA should examine the possibility of rule conflict as a factor in decision making. Principal Agent Theory provides a model of the individual to frame and formalize assumptions about secrecy costs and benefits. Agency theory is ess entially a relational construct between individuals placed in the role of principal and agent (Eisenhardt, 1989). At the simplest level, both the principal and the agent have different utility functions that create the opportunity for variance between the his/her own by presenting incentives and inducements and must monitor the agent to identify deviation. Conversely, the agent seeks to maximize his/her own utility and desires wide discretionary aut hority to do so. As a result, information is a critical aspect of principal agent relationships (Eisenhardt, 1989). The principal seeks information to monitor the activities of the agent, but this information is not costless, nor does the principal have li mitless capacity for processing information (Simon, 1957). The agent has rational cause to restrict availability of information to the principal and enjoy the benefits of a deliberate information asymmetry. Both principal and agent maintain a subjective an d boundedly rational perceptive of the costs and benefits of information disclosure (Stigler, 1980).

PAGE 38

26 The literature on the costs and benefits of secrecy and transparency suggests some general suppositions can be made using the principal agent model. In general, principals will likely overvalue information disclosure and agents will likely overvalue inform ation secrecy. In organizations with a high degree of cultural secrecy or an institutional bias toward non disclosure, the perceived costs of information disclosure might be quite high with relatively low benefits. Conversely, because of the relative lack of information about agent activities for secret organizations, principals might place a premium on information disclosure from these agencies. Regardless, both principal and agent are likely to base decisions on the perceived, rather than objective, cost s and benefits. This leads to Assumption 1: Principals and agent s will select desired information disclosure functions so that the perceived benefits exceed the perceived costs. Though not directly testable, this assumption suggests that individual decisions likely play a significant role in determining the quantity and quality of transparency provided by an agency. Moreover, these decisions are likely affected by potentially competing goals and institutional influences. In short, administrative disc retion is a key element in secrecy and transparency programs and bears examination. Discretion and its Application to Secrecy and Transparency Decisions The competing conditions of secrecy and transparency in organizations suggest that FOIA administrators are confronted with potentially conflicting policy objectives which necessitate the exercise of considerable discretion. Understanding discretion within the specific context of secrecy and transparency decision making and rule sets directly informs methodo logical approaches for addressing the research questions by suggesting observable indicators of decision making.

PAGE 39

27 At the heart of the administrative discretion condition is a potential for variance between policy objectives and program outcomes caused by bo undedly rational humans operating in a complex social context Work on discretion by Lipsky (1980), Riccucci (2005) and Sandfort (2000 ) shows that policy inputs do not necessarily translate into outcomes as intended, a condition expected to be present in F OIA programs as well. In the case of FOI policy for national security organizations, the problem is compounded by the competing guidance to both conceal and disclose and the potential for value conflict experienced by the individual administrator. T he con dition of administrative discretion is rooted in the organization as a social construct consisting of individuals. Viewed through the lens of principal agent theory, discretion may be seen as a bundle of principal agent situations. At the macro level, gove rnment organizations are agents to multiple principals, at once beholden to the chief executive as chief administrator, to the legislature as representatives of the people and originators of policy and to the people themselves as both ultimate shareholders and ultimate beneficiaries of services (Finer, 1941; Mosher, 1968 ). In the event that these principals conflict, the organization faces an ordering problem to determine which set of responsibilities take precedence. At the micro level, principal agent dyn amics manifest themselves between individuals as managers and workers. In both macro and micro cases, the principal faces monitoring and enforcement costs which ict agents (both in individual and organizational forms) will have different preferences from their principals and can be expected to make decisions which deviat decisions in favor of obtaining some advantage on the part of the agent (Maynard Moody & Musheno, 2000). For organizations, this might mean seeking more control of resources or

PAGE 40

28 information as Niskanen (1968) suggests, or expr work indicates. results in potential costs for the agent leads to the conclusion that agents will attempt to minimize costs by di sclosing less or resisting transparency initiatives. In the case of FOIA, this literature suggests that agencies will adopt imperfect realizations of policy, opting for more secrecy than desired by their multiple principals. However, principal agent theory is limited in perspective and does not account for the social influences on the agent or the possibility of socially beneficial discretion (Maynard Moody & Musheno, 2000) In practice, b oth principal and agent are embedded in a social construct (i .e. team s, groups and organizations) that produces informal rules, establishes collective norms and ultimately influences administration discretion. FOIA administrators in the modern federal bureaucracy operate in a complex environment with the potential for conf work on Red Tape indicates that administrators respond to the complexity by adding endogenously created rule sets on top of externally imposed requirements and guidance. This creates further complexit y for individuals in the workplace. T he se rule sets constrain the individual exercise of discretion, particularly by workers and street level bureaucrats who may be forced to adhere to sequences or protocols that serve contrary purposes (Maynard Moody & Mu sheno, 2000; Riccucci, 2005; Scott, 1997). The effect is felt by organizations as well as individuals. Red Tape Theory indicates that organizations may become overwhelmed by the density of rules and the compliance burden they impose (Bozeman, 1993). Dealin g with these rule sets often requires mandatory training, compliance monitoring and other costs, diverting 2000). Similar to having multiple principals, a sufficien t density of rules can result in conflicting rule sets, presenting managers and workers with a discretionary choice of which rules to adhere

PAGE 41

29 to and which rules merit disregard or deferred compliance (Maynard Moody & Musheno, 2000; Sandfort, 2000). In ad dit ion to the formal rule sets workers develop short cut practices and tools to assist in dealing with day to day tasks without referencing statute or formal procedures. These informal constructs form a second phalanx of rules and procedures that offer the o pportunity for discretion to be exercised. In the case of FOI, agency and department level rule sets have the potential to influence FOI outcomes in ways unintended by policy as codified in law. Organization norms may also influence discretion. Organizatio ns are social constructs, and as such, work is a social occasion. Work by Riccucci (2005), Sandfort (2000), Sowa and Selden (2003) shows that elements of organizational culture, such as shared values, collective norms and perception of goals and prioritie s influence individual worker behavior. While organizational rules may indicate the range of actions that are technically permitted or restricted, organizational norms influence what actions are socially acceptable, what behaviors are informally rewarded a nd what attitudes are expected. These norms help workers make sense of complex cases and scenarios in a way that statute or formal direction cannot (Sandfort, 2000). This social process helps workers establish collective perceptions of their work role and Simmel and Weber suggest that secrecy can have a significant impact on social processes in organizations, leading to more secrecy. a social construct, Anand and Rosen (2008) and Keane (2008) find that secrets can be embedded in and reinforce organizational culture. Keane (2008) suggests that secrets can become symbolically important for organizations and increase the social cohesion of the organization by providing a clear boundary. Anand and Rosen (2008) make a distinction between who has

PAGE 42

30 sanctioned a secret, arguing that secrets sanctioned by organizational insiders are more likely to transgress ethical boundaries that secrets sanctioned by societal norms or law. Additionally, this process may lea d to acculturation of unethical behavior or excessive secret keeping (Anand & government developed an unreasonable amount of secrecy for a liberal democra cy due the acculturation of secrecy as a value (Bennett, 1985; Christoph, 1975; Galnoor, 1975; Griffith, 1989) Expanding the principal agent dyad to the perspective of relations between organizations rather than individuals provides a frame for viewing the relationships between federal agencies and their stakeholders through the lens of agency theory. Both principa l and agent will create institutional relationships and mechanisms to achieve their goals, the former to obtain relevant low cost information about the agent, and the latter to restrict information disclosure. In organizations, both principal and agent are affected by institutional and social forces that shape the desired selection of disclosure functions. The secrecy literature indicates that secrecy can exert a powerful psychological force and creates social pressure to maintain collective secrets. Inform al norms (Sandfort, 2000) and endogenously created rules (Bozeman, 1993; 2000) influence the exercise of individual administrator discretion. For example, federal agencies may selectively use a combination of official and administrative secrecy rules to ju stify disclosure decisions. This leads to Assumption 2: In organizations, the disclosure function of the principal agent dyad is affected by institutional forces. These institutional forces are likely to shape both program level and discrete outcomes and o ffer the potential for creating unintended effects and variance from policy goals.

PAGE 43

31 While social norms do help to reduce individual variation within a given organization through an acculturation process which inculcates shared norms and hence priorities, t he nature of norm formation indicates that significant organization to organization variation may still be present, suggesting additional analysis is needed to understand variations between organizations (Sandfort, 2000). The implication from the literatur e is that secrecy can become inculcated in organizational norms and some organizations will be more secretive than others, opening up the possibility to developing a scale of secrecy or transparency which might inform the principal research question. In s um, the literature on discretion suggests that understanding of FOI in practice requires a deeper look at specific agency rule sets and cultural factors that might influence individual and organizational discretion in the execution of agency FOI programs. Taken with the literature on secrecy as a phenomenon and instrumental transparency, this set of literature suggests that the effects of secrecy and transparency institutional influences should be observable in individual decisions as well as program outcom es, as individual administrators make disposition choices as influenced by organizational norms and internal rule sets. Additionally, this literature leads to the conclusion that variance might be observable not just in disposition decisions, but also in m ore nuanced instrumental choices such as the choice of phrasing or form of response, suggesting a qualitative component to transparency research is warranted. Conclusion s from the Literature Collectively, the literature helps illuminate the topic of FOI in practice, pointing to several areas for examination and research. The literature on secrecy indicates that withholding information is linked to organizational power dynamics and can be a significant force in

PAGE 44

32 organizational culture. This directly informs t he principal research question How does secrecy in government organizations impact execution of transparency initiatives ? and suggests that research is likely to discover meaningful differences between secret agencies and their less secretive counterpar ts. The transparency literature highlights the role of institutional rules and individual discretion on the part of agents as influences on transparency outcomes. Viewing transparency as an instrument leads to the need to examine not just the coarse output s of transparency programs, but also some of the more nuanced implementation aspects of transparency programs. Literature on secrecy in practice indicates that the costs and benefits of secrecy and transparency are important factors to consider, as heurist ics are used in day to day decision making. Scholarship on administrative discretion leads to the conclusion that examination of agency internal rules sets is warranted, as the possibility for conflicting guidance is substantial. Additionally, the discreti on literature suggests that social norms play a significant role in individual decision making and offer a possible explanation for systematic variances between organizations. A major conclusion from the literature is that there is a strong potential for over production of secrecy, particularly in government organizations (Vaughn, 1997/2009; Weber, 1920/2009). This overproduction is caused in part to administrative discretion decision, such as over classification (Roberts, 2006; Rourke, 1960; Rowat, 1965) but also in part due to the compartmentalization that the bureaucracy uses as an organizing technique which creates intra firm organizational secrecy (Dufresne & Offstein, 2008; Weber, 1920/2009 ). In some cases this allocative inefficiency can be signific ant, increasing program costs by 25% (Aftergood, 1999/2009). In the case of FOI programs, over production of secrecy can have important consequences for democratic v alues and citizen participation by diminishing transparency precisely on those agencies tha t need it most. This conclusion provides both impetus for

PAGE 45

33 additional public administration research and directionality for hypotheses regarding secrecy and its impacts on FOIA programs. In sum, the literature provides the broad strokes for understanding th e transparency dynamic. Stakeholders acting as principals desire information about organizations as agents through a medium. The literature suggests that secret keeping can impact both the principal and agent, chiefly through social norms and constructs. H owever, the literature makes it clear that our collective understanding of secrecy is limited, particularly in how secrecy regimes and practices impact transparency outcomes and mechanisms. This research addresses that gap by exploring how the US intellige nce community as a specific instance of secrecy impacts the transparency outcomes outlined in the FOIA.

PAGE 46

34 CHAPTER III FREEDOM OF INFORMATI ON: RIGHTS AND RESEA RCH Introduction The literature summarized in the previous chapter provides a conceptual basis for understanding secrecy and transparency. To understand how these concepts are applied in practice in the case of FOIA, a deeper look is needed. This chapter begins with a brief glance at the historical origins of the Act to establish the intent of the Act An analysis of the institutional grammar of the contemporary version of the Act and current executive orders helps shed light on how policy intent aligns to specific actions Additionally, this analysis reveals actors a nd measures for potential observation and comparison. The chapter concludes with a review of contemporary research and implications for this research effort. Freedom of Information as a Redress for Government Secrecy The Freedom of Information Act of 1966 emerged as a result of a decades long struggle by the press and sympathetic legislators to peel back the veil of government secrecy described during and follow ing World War II, causing concern amongst legislators, journalists and scientists (Rourke, 1960). Archibald describes a portion of the letter to Representative John Moss creating a Special Sub d (1979, p. 313). The text of this paragraph succinctly frames the problem that the FOIA would eventually be enacted to address: An informed public makes the difference between mob rule and democratic government. If the pertinent and necessary information on governmental activities is

PAGE 47

35 denied [to] the public, the result is weakening of the democratic process and the ultimate atrophy of our form of government. (Archibald, 1979, p. 313) The Moss commission began a ten year effort to examine the extent of secrecy in the federal government and propose changes. Unsurprisingly, the commission found an over abundance of secrecy, both official and administrative. The root of executive branch se crecy laid the 1946 Administrative Procedures Act which afforded agencies and departments sufficient ald, 1979; Rourke, 1960). Given this leeway, administrators and bureaucrats performed as the Weber (1920/2009) anticipated, over producing secrecy and concealing far more than one would anticipate in the public interest. The commission highlighted a few eg After more than a decade of what Archibald desc ribes as a public relations campaign, the Moss subcommittee secured the passage of a deliberately weakened FOIA in 1966, with an effective date of July 4, 1967. Executive departments, despite testifying on the benefits of full disclosure, resisted the spir it of the act with vigor, developing internal rule sets and mechanisms to defeat the purpose of the law (Archibald, 1979; Relyea, 1975; Roberts, 1979). In some cases, agencies began to delay responding to the requests, creating secrecy by delay, in other c ases they charged fees intended to make completion of the request cost prohibitive by the requester amendments to the FOIA in 1974, correcting these deficiencies and making the law more enforceable by both the executive and judicial branches (Relyea, 1979). Though the act would

PAGE 48

36 see additional modifications in subsequent decades, the 1974 version established the essential rule set in force today. Substantial updates were mad e again in 1996 to bring the law in line with the digital age with the passage of the Electronic Freedom of Information Act (E FOIA) which added requirements pertaining to the agencies posting and receiving information via electronic means. Additional upda tes were made in 2007 to incorporate practices established by executive order and add regulatory language to further define agency FOIA activities. In the 2007 update, (OPEN) government act of 2007 Congress reaffirmed the intent of FOIA, linking constitutional democracy to informed consent and citing Supreme Court rulings (OPEN (O PEN 2007). FOIA Establishes Public and Government Rights to Information The Freedom of Information Act is instantiated in law as Section 552 of Title 5, Government Organizati on and Employees, of the United States Code (5 U.S.C. § 552). Containing both constitutive and regulatory clauses, the Act establishes the mechanisms for citizens to request information and enumerates the responsibilities of agencies in responding to those requests. Broadly speaking, the Act has four complementary aims. First, in order to increase transparency, it contains some requirements for agencies to make a minimal set of information available to the public. Second, the Act details administrative proce dures and required conditions for agencies to appropriately service information requests. Third, the Act establishes 5 U.S.C. § 552 (b)) and in the case of the intelligence community, compels those agencies to withhold records from

PAGE 49

37 foreign government entities (5 U.S.C. § 552 (a)(3)(E)). Fourth, the Act contains performance and reporting requirements for agencies to ensure appropriate congressional and executive oversight of agency FOIA p rograms. The Institutional Grammar of FOIA To examine how the intelligence agencies implement the Act, a thorough examination of the text of the legislation is needed, particularly in comparison or contrast to executive instructions that also have a poten tial to impact agency discretion. The Institutional Grammar Tool provides a methodology for analyzing institutional statements in the law and producing insights to help frame research efforts. In particular, use of the institutional grammar approach allows for the identification of clauses requiring action or compliance from specific actors. Of particular interest for this research effort are institutional statements involving federal agencies that provide standards for performance or compliance that can be used to create testable hypotheses and observations for research. The Act contains both definitive clauses, to establish the meaning of particular elements, and institutional statements, providing instructions for execution. The coding method proposed by Basurto et al. (2010) provides a baseline for evaluating the syntax of regulatory Aim (I), Condition (C) and Or Else (O). These five components indicate what actions are required, permitted or forbidden by which actors for what purpose under what conditions and with what penalties for non compliance/non performance. A modification proposed by Siddiki, Weible, Basurto and Calanni (2012) includes an object (B) code to i dentify animate and inanimate targets of legislation and is roughly analogous to the direct object of sentences. This modification reduces ambiguity in attribute coding and makes it easier to distinguish between

PAGE 50

38 two statements that have identical attribute deontic and aim elements but differing objects. as modified by Siddiki, Weible, Basurto and Calanni (2012). Coding of the section yielded 98 observable units of institutional syntax. Statements containing the five ADICO elements are categorized as rules, statements with the first four (ADIC) are norms and those without a Deontic are categorized as strategies. All 98 units of observation within the FOIA section were categorized as norms, meaning that all institutional Most statements pertained to the agency or agency officials (The/Each Agency, Head of Agency, and Chief FOIA Officer). Seventy of t he 98 institutional statements are directed at the agency or agency official level, reflecting an administrative law emphasis (See Table 3.1). Other actors include the US District Courts (n=10), the Attorney General (n=7) and Special Counsel (n=4). The 200 7 amendments established the Office of Government Information Services (OGIS) within the National Archives and Records Administration and created obligatory statements for OGIS (n=3) and the General Accounting Office (GAO) (n=2). Table 3.1 Institutional Grammar Analysis Results 5 U.S.C. 552 Deontic Actor Obligatory Permissive Restrictive Total Agency 47 8 8 63 The Courts 3 6 1 10 Attorney General 6 1 0 7 Chief FOIA Officer 6 0 0 6 Special Counsel 4 0 0 4 OGIS 3 0 0 3 GAO 2 0 0 2 Person 1 0 1 2 Head of Agency 1 0 0 1 Total 73 15 10 98

PAGE 51

39 Over seventy four percent of the statements were obligatory, using the language 98) statements. A minor share of statements were permissive (15 of 98), offering actors the latitude to take an optional course of action. Of these permissive statements, 6 of 15 were directed at the courts, offering latitude for the courts to enforce rul ings. Only 10 of 98 statements were restrictive, and 8 of these 10 were aimed at limiting the actions of agencies. A Closer Look at t he 20 Day Rule Needed the individual appears in only two statements, and in neither case is the syntax permissive. The first statement is in awkwardly worded restrictive syntax which ind icates that: Except to the extent that a person has actual and timely notice of the terms thereof, a person may not in any manner be required to resort to, or be adversely affected by, a matter required to be published in the Federal Register and not so p ublished. (FOIA, 2009) The second statement involving an individual person requires that individuals requesting expedited processing show compelling need. As written, the law does not expressly permit individuals to submit requests, but rather obligates ag encies to receive them. The presumption of disclosure, while stated clearly in congressional findings, is not strongly represented in regulatory statements. Instead, the bulk of the burden for transparency falls on the citizenry to request rather than the agencies to provide.

PAGE 52

40 Though the law establishes a number of temporal conditions designed to prevent secrecy by delay and compel agencies to perfect requests within 20 working days, the section r non compliance. Additionally, although agencies are obliged to report the number of requests determined within 20 days, there is no syntax compelling agencies to report their overall compliance with this standard. The information specified in the annual reports provides only an incomplete benchmark of compliance, making detection of non compliance challenging. Moreover, the law does not require agencies to report how often they invoked a delay or for what purposes, leaving open the possibility of abuse of § 552 (a)(6)(B)(i) which allows agencies to delay perfecting the request if unusual circumstances are present. This gap in the legislation suggests that a more detailed look at agencies performance u nder the 20 day standard is warranted and may yield useful insights. Executive Order 13392 Supplements Legislation The FOIA is not only the direction that agencies have received for execution of freedom of information. Periodically, Presidents issue execut ive orders aimed at improv ing FOIA programs in federal agencies information disclosure consistent with NP M concepts. The order establishes FOIA as an essential service of the government and directs agencies to establish service centers and public liaisons for handling requests. Additionally, the order required agencies to submit and post plans containing mile stones for reducing the backlog of unanswered requests and improving the quality of FOIA service. The order further directed agencies to conduct follow up analysis of those milestones and issue reports on milestone progress and deficiency correction.

PAGE 53

41 The Institutional Grammar of EO 13392 The text of this order was also coded using the same method described above to identify institutional statements of interest. In contrast to the FOIA, EO 13392 contains both norms (n=60) and strategies (n=7). Five of the s even strategies can be found in the first few paragraphs of the order and express a desire for a well informed citizenry and articulate how agencies should view FOIA requesters as customers of the agencies and respond accordingly. Like the FOIA, the norms in the Executive Order are primarily obligatory statements directed at the agencies or agency officials. Forty nine of 60 norms fit this description (see Table 3.2). The order emphasizes the role of the Chief FOIA Officer, FOIA Public Liaisons, and FOIA R equest Service Centers; directing 30 statements at these three actors. This emphasis on this structure removes some discretion from agencies and directs a degree of uniformity in administrative practices. The order also establishes new reporting criteria, requiring the Chief FOIA Officer to review agency practices and issue a report to the Attorney General. This report may be a valuable source of information on how the agencies perceive their FOIA programs and their ability to service the public.

PAGE 54

42 Tabl e 3.2 Institutional Grammar Analysis Results EO13392 Deontic Actor Obligatory Permissive Restrictive Strategy Total Chief FOIA Officer 21 0 0 0 21 Agency 12 0 0 0 12 Head of Agency 9 0 0 0 9 FOIA Public Liaisons 5 2 0 0 7 Public 1 0 0 5 6 Attorney General 4 1 0 0 5 All 1 0 1 0 2 President 0 0 0 2 2 FOIA Requester Service Center 2 0 0 0 2 OMB Director 0 1 0 0 1 Total 55 4 1 7 67 Shared Rules, Similar S yntax The syntax between the FOIA and EO 13392 has a number of similarities. Neither source compliance unstated. Six statements match exactly between the two texts, as phrasing from the 2005 Executive Order was incorporated into the 2007 update for FOIA. Neither document con tains much in the way of permissive statements for FOIA requesters, instead burdening the agencies with compliance. In fact, as many executive orders do, EO13392 closes with a statement that the order does not create a right or benefit enforceable in law a gainst the government. As a consequence, though the order is clearly intended to improve the quality of FOIA servicing by federal agencies, access to information falls short of being enshrined as a right. The order specifies some additional reporting crite ria for the agencies, including narrative reports and plans from Chief FOIA Officers that address agency performance. These plans offer a potential source of insight into agency issues with providing transparency and servicing FOIA requests. The strategies included in the order are also significant. In the first few

PAGE 55

43 paragraphs, the order specifies that FOIA is to be treated as a service. This framing highlights the need to examine FOIA outcomes through a service lens, looking not just at requests granted or denied and records released or withheld, but also examining service performance issues like timeliness and quality of response. This suggests a qualitative coding approach to agency responses and interactions might yield useful insights. Institutional Gr ammar Conclusion Analysis of 5 U.S.C. § 552 and EO 13392 using the Institutional Grammar Tool proved useful to this research effort. The coded syntax of institutional statements provides a baseline for determining agency compliance with codified norms. Add itionally, analysis of the FOIA text leads to the conclusion that agency performance against the 20 day rule bears additional scrutiny and may be a key measure of performance. Analysis of the executive order suggests that service quality issues may also be worthy of examination. The literature suggests that the secret agencies will exhibit meaningful differences in their FOIA programs compared to their counterparts. Analysis using the institutional grammar coding suggests that examining differences in servi ce timeliness and quality may also yield fruitful results. Contemporary FOIA Research Contemporary research on the US FOIA has explored several hypotheses related to the effects of the administration and the effects of the NPM movement on FOIA execution by federal agencies. This dissertation builds upon this contemporary research base to explore new territory.

PAGE 56

44 Problems with FOIA: The National Security Archive at George Washington University ive (TNSA) performs monitoring of and selected research on FOIA results from across the federal government. Since 2003, the archive has periodically conducted the Knight Open Government Survey to query selected agencies on their oldest pending FOIA reques ts, and, since 2006, compares those results to the information contained in agency annual reports (TNSA, 2003; 2010 ; 2011 ). The survey information indicates that backlog of FOIA requests is a sizeable problem, particularly in agencies with a national secur ity mission and some requests remain open after more than a ; 2011 ). TNSA deduces that the referral process is partially to blame for delayed requests, as the receiving agency refers some or all of the requ est for records to another agency, adding processing delays and significantly complicating the records retrieval and declassification process (TNSA, 2003; 2010). Although useful as a potential indicator of problem areas, the methodology of the survey does not appear sufficiently stable or rigorous enough to draw firm empirical conclusions about agency performance nor were the conclusions subject to peer review. In addition to the Knight survey, TNSA conducted target research to examine the effects of change s in administration on FOIA implementation. Within the Federal Government, the Department of Justice, headed by the Attorney General, is responsible for the oversight of the FOIA program by executive agencies and departments. However, FOIA execution is pri marily decentralized, leaving considerable opportunity for discretion in interpreting the letter of the FOIA as written. In order to assist agencies and establish clear policy, the Attorney General issues guidance to the executive branch on how to apply FO IA, consistent with the policy goals of the President. During the Clinton administration, Attorney General Janet Reno issued

PAGE 57

45 guidance aimed at improving transparency to the effect that agencies should presume disclosure to the maximum extent possible (Kim 2007; TNSA, 2003). In 2002, under the George W. Bush administration, John Ashcroft issued a memorandum reversing this presumption and encouraging agencies to protect unclassified, but potentially sensitive information from disclosure by using the exempti ons established by FOIA (Kim, 2007; TNSA, 2003). In January of 2009, President Obama rescinded this guidance in place of direction similar to the Clinton administration, encouraging disclosure (WhiteHouse, 2009). TNSA conducted some analysis of annual repo rts and interviewed some agency FOIA personnel to assess whether or not these memoranda had substantive effects on federal government FOIA programs (TNSA, 2003; 2010). Additionally, TNSA work revealed inconsistencies in how agencies promulgated executive g uidance. Consistent with the administrative discretion literature, this conclusion suggests the possibility that agency FOIA processes and rules might vary significantly from established policy. The work done by TNSA points to the unmistakable conclusion t hat FOIA in the federal government is not functioning as intended, with exceptional delays in processing and was not eliminated by the 1974 amendments, but is, instea d, a persistent problem found in contemporary agency FOIA programs, although the extent and impact of agency backlogs is not immediately apparent. This dissertation will examine backlog data as a possible point of disparity between intelligence agencies an d their less secretive peers. Effects of A dministration C hange on FOIA outcomes: Kim (2007) Like TNSA, Kim (2007) also selected the effect of a change in administration as a research area, though with more rigorous methodology based on data from annual FO IA reports rather than documentation released under request. Kim evaluates the claim that the Ashcroft

PAGE 58

46 memo produced changes in agency disclosure patterns using data available in FOIA annual reports. Kim selected 25 agencies, including the 16 departments a nd 9 major independent agencies, including the CIA, using data from annual reports from 1998 to 2005. Kim found that agencies were less effective in processing requests under the Bush administration and backlogs increased. Additionally, Kim found that full grant rates had indeed decreased during the Bush administration and that the use of exemptions had increased, indicating the Bush executive guidance had indeed made the government more secretive compared to the Clinton Administration. sound, but is not readily repeatable due to reporting changes in 2008. Indeed, boundaries (i.e. 2004 vs 2010) may include unwanted variation due to administration change. This sugge sts that using data from only a single administration (i.e. the Obama Presidency, 2008 present) may help focus only on variables of interest and avoid problems with comparing rule ual report data provides an exemplar for using this source of data to draw empirically sound conclusions about FOIA programs. The E ffects of New Public Management on FOIA Programs : Piotrowski and Roberts As a reform philosophy, New Public Management (N PM) has had the potential to impact a wide range of government programs and functions. Piotrowski (2003; 2007) and Roberts (2000; 2002; 2006) independently examined the effects of New Public Management (NPM) on FOIA programs in the federal government, find ing that FOI programs were adversely affected by the reform movement. Roberts concluded that because the reforms included paring non essential functions and spending, FOI programs were a potential target of cost savings

PAGE 59

47 measures. Roberts found evidence of reduced effectiveness in Canadian FOI programs (2002) and that NPM had weakened the ability of citizens to monitor public institutions (2002), particularly weakly organized sectors (2000). Additionally, Roberts found national security networks to be of par Piotrowski adopted a case study approach, examining the effects of the National Performance Review (NPR) on FOIA. While NPR did not foc us specifically on FOIA, Piotrowski (2007) found unintended consequences of the reform activity on FOIA outcomes. Piotrowski found that FOI was often left out of performance management objectives (2003) and emphasis by agencies was on cost effectiveness ra ther than customer effectiveness (2007). Additionally, Piotrowski (2007) found variance in how FOIA programs perceived the need for customer service, and whether or not FOIA officers viewed requesters as agency customers. Given the NPM emphasis on contract ing, Piotrowski raises the question of how contractor records produced on behalf of the government, but not under government control are potentially out of reach for FOIA (2007). Additionally, Piotrowski found agencies applying administrative procedures to administrative procedures and look at contract resources in the FOIA offices as variabl es of interest. Implications for FOIA research Analysis of the institutional grammar of the FOIA and recent research provides a point of departure for examining FOI programs in the intelligence community. The syntax of the law indicates that compliance wi th the 20 day response rule bears examination. A careful look at

PAGE 60

48 EO13392 suggests that researchers should examine not just response outcomes, but also service backlogs is an important measure of the effectiveness of FOIA programs and should be considered in evaluating program health. Additionally, TNSA work reveals that the way agencies able as the records returned, leading to the conclusion that an experimental methodology might yield disposition rates across time periods and between agencies. Piotrows ki (2003; 2007) uses qualitative methods, particularly in examining performance measures, which might also be applied to analysis of responses and agency planning materials to generate insights on ies consider FOI programs to be non mission essential might be evaluated by examining the effects of alterations in program staffing mix on FOIA program performance. Absent from this contemporary research is any detailed analysis of how the intelligence co mmunity performs with respect to FOIA. Additionally, although discretion is clearly present as a factor in FOIA decision making, no extant research examines which or how factors affect the exercise of discretion in FOIA programs. Although the literature cl early frames FOI as a secrecy problem and some scholars identify the national security environment as an exemplar of secretive culture (Roberts, 2006; Shils, 1956), no research can be found that examines how FOI execution differs by national security agenc ies versus other agencies. These gaps point to the the exercise of administrative discretion.

PAGE 61

49 CHAPTER IV CONCEPT DEVELOPMENT AND HYPOTHESES Introduction This chapter connects the literature and research summarized in the Chapters II and III to the research methodology in Chapter V by presenting a model of transparency to be used for dev eloping operational definitions The literature referenced in Chapter II on secrecy, transparency and discretion provides the conceptual basis needed to convert principal agent assumptions into guiding propositions for research. The extant research in Chapter III on FOIA helps identify operational indicators that can be used to de velop testable hypotheses. This chapter offers a description of selected hypotheses and a rationale for selection. Transparency Model The literature provides the essential elements needed to construct a micro level model of transparency. Transparency can b e described as a mediated interchange between secret keeper/agent and stakeholder/principal (see Figure 4.1) One or more actors, acting as principal, desires information about the availability and workings of another actor in order to monitor the performance (Eisenhardt, 1989; Meijer, 2013). The organization, acting as agent, has cause to moderate and meter the availability of information about its activities to its principals. Individuals in secret keeping organizations are subject to institution al influences, including formal and informal rules (Bozeman, 1993; 2000), social norms (Sandfort, 2000) and collective biases (Dufresne & Offstein, 2008). Similarly, stakeholders, either collectively or individually, may hold a different appreciation of th e value of the information than the secret keeper, even without a specific knowledge of the information (Prat, 2006; Stigler, 1980).

PAGE 62

50 The observation of the agent by the principal is rarely direct, but rather occurs through a medium (Meijer, 2009). In the case of the federal government, this medium is largely through official records (Piotrowski, 2007; Roberts, 2006). The interchange is mediated in terms of both content (the what) (Pasquier & Villeneuve, 2007; Piotrowski, 2007; Roberts, 2006) and mechanism /method (the how) (Fung, Weil & Graham, 2007; Meijer, 2009). The literature suggests that secrecy restricts the availability of information, acting as a barrier between principal and agent (Weber, 1920/2009). These elements can be placed together in simpli fied model of transparency (see Figure 4.1) that illustrates the relationship between observed and observing actors. This simple model belies a set of complex interactions and potentia lly observable indicators. This research applies this model as a basis for understanding US int elligence agency FOIA outcomes and contributes towards refining this model. In particular, this research helps refine understanding of how secrecy affects transparency outcomes. The following section describes how the key conc eptual elements of the model will be operationalized for study.

PAGE 63

51 Refining Key Concepts into Operational Definitions Operationalizing Secrecy The work of Weber, Simmel, Shils and Bok provides the conceptual basis for secrecy. As the phenomenon of intentional ly concealing information, secrecy has several direct and indirect influences on information release. Secrecy is institutionalized by formal policies and rule sets to classify and restrict dissemination of categories and classes of information, but it can also create social influences on shared norms with the broad potential to impact information release beyond formally classified information (Simmel, 1906; Weber, 1920/2009). These social influences, in turn, directly influence the exercise of discretion by individual administrators. Detecting the presence of secrecy and distinguishing between degrees of secrecy can be difficult, however. The concept of official secrecy, i.e. state condoned or required secrecy, is difficult to operationalize, as there are se Oversight Office compiles an annual report to the President containing an estimate of the cost pri ncipal intelligence agencies is classified and excluded from the public report. Similarly, the budget (including the total size) for individual intelligence agencies is considered classified and rarely released and only then as an aggregate number represen ting multiple agencies. The exclusion of this type of data makes the task of operationalizing secrecy difficult at the organizational level. Nonetheless, an alternative way to measure the concept of secrecy is by selecting the core membership of the US in telligence community as the test population for being presumably more secretive than other domestic civil agencies. This assumption is critical

PAGE 64

52 to the definition of hypotheses as well as selection of research methods as it calls for a comparative approach to both. Testing Transparency elements: the mechanism or means for the principal to compel or request information, the compulsory or voluntary disclosure of information by the agent and the medium through which information is disclosed. FOIA provides a portion of the means for requesting information and nominal guidelines for disclosure decisions, but the agencies are substantially left to their discretion to select the medi um and make a host of decisions influencing disclosure. These decisions result in indicators observa ble at the request level by examining individual request responses and dispositions and at the program level by examining aggregate program results. The con temporary work on FOIA and analysis of the institutional grammar of the law helps identify potential operational indicators which can be used to test hypotheses. Comparing these indicators, using the intelligence agencies as the subject group and a sample of non intelligence agencies as the control group provides a relative method of comparing agency transparency. Program grant and denial rates provide one measure of relative transparency, although these are likely affected by additional variables. Examin ing quality aspects such as timeliness of response and backlog rates provides another measure of transparency as service. Many of these indicators derived from analysis of 5 U.S.C. § 552 and EO 13392 can be combined to form a relative scale of FOIA responsi veness.

PAGE 65

53 evaluation. Some observations can be captured by examining individual request responses, while others can be gathered by looking at program level aggregate data. Both data sources are useful in painting a fulsome portrait of agency transparency. Dealing with Discretion Discretion manifests itself in disposition decisions (e .g. fully grant, partially grant) and timeliness of response, but can also be observed in secondary decisions such as the assignment of fee category, determination of what artifacts constitute a record and the emphasis of rights individual response letters As the literature indicates, individual discretionary decisions are affected by the institutional context of the decision maker. Agency internal rule sets, which are typically not widely disseminated, can be obtained via FOIA request and provide a partia l glimpse into the complex system of internal rules affecting discretion. While FOIA processing rules and directions (such as guidelines for multi track processing or use of exemptions) can be expected to have a direct effect on FOIA decisions and outcomes other agency level rules such as record management schedules may also exert influence, perhaps unintentionally. transparency posture. This research aims to examine s ome of the aspects affecting discretion and individual decision makers by examining selected agency policies. Hypotheses The assumptions identified in Chapter II help influence development of guiding propositions that connect the research questions to ope rational level hypotheses. The conclusion from the literature that intelligence agencies are likely to over protect unclassified information provides directionality for most hypotheses. If over production of secrecy exists as suggested; it should be observ able in both program outcomes and individual decisions,

PAGE 66

54 suggesting a multi part research design to address research questions. The hypotheses below are grouped by research question and foreshadow selection of the supporting research design. As indicated ab ove, secrecy is operationalized by selecting a group of intelligence agencies as the subject group and non intelligence agencies as the control group. Evaluating Program Level Outcome Differences The first research question How do Intelligence Agency FOI A program processes and outcomes differ from their less secretive counterparts? explores the assumption that the institutional influence of secrecy will result in observables differences in program level outcomes. The core presumption is that the intellig ence agencies, acting as agent in a principal agent relationship, will select disclosure practices that are more restrictive than their less secretive counterparts. This function is not directly observable, however, but may be inferred by observing program outcomes. From the literature and extant FOIA research, it can be presumed that intelligence agencies would have lower information disclosure rates; delay disclosure decisions more and be more secretive about information disclosure. These presumptions are expressed as guiding propositions. The review of contemporary research in Chapter III provides operational indicators to convert these guiding propositions into testable hypotheses as shown in Table 4.1.

PAGE 67

55 Table 4.1 Research Question 1 Concept Summary Research Question Guiding Propositions Operational level Hypotheses How do Intelligence Agency FOIA program processes and outcomes differ from their less secretive counterparts? Intelligence agencies will restrict information disclosure more than other agencies H1: Intelligence agencies will have a lower percentage of fully granted requests H2: Intelligence agencies will exhibit higher partial grant and admin denial rates Intelligence agencies will delay disclosure more than other agencies H3: Intelligence agencies will have longer mean processing times and higher backlogs Intelligence agencies will be more secretive about information disclosure H4: Intelligence agencies will be less transparent about their FOIA program All four hypotheses are comparative in nature, with expected variation between intelligence agencies and other non intelligence agencies. These four hypotheses evaluate the variation at the program outcome level, i.e. the aggregate effects of multiple requests processed follows. a lower percentage of full y grant ed r equest s than other agencies even facto ring for exemption 1 and exemption 3 denials. Simmel and Weber suggest that secrecy exerts a powerful sociological force on organizations, effectively creating a barrier between keepers of secrets and those wanting to discover information. If this is corr ect, then agencies in which secrecy is a normalized organizational practice would be expected to use a variety of means to deter information seekers, resulting in lower than average full grant rates. At least by this one measure, these agencies would be co nsidered less transparent than their non secretive peers. To account for the possibility that requesters have asked for classified information, Exemption 1 (national

PAGE 68

56 security information) and Exemption 3 (other statutorily exempted information) denials can be accounted for as control variables, presumably leaving only requests for unclassified information for comparison. H2: Intelligence agencies will use partial grants and administrative denials to protect more information than other agencies. Among the m eans used by agencies is the ability to partially grant or use administrative denials to effectively deny information requests without appearing to rely on authorized exemptions. This hypothesis builds on H1 by exploring the means of denial other than deni al by exemption. This type of denial may be preferable to agencies over a denial by exemption, which can be easier to challenge through the appeals process or judicial review. mean processing times and higher backlog rates than other agencies complying with the letter of the law, while resisting the spirit (Archibald, 1979). Work by TNSA (2003; 2010) suggests tha t this is still a problem area. Extending the assumption that secret agencies would use a variety of means to deter information seekers, this hypothesis tests the hypo thesis will require accounting for program resources and request volumes to isolate secrecy the variable of interest H4: Intelligence agencies will be less transparent about their FOIA program than peer agencies.

PAGE 69

57 One of the potential variations between intelligence agencies and their peers is how transparent they are about transparency. The principal agent logic assumes that intelligence agencies will perceive the costs of transparency to be high and the benefits t o be low. Although this maybe evident in a number of areas, examining disclosure of information about transparency programs provides a useful comparison between agencies. Evaluating Individual Request Disposition Variances The second research question Ho w do Intelligence Agency disposition decisions differ from th eir less secretive counterparts explores the secrecy transparency problem at the individual request level. Comparing program grant and denial rates (as proposed for hypotheses 1 and 2) relies o n the assumption that the requests submitted to different agencies are similar enough to produce different outcomes based on variations between the programs. This assumption is suspect, at best, and difficult to verify. An experimental research design, whe re the requests are identical to different agencies, might more closely reveal variations in outcomes due to agency internal dynamics including norms and rule sets The same set of principal agent assumptions used for the first research question apply here and lead to similar guiding propositions as shown in Table 4.2 with the addition of a service related guiding contained in EO13392 for agencies to consider FOIA a key service provided by the government, exploring differences in how agencies provide FOIA as a service is appropriate. The literature supports the presumption that the FOIA services offered by intelligence agencies will differ markedly from the FOIA services provided by non intelligence agencies. Table 4.2 below presents the three hypotheses related to evaluating individual requests.

PAGE 70

58 Table 4.2 Research Question 2 Concept Summary Research Question Guiding Propositions Operational level Hypotheses How do Intelligence Agency disposition decisions differ from their less secretive counterparts? Intelligence agencies will restrict information disclosure H 5 : Intelligence agencies will issue more denials and less full disclosures in response to requests than the control agency set Intelligence agencies will delay disclosure H 6: Intelligence agencies will be more likely to issue delays and take longer to make release decisions than the control group Intelligence agencies will not offer high quality FOIA services H7: Intelligence agencies will perform worse on service quality indicators than the control group. H5: I ntelligence agencies will issue more denials and less full disclosures in response to requests than the control agency set. Comparable to hypotheses H 1 and H 2, this hypothesis tests the premise than secret agencies will restrict the release of unclassified information more than non secret agencies. By controlling for request variations, the experimental results should help isolate the effects of secrecy in practice in these agencies. The primary indicator is the disposition status and records released. H6: I ntelligence agencies will be more likely to issue delays and take longer to make information release decisions than the control agency set again Intelligence agencies are expected to take longer to res pond with a final decision to the experimental requests. This hypothesis supports h ypothesis H 3, which tests the effect at a program outcome level, by controlling for variation in the substance of the requests themselves.

PAGE 71

59 H7: Intelligence agencies will pe rform worse on service quality indicators than the control group. EO13392 raises the interesting question of how well agencies are providing FOIA as a service. Extending the principal agent logic to service provisioning, it is expected that intelligence ag encies would be less likely to prioritize FOIA as service as a key function. Some service aspects, such as time to respond can be easily evaluated. Other aspects, such as response quality, are more complicated, but can be evaluated by coding responses rece ived. Service quality has the potential to vary request by request and should be evaluated using individual requests rather than program level indicators. Exploring Institutional Rules By testing Hypotheses 1 7 the effects of secrecy can be evaluated, but not necessarily the causal mechanisms. The third research question seeks to understand the intervening influences of agency rule sets and norms on FOIA programs. Red Tape theory (Bozeman, 1993; 2000) and the discretion literature (Maynard Moody & Musheno, 2000; Sandfort, 2000) indicate that organizations will develop internal rule sets that may result in different effects than intended by policy. Cultural conditions and rules that deviate from established FOIA policy are expected to be present, but have the potential to influence disposition either positively (more information released) or negatively (less information). For the first two research questions, the literature and existing FOIA research supported a clear connection between guiding principles and operational indicators that allowed for hypotheses generation. This connection is more tenuous for the third research question, however, as a connection between specific agency policies and FOIA programs is not well

PAGE 72

60 represented either in the literature or in previous research. Some logical inferences can be made, however, based on the dependency on FOI execution on official records and the exclusion of classified information. In the case of the former, what an agency designates as a record directly impacts what the FOIA offices can release to requesters. More indirectly, an These inferences merit additional exploration. As Table 4.3 indicates, this leads to ex ploratory questions to further develop our understanding of how these specific agency policies relate to FOIA programs. Table 4.3 Research Question 3 Concept Summary Research Question Guiding Propositions Exploratory Questions How do FOIA related rules differ between intelligence agencies and their less secretive counterparts? FOIA programs are influenced by Records Management policies EQ1: How do Intelligence agency records management policies differ from non intelligence agencies? FOIA programs are influenced by CUI policies EQ2: How do intelligence agency CUI policies differ from non intelligence agencies? EQ1: How do intelligence agency records management policies differ from non intelligence agencies? This question examines the possibility of influence by a specific rule set namely, records management rules. Agencies maintain record management instructions, directives or policies to define what constitutes a record and direct employees on storage and management of records and records disposition schedules that indicate how long certain records should be kept. Not all of these policies or directions are commonly available via agency web sites, limiting the amount of a priori information available. Explo ration is needed to uncover potential areas of

PAGE 73

61 difference in polices and records management schedules between intelligence agencies and non intelligence agencies. EQ2: How do intelligence agency Controlled Unclassified Information policies differ from non intelligence agencies? Information in the federal government may be broadly binned into three major categories: Classified National Security Information, governed by EO13526, unclassified information and Controlled Unclassified Information (CUI) governed b y EO13556. This former category is a clear representation of official secrecy and explicitly exempt from disclosure under FOIA (although information may be declassified under certain circumstances). The middle category may be broadly construed as either in formation that has yet to be classified (literally un classified) or information designated for public release. The latter category is a gray area between the two and has a special relationship with FOIA. According to Department of Justice and NARA guidan ce, designation of a record as CUI does not automatically exempt it from disclosure under FOIA. However, it is unclear whether individual agencies and administrators adhere to this guidance. The expectation from the literature is that intelligence agencies will use a CUI policy to restrict release of information. Even more than records management policies, agency implementation of EO13556 and CUI rules is something of a black box. In fact, it cannot be readily determined whether or not agencies even have a President on CUI leads to the impression that CUI categories are widely used throughout the federal government, but efforts to create a comprehensive catalog of markings and the agencie s that use them is incomplete. Additional exploration is needed on this topic to identify areas of interest for potential research.

PAGE 74

62 CHAPTER V RESEARCH DESIGN Overview This chapter presents a research design for addressing the seven hypotheses and two explo ratory questions. The principal design for this research consists of three parts, a quantitative portion drawn from annual report data designed to evaluate FOIA program outcomes, an experimental portion to evaluate differences in agency responses to discre te requests and a qualitative portion to code results from the experiment and information from agencies websites The three part approach addresses different levels of questions, investigating program outcome level and individual request level questions In both cases, the unit of analysis is the agency, with the five core intelligence agencies (described below) and the Office of the Director of National Intelligence (ODNI) as the subject population representing secretiveness as concept. Hypotheses 1 3, w hich focus on FOIA program outcomes, will be addressed by examining data self reported by the agencies in their annual reports. This data set contains program level fields such as the number of fully granted or partially granted requests that can be normal ized into rates for between agency comparison. Hypothesis 4, which examines agency transparency about FOIA programs, requires between agency comparisons of a different sort, examining information available on agency web sites. Evaluating hypotheses 5 7, wh ich explore disposition decisions requires a methodology that targets the individual request level. An experimental approach in which the researcher sends identically worded requests to both subject and control agencies will generate information not avail able in the annual reports. The first order data gathered includes easily

PAGE 75

63 measurable information, such as disposition result (i.e. granted, denied) and time to return a disposition. The experiment will also yield second order qualitative information such a s the phrasing of request responses and content of informal exchanges between the experimenter and FOIA liaisons. Additionally, the experimental requests are expected to yield records that can be used for evaluating exploratory questions 1 and 2 related to FOIA related rules. Table 5.1 below provides a summary of the principal evaluation method in relationship to hypotheses. Table 5.1 Research Design Summary Research Question Area Hypotheses Principal Method RQ1: How do Intelligence Agency FOIA program processes and outcomes differ from their less secretive counterparts? H1: Intelligence agencies will have a lower percentage of fully granted requests Quantitative H2: Intelligence agencies will exhibit higher partial grant and admin denial rates Quantitative H3: Intelligence agencies will have longer mean processing times and higher backlogs Quantitative H4: Intelligence agencies will be less transparent about their FOIA program Qualitative RQ2: How do Intelligence Agency disposition decisions differ from their less secretive counterparts? H5: Intelligence agencies will issue more denials and less full disclosures in response to requests than the control agency set Experiment H6: Intelligence agencies will be more likely to issue delays and take longer to make release decisions than the control group Experiment H7: Intelligence agencies will perform worse on service quality indicators than the control group. Experiment RQ3: How do FOIA related rules differ between intelligence agencies and their less secretive counterparts? EQ1: How do Intelligence agency records management policies differ from non intelligence agencies? Qualitative EQ2: How do intelligence agency CUI policies differ from non intelligence agencies? Qualitative

PAGE 76

64 Annual Report Data used to Evaluate Program Outcomes Data Source FOIA Annual Reports The data for quantitative analysis of hypotheses 1 3, evaluating program outcomes, come from annual reports submitted by federal agencies. The FOIA requires agencies to provide to Congress each year a report containing basic, aggregate counts on the dispos ition of FOIA requests and backlog information. The reports contain data from independent agencies, some quasi governmental agencies, and components or subordinate agencies for all 16 federal departments. In the latter case, Department by Department variat ion exists in which elements submit reports In some cases, such as the Department of Defense, the reporting components are subordinate operating agencies while other departments use administrative structures (such as regional designations or a central off ice) to report data. For each fiscal year, each agency or reporting component with a FOIA program reports the number of FOIA requests pending at the beginning of the year, the number received, the number processed, the number fully granted, the number par tially granted, the number denied due to exemption, by exemption category (1 through 9) and the number denied due to administrative reasons, by cause. The reports are a rich source of raw information, providing request and disposition volumes, exemption an d administrative denial counts, and quality/efficiency measures such as backlog volume, median processing delays, and FOIA program resources ( see http://www.justice.gov/oip/reports.html for examples ). The reports provide information used to assess program outcomes, but do not shed light on individual requests.

PAGE 77

65 Since 2008, agency and component annual report data have been compiled on www.foia.gov a website administ ered by the Department of Justice to offer key FOIA information in a centralized location. This database provides organization year observations from 423 reporting components which can be used to construct a profile of federal government programs in aggre gate for comparative purposes. This source offers a complete data set, with all reported fields available. Data prior to 2008 cannot be easily compared to post 2008 data (M. Kim, personal communication, March, 29, 2013). Prior to 2008, some agencies report ed aggregate values for FOIA requests and Privacy Act requests together. Some agencies separated the values, but this was not consistently done. In FY2008 and beyond, the reports pertain to the FOIA program only Additionally, individual agency disclosure and posting of FOIA annual reports prior to 2008 is inconsistent and incomplete. For this reason, FY2008 was selected as the starting period for annual report data. The foia.gov record set consists of 1788 organization year dyads from 423 different reporti ng components for fiscal years FY2008 through FY2012. The data are partially unbalanced, as some entities did not report FY2008 data (the first year of data) and some departments reorganized during the data period, eliminating agencies or components. To av oid double counting, aggregate department values have been excluded (i.e. department level totals). The data have been cleaned to remove null observation rows (N=9) and eliminate department totals. The resultant data set consists of 1715 observations (orga nization year dyads) from over 400 reporting components for most of a five year period. The primary observations of interest consist of organization year dyads from the C entral I ntelligence A gency (CIA) Defense Intelligence Agency ( DIA ) the National Geospatial

PAGE 78

66 Intelligence Agency ( NGA ) the National Security Agency ( NSA ) and the National Reconnaissance Office ( NRO ) plus the Office of the Director of National Intelligence (ODNI). The first five agencies described the intelligence community and have statutorily specified intelligence gathering and analysis functions. Although nominally a headquarters organization, the ODNI acts as the center of the intelligence community and operates some multi agency operational f unctions, such as the National Counter Proliferation Center and the National Counter Terrorism Center. As such, these six agencies are presumed to have internalized secrecy more than peer organizations, owing to their unique intelligence mission and are th The other 417 reporting components in the data set serve as the comparison group. Though some of these components have national defense or homeland security functions, the set should sufficiently represent Variables Used The annual report data collected from www.foia.gov contains a number continuous and discrete variables for each agency year pair. Dummy categorical variables were created to represent Intelligence Community components. The key explanatory variable is a dummy variable representing the six intelligence agen cies described above. Six dependent variables were created to test the hypotheses. Table 5.2 summarizes the dependent variables selected. To test the premise of Hypothesis 1 that intelligence agencies grant fewer and deny more requests than their non intel ligence peers, two ratio variables were created to serve as dependent variables. The first, Released in Full Ratio, is the percentage of requests processed that were returned as fully granted or released in full, indicating that the

PAGE 79

67 agency determined that they had satisfied this request. Higher values represent more transparency, at least in the form of requests categorized as granted. The second ratio variable the Denied in Full Ratio, is the percentage of requests processed that were denied by asserting the rights of the government to protect the record of interest using a FOIA or other statutory exemption. Higher denied in full ratios means more information protected by the agency and less information released to requesters. These two measures are concep tually diametrically opposed, the first a measure of transparency, the second an assertion of official secrecy. Converting raw request counts into ratios allows for normalization between programs of different sizes and simplified reporting of coefficients. The dependent variables for Hypothesis 2 test the premise that intelligence agencies may make use of partial grants and administrative denials as a way of restricting information release. Partial grants represent selective release of information by partia lly fulfilling the request. For example, if several records are requested in a single request, the agency might consider the request partially granted if one or more records were released (e.g. documents from a single year) or if a portion of a record were released (even if the rest of the document was redacted). The annual reports reveal very little about this disposition category and leave much open to interpretation. As such, the part ial grants may also be viewed a s partial denials and represent a decisi on by the agencies to restrict disclosure. A Partial Grant R ate ratio variable was defined as the percentage of requests processed that were returned as partially granted. Administrative denials consist of discretionary rejections that are not based on exe mptions. In this case, the government is not asserting its right to protect information (as it is when invoking an exemption), but the agency has selected another reason for not granting the

PAGE 80

68 choice, the secrecy literature suggests that administrative denials would be used by intelligence agencies as a tool to limit information disclo sure. Administrative denials are counted differently than other denials and may be perceived differently by program administrators than a denial using an exemption. A n Administrative Denial R ate ratio variable was defined as the percentage of requests proc essed that were returned as partially granted. Hypothesis 3 is temporal in nature, testing the assumption that intelligence agencies may practice secrecy by delay. Two potential indicators of this assumption can be constructed FOIA request backlog, defined as a ratio of the number of requests not completed to the number of requests received. Backlog, the amount of requests agencies carry over from year to year, has been a source of concern for administrations seeking to improve transparency ( TNSA, 2003; 2010 ). Ideally, the ratio is low, reflecting only requests exceptionally difficult and/or time consuming to perfect and requests received late in the fiscal year. In a few cases (n=32), the ratio is greater than one indicating that the agency or component carries forward into the next fiscal year a queue of unperfected requests greater than the number of requests received that year by the agency. Higher ratios may be indicative of poor program performance or secrecy by delay. Th e second indicator is a continuous variable representing the average time taken by agencies to complete requests. Agencies do not directly report this value. Instead, the data contain mean values for completed simple requests and complex requests separatel y. However, a ratio of simple to complex requests can be calculated allowing an approximate mean to be constructed. The Mean Request Duration dependent variable is this approximate mean duration in business days taken to disposition requests. The work by G eorge Washington University

PAGE 81

69 suggests a relationship between the time to perfect requests and their disposition, with the implication that longer processing times were likely correlated to increased denials (TNSA, 2003; 2010). Table 5.2 Dependent Vari able Definitions Concept Variable Name Definition Hypothesi s Relative transparency Released in Full Ratio Ratio of requests fully granted (disclosed) to total requests dispositioned 1 Non disclosure Denied in Full Ratio Ratio of requests withheld using exemption to total requests dispositioned 1 Non disclosure Partial Grant Ratio Ratio of requests partially granted to total requests dispositioned 2 Non disclosure Admin Denial Ratio Ratio of requests not fulfilled for administrative reasons to total requests dispositioned 2 Secrecy by delay Backlog Ratio Ratio of requests backlogged at the end of the FY to requests received that FY 3 Secrecy by delay Mean Request Duration Mean duration in business days taken to disposition requests 3 A number of control variables were used to attempt to isolate the effect of the explanatory variable (See Table 5.3). To account for the possible influence of labor on the dependent variables, the Number of Full Time Employees and Equivalent number of Full Time Employees were added as independent variables. The first value captures the number of full time employees dedicated to processing FOIA requests. The second value captures non full time employees involved in FOIA processing, including part time employ ees or full time personnel

PAGE 82

70 who have FOIA responsibilities as a fraction of their job (Justice 2009). Conceptually, more employees on the program should expedite processing, resulting in lower Backlog ratios and shorter mean durations. Table 5.3 Independent Variable Definitions Variable Name Definition Use Intelligence Group Dummy variable for test group Explanatory variable No. FTE Number of full time employees employed by FOIA program Control variable No. Equivalent FTE Number of full time equivalent employees employed by FOIA program Control variable Ratio of Complex Requests Ratio of complex requests to total requests Control variable Processing Costs In dollars, agency reported costs of their FOIA program Control variable Litigation Costs Ratio of agency reported FOIA related litigation costs to total costs Control variable Mean Request Duration Mean duration in business days taken to disposition requests Control variable P revious research by George Washington University (2003) suggested that the complexity of a request could significantly impact both its disposition (i.e. granted, partially granted, denied) and the amount of time required to perfect the request. The annual report data reveals little about the nature of requests made to the agencies. One proxy measure available is the categorization of requests into simple and complex requests by agencies as part of their multi track processing rule set, though it appears that there are no standard criteria for this categorization Though agencies do not report this figure directly, it may be derived from other fields in the annual report data. The Complex Request Ratio is defined as the percentage

PAGE 83

71 of non 2 to total requests rec eived. This variable partially captures the concept of differential complexity, that some agencies routinely receive requests that are more difficult to satisfy than other agencies. The data contain values for processing and litigation costs incurred by ag ency FOIA programs. The value for processing costs is the sum of all labor, technical and administrative costs incurred by the agency to process requests and administrative appeals (Justice, 2009) This variable tested high for multicollinearity (likely du e to the interaction with the labor variables) and was dropped from regression models after diagnostics. The value for litigation costs is a ratio of the costs related to litigating FOIA requests to the total program costs (Justice Handbook). This variable captures an aspect of complexity in that higher litigation cost percentages likely represents increased contestability of agency decisions and may influence disposition results. These control variables should help isolate the effects of the explanatory va riable by accounting for the influence of resources and complexity on the dependent variables. Analy tical Techniques The initial analysis consists of interpretation of descriptive statistics, difference of means tests and correlations between the explanato ry variable and dependent variables. This analysis provides first order indicates of the validity of the hypotheses by estimating the size and significance of the effect of the explanatory variable (intelligence group membership). Stata software was used to perform all statistical analysis 2 may vary from agency to agency.

PAGE 84

72 To isolate the effects of the explanatory variable on the dependent variables, the researcher constructed Ordinary Least Squares (OLS) regression models containing the control variables. Each model was tested for multi collinearity, heteroskedasticity and influential outliers. Independent variables with Variance Influence Factor (VIF) scores higher than 4 were eliminated. Models were run with cluster ed standard errors to mitigate the effects of heteroskedascity. After th ese adjustments and the removal of influential outliers with a dfbeta score above |.25|, each model was re run. Models with bounded ratios as dependent variables were subject to an additional treatment and re run as fractional logit models. Each hypothesi s was evaluated by examining the initial analysis and the regression model outputs. While high R squared values for the regression models will provide confidence that the model adequately captures the variation; the direction, strength and significance of coefficients, particularly for the explanatory variable, will provide the best indication of the validity of the hypotheses. E xperimental Requests used to Evaluate Disposition Variances Experiment Design While the annual reports provide program outcome d ata, they do not shed light on how individual requests are handled, which more directly illuminates the role of administrative discretion in FOI decision making. Some agencies produce FOIA logs, detailing individual requests, but this is not a consistent p ractice. Within the intelligence community, only the ODNI and DIA voluntarily disclose logs on their website, and DIA logs do not contain disposition

PAGE 85

73 information. 3 As a result, there is no source available that provides stable data on individual requests t hat can be used for either case study or quantitative analysis. Additionally, comparing program outcome data as reported in the agency annual reports forces the researcher to make the assumption that the requests between agencies are comparable. This assum ption is not likely to be true and creates a challenge to the validity of conclusions drawn solely from the annual report data. Using an experimental approach that holds the requests between agencies constant can help supplement findings from the quantitat ive analysis portion of the design. The annual report data does not, providing insight on individual agency idiosyncrasies and process variances. To collect informat ion on the request process, three rounds of experimental requests were sent to the five core IC agencies and the ODNI as the subject group and 10 comparable agencies without an i ntelligence function and little to no national security function as the contro l group ( See Table 5.4). The ten control agencies were selected using stratified random Comparable programs were selected using the following parameters: FY2012 FOIA request volume between 80 and 4119 (+/ 10% from subject agencies), FOIA full time staff greater than or equal to 1, designation as a key component or element of a department on www.usa.gov and a status other than quasi official agency or government corporation 3 ODNI logs are available from 2006 to 2012 and contain request ID, requester name, subject, disposition and disposition date. DIA logs are available from 2001 to 201 0 and contain request ID, requester name and subject only.

PAGE 86

74 Table 5.4 Agency Selection Experimental Portion Test Group Abbr Control Group Abbr Central Intelligence Agency CIA Transportation Security Administration DHS TSA Defense Intelligence Agency DIA National Oceanic and Atmospheric Administration DOC NOAA National Geospatial Intelligence Agency NGA Defense Contract Management Agency DOD DCMA National Reconnaissance Office NRO United States Marshals Service DOJ USMS National Security Agency NSA Mine Safety and Health Administration DOL MSHA Office of the Director of National Intelligence ODNI Administration for Children and Families HHS ACF National Aeronautic and Space Administration NASA Office of the Comptroller of Currency Treas OCC Rural Development USDA RD United States Geological Survey USGS A three round design provides an opportunity for multiple observation points without being overly cumbersome to the agencies. These rounds use records returned, disposition decisions and time to return responses as variables for evaluating hypotheses H5: I ntelligence agencies will issue more denials and less full disclosures in response to requests than the control agency set, H6: I ntelligence agencies will be more likely to issue delays and take longer to make information release decisions than the control agency set and H7: Intelligence agencies will perform worse on service quality indicators than the control group. A dditionally, the records returned will be used to address exploratory questions EQ1: How do intelligence agency records management policies differ from non intelligence agencies? and EQ2: How do intelligence agency Controlled Unclassified Information policies differ from non intelligence agencies? Evaluation of these hypotheses and q uestions will help formulate an assessment of how

PAGE 87

75 transparent agencies are about transparency and how agency rule sets influence disclosure in practice. The request rounds were staggered by approximately 45 days to space out the requests and avoid the appe responding to requests is a non trivial action on the part of agencies (the average cost per request is $637 (Justice, 2012)), the requests were for records that are of inherent value in addressing the exploratory questions. The first round of requests was released on 25 Sept 2013, with rounds two and three approximately 45 and 90 days later, respectively. Each experimental round concluded 100 business days from the request date, allowing roughly five months for agencies to render a response. Table 5.5 Experimental Portion Summary Round Number Request Request Date 1 Records Management Policies and Schedule 25 Sept 2013 2 CUI Policy 12 Nov 2013 3 FOIA Logs 8 Jan 2014 Round 1 Records management policy and schedule. This round requests that agencies provide their records management program policy documents and their records disposition schedule. See Appendix A for the full text of the request. When FOIA was originally enacted, the transactions and decisions of American government were largely captured in the form of paper records. The access that FOIA provided to these records offered citizens and stakeholders significant insight into the result s, if not the working processes of government. In the four decades since the original FOIA enactment, the way in which government business is conducted and documented has changed significantly. The

PAGE 88

76 digital revolution facilitated a proliferation of electron ic communications that may fail to meet the threshold of record as originally envisioned by FOIA ( Roberts, 2006 ). Though storage of electronic documents and files is considerably cheaper than storage of an equivalent amount of hardcopy documents, storage c osts are non trivial and may be considerable for large agencies. As a result, not all information is designated as a record, and few records are designated for permanent storage by the National Archives and Records Administration (NARA). Informal interviews revealed that conflicting guidance regarding storage/retention of FOIA records could lead to materials being destroyed. For example, NARA publishes the General Records Schedule (GRS), a prescriptive guide indicating which records are designated for permanent storage and how long to retain temporary records. Per the GRS, FOIA annual reports are designated as permanent records scheduled to be transferred to NARA for safekeeping. However, both DIA and NGA FOIA officers indicated that their respectiv e agency policies designated these records as temporary and allowed them to destroy them after a few years. This indicates that there is the potential for variance in agency specific records management policies which may have direct impacts on transparency as agencies cannot release records which do not exist or have been destroyed. Agencies typically publish instructions or directives to their workforce to implement their records management policy and specify disposition instructions and retention duratio ns for specific records. Directive documents, if provided, provide d insight into directions to employees for the maintenance and retention of records and were reviewed for their connection to agency transparency efforts. The records management schedules fo r each agency, if provided, were used to identify agency specific policy variations. Though the records management schedules and policy documents are presumably unclassified and subject to full

PAGE 89

77 release, intelligence agencies are expected to delay and restr ict the release of these documents more than the control group. Round 2 Controlled Unclassified Information (CUI) p olicy. This round requests that agencies provide documents that describe t he agency policy or on use of Controlled Unclassified Informatio n markings such as For Official Use Only (FOUO), Sensitive But Unclassified (SBU) or Law Enforcement Sensitive (LES). See Appendix A for the full text of the request. Though not every agency maintains classified information, many agencies potentially mak e use of CUI markings to restrict information dissemination. These markings represent a identified that there were over 100 agency specific rule regimes governing CUI in the Federal government. The 2009 report by the inter agency Task Force on Controlled Unclassified Information (CUITF, 2009) indicated that several agencies had used CUI markings incorrectly to deny release under FOIA. Though E xecutive O rder 13556, issued in 2010, directs agencies not use CUI markings as a justification for denial of FOIA requests, some agencies may still do so and may invoke an exemption to withhold documents marked with CUI. This round tests the premise that intelligence agencies a re more likely to make use of internally generated rule sets, like CUI, to restrict the release of information than agencies in the control set. Round 3 FOIA logs. This trial round requests that agencies disclose FOIA logs from FY2008 FY2012. See Appendi x A for the full text of the request. Agency logs will possibly help identify patterns in FOIA requests for a given agency and the prevalence of requests from different sectors (i.e. academic, commercial, press, etc.). Additionally, the records requested in this round all have the potential to be exempted from release using exemption 2, which allows agencies to withhold records they think are solely

PAGE 90

78 related to internal practices. This exemption is highly discretionary, however, and variation is expected be tween the intelligence agencies and control groups, with the former using exemption 2 more frequently to deny release of requested records in this round. Anticipated experiment outcomes. The experiment yield ed dozens of individual responses. The responses to the request are just as significant as any records released. Of primary importance is whether or not the request was granted, but when and how the agency responds is also usefu l data. The time an agency took to respond provided one indicator of how well the agency is pro viding FOIA as a service and determine d whether or not the agency is compliant with the requirement to respond to requests within 20 business days. The medium of the response and any records returned were also viewed as indicators of service Additionally, statements in the response were coded to measure service quality dimensions. In addition to the responses, the experiment is expected to yield some records for qualitative evaluation. Not all agencies are expected to return records, however, resulting in an incomplete set. Experimental Analysis All responses and correspondence from agencies were logged. For each trial round, results from the test group, including response times and request disposition, were compared against results from the control group. The primary analytical technique is the comparison of results from the test group against results from the control group. Some aspects, like time to return a response, lend themselves to easy direct comparison, while other aspects, such as record responses, require a more nuanced interpretation.

PAGE 91

7 9 Hypothesis 5 Intelligence agencies will issue more denials and less full disclosures in response to requests than the control agency set will be evaluated by comparing disposition results between the intelligence agencies and the control group. The intelligence agencies are expected to issue more denials and partial grants than the control group. Hypothesis 6 Intelligence agencies will be more likely to issue delays and take longer to ma ke release decisions than the control group ; will be evaluated by comparing response times between the intelligence agencies and the control group. Intelligence agencies are expected to have a higher mean duration and a more frequent use of delay notices t han the control group. The written responses from the agencies contain valuable information, regardless of any records supplied. Agencies we re expected to issue at least one response, but in many cases released two or three responses to each request. Agenc ies have broad latitude in determining how to respond to individual requests, particularly in selecting the medium for communication hypothesis 7 Intelligence agencies will perform worse on service quality indicators than the control group, responses were coded for a number of quality indicators derived from rules found in the FOIA text, in EO 13392 and OGIS best practices (See Table 5.6). These indicators measure a number o f aspects of service quality. The first indicator assesses whether or not the agency provided either an initial or final response within the 20 business day period required by the Act. The second indicator assesses the relatively simple requirement to pro vide requesters with a tracking number to reference the request. The tracking number is used internal to the agency for tracking and record keeping purposes. The third and fourth indicators assess the responsiveness using electronic means. If the agency provided any human generated correspondence via e

PAGE 92

80 means of tracking the requ est status. The final two indicators assess if the agencies provided the fifth indicator if they provided information related to the request in the response tha t supplemented their statement of disposition (granted, denied, released in part). Agencies than an agency mailing address) for requester questions or issues. T able 5.6 FOIA Service Scale Indicator No. Question Source 1 Did the agency respond with 20 days? 5 U.S.C 552 (a)(6)(A)(i) 2 Did the agency offer information along with the disposition? EO13392 (1)(b), OGIS Best Practices 3 Did the agency respond electronically? 5 U.S.C 552 (a)(3) 4 Did the agency provide a tracking number? 5 U.S.C 552 (a)(7)(A) 5 Did the agency provide electronic means to obtain status? 5 U.S.C 552 (a)(7)(B) 6 Did the agency provide a point of contact for questions? OGIS Best Practices The researcher coded responses to requests received in each round against the questions in the service scale and compared the intelligence agency performance to the control group. Results of the experiment are detailed in Chapter VI I. Each round was analyzed as a discrete event. The analysis concludes with a summary of key findings from the experiment as whole. Agency Records and Artifacts Support Exploratory Questions Measuring the quantity and quality of information that federal agencies s hare with the public is a challenging affair that must be addressed piecemeal. Annual report data provides an

PAGE 93

81 aggregate measure of how often requested records are being released, but not necessarily the quantity of material released or the quality of servi ce provided. The experiment provides a partial glimpse into both quantity and quality of disclosure, but only in response to requests. Qualitatively examining other aspects of transparency can help supplement the quantitative and experimental findings by p roviding more insight into the instrumental nature of transparency. Transparency about Transparency How do Intelligence Agency FOIA program processes and outcomes differ from their less secretive counterparts? for hypotheses H1, H2 and H3 uses a quantitative approach Intelligence agencies will be less transparent about their FOIA program qualitative coding approach. The literature referen ced in Chapter II leads to the conclusion that FOIA programs in government agencies are placed in the special position of being both the object and instrument of transparency as records might be held by contractors or organizations within the agency. This role offers an opportunity to compare how transparency agencies are about their transparency programs, adding another perspective towards answering the first research question. T he norms in the FOIA text, EO 13392 and best practices documented by the Offic e of Government Information Services (OGIS) leads to a set of questions that can help provide a scale of transparency. This scale (shown in Table 5.7) uses seven indicators to assess the relative transparency of individual agencies on the topic of transpar ency. The first two indicators, derived from the Act and EO 13392 assess whether or not the agency posts the most recent copy of their annual report and their Chief FOIA Officer report on their public facing website. These reports provide the public, press and oversight stakeholders with information about the

PAGE 94

82 agency maintains an electronic reading room accessible to the public and whether or not that reading room pr ovides frequently requested records. In an effort to promote proactive transparency, the Act requires agencies to post certain information and frequently requested records in an electronic reading room. Indicators 5, 6 and 7 assess whether or not the agenc ies provide information that might help requesters submit properly formatted requests for valid records. Table 5.7 Transparency about Transparency Scale Indicator No. Question Source 1 Does the agency make the most recent annual reports available online? 5 U.S.C 552 (e)(3) 2 Does the agency make the most recent Chief FOIA Officer reports available online? EO13392 (3)(c)(i) 3 Does the agency have an Electronic Reading Room accessible to the public? 5 U.S.C 552 (a)(2) 4 Are frequently requested records displayed? 5 U.S.C 552 (a)(2)(D) 5 Is the agency policy/regulation for FOIA available online? OGIS best practices 6 Does the agency makes records management policies available online? OGIS best practices 7 Does the agency provide a handbook for requesters? 5 U.S.C 552 (g)(3) The researcher used the same sets of agencies selected for the experimental rounds to sites. Agencies were ranked by the total number of positive answers. The expected direction for the hypothesis is that intelligence agencies will rank lower on the scale than agencies in the control group. Records Analysis Answering EQ1 How do Intelligence agency records management policies differ from non intelligence ag encies? and EQ2 How do intelligence agency CUI policies differ from

PAGE 95

83 non intelligence agencies? requires records returned during the experimental portion. However, it cannot be assumed that all agencies will return records. Although the records requested are presumed to be unclassified and eligible for release, this may not be true of every agency (which is itself, a valuable data point). In fact, the literature suggests that agencies in the test group are less likely to release information or have a high er probability of delaying release outside of the experiment period. As a result, analysis of the records is likely to be inconsistent. Records returned from trial round #1 (Records Management Policies and Schedules) offer several points for evaluation. The researcher compared purpose statements in the policies, looking for demonstrable differences in policy scope or framing, particularly in terms of how records management relates to information disclosure. Items of interest in the records management sche dule include retention durations of FO IA related materials, such as annual reports and FOIA logs The goal of this analysis is to identify potential areas of examination for future research. A similar method was used to evaluate records returned from trial round #2 (CUI Policy). The relationship between CUI policies and information disclosure in general and FOIA in particular is unclear. Analyzing records returned in this round helps shed light on potential aspects of the relationship that merit deeper inv estigation. Records returned from trial round #3 (FOIA logs) offer great promise, but also have the possibility for significant between agency variation. Agency logs were compared against each other for relative transparency, examining how much information is revealed by the logs.

PAGE 96

84 Threats to D esign Secrecy Operationalization The research design rests on the assumption that intelligence agencies are indeed more secretive than non intelligence agencies. This is a challenging assumption to test. Because of the organizational structure of the intelligence community, it is impossible to construct a complete community aggregate set (see Appendix B ) as only 7 of the 17 members produce discrete FOIA data or reports. As such, the set of the five core ag encies plus the Office of Director of National Intelligence provide the best opportunity to observe intelligence agencies as a subject population. Measuring Transparency The research design attempts to assess the concept of transparency in a number of ways The design rests heavily on the presumption that responses to FOIA requests can be used as a partial measure of transparency. Several facets of these responses are captured, examining definition. The FOIA responses capture well elements of what Heald (2006) described as event transparency. Process transparency is much more difficult to evaluate, but is partially captured through the indicators on the FOIA servic e scale. The use of a transparency about transparency scale provides another possible measure of transparency, partially capturing the concept of proactive transparency. Taken together, these measures provide a defensible representation of transparency as a concept.

PAGE 97

85 Bias and the use of Mixed Methods FOIA program outcomes, as measured by annual reports from the agencies, likely include bias from the requesters themselves. That is, some agencies might attract more frivolous or difficult requests than other agencies, potentially skewing results at the aggregate level. While this bias is not correctable on a large scale, the use of an experimental design, which holds the requests constant across agencies, provides some small scale triangulation. Additionally, accounting for Exemption 1 and Exemption 3 denials offers the possibility of removing the influence properly classified information from the results, offering a more apt comparison between agencies. Each experimental request bears the possibility of idiosy ncratic responses from individual FOIA officers, though the use of multiple rounds provides multiple observation points and lessens the possibility that a single outlier response will influence the results. The use of qualitative coding will assist in fill ing in parts of the picture not provided by quantitative analysis alone. Taken as a whole, the research design will evaluate program outcomes, individual disposition responses and a measure of transparency in organizational norms, providing a fulsome portr Data and Information Accuracy The research design is heavily dependent on source data self reported by the agencies. In the case of the annual reports, the reported values are assumed to be accurate. Responses return ed from experimental requests are assumed to be valid, regardless of their accuracy. For instance if an agency responses that no records are available, this is a considered a valid and useful response, regardless of its factual accuracy. Experimental resp onses will not be appealed.

PAGE 98

86 The experimental nature of this portion involves some unknowns and may result in unexpected outcomes. Small n Evaluation Selecting only six agencies/offices as representative of the intelligence community reduces the statistical power of the multivariate analysis. However, the selection is justified given the impossibility of constructing an intelligence community wide set. By using multiple years of data, the small n problem is ameliorated, though not completely eliminated. The use of mixed methods provides additional confidence in the results and will help strengthen analytical conclusions.

PAGE 99

87 CHAPTER V I TESTING HYPOTHESES U SING ANNUAL REPORT D ATA Introduction A portion of the research design calls for comparing intelligence age ncy FOIA program Data from annual federal agency reports provide d the basis for this comparison. The data consist of 1715 organization year dyads from agencies and co mponents across the federal government from the fiscal years 2008 to 2012. The analysis contained in this chapter uses this data set to evaluate relationships of interest and draw preliminary conclusions. Specifically, the data highlight the relationship b etween intelligence agency membership and aggregate disposition outcomes (e.g. percentage of requests fully granted, denied or partially granted) as well as aggregate processing results (e.g. backlog percentages and average time to disposition).

PAGE 100

88 Table 6.1 Dependent Variable Overview Hypotheses Variable Name Definition H1: Intelligence agencies will have a lower percentage of fully granted requests Released in Full Ratio Ratio of requests fully granted (disclosed) to total requests dispositioned Denied in Full Ratio Ratio of requests withheld using exemption to total requests dispositioned H2: Intelligence agencies will exhibit higher partial grant and admin denial rates Partial Grant Ratio Ratio of requests partially granted to total requests dispositioned Admin Denial Ratio Ratio of requests not fulfilled for administrative reasons to total requests dispositioned H3: Intelligence agencies will have longer mean processing times and higher backlogs Backlog Ratio Ratio of requests backlogged at the end of the FY to requests received that FY Mean Request Duration Mean duration in business days taken to disposition requests This chapter begins with a discussion of the descriptive statistics from the data and a preliminary analysis of means differences and correlations of interest. The chapter continues with analyses of the multivariate regression models performed for each of the dependent variables listed in Table 6.1. The chapter concludes with a discussion of the limitat ions of the data and a summary of findings. Descriptive Statistics and Preliminary Analyses The descriptive statistics from the annual report data provide insight into the federal wide variance between agencies. Some agencies report receiving no requests, while others receive hundreds of thousands each year. As a panel set containing organization year dyads for a five year period,

PAGE 101

89 the data contain values for key variables to help as sess how FOIA programs in intelligence agencies differ from programs in non intelligence agencies. Mean values of the dependent variables, as measured through annual report data, DIA, NGA, NSA, NRO and ODNI and the reporting population as whole. Table 6.2, shown below, displays the mean values for the whole reporting population, the control group and the intelligence group for the selected set of dependent variables. The table di splays some significant differences between released in full and denied in full rates as well as backlog percentages between the control group and the intelligence group consistent with the expectations of the hypotheses. Table 6.2 Descriptive Statistics Annual Report Data Population Control Group Intel Group Relevant Hypothesis N (number of organization year dyads) 1715 1688 27 Mean Released in Full Rate 32.28% 32.26% 9.20% 1 SD 0.2324 0.2323 0.0554 Mean Denied in Full Rate 4.77% 4.53% 19.28% 1 SD 0.0788 0.0741 0.1724 Mean Released in Part Rate 23.79% 23.71% 28.59% 2 SD 0.1864 0.1871 0.1384 Mean Admin Denial Rate 39.13% 39.07% 42.91% 2 SD 0.2273 0.2279 0.1845 Mean Backlog Ratio 16.64% 15.45% 89.73% 3 SD 0.3273 0.2869 1.0590 Mean Request Duration (in days) 54.7233 51.8569 233.9276 3 SD 90.4104 82.7687 245.5190

PAGE 102

90 For the population as a whole (i.e. all federal agencies during the period), most requests were granted, with an average Released in Full Ratio of 32.28% and a mean Released in Part Ratio of 23.79%, resulting 56.07% of requests returned with some amount of records release 4 However, the mean values of the dependent variables for the intelligence group tells a different story, with just 9.20% of requests released in full and 28.59% of requests partially granted for a combined ratio of 37.79% of requests returned with some information. The intelligence group uses FOIA statutory exemptions to deny requests at a h igher rate than their peer agencies, returning 19.28% of requests as Denied in Full, compared to 4.77% of requests denied by the total population. The intelligence group also carries a higher mean backlog rate equal to 89.73% of requests received annually, compared to 16.64% for the population as a whole. An independent sample t test was conducted to compare the differences between intelligence agencies and non intelligence agencies in the dependent variables. The t test results (shown in Table 6.3) show that for the Released in Full and Denied in Full Rates the differences between the intelligence group and a control group sample was significant at the p<.001 level, lending support to Hypothesis 1. Similarly, the differences for the Backlog Ratio and the Mean Duration of Requests were also significant at the p<.01 and the p<0.001 level, respectively, supporting Hypothesis 3. Contrary to the expectations for Hypothesis 2, however, neither the Released in Part or Denied using Administrative reasons rates showed a statistically significant difference between groups. This comparison of means point to significant variances between the intelligence group and the rest of the population for variables supporting Hypothes es 1 and 3. 4 Interestingly enough, the DOJ and the www.foia.gov website exclude Administrative Denials from the calculations of Released in Full, Released in Part and Denie d in Full Ratios, resulting in much higher apparent Released in Full and Released in Part ratios.

PAGE 103

91 Table 6.3 Difference of means Intelligence Group vs Control Sample t df P Released in Full Rate 5.5273 43.0649 0.0000 Denied in Full Rate 3.9115 26.1575 0.0004 Released in Part Rate 0.9872 27.5781 0.3283 Admin Denial Rate 1.8090 27.3153 0.0763 Backlog Ratio 3.6136 26.0623 0.0012 Mean Request Duration 3.6928 26.0947 0.0009 Note: T test performed using N=30 sample control group with unequal variances A point bi serial correlation was performed to identify the relationship between the explanatory variable (a dummy variable for the intelligence group) and dependent variables. Table 6.4 displays the results. Consistent with expectations for Hypothesis 1, the correlation for Released in Full Rate is negative and significant at the p<.001 level and the correlation for Denied in Full Rate is positive and significant at the p<.001. The results for Released in Part Rate and Admin Denial Rate show weak coeffici ent values and no statistical significance, likely indicating that Hypothesis 2 is incorrect. Consistent with expectations for Hypothesis 3, the correlation for Backlog Ratio and Mean Request Duration are positive and significant at the p<.001.

PAGE 104

92 Table 6.4 Point Bi serial Correlation Results versus Intelligence Group Dummy Variable r pb t P Released in Full Rate 0.1270 5.2402 0.0001 Denied in Full Rate 0.2355 9.9125 0.0001 Released in Part Rate 0.0329 1.3485 0.1779 Admin Denial Rate 0.0212 0.8688 0.3851 Backlog Ratio 0.2854 12.7953 0.0001 Mean Request Duration 0.2507 10.7176 0.0001 Both the difference of means tests and the point bi serial correlation results provide support to Hypothesis 1 and 3 and indicate that the null hypothesis is more likely true than Hypothesis 2. However, these tests do not account for the possible influence of other variables such as resources expended or complexity of requests. Regression models that include independent variables to isolate the primary relationships of interest will be used to examine each hypothesis to isolate the effects of the explanatory variable. Analyses for Hypothesis 1 Intelligence Agencies Grant Fewer Requests programs will have a lower percentage of fully granted requests than other agencies, even factoring for exemption 1 and exemption 3 denials. The descriptive statistics indicate a strong variance between the intelligence group and the rest of the reporting population with respect to requests Released in Full and Denied in Fu ll due to Exception. The data show strong significant correlations in the direction that supports H1. As described in Chapter V, these dependent variables may also be influenced by the le vel of resources allocated by agencies to service requests, the complexity of the requests and the time

PAGE 105

93 taken to process requests. Multivariate models will help account for this potential influence and isolate the effects of the explanatory variable. Intel ligence Agencies Exhibit Lower Released in Full Rates The point bi serial correlation suggests that membership in the intelligence group negatively affects the Released in Full Rate by 12.7% (p<.001). To confirm the effect of the explanatory variable on th e Released in Full Ratio as a dependent variable, a regression model was developed that incorporates the independent control variables discussed in Chapter V. The model was tested using Ordinary Least Squares (OLS) regression clustered at the component le vel. The model passed tests for multicollinearity in independent variables. Heteroskedasticity is present and partially addressed by using robust standard errors. Multiple outliers were identified during diagnostics and these organization year dyads were r emoved in a subsequent run 5 Table 6.5 below shows these results. The regression model shows effect sizes and directions similar to the point bi serial correlation. With the outlier dyads removed, the regression model indicates that the explanatory variab le has a moderate negative effect of 17.75% on the Released in Full Rate, with strong significance (p=.000). 5 Outliers were selected by examining DFBETA results across multiple variables. Results >.25 were removed.

PAGE 106

94 Table 6.5 Regression Results versus Released in Full Rate Clustered OLS Outliers Removed Dummy: Intelligence Group 0.1522 *** 0.1775 *** (0.0418) (0.0381) No. Full Time Employees 0.0010 ** 0.0010 *** (0.0003) (0.0003) Equiv. Full Time Employees 0.0018 ** 0.0021 ** (0.0006) (0.0007) Litigation Related Costs 0.2351 *** 0.2925 *** (0.0629) (0.0685) Complex Request Ratio 0.0371 0.0348 (0.0252) (0.0268) Average Processing Time in Days 0.0004 *** 0.0005 *** (0.0001) (0.0001) _cons 0.3643 *** 0.3657 *** (0.0159) 0.0166 N 1606 1562 R sq 0.0876 0.0915 Standard Errors in parentheses p<.05, ** p<.01, *** p<.001 The model shows relationships with the independent variables that match expectations. The low r squared value for the OLS model indicates that this model has relatively little utility as a prediction aid and that the variables included in the model explain only a small portion of the variance present in the dependent variable. The use of a bounded percentage as the dependent variable violates OLS assumptions. To counter this violation, a fractional logit regression was performed to verify the impact and si gnificance of the explanatory variable. The fractional logit results show a negative odds

PAGE 107

95 relationship ( = 1.1 502 ) between the explanatory variable and the dependent variable significant at the p<.001 level. Expressed simply, being in the intelligence gr oup decreases the odds of a request being released in full by 31.66% compared to the control group. All three of these results support H1. The dummy variable for the Intelligence Group shows negative coefficients in the expected direction across the point bi serial correlation, the OLS regression and the fractional logit regression, all with moderate to strong significance. This suggests that being a member of the intelligence group exerts a moderate and negative effect as Hypothesis 1 anticipated. In sum, the data show that Intelligence Agencies have lower Granted in Full rates than other agencies controlling for resources, complexity and time to process. Intelligence Agencies Exhibit Higher Denied in Full Rates Logically, the lower release rates should be associated with an increase in denials, partial grants or administrative denials. The data indicates that Intelligence Agencies have a greater proportion of Denied in Full requests compared to other agencies. The point bi serial correlation from table 6.4 shows a positive relationship between the dummy explanatory variable and the Denied in Full rate. As expected, this effect is strongly positive (23.55%) and significant at the p<.001 level. The regression model shows a slightly weaker effect, equivalent t o a 18.28% increase in Denial Rates for the Intelligence Group with moderate significance (p=.010). This finding supports H1 as expected.

PAGE 108

96 Table 6.6 Regression Results versus Denial Rate Clustered OLS Outliers Removed Dummy: Intelligence Group 0.1478 0.1828 ** (0.0660) (0.0705) No. Full Time Employees 0.0001 0.0001 (0.0001) (0.0001) Equiv. Full Time Employees 0.0003 0.0002 (0.0002) (0.0002) Litigation Related Costs 0.0352 0.0383 (0.0197) (0.0195) Complex Request Ratio 0.0118 0.0001 (0.0086) 0.0069 Average Processing Time in Days 0.0000 0.0000 (0.0000) (0.0003) _cons 0.0410 *** 0.0417 *** (0.0046) (0.0043) N 1606 1565 R sq 0.0660 0.0930 Standard Errors in parentheses p<.05, ** p<.01, *** p<.001 Removing outlier organizations improved the model slightly. Like the previous regression results, the low r squared value indicates that this model captures a small percentage of the variation in the dependent variable. Unlike the previous models, the independent variables do not display reportable significance in this model. While resources and time to process affect Released in Full Rates these variables did not have a significant effect on Denied in Full rates.

PAGE 109

97 The fractional logit regression indicates a strong positive relationship between the explanatory variable and dependent variable with a log coefficient of 1. 836 (p<.001). This can be interpreted as the odds of an intelligence agency denying a request are 6.27 times higher than a non intelligence agency with identical characteristics. This result is consistent with expectations and suggests that intelligence agencies are more likely to make sure of statutory exemptions to deny requests than other agencies with similar characteristics. Discussion and Assessment The data strongly support confirmation of Hypothesis 1 as intelligence agencies exhibit lower released in full rates and high er denied in full rates than non intelligence agencies. The findings are consistent with the expected direction indicated from the literature. The regression results on Tables 6.5 and 6.6 reflect all grants and denials, including requests made for classif ied information or information protected by law. With these requests included the data, the results reflect official secrecy as well as administrative secrecy and do not necessarily improve our understanding of secrecy as phenomenon. However, factoring out these requests proved problematic. The annual report data includes counts of the exceptions used in fully denying or denying requests. The Department of Justice guidance (Justice, 2008, 2013) to agencies is to report the count of requests with the exempti on in question used, rather than the count of the number of times an exemption was applied. Some over reporting is present, castings doubts that the counts are accurate and indicating that some agencies may have used the latter reporting method. Regression models performed using a modified denial rate that excludes exemption 1 (classified national security information) showed extremely poor measures of fit, with adjusted R squared values less than 0.0171 and wide confidence intervals.

PAGE 110

98 These results are not statistically relevant and should not be used to inform assessment of the hypothesis. The regression models provide some support for Hypothesis 1 with t he caveat that the data include requests for information that agencies are prohibited by law from disclo sing. Thus the effects show in Tables 6.5 and 6.6 are valid, that intelligence agencies grant fewer and deny therefore only partially capture the question of int erest. Analyses for Hypothesis 2 Intelligence Agencies use Partial Grants and Administrative Denials Hypothesis 2: Intelligence agencies will use partial grants and administrative denials to protect more information than other agencies. The descriptive statistics show only a few percentage points of difference in Released in Part and Administrative Denial rates between the intelligence group and the reporting population, suggesting this hypothesis may be incorrect. Likewise, the point bi serial correlations are weak with no statistical significance. The multivariate regression models used for testing Hypothesis 1 were used to evaluate possible relationships of interest for Hypothesis 2, using Released in Part (Table 6.7) and Administrati ve Denial (Table 6.8) rates as dependent variables. Analysis Released in Part as a Dependent Variable Analysis of Released in Part ratio as a dependent variable produced very weak results. The model produced extremely low adjusted R squared values, sugge sting that the model variables are not adequately capturing variation of interest. The dummy variable for the

PAGE 111

99 intelligence group has a small coefficient and does not show reportable significance (p=.203), suggesting little to no influence. The fractional l ogit model also failed to show significant results for the explanatory variable (p=.173). These results suggest that no influence of interest is likely present and the null hypothesis is more likely to be true. Table 6.7 Regression Results versus Partial Grant Rate Clustered OLS Outliers Removed Dummy: Intelligence Group 0.0563 0.0605 (0.0466) (0.0474) No. Full Time Employees 0.0004 0.0001 (0.0005) (0.0006) Equiv. Full Time Employees 0.0010 ** 0.0007 (0.0003) (0.0003) Litigation Related Costs 0.0412 0.0737 (0.0547) (0.0487) Complex Request Ratio 0.0254 0.0185 (0.0189) (0.0192) Average Processing Time in Days 0.0000 0.0000 (0.0001) (0.0001) _cons 0.2377 *** 0.2380 *** (0.0121) 0.0120 N 1606 1552 R sq 0.0113 0.0100 Standard Errors in parentheses p<.05, ** p<.01, *** p<.001 Analysis Administrative Denials as a Dependent Variable As Table 6.8 indicates, regression models using Administrative Denial as the dependent variable showed similar results to the Partial Grant model. The regression models did not indicate a significant relationship between the Intelligence Group (p= .186) and Administrative

PAGE 112

100 Denials. This model did not exhibit multicollinearity or heteroskedas ti city. The fractional logit model showed similar results for the explanatory variable (p=.192), also suggesting that no influence of interest is likely present. Table 6.8 Regression Results versus Admin Denial Rate Clustered OLS Outliers Removed Dummy: Intelligence Group 0.0520 0.0552 (0.0421) (0.0417) No. Full Time Employees 0.0008 0.0012 (0.0006) (0.0009) Equiv. Full Time Employees 0.0011 0.0016 (0.0005) (0.0008) Litigation Related Costs 0.2411 *** 0.2717 *** (0.0723) (0.0734) Complex Request Ratio 0.0001 0.0124 (0.0255) (0.0265) Average Processing Time in Days 0.0004 *** 0.0005 *** (0.0001) (0.0001) _cons 0.3569 *** 0.3519 *** (0.0133) 0.0136 N 1606 1546 R sq 0.0669 0.0671 Standard Errors in parentheses p<.05, ** p<.01, *** p<.001 Discussion and Assessment These regression models fail to show significance for the explanatory variable. The very low adjusted R squared value of the models and high size of the constant relative to the coefficient of the dummy variable strongly suggests that the explanatory variable does not exert

PAGE 113

101 a meaningful effect on the dependent variable. These results, taken with the finding of no significance on the point bi serial correlations, indicates that Hypothesis 2 is likely incorrect and the null hypothesi s is more plausible. A number of explanations are plausible for explaining the null hypothesis. It is possible that partial grant and administrative denial criteria vary idiosyncratically by agency. It is also possible that the requests submitted to intel ligence agencies lend themselves more towards full denials than partial grants or administrative denials. A more likely explanation is that intelligence agencies are comfortable using full denials rather than partial denials and may withhold the entire req uest rather than redact portions. Analyse s f or Hypothesis 3 Intelligence Agencies have Higher Backlog and Longer Response Times Hypothesis 3 : Intelligence agencies will have longer mean processing times and higher backlogs than other agencies. Hypothesis 3 tests the premise that intelligence agencies are more likely to practice secrecy by delay, metering the disclosure and release of information. The annual report data contains information about agency backlogs and average processing times that provide so me insight into agency responsiveness. Intelligence Agencies exhibit Higher Backlog Ratios The point bi serial correlation (r= 0.2854 p<.001) indicates that the explanatory variable likely exerts a significant influence on the Backlog Ratio as a dependent variable. The regression models help clarify this relationship and highlight other relationships of interest (see Table 6.9). At first run, the intelligence community dummy variable shows a large coefficient equivalent to

PAGE 114

102 a 40.37% increase in backlog ratio significant at p=.071. However, this result is significantly skewed by DIA observations. Once these and other outlier observations were removed, the effect size (12.13%) and significance (p=.170) of the explanatory variable decreased. The model shows a si gnificant relationship between processing time and backlog ratio, as expected.

PAGE 115

103 Table 6.9 Regression Results versus Backlog Ratio Clustered OLS Outliers Removed Dummy: Intelligence Group 0.4037 0.1213 (0.2232) (0.0882) No. Full Time Employees 0.0001 0.0000 (0.0003) (0.0002) Equiv. Full Time Employees 0.0005 0.0003 (0.0002) (0.0002) Litigation Related Costs 0.1743 ** 0.1038 (0.0635) (0.0518) Complex Request Ratio 0.0362 0.0169 (0.0216) (0.0150) Average Processing Time in Days 0.0030 *** 0.0034 *** (0.0004) (0.0003) Square of Average Processing Time 1.97E 06 ** 2.84E 06 *** 0.0000 (0.0000) _cons 0.0063 0.0143 (0.0088) (0.0067) N 1603 1542 R sq 0.4129 0.4841 Standard Errors in parentheses p<.05, ** p<.01, *** p<.001 The model also shows a moderately strong relationship between litigation costs and backlog ratios, perhaps suggesting that agencies with a high percentage of litigation expenses are more likely to accrue higher backlog rates than other agencies. This model shows a relatively decent goodness of fit with R squared values approaching 0.5. The Backlog Ratio presents a special regression modeling case. The ra tio can exceed one (in cases

PAGE 116

104 though these are unusual cases. In this data set, 36 of the observations (2.2 %) had backlog ratios exceeding one. Thus, technically, models using Backlog Ratio as a dependent variable should not violate OLS assumptions. However, as a precaution, a fractional logit regression was performed after removing observations where the Back log Ratio exceeded one. The logit results differ from the OLS results. With the same outliers removed, the fractional logit model showed the explanatory variable to have moderate effect size and strong significance (p=.000). The coefficient of .5552 can be interpreted by an odds ratio that indicates intelligence agencies are 1.74 times more likely to have a backlog ratio approaching one than non intelligence agencies with similar resource characteristics The point bi serial correlation and the fractional logit model both suggest that the intelligence community dummy variable has a significant effect on the Backlog Ratio; however the OLS model with outliers removed shows a weaker relationship. This is a mixed finding that generally supports Hypothesis 3. Intelligence Agencies show Longer Mean Durations As Table 6.10 indicates, first order bivariate regression models suggest that membership in the Intelligence Community results in longer processing times. In fact, the combined (simple and complex requests) model shows that the mean request duration is 182 days longer for members of the intelligence group than for other agencies. However, these models have very low R squared values and large standard errors for the explanatory variable which suggests that the simple bivariate model captures little of the variation of interest.

PAGE 117

105 Table 6.10 Regression versus Time to process requests Mean Delay Simple Requests Complex Requests Combined Dummy: Intelligence Group 1.1457 173.7060 *** 182.0706 *** (10.0378) (99.8918) (94.7449) _cons 23.0980 *** 35.9770 *** 51.8560 *** (1.4698) (2.8345) (3.0383) N 1527 1497 1715 R sq 0.0000 0.0781 0.0629 Standard Errors in parentheses p<.05, ** p<.01, *** p<.001 Using a regression model with the combined mean duration as the dependent variable and independent variables that control for resource loading in the form of employees and program costs and the ratio of complex requests improves the model fidelity. The reg ression model shown in Table 6.11 moderates the influence of the explanatory variable somewhat compared to the bivariate model. Accounting for the independent variables, the size of the dummy explanatory variable coefficient drops to 167.90 days with signi ficance approaching the p<.05 threshold (p=.052). The model passed multicollinearity tests and improved in fit with the removal of outlier organizations. Though the adjusted R squared values for the models fall short of a strong fit, the size and significa nce of the effect provided by the explanatory variable provide a good indication that this relationship is likely valid. The independent variables match expectations and indicate that resources and complexity are key influencers in processing time.

PAGE 118

106 Tab le 6.11 Regression versus total Time to process requests Clustered OLS Outliers Removed Dummy: Intelligence Group 169.9509 167.9099 (87.1518) (86.2307) No. Full Time Employees 0.4345 1.0417 ** (0.1979) (0.3622) Equiv. Full Time Employees 0.4274 *** 0.6689 *** (0.1231) (0.2065) Litigation Related Costs 84.9484 ** 115.9076 *** (27.1227) (25.9753) Complex Ratio 58.8227 *** 50.0869 *** (9.4049) (7.7750) _cons 29.9268 *** 29.521 *** (3.1419) (3.3144) N 1606 1538 R sq 0.1403 0.1749 Standard Errors in parentheses p<.05, ** p<.01, *** p<.001 Discussion and Assessment Evaluating backlog and mean processing time provides a measure of program responsiveness. The regression results indicate a statistically valid relationship between members of the Intelligence Group and larger backlogs and longer mean request durations that supports Hypothesis #3. Addition of the Co mplex Ratio variable greatly improved results, suggesting that the intuitive premise that complex requests take longer to complete is probably true and a factor in program responsiveness. There is the possibility that complex requests to intelligence agenc ies are much more complex than requests to other agencies, but this cannot be determined from annual report data. Taken as a whole, these results provide moderate confidence that requests to the intelligence agencies take longer to complete and that those

PAGE 119

107 agencies carry a higher backlog than other agencies with similar characteristics, though the weaknesses in the models leave open the possibility of other influencing variables. The data show a correlation between the mean combined processing time (. 0875) and the percentage of requests withdrawn significant at the p<.001 level. Although causality cannot be inferred, this correlation indicates a relationship between agency responsiveness and requesters withdrawing FOIA requests, suggesting that the longer i t takes to process requests, the more likely requesters are to voluntarily withdraw from the process. If this relationship is indeed causal, then processing lengths have a direct influence on agency transparency. If this relationship is indeed causal, the n processing lengths have a direct influence on agency transparency. Chapter Summary Analysis of the annual report data supports confirmation of hypotheses 1 and 3, subject to the limitations of the data. The regression models confirm the expected conclus ion that intelligence agencies grant less and deny more requests than other agencies. However, the models used are unable to account for requests for information protected by law and show low overall goodness of fit. Nevertheless, the strength and signific ance of the coefficients of the explanatory variable in both OLS and fractional logit models provide confidence that Hypothesis 1 is likely correct. The data showed that hypothesis 2, which predicted higher partial grant and administrative denial rates fr om the intelligence group, is likely incorrect. First order correlations and regression models offer no support to Hypothesis 2. The analysis provides some qualified support to Hypothesis 3. The models provide a mixed result on the relationship between int elligence agencies and backlog ratio. The OLS model

PAGE 120

108 shows a weak effect and significance, while the fractional logit model shows a stronger effect and significance. The models for processing time provide some confidence of a relationship between intelligen ce agencies and increased processing durations. Taken together, these findings cautiously indicate that Hypothesis 3, that intelligence agencies have higher backlog ratios and longer mean processing durations is likely valid. As mentioned in Chapter V, the data are subject to some caveats and limitations. The annual report values must be taken at face value and are potentially subject to misreporting and error by agencies. Without a standard, government wide definition of complexity, the construction of a complexity ratio is less a measure of true complexity and more a measure of perceived complexity. Some agencies treat all requests as simple and others treat all requests as including the possibility that the mere presence of classified information systems creates additional complexity that non intelligence agencies do not experience. In a macro sense, this data set provides a good source to evaluate program l evel effects, but cannot shed much light on how similar requests might be handled differently by the intelligence agencies. The experimental portion, detailed in the next chapter, helps inform our understanding of agency responses to individual requests an d adds additional data to help address the principal research question.

PAGE 121

109 CHAPTER V I I EXPERIMENTAL RESULTS Overview The experimental portion addresses RQ2: How do Intelligence Agency disposition decisions differ from their less secretive counterparts? The experimental results will help address Hypotheses 5, 6 and 7 which test the differences between intelligence agencies and a control group in responding to individual requests. Hypothesis 5 tests the premise that intelligence agencies will issue fewer full grants and more denials than the control group. Hypothesis 6 tests the premise that intelligence agencies will take longer to respond to requests than the control group. Hypothesis 7 assumes that intelligence agencies will perform worse on service quality indicators than agencies in the control group. To compare disposition results between intelligence agencies and their less secretive counterparts, the researcher constructed an experiment in which three rounds of requests were sent to an intelligence grou p consisting of six intelligence agencies and a control group consisting of ten non intelligence peer agencies. For each round, the researcher submitted an identically phrased request and waiver for fees to each agency. Request responses were logged for da te of correspondence, official disposition, fee status and records returned (if any). To compare service quality and execution, each response was coded for phrases/components of interest in seven areas: acknowledgement, contact info, fee statement, request er rights, requester responsibilities, government rights and statements of workload/backlog. This chapter provides results for each round and concludes with key findings and observations from the experiment as a whole.

PAGE 122

110 Round 1 Results Objective Rou and disposition schedules. Records management is directly related to instrumental transparency es to FOIA. The federal government has an inclusive definition of what constitutes a record: Records include all books, papers, maps, photographs, machine readable materials, or other documentary materials, regardless of physical form or characteristics, made or received by an agency of the United States Government under Federal law or in connection with the transaction of public business and preserved or appropriate for preservation by that agency or its legitimate successor as evidence of the organizatio n, functions, policies, decisions, procedures, operations, or other activities of the Government or because of the informational value of the data in them. (44 USC 3301) This broad definition leaves significant discretion for agencies to determine what may be considered a record. To guide employees in determining what should be considered a record and how long it should be retained, agencies use a records management schedule that combines defining terms and disposition instructions. The National Archives and Records Administration issues a General Records Schedule (GRS) that p rovides a recommended schedule for retention and destruction. Some agencies use the GRS as their primary reference, while other agencies use a customized schedule. This is an

PAGE 123

111 administrative detail that is critical to transparency as a citizen cannot reques t and an agency cannot return records that no longer exist. The researcher requested that agencies provide both the records management policy document and the records management schedule used by the agency. As with subsequent rounds, the service provided b y the agencies in responding to the request is as significant as the records returned (if any). For this round, intelligence agencies were expected to withhold more records or portions of records and take longer to respond to requests than their non intell igence agency counterparts. Response Summary As Table 7.1 indicates, 12 of 16 agencies completed the request within 100 business days. Four of six intelligence agencies completed the request, returning records of some sort. The dispositions were split, wit h two agencies responding by fully granting the request and two agencies partially granting the request, invoking an exemption to withhold or redact information. In contrast, eight of 10 control agencies completed the request within 100 business days. All eight of these returned records or information without redactions or withholds. Even in the GRS and a statement indicating that they follow the GRS and have no agency specific records policies.

PAGE 124

112 Table 7.1 Round 1 Disposition Results Records Returned Agency Request Disposition Exemptions Used Policy or Instruction Disposition Schedule Intelligence Group CIA Not Received* DIA Granted Yes Yes NGA Granted Yes Yes NRO Partially Granted b3, b6 Yes Yes NSA Not Received* ODNI Partially Granted b3 Yes Yes Control Group DHS TSA Granted Yes No DOC NOAA Granted Yes Yes DOD DCMA Not Received* DOJ USMS Granted No Yes DOL MSHA Granted Yes Yes HHS ACF No Responsive Records No Yes NASA Granted Yes Yes Treas OCC Granted No Yes USDA RD Not Received* USGS Granted Yes Yes *As of 100 business (145 calendar) days from date of request A total of 25 responses were coded for service quality aspects. A marked difference between intelligence and non intelligence agencies was observed when examining service quality and performance indicators. As Table 7.2 indicates, agencies in the intelligence group issued more delay notices and took about twice as long to respond. The mean duration for processing for the intelligence group was 50 days compared to 21.6 days for the non intelligence group. There was not a major difference between the intelligence and control group regarding co mpliance with the 20 day rule, which specifies agencies to either issue a final response or a delay notice within 20 business days of receiving the request. Four of six intelligence agencies were compliant, compared to five of 10 control agencies.

PAGE 125

113 Table 7. 2 Round 1 Timing of Correspondence Delay notice Final Disposition Number of Days to Respond Compliant with 20 day rule? Agency Issued? Number of Days to Respond Intelligence Group CIA Yes 25 TBD 100+ No DIA Yes 11 92 Yes NGA No --1 Yes NRO Yes 0 85 Yes NSA Yes 43 TBD 100+ No ODNI Yes 19 22 Yes Control Group DHS TSA No --38 No DOC NOAA No --1 Yes DOD DCMA No --TBD 100+ No DOJ USMS No --1 Yes DOL MSHA Yes 20 51 Yes HHS ACF No --19 Yes NASA Yes 2 20 Yes Treas OCC No --22 No USDA RD No --TBD 100+ No USGS No --21 No Note: Date measured in business days (excluding federal holidays) from date of submission to date of correspondence A clear difference emerged in the way records were returned. Recalling instrumental definition of transparency, effective transparency becomes as much about the how as the what. The request indicated that electronic records were preferred, with posting to the ibed as an acceptable method of satisfaction. Round 1 indicates a clear difference between the intelligence group and the control group on the ability and/or willingness to return both correspondence and records via electronic means. As Table 7.3 indicates all responses from the intelligence group were provided non electronically via postal mail. In clear contrast, most (14 of 15) responses provided by the

PAGE 126

114 control group were provided by email. Similarly, Intelligence agencies returned few records electroni cally, opting to send paper copies of the records via mail. In contrast, seven control agencies provided responses electronically, as requested, with only one agency returning records in a hardcopy format. Table 7.3 Round 1 Correspondence Medium Response Medium Record Medium Agency Initial Final Directive Schedule Intelligence Group CIA Mail ------DIA Mail Mail URL Hardcopy NGA --Mail Hardcopy Hardcopy NRO Mail Mail Hardcopy Hardcopy NSA Mail ------ODNI Mail Mail Hardcopy URL GRS Control Group DHS TSA --Email Electronic N/A DOC NOAA Email Email URL URL DOD DCMA Email ------DOJ USMS --Email N/A+ URL GRS DOL MSHA Email* Mail Hardcopy Hardcopy HHS ACF --Email* N/A URL GRS NASA Email Email URL URL Treas OCC Email Email N/A Electronic, URL GRS USDA RD Email ------USGS Email Email URL URL Notes: Letter sent by both email and mail + USMS sent a document titled "Policy Directives" that was not specific to records management Service Summary The responses returned allow for some assessment of FOIA service. The six service quality indicators shown in Table 7.4 were derived from statute, executive order and OGIS best

PAGE 127

115 practices as indicated in Chapter V. Possible scores on the scale range from ze service. As the table indicates, on average, the intelligence agencies performed slightly worse on service indicators (mean=2.5) than the control age ncies (mean=3.7). Intelligence agencies scores ranged from 1 to 3 with a mode of 3. The CIA scored the lowest on the scale, with only one Table 7.4 Round 1 FOIA Service Scale Agency Respond within 20 days? Provide a tracking number? Respond electronic ally? Provide electronic means to get status? Offer additional information? Provide a point of contact for questions? Total Intelligence Group CIA No Yes No No N/A No 1 DIA Yes Yes No No No Yes 3 NGA Yes Yes No No No Yes 3 NRO Yes Yes No No No Yes 3 NSA No Yes No No N/A Yes 2 ODNI Yes Yes No No No Yes 3 Total n=Yes 4 6 0 0 0 5 15 Control Group DHS TSA No Yes Yes No No Yes 3 DOC NOAA Yes Yes Yes Yes Yes No 5 DOD DCMA No Yes Yes No N/A Yes 3 DOJ USMS Yes Yes Yes No No No 3 DOL MSHA Yes Yes Yes Yes No Yes 5 HHS ACF Yes Yes Yes No Yes No 4 NASA Yes Yes Yes No No Yes 4 Treas OCC No Yes Yes Yes Yes No 4 USDA RD No Yes Yes No N/A No 2 USGS No Yes Yes No Yes Yes 4 Total n=Yes 5 10 10 3 4 5 37

PAGE 128

116 The conclusion from this round is that the intelligence agencies lag behind their counterparts in either the ability or desire to provide service electronically, both in terms of responses and in terms of request tracking. Additionally, the intelligence agencies volunteered very little supplemental information with the request disposition. Four of 10 control agencies provided information outside of the required disposition statu contrast, none of the intelligence agencies provided supplementary information beyond the disposition result or the records retu rned. Other recommended practices include offering an estimated disposition date and recommending the Office of Government Information Services as an additional resource for resolving difficulties (OGIS, 2014). Of the seven agencies providing a delay notic e, only two, NASA and DOL MSHA, provided an estimated date of completion. Only one agency, USGS, recommended the OGIS to requesters as a possible source of information/assistance in resolving FOIA problems. Material Summary The records returned in this rou nd allow addressing of the first exploratory question : How do Intelligence agency records management policies differ from non intelligence agencies? The experiment yielded nine sets of records management policy documents and nine records management schedul es. Collectively, this volume represents thousands of pages of material. Records Management Program Directives. The policies, directives and instructions provided by the agencies take different forms, making direct textual comparison challenging. In general, these documents are designed for agency internal use, providing instructions to agency

PAGE 129

117 components and specifi program. The researcher reviewed the directives provided with an eye towards understanding the relationship between records management and transparency and seeing if and how the intelligence agencies differ from the non intelligence agencies in this respect. There is some apparent variation in the purpose of the records management programs from agency to agency. Three groups appear to be present. The first group of directives emphasized the co ntrol and Records and Information Life procedures and assign responsibilities governing control, filing, destruction and archiving (that is, life group. The second group mentioned use of the records retained, but stopped short of referencing records release as a goal of the r ecords management program. DIA, TSA and NRO use the statutory definition contained in 44 U.S.C. § 2901 (2) which describes Record Management as: The planning, controlling, directing, organizing, training, promoting, and other managerial activities related to the creation, maintenance and use, and disposition of records to achieve 6 adequate and proper documentation of Federal policies and transactions and effective and economical management of agency operations. NASA and DOL/MSHA have record management guida nce that expresses a similar philosophy. 6 The TSA directive substitutes (intentionally or n

PAGE 130

118 The third group clearly specifies records release in the purpose of the directive. As an guidelines for the creation, maintenance, use, protection, similar lang uage referencing records release. The connection between records management and FOIA is less than clear for most agencies. Few of the directives mention FOIA directly at all in the body of the document. Five agencies, including NASA, NGA, NOAA, NRO, and USGS have directives which do not reference FOIA at all. Two agencies, DIA and DOL/MSHA, make references to retaining records requested via FOIA, but with an emphasis on retention rather than release. Fo Act (FOIA) requests, all documents (record and non record materials) cannot be destroyed or Of the nine directives receiv ed, only two reference the need to maintain records for the purpose of responding to FOIA requests. The TSA directive provides a statement clearly TSA shall make reas onable efforts to maintain recor ds in formats or media that are reproducible for purposes of the Freedom of Information Act (FOIA). also includes a paragraph for FOIA, and assigns responsibilities for records with respect to FOIA. T his small sample leads to the conclusion that for most agencies, FOIA and records management are unconnected. The disconnect means that records are created, maintained,

PAGE 131

119 stored and destroyed substantially without concern for release or dissemination to the public. The intelligence agencies appear no worse than other agencies in this respect, and in fact, the management policy designed with transparency in mind. Records Di sposition Schedules. Approximately half of the agencies contacted use the general records schedule as their reference for retention durations while the other half maintains an agency specific set of disposition schedules. In the seven disposition schedules provided, there was little noticeable variance between agency specific rules and the General Records Schedule. Looking at the retention of FOIA related records, in general, most agencies retain these files for two years, with a few agencies maintaining so me files for up to six years. Round 2 Results Objective Round 2 tested agency responses to a request to disclose agency specific records pertaining to the Controlled Unclassified Information policies or directives issued by the agencies. As described in C hapter V, CUI is a class of information that is not classified, but may be restricted from public release. Understanding variances in policy for this category of information contributes to understanding how intelligence agency FOIA related rules differ fro m non intelligence agency rule sets. Response Summary As Table 7.5 indicates, only 11 of 16 agencies completed the request within 100 business days. Of these 11 agencies, only three returned the request as granted or partially

PAGE 132

120 granted. Most agencies (7 of 11) did not locate any responsive record s. Only two denials were issued; HHS ACF withheld 36 pages and NASA withheld 15 pages of records. Both HHS ACF and NASA invoked the (b)(5) exemption, which allows agencies to withhold inter and intra agency memora nd ums. For this round, there were no meaningful difference s in disposition status between the intelligence agencies and the control group. Table 7.5 Round 2 Disposition Results Agency Request Disposition Exemptions Used Records Returned Intelligence Group CIA No responsive Records No DIA Not Received* --NGA Granted Yes NRO Not Received* --NSA Not Received* --ODNI No responsive Records No Control Group DHS TSA No responsive Records No DOC NOAA Not Received* --DOD DCMA Granted Yes DOJ USMS No responsive Records No DOL MSHA No responsive Records No HHS ACF Denied b5 No NASA Partially Granted b5 Yes Treas OCC No responsive Records No USDA RD Not Received* --USGS No responsive Records No *As of 100 business (145 calendar) days from date of request For round 2, the intelligence agencies displayed similar service and performance characteristics to round 1. A total of 21 responses were coded for content and service factors. As Table 7.6 indicates, the intelligence agencies issued a slightly higher frac tion of delay notices than control agencies and were roughly equivalent in terms of compliance with the 20 day rule.

PAGE 133

121 The intelligence agencies that completed the request took an average of 11 days to disposition the request, compared to a control group ave rage of 29.6 days. However, when the still pending responses are accounted for, the intelligence group has a higher average time to process (61 days) than the control group (49.6 days). Table 7.6 Round 2 Timing of Correspondence Delay notice Final Disposition Number of Days to Respond Compliant with 20 day rule? Agency Issued? Number of Days to Respond Intelligence Group CIA Yes 22 52 No DIA No --TBD 100+ No NGA Yes 6 6 Yes NRO Yes 82 TBD 100+ No NSA Yes 11 TBD 100+ Yes ODNI No --8 Yes Control Group DHS TSA No --55 No DOC NOAA No --TBD 100+ No DOD DCMA Yes 2 59 Yes DOJ USMS Yes 26 84 No DOL MSHA No --18 Yes HHS ACF No --18 Yes NASA Yes 6 29 Yes Treas OCC No --31 No USDA RD No --TBD 100+ No USGS No --2 Yes Note: Date measured in business days (excluding federal holidays) from date of submission to date of correspondence A clear difference can be found in how the intelligence agencies returned correspondence and records compared to the control group. Table 7.7 below shows that most intelligence agency correspondence occurred by mailed letter while the control group agencies overwhelmingly rely on email as a means of communicating with requesters.

PAGE 134

122 Table 7.7 Round 2 Correspondence Medium Response Medium Agency Initial Final Record Medium Intelligence Group CIA Letter Letter N/A DIA No correspondence received NGA --Email PDF NRO Letter ----NSA Letter ----ODNI --Letter N/A Control Group DHS TSA --Email --DOC NOAA Email ----DOD DCMA Email Email PDF DOJ USMS Email Email N/A DOL MSHA Email Letter N/A HHS ACF Letter* Letter N/A NASA Email Email PDF Treas OCC --Email N/A USDA RD Email ----USGS Email Email N/A Notes: Letter sent by both email and mail Service Summary Like round 1, the intelligence agencies on average performed slightly worse on service performance indicators (mean=2.3) in round 2 than their control group counterparts (mean=3.5). However, this low average for the intelligence group belie s a wider range of scores than were found in round 1. The Defense Intelligence Agency scored a zero on the service scale by failing to return a response. In contrast, the National Geospatial Intelligence Agency tied three control agencies for the highest score (5) on the scale.

PAGE 135

123 Table 7.8 Round 2 FOIA Service Scale Respond within 20 days? Provide a tracking number? Respond electronic ally? Provide electronic means to get status? Offer additional information? Provide a point of contact for questions? Total Intelligence Group CIA No Yes No No No No 1 DIA No N/A No N/A N/A N/A 0 NGA Yes Yes Yes No Yes Yes 5 NRO No Yes No No No Yes 2 NSA Yes Yes No No No No 2 ODNI Yes Yes No No Yes Yes 4 Total n=Yes 3 5 1 0 2 3 14 Control Group DHS TSA No Yes No No No Yes 2 DOC NOAA Yes* Yes Yes Yes No No 3 DOD DCMA Yes Yes Yes No Yes Yes 5 DOJ USMS No Yes Yes No No No 2 DOL MSHA Yes Yes No Yes Yes Yes 5 HHS ACF Yes Yes Yes No Yes Yes 5 NASA Yes Yes Yes No No Yes 4 Treas OCC No Yes Yes Yes No No 3 USDA RD Yes* Yes Yes N/A No N/A 2 USGS Yes Yes Yes No No Yes 4 Total n=Yes 8 15 9 3 5 9 49 These agencies issued an initial response but did not request a processing delay Some of the intelligence agencies responded differently in round 2 than in round 1. Both NGA and ODNI offered supplementary information with their disposition and NGA responded electronically, providing not only disposition information but also records via email. Like the first round, the intelligence agency group on average performed worse on the service scale than the control group on average. However, there was a greater spread between

PAGE 136

124 agencies on round 2 and two intelligence agencies, NGA and ODNI, sco red higher on the service scale than the control group average. Materials Summary Overall, the only a few agencies returned records, limiting the information that could help inform the exploratory questions. Three of sixteen agencies returned records of s ome sort. Of these, only one, DCMA, returned a record matching the request. NGA responded by sending a PDF containing a copy of the executive order and a memorandum received by the agency from e released was a letter designating an individual as the CUI lead for the agency. instruction for handling of CUI provides a questionable connection to FOIA. The instruction states: FOUO is a dissemination control applied by DoD to unclassified information when disclosure to the public of that particular record, or portion thereof, would reasonably be expected to cause a foreseeable harm to an interest protected by one or more exemptions of the Freedom of Information Act (FOIA) (section 552 of Ti tle 5, United Staets [sic] Code (Reference (bc))). This sentence stands in apparent contradiction to executive branch guidance that CUI should not be used as a reason to invoke a FOIA exemption (Executive Order 13556) Further, the exemptions provided in instruction, but rather to allow agencies the discretion to withhold specific categories of information. The potentially circular approach adopted by DCMA has the potential to overly On a

PAGE 137

125 broader scale, the lack of returned responses is a likely indicator that CUI policy has not been adopted by individual agencies and remains substantially incomplete at the federal level. Round 3 Results Objective 1/1/12 to 12/31/13. The logs capture a record of FOIA requests received by the agency. Returned logs help inform assessme nt of how FOIA related rules differ by agency. Response Summary As Table 7.9 indicates, 10 of 16 agencies fulfilled the request within 100 business days. Three of six intelligence agencies completed the request, compared to seven of ten control agencies. D records. The intelligence group did not show a major difference in disposition results from the control group.

PAGE 138

126 Table 7.9 Round 3 Disposition Results Agency Request Disposition Exemptions Used Records Returned Intelligence Group CIA Granted Yes DIA Partially Granted Yes NGA Not Received* NRO Not Received* NSA Not Received* ODNI Granted Yes Control Group DHS TSA Administratively Closed Yes++ DOC NOAA Granted Yes+ DOD DCMA Not Received* DOJ USMS Granted Yes DOL MSHA Granted Yes HHS ACF Not Received* NASA Granted Yes++ Treas OCC Partially Granted b6 Yes USDA RD Not Received* USGS Granted Yes++ Note: *As of 100 business (145 calendar) days from date of request + NOAA's response provided instructions on how to search the online database ++ Records did not fully address request. The timing of responses for round 3 reflects a trend found in round 1 and 2 results. All six intelligence agencies responded with an intent to delay responding (see Table 7.10). By contrast, only one control group agency issued a delay notice. However, half of the control agencies exceeded the 20 day response window without issuing a delay notice. On average, intelligence agencies took longer to issue a final response (57.3 days) than control agencies ( 25.2 days). However, both groups still have a significant number of final responses pending at the end of the 100 day round duration.

PAGE 139

127 Table 7.10 Round 3 Timing of Correspondence Delay notice Final Disposition Number of Days to Respond Compliant with 20 day rule? Agency Issued? Number of Days to Respond Intelligence Group CIA Yes 43 56 No DIA Yes 3 78 Yes NGA Yes 4 TBD 100+ Yes NRO Yes 50 TBD 100+ No NSA Yes 1 TBD 100+ Yes ODNI Yes 5 38 Yes Control Group DHS TSA No --99 No DOC NOAA No --1 Yes DOD DCMA No --TBD 100+ No DOJ USMS No --1 Yes DOL MSHA Yes 3 41 Yes HHS ACF No --TBD 100+ No NASA No --1 Yes Treas OCC No --23 No USDA RD No --TBD 100+ No USGS No --11 Yes Note: Date measured in business days (excluding federal holidays) from date of submission to date of correspondence Like previous rounds, intelligence agencies responded primarily by non electronic means. By contrast, the control group agencies responded primarily by electronic means. These results are consistent with the outcome of round 1 and 2.

PAGE 140

128 Table 7.11 Round 3 Correspondence Medium Response Medium Agency Initial Final Record Medium Intelligence Group CIA Letter Letter Hardcopy DIA Letter Letter Hardcopy NGA Email --NRO Letter --NSA Letter --ODNI Letter Letter Mixed++ Control Group DHS TSA Email+ Email URL DOC NOAA Email Email URL DOD DCMA Email+ --DOJ USMS Email URL DOL MSHA Email Letter DVD HHS ACF No Correspondence Received NASA Email Electronic Treas OCC Email Email Electronic USDA RD Email+ --USGS Email Electronic Notes: Letter sent by both email and mail + Acknowledgement received via email ++ ODNI returned a portion of the records hardcopy and a portion via URL Service Summary In round 3, as a group the intelligence agencies performed slightly worse on service indicators than the control group. Like prev ious rounds, a smaller proportion of intelligence agencies provided electronic responses or a means to get status than control agencies. In other areas, such as responsiveness or offering additional information or contacts for questions, the difference bet ween intelligence agencies and control agencies was less marked.

PAGE 141

129 Table 7.12 Round 3 FOIA Service Scale Agency Respond within 20 days? Provide a tracking number? Respond electronic ally? Provide electronic means to get status? Offer additional information? Provide a point of contact for questions? Total Intelligence Group CIA No Yes No No No No 1 DIA Yes Yes No No N/A No 2 NGA Yes Yes Yes No N/A No 3 NRO No Yes No No N/A Yes 2 NSA Yes Yes No No N/A Yes 3 ODNI Yes Yes No No Yes Yes 4 Total n=Yes 4 6 1 0 1 3 15 Control Group DHS TSA Yes* No Yes No No No 1 DOC NOAA Yes Yes Yes Yes Yes No 5 DOD DCMA No Yes Yes No No Yes 3 DOJ USMS Yes Yes Yes No No No 3 DOL MSHA Yes Yes Yes Yes Yes Yes 6 HHS ACF No 0 NASA Yes Yes Yes No No Yes 4 Treas OCC Yes* Yes Yes Yes No No 3 USDA RD Yes* Yes Yes No No No 2 USGS Yes Yes Yes No No Yes 4 Total n=Yes 5 8 9 3 2 4 31 Acknowledg ment received within 20 days Materials Summary A total of nine sets of logs were received in r ound 3. The logs can be roughly categorized into two types. The first type is a minimal log consisting only of the tracking number, requester name, subject title and date received. The second type of log is mo re fulsome and includes disposition status and may have additional information. Of the three logs received from intelligence agencies, all three fell into the first category. Of the seven logs received from control agencies, five of the seven fell into the second type. Three control agencies provide d

PAGE 142

130 additional information in their log files. USMS adds the number of pages released to their logs and NASA tracks which exemptions were used, a helpful reference for transparency scholars. NOAA uses foiaonline.go v which allows the public to see not only the full disposition information available, but also any records provided to requesters in response. The minimal log format used by the intelligence agencies provides little in the way of process transparency. If requesters were using this log information to judge whether or not they should submit a request on a similar topic or for a similar record, they would have no information to aid in their decision making. By contrast, the log format used by five of seven co ntrol agencies provides for greater process transparency, and in the case of NOAA, may allow potential requesters to view records provided to other requesters. Discussion and Key Findings Hypothesis 5 Little Difference in Disposition Outcomes Hypothesi s 5 tests disposition outcomes by asserting that i ntelligence agencies will issue more denials and less full disclosures in response to requests than the control agency set The experimental results show a bit more nuanced outcome, influenced in part by delays by intelligence agencies in completing requests. For round1, H5 appears to be valid, as two intelligence agencies and no control agencies used exemptions to withhold portions of records. For round 2, however, the opposite was true, as one control ag ency denied the request and another used exemptions to withhold records. Round 3 offered mixed results, leaning slightly towards support of H5. Taken across all three rounds, intelligence agencies issued full grants at a slightly lower rate (41.6%) than th e control agencies (50%) but when partial grants were included, the gap closed to 44.3% versus 46.6%.

PAGE 143

131 Table 7.13 Consolidated disposition results Agency Round 1 Round 2 Round 3 Intelligence Group CIA Not Received No Records Granted DIA Granted Not Received Partially Granted NGA Granted Granted Not Received NRO Partially Granted Not Received Not Received NSA Not Received Not Received Not Received ODNI Partially Granted No Records Granted Control Group DHS TSA Granted No Records Admin Closed DOC NOAA Granted Not Received Granted DOD DCMA Not Received Granted Not Received DOJ USMS Granted No Records Granted DOL MSHA Granted No Records Granted HHS ACF No Records Denied Not Received NASA Granted Partially Granted Granted Treas OCC Granted No Records Partially Granted USDA RD Not Received Not Received Not Received USGS Granted No Records Granted Intelligence agencies were expected to use exemptions more than control agencies. Three Intelligence agencies used the b3 and b6 exemptions to withhold information while three control agencies used the b5 and b6 exemptions to withhold information. This result does not provide a strong indication in support of the hypo thesis. The experimental results suggest roughly similar disposition decisions by intelligence agencies and control agencies, though this result is impacted somewhat by the delay in responses from the intelligence agencies. The results fail to show convin cing support for Hypothesis 5, indicating that intelligence agencies are not more likely to issue denials and less likely to fully grant requests than control agencies given identically worded requests.

PAGE 144

132 Hypothesis 6 Delays by Intelligence Agencies Hypot hesis 6 tests the assumption that intelligence agencies will be more likely to issue delays and take longer to make release decisions than the control group The experimental results strongly support H6. Intent to delay by intelligence agencies The FOIA requires agencies to inform requesters if final disposition of a request cannot be fulfilled within 20 working days of receipt. Of the six intelligence agencies, three issued delay notices for each round of requests and the other three issued delay notices for two of three rounds. Of the total of 18 requests sent to intelligence agencies, 15 delay notices were received (83.3%). Of the 15 delay notices received, 9 were issued within the 20 business day period required by law. In two of the three cases in whi ch a delay notice was not received, the agencies were able to finish the request within 20 business days, obviating the need for a delay notice. Control agencies issued fewer delay notices, but this finding bears closer examination. Of the 30 requests sen t to control agencies, only six delay notices were received, though this result does not indicate better performance. The researcher found 14 cases in which a delay notice was not issued but should have been, indicating that these control agencies failed to communicate that the request would exceed the 20 business day period. Five of the six delay notices received were issued within the required 20 day period. Longer final response durations by intelligence agencies. The delay notices issued by intelligence agencies appear to be well justified. Table 7.14 presents a consolidated summary of how long it took agencies to issue a final response. Among agencies that completed the request within a 100 business days, intelligence agencies took an average of 15.5 day s longer to

PAGE 145

133 complete the request than agencies in the control group. Additionally, a higher percentage of intelligence agencies were unable to complete the request within the 100 day window. Table 7.14 Final Disposition Response Durations in Business Days Final Response Agency Round 1 Round 2 Round 3 Intelligence Group CIA TBD 100+ 52 56 DIA 92 TBD 100+ 78 NGA 1 6 TBD 100+ NRO 85 TBD 100+ TBD 100+ NSA TBD 100+ TBD 100+ TBD 100+ ODNI 22 8 38 Control Group DHS TSA 38 55 99 DOC NOAA 1 TBD 100+ 1 DOD DCMA TBD 100+ 59 TBD 100+ DOJ USMS 1 84 1 DOL MSHA 51 18 41 HHS ACF 19 18 TBD 100+ NASA 20 29 1 Treas OCC 22 31 23 USDA RD TBD 100+ TBD 100+ TBD 100+ USGS 21 2 11 20 day compliance an issue for all agencies. The law requires agencies to either perfect a request within 20 business days or issue a notice of delay with justification. Compliance with this aspect of the law was lower than expected. Across all three rounds, less than 55% of agencies were compliant. However, the intelligence agencies were slightly more compliant than control agencies, in large part because the issuance of a delay notice within 20 days is considered compliant. The law also requires agencies to provide an estimate of when the request is expected to be met. Only two agencies requesting a delay provided this information in any round.

PAGE 146

134 Hypothesis 6 supported. The experimental results support Hypothesis 6, suggesting that intelligence agencies are more likely to issue delays and take longer to respond to requests than other agencies given identically worded requests. Hypothesis 7 Intelligence Agencies lag on some Service Indicators Hypothesis 7 tests the assertion that intelligence agencies will perform worse on service indicators than age ncies in the control group. The experimental results partially support this hypothesis and provide some insight into FOIA service provisioning. As group across all three rounds, the intelligence agencies performed worse than control group agencies on sever al service performance indicators. However, some intelligence agencies performed relatively well on the service indicators and outperformed the control group average. In general, intelligence agencies performed similar to their peers on a few indicators, s uch as responding to requests within 20 days, providing a tracking number for requests and providing contact information for requesters. Intelligence agencies performed only slightly worse than control agencies when it came to providing supplementary infor mation. The gap in service performance, however, was more pronounced when it came to electronic responses and delivery of material. Intelligence agencies provide less information with the response. FOIA does not require agencies to provide an information s ervice. In other words, agencies do not have to answer questions posed in a request. Instead, agencies are obligated only to respond to properly formatted requests for records. This poses as interesting problem for requesters, as the existence of or title of a record may not be readily apparent, nor is there an information service provided by the agencies to assist in discovery of records which could possibly be requested. For

PAGE 147

135 eived places a burden on requesters to have knowledge of the existence and proper description of the records. To ameliorate this burden, EO 13392 encourages agenci es to provide information beyond the disposition results and OGIS recommends it as a best practice to encourage transparency (OGIS, 2014) Across all three rounds, out of 18 requests, intelligence agencies provided supplementary information only three time s. Control group agencies were more forthcoming, offering additional information 10 out of 30 requests. Both intelligence and control agencies provided a point of contact in their response approximately half the time. Taken as a whole, it means that even g ranted requests from an intelligence agency might contain very little information. A typical interim response from an intelligence agency might consist only of an acknowledgement of the request, assignment of a tracking number and a statement informing the requester of a delay. A final response might not include much more information, offering only the final disposition result and attached records. While some control agency responses were equally terse, a few more provided richer detail and a point of conta ct to address questions. From the perspective of providing transparency as a service, this additional information and a point of contact for questions is major differential and may make an agency that denies a request but provides information about the req uest effectively more transparent than an agency that returns the requested records but offers no contextual information to help make sense of the record. Intelligence agencies mail responses back, control agencies e mail. Each round made the request for r ecords in an electronic form with the note that posting the record(s) to the

PAGE 148

136 electronic reading room with notice to the requester would satisfy the request. A ll requests were made electronically, either through email or web form submission. There was a mar ked difference between intelligence and non intelligence agencies in meeting this aspect of the request. In only a few cases, the intelligence group responded electronically, electing instead to provide records in hard copy format. By contrast, far more no n intelligence agencies responded electronically, some posting or pointing to the records online and others emailing the requested records. In both intelligence and control groups, few agencies provide online tracking mechanisms, and many of these only of fer the status as open or closed. This simple status indicator offers little insight as to where the request is in the process or when to expect a final response. When it comes to FOIA requests, there is very little process transparency in general and even less e transparency. There is an interesting asymmetry here with respect to intelligence agencies, as citizens can submit electronic requests, but should have little expectation of receiving an electronic response in return. The experiment shows that, as a group, intelligence agencies do worse at servicing FOIA requests electronically than other agencies. The results strongly suggest a bias by intelligence agencies against electronic dissemination means. Though the experiment does not directly reveal why the bias exists, several related explanations are plausible. The intelligence agencies likely have a more complex information technology environment than many other agencies, making dissemination of unclassified information more challenging. However, g iven that the experiment requested records that were presumably unclassified (rather than needing declassification), this explanation seems insufficient. A more plausible explanation is that the intelligence agencies have underinvested in e transparency capabi lity or th at FOIA officers in

PAGE 149

137 these agencies eschew electronic dissemination for security or cultural reasons. Some of the intelligence agencies have robust public web presences, suggesting the latter explanation may be more likely. F orm letters show fram ing emphasis Not surprisingly, many agencies issue form letters as responses, typically with customization in the introduction to acknowledge the specific request. Three rounds of requests provided plentiful samples and, in a few cases, interesting fodder for analysis. FOIA contains a bundle of rights, a few explicit rights granted to the government to withhold records and a few implicit rights allowed to the citizen to request records and appeal determinations. In responding to requests, some agencies cho se to rights (such as the right to appeal or seek additional assistance) and a few presented both bundles of rights equally. The experiment did not yield enough respon ses to form a conclusion about patterns of rights emphasis, but the data point to the possibility of exploring this framing in future research. Some of the insights from form letters were a bit more trivial and idiosyncratic. Correspondence from the DIA inexplicably often contained the incorrect gender pronoun. Delay notices from DIA cited the reason for delay as a ba cklog o f 1139 requests a number that remained constant across all three rounds Delay notices from the NSA stated that delays were despite the fact that request volume has been relatively steady since 2008 and even decreased by more than 10% in FY2012 A few agencies relied less on form letters, providing correspondence with personalized statements or emails that addressed the request directly

PAGE 150

138 Request fees and the cost of transparency. For each round, the researcher asked that the request be placed in the educational category and granted a waiver of fees. The criteria for request to request. Some agencies granted the request, while others placed the request in the which category the request had been placed in. In a very few cases, a waiver of fees was granted. Several agencies did not grant a waiver, totaling fees for hardcopy reproduction. However, in every case but one, the assessed fees fell below billable thres holds and did not require remuneration. For the Round 3 request, the CIA billed the requester for $13.60 in reproduction fees. Overall, the experiment did not yield enough observations about fees and waivers to make a generalizable assessment, but the use of hardcopy reproduction by intelligence agencies appears to make it more likely that requesters will be assessed fees by intelligence agencies than control agencies. Records Shed Some Light on Exploratory Questions The experiment provided some material to inform RQ3: How do FOIA related rules differ between intelligence agencies and their less secretive counterparts? In round 1, nine agencies returned records management directives that could be used to examine the re lationship between FOIA and records management. Few agencies connected FOIA to their records management program or mentioned public release or transparency as a key aim of records management. The records returned suggest that for most agencies, FOIA and re cords management are largely independent. As a group, intelligence agencies appear no better or clear reference to FOIA and public release of information as key task. Though a direct

PAGE 151

139 connection between records management instructions and FOIA performance is difficult to establish, the ODNI performed above average on service indicators in all three rounds. Round 2 of the experiment yielded only one relevant record. This result suggests that many agencies have likely not produced or promulgated instructions for the handling and management of Controlled Unclassified Information. Round 3 showed that intelligence agencies release less information in their FOIA logs than age ncies in the control group. Though no firm conclusions can be drawn from this result, it suggests that the underlying norms or rules for intelligence agencies might be different than norms for other agencies. Summary of Experimental Round Overall, the exp eriment successfully provided useful information to inform assessment of hypotheses 5, 6 and 7, which anticipated differences between intelligence agencies and a control group of peer agencies in disposition decisions, response timing and service performan ce indicators. The three rounds of requests provided observations to draw conclusions regarding the validity of the hypotheses. The experiment results do not provide strong support for Hypothesis 5. Given identical requests, intelligence agencies issued di sposition decision similar to agencies in the control group. The experiment does support Hypothesis 6 and 7, however. As a group, intelligence agencies took longer to render final disposition decisions and responded to most requests with a delay notice. In telligence agencies performed substantially similarly to agencies in the control group on half of the service indicators. On a service indicator measuring whether or not the agency provided additional context information with the response, agencies in the intelligence group performed slightly worse than agencies in the control group. On service indicators

PAGE 152

140 measuring electronic status and electronic responses, the intelligence group performed considerably worse than the control group. Though intelligence agen cies rendered similar disposition decisions, they took longer on average to respond and scored lower on service scale indicators than agencies in the control group.

PAGE 153

141 CHAPTER V I II SUMMARY OF FINDINGS AND RECOMMENDATIONS Introduction The principal resear ch question asks: How does secrecy in government organizations impact execution of transparency initiatives? This research addresses that question by comparing the FOIA program and individual request outcomes of six agencies from the US intelligence commun ity to a control group of non intelligence agencies. Analysis of annual report data provided by federal agencies reveals significant differences between intelligence agencies and non intelligence agencies in disposition results, backlog ratios and average processing times after accounting for complexity of requests and program resources. To control for possible differences in the nature of the requests themselves, the researcher constructed an experiment consisting of th re e rounds of identically worded req uests sent to six intelligence and ten control agencies. The results of the experiment show that intelligence agencies make similar disposition decisions and provide similar records to non intelligence agencies, but take longer to respond, offer less infor mation with responses and do not respond well electronically. The findings inform refinement of the micro level model of transparency under study. In particular, the research adds to the understanding of the effects of secrecy on the medium of transparenc y. This chapter summarizes the key findings from Chapters VI and VII and presents implications for theory and recommendations for practice based on these results. The chapter concludes by revisiting the primary research question in light of the key finding s. A discussion of the limitations of the research and contributions to theory and practice are reserved for Chapter IX.

PAGE 154

142 Summary of Findings Analysis of Annual Report Data Analysis of annual report data provides evidence to answer RQ1: How do Intelligence Agency FOIA program processes and outcomes differ from their less secretive counterparts? In sum, the annual report data reveal some significant differences between intelligence agency FOIA programs and non intelligence agency programs. As Table 8.1 summarizes below the data strongly support Hypothesis 1, indicating that after accounting for resource loading and complexity, i ntelligence agencies grant a lower proportion (17.8%) of requests and deny a higher proportion (18.3%) of FOIA re quests than other federal agencies. This finding is based on the strength and significance of the explanatory variable coefficient in multiple regression models. This finding is tempered somewhat by the inability to account for requests that truly involve classified information or may warrant declassification review. Table 8.1 Summary of Results Analysis of Annual Report Data Hypothesis Dependent Variable Finding H1: Intelligence agencies will have a lower percentage of fully granted requests Full Grant Ratio Strong Support Denial Ratio Strong Support H2: Intelligence agencies will exhibit higher partial grant and admin denial rates Partial Grant Ratio Rejected Admin Denial Ratio Rejected H3: Intelligence agencies will have longer mean processing times and higher backlogs Backlog Ratio Moderate Support Mean Request Duration Moderate Support

PAGE 155

143 No evidence was found to support Hypothesis 2, leading to the conclusion that intelligence agencies do not exhibit higher partial grant or administrative denial rates than other agencies. The explanatory variable failed to show significance and regression models displayed extremely low goodness of fit values indicating that being a member of the intelligence agency group did not exert a meaningful effect on partial grant or administrative denial ratios. Though the data provide little direct insight, a poss ible explanation for this finding might be found in the nature of the disposition themselves. Partial grants are likely highly discretionary, as it requires a judgment on the amount of material to release and how many redactions or withholds might be consi dered a full denial rather than a partial release. The criteria for an administrative denial are likely to vary between agencies, suggesting idiosyncratic variances may be present. In both cases, it may be reasonable to assume that agency idiosyncratic var iances are more influential than variances due to membership in the intelligence community. The data provide moderate support to Hypothesis 3, that intelligence agencies exhibit higher backlog rates and longer mean durations to perfect requests than other agencies. The regression models indicate that processing time exerts a moderate and strongly significant effect on the backlog ratio and that being an intelligence agency exerts a weakly significant but large effect on mean processing durations compared t o other agencies. These models have relatively decent goodness of fit measures that indicate that the models are accounting for a moderate percentage of variation in backlog ratio and mean request duration. The finding is tempered somewhat by the lack of s trong significance from the explanatory variable. Some explanation may be found in the strong and significant relationships between litigation costs and complexity and the average time to process requests. The complex ratio variable only captures a portion of the concept of request complexity. It is possible that requests to

PAGE 156

144 agencies. The conceptual relationship between litigation costs and processing time is not immediately apparent and may merit additional research. Presumably, litigation costs are incurred as a response to request dispositions and do not have a direct causal influence on individual request response times. However, the threat of litigation might cause agencies to act more cautiously and therefore spend more time responding to the requests. Or, litigation could divert resources away from addressing new requests and cause resource related delays. The data cannot provide any insight to address this s peculation, however. This finding indicates that FOIA service performance may be a problem area for intelligence agencies and additional scrutiny is needed. Comparing Transparency about Transparency The data from annual reports suggest some key differenc es between intelligence agency FOIA programs and programs in other agencies. The data support the hypothesis that intelligence agencies grant fewer and deny more requests than other agencies, even accounting for program resources and request complexity. Ho wever, the question remains if this finding means that intelligence agencies are in fact less transparent than other agencies. The researcher used the indicators referenced in Chapter V to assess how transparent intelligence agencies were about their FOIA programs relative to a set of control agencies. Hypothesis 4 anticipated that intelligence agencies would be less transparent about their FOIA and transparency programs than other agencies. Table 8.2 shows that, on average, the opposite was true. The int elligence group (mean = 4.16) averaged slightly higher on the scale than the control group (mean = 3.9). Although one intelligence agency scored a zero on the scale, most intelligence agencies performed similarly to their control group counterparts on the scale by providing public access to transparency related records and information.

PAGE 157

145 The lack of apparent difference between intelligence agencies and control agencies provides little insight to address the primary research question. In general, agencies di d well at providing an electronic reading room and frequently requested records. Improvement is needed in providing access to the most recent copies of Annual Reports and Chief FOIA Officer Reports and providing greater access to key policy documents. Table 8.2 Transparency About Transparency Scale Indicators Agency 1 Annual Reports 2 Chief FOIA Officer Reports 3 Electronic Reading Room 4 Frequently Requested Records 5 Agency FOIA Policy 6 Agency Records Policy 7 Requester Handbook Total Intelligence Group CIA No No Yes Yes No No Yes 3 of 7 DIA No No Yes Yes Yes Yes Yes 5 of 7 NGA No No No No No No No 0 of 7 NRO Yes* No Yes Yes Yes No Yes 5 of 7 NSA Yes* No Yes Yes Yes Yes Yes 6 of 7 ODNI Yes Yes Yes Yes Yes No Yes 6 of 7 Total 3 1 5 5 4 2 5 25 Control Group DHS TSA Yes* No Yes No No Yes Yes 4 of 7 DOC NOAA Yes Yes Yes Yes Yes Yes Yes 7 of 7 DOD DCMA No No Yes Yes No No Yes 3 of 7 DOJ USMS No No Yes Yes No No Yes 3 of 7 DOL MSHA Yes* Yes* Yes++ No No No No 3 of 7 HHS ACF No No Yes No No Yes No 2 of 7 NASA Yes Yes Yes Yes Yes Yes Yes 7 of 7 Treas OCC No No Yes Yes No No Yes 3 of 7 USDA RD Yes* Yes* No No No No No 2 of 7 USGS Yes* No Yes No Yes Yes Yes 5 of 7 Total 6 4 9 5 3 5 7 39 Agency websites link to parent department annual reports ++ L ink provided to parent DOL reading room The scale does not lead to any clear conclusions and indeed, presents some counter intuitive results The wide range of values on the scale suggests idiosyncratic variance is a likely

PAGE 158

146 explanation. Further, the indicators chosen, though selected from law, executive order and OGIS best practices, only partially represent the concept of proactive transparency These indicators are measures of directed transparency and likely poorly reflect discretionary transparency decisions. Similarly, s electing indicators involving agency FOIA programs allowed for comparisons between agencies with very different missions, b ut is a poor indicator of how transparent those agencies are on their substantive topic areas. Capturing both the quantity and quality of content proactively provided by agencies proved challenging and may ultimately be entirely subjective. Comparing Experimental Results Responses returned from FOIA requests provided the raw material to address RQ2: How do Intelligence Agency disposition decisions differ from their less secretive counterparts? As table 8. 3 shows, the results from the experiment indicat e that intelligence agencies take longer to make disposition decisions and lag behind peers in a number of service performance indicators. Table 8. 3 Summary of Results Experimental Portion Hypothesis Indicators Finding H5: Intelligence agencies will issue more denials and less full disclosures in response to requests than the control agency set Disposition results No t Supported H6: Intelligence agencies will be more likely to issue delays and take longer to make release decisions than the control group Delay Notifications, Time to final response Strong Support H7: Intelligence agencies will perform worse on service quality indicators than the control group. Response Timeliness, Electronic Responsiveness, Information Provided Partially Supported

PAGE 159

147 No clear evidence was found to support Hypothesis 5. As a group, given identically specified requests, the intelligence agencies made substantially similar disposition decisions to their control group peers. The intelligence agencies released similar recor ds with substantially similar use of exemptions to withhold or redact specific items as the control group. This finding stands in apparent contradiction to the results from hypothesis 1 which showed statistically significant differences in dispositions bet ween intelligence agencies and non intelligence agencies. The initial conclusion is that a large portion of the variance shown in hypothesis 1 may be attributable to the nature of the request. By holding the request content static, the variance in disposit ion results appears to vanish but this is possibly due to the records requested during the experiment Selecting records that might be common to multiple agencies with different missions proved challenging and it is possible that the records selected offe r poor discriminatory power. In sum, the experimental results show that intelligence agencies are similarly transparent about similar things to control agencies but the annual report data suggests that they are likely dissimilarly transparent given dissimi lar topics. The experimental results strongly support Hypothesis 6. Intelligence agencies issued a higher proportion of delay notices and on average took longer to release a final decision than agencies in the control group. These results provide additiona l support to the findings from hypothesis 3 This combined finding suggest s that delays have become routine practice for intelligence agencies and requesters should not expect a timely response from these agencies. Multiple explanations are possible. Intelligence community FOIA programs could lack the technical infrastructure needed to quickly search and retrieve records. The need or habit of reviewing responses for classified content could substantially account for delay s. Although it is by agencies would have policies or procedures in place that support intentional delay, this

PAGE 160

148 explanation seems less likely than other potential ca uses, including informal social norms that tacitly condone delaying responses. The experiment produced mixed evidence on service indicators. In several areas, the intelligence agencies performed similarly to agencies in the control group. In a few areas, h owever, such as electronic responses and inclusion of supplementary information, the intelligence agency group performed worse that the control group. Though idiosyncratic causes are poss ible, it is more likely that this difference in service performance, particularly with respect to electronic responsiveness, is a result of institutional differences between the intelligence agencies and other agencies. The gap could be caused by under investment in electronic response capabilities, outmoded processes or de liberate policies by FOIA found reduced emphasis on FOIA in some agencies as a result of NPM reforms that supports the under Theory (1993, 2000) could support a procedural cause and the work of Weber (1920/2009) and Niskanen (1968) could support the deliberate policy choice premise. For some rounds, the distribution of scores between intelligence agencies indicated that agency by agency variation is present. Why some intelligence age ncies provide better service than others might be an appropriate topic for future research. Implications for Theory Simmel (1906) Weber (1920/2009), Shils (1956) and Bok (1982) find secrecy to be a powerful social construct, creating a division between s ecret keepers and the uninitiated. Six decades later, FOIA advocates took aim at one aspec t of this second world, targeting (Archibald, 1979) of federal go vernment agencies in general and the growing national security

PAGE 161

149 establishment in particular. This research adds to the knowledge of the contemporary US Intelligence Community as a specific example of this second world and reaffirms the persistent presence o f the paper curtain identified five decades earlier. The research also supports refinement of the model of transparency, adding detail to the understanding of how the mechanisms, timing and content of transparency are affected by secrecy. The support for h ypothesis 1 shows even after accounting for other variables of interest, the US intelligence community grants less and denies mor e requests than other agencies. Though impossible to quantify in terms of the amount or quality of information withheld, the da ta confirm that with a grant rate over 20% less than other agencies and a full denial rate 15% higher than other agencies, intelligence agencies exercise the option to withhold records frequently. Viewed from the lens of transparency as an intrinsic value, this percentage represents lost opportunities for citizen engagement. Viewed through the more temperate lens of instrumental transparency, the finding for hypothesis 1 suggests something categorically different about intelligence agencies as guarantors of transparency that bears closer examination. The findings for hypothesis 1 partially inform our understanding of directional transparency as described by Heald (2006). The intelligence community has several stakeholders and potential beneficiaries of tran sparency. The strong support for hypothesis 1 suggests impairment to transparency to one set of stakeholders, namely the subset of general public individuals, commercial companies, media organizations and academic institutions that submit FOIA requests What the data cannot reveal is how restrictive the intelligence community is to other potential stakeholders such as the legislative branch or executive branch management offices. Additional research can shed more light in this area, disaggregating the

PAGE 162

150 pub lic/commercial/medic/academic category to see if responses from one group (e.g. media) are treated differently than requests from other groups. The findin gs for hypothesis 3 provide some insights on how secrecy impacts the mechanism s of transparency, at le ast as implemented by FOIA programs. The moderate support for hypothesis 3 shows contemporary intelligence community. The finding s for hypothesis 6 provide additional confirmation fo r hypothesis 3. Based on both the annual report data and the experimental results, there is strong evidence to conclude that intelligence agencies take longer to respond to requests, even after accounting for resource and complexity and holding the nature of request constant. There is a potentially a lot to unpack here that generates more questions than answers Does secrecy produce an inherent slowing effect of the transparency medium? Do agencies intentionally delay requests to reduce the potential value of the information released or to discourage requesters? Is there something about the technical capacity or internal processes that can be viewed as a proximate cause of delays? How do delays impact effective transparency? Should transparency delayed be co nsidered similar to transparency denied? From a theoretical perspective, it introduces another element to consider in modeling transparency. In addition to media/content (what) and medium/mechanism (how), it is necessary to add a temporal element (when). M modification may be important to distinguish contemporary transparency (information available now) from historical transparency (information available after the fact) In a practical sense,

PAGE 163

151 declassification of historical artifacts by intelligence agencies should not be treate d the same as transparency, as the former offers little insight into the active workings of the actor. Meijer (2009) states that most transparency is mediated, often through electronic means. The findings for h ypothesis 7 show that the intelligence agencie s largely avoid use of the electronic medium. From an operational perspective, there is a lot to discover related to this problem including the need for better understanding of both the causes and effects of not using electronic means. From a theoretical p erspective, it raises questions about transparency brokerage and direct citizen access. When FOIA was enacted, it was envisioned that the media would play a central role as a broker of transparency between federal agencies and the general public by request ing, receiving, analyzing and using a mass media platform to communicate results. The rise of digital documentation and e government initiatives makes it easier for citizens to directly consume records from federal agencies or for citizens to act as broker s themselves, using electronic platforms like blogs to re disseminate information gathered from agencies. The continued use of hard copy records by the intelligence community makes the latter model difficult, requiring citizens to invest capital and or lab or into converting hard copy documents into a format for digital consumption. The digital barrier means that effectively, transparency from the intelligence community requires a broker, introducing another actor into the model and extending principal/agent trust problems. These results help shape modifications to the conceptual model of transparency as shown in Figure 8.1. Specifically, they highlight the need to understand the medium of transparency as a combination of three aspects: content, mechanism and timing. The results also show that secrecy, rather that acting as a pure barrier, appears to exert a moderating effect on all three aspects of the medium.

PAGE 164

152 In short, the findings help inform refinement of theory, particularly in understanding how secret keeping can impact the medium and timing of transparency. It also highlights the need for additional work to refine the social construction aspects of transpa rency, exploring if the content and mechanisms of transparency are different for different recipient groups. The concept of brokered transparency also needs additional work to discover more about potential brokerage roles in mediated transparency. Recommen dations for Practice Recommendations fall broadly into two categories, recommendations for policy makers in the executive and legislative branches and recommendations for FOIA program administrators in the executive branch. The former center on changes in reporting requirements that can increase the transparency of FOIA performance and allow for easier monitoring of key performance aspects. The latter set of recommendations include suggestions to improve service performance. Recommendations for future resea rch are captured in the concluding chapter.

PAGE 165

153 Policy Recommendations Require reporting of true mean durations. As currently specified by the Department of Justice (2013), agencies do not calculate or report a true mean duration to perfect requests instead r eporting averages of three categories: Simple, Complex and Expedited Processing. The use of these categories for request tracking is at the discretion of the agency. The reports provide a count of simple and complex requests that fall into temporal bins. T he first ten bins are equally spaced 20 day increments, the final three bins are 201 300 days, 301 400 days and 400+ days, respectively. While this distribution is potentially useful to researchers, direct reporting of a true mean value would make comparat ive evaluations of agency performance much easier. Provide c ommon definitions for simple and complex requests. The handbook for filing nition common to federal agencies would help standardized reporting across components and allow for better identification of agency differences. Require agencies to provide data to foia.gov. Using foia.gov to export data about agency programs greatly aided in research, making possible large scale comparisons that would have been difficult or impossible prior to 2008. Foia.gov should continue to be resourced and operated to provide citizens and academics with access to FOIA program performance information to increase transparency. Focus on E transparency in intelligence agencies. As the experiment shows, intelligence agencies issue fewer responses and records electronically than their peer control agencies. Though the law requires agencies to provide a means for obtaining electronic status, the

PAGE 166

154 experiment indicates that intelligence agencies in particular have not adopted the required technology or processes to support this mandate. Adoption of electronic status technology or participation in programs such as foiaonline.gov would substantially improve FOIA service delivery by intelligence agencies. This should continue to be a focus area of Chief FOIA Officer Reports. Few agencies in general and no intelligence agency currently make use of FOIA Online, the con solidated Federal government portal for receiving and responding to FOIA requests. Broader adoption of foiaonline.gov by agencies would help improve FOIA service performance and contribute towards increased process transparency. This finding also raises a number of questions for future investigation. Do intelligence agencies intentionally respond via non electronic means? If so, why? When considering transparency outcomes, should non electronic transparency be given the same weight as electronic transparen cy? Require r eporting of 20 day compliance. Annual reports currently obscure whether or not agencies comply with the 20 day rule. As the experimental indicates, less than 55% of agencies were compliant in providing a response or notice of delay within 20 b usiness days of receiving the request. This pattern is likely present across the federal government. Modification of the annual reporting requirements by Congress or the Executive Branch to include a summary of how many requests were compliant with this ru le would allow for efficient monitoring of this performance requirement. Require reporting from primary intell igence agencies Annual FOIA reports are only explicitly required for two intelligence community components, the CIA and the ODNI. While it may b e impractical to expand reporting requirements to all sixteen components, the DIA, NGA,

PAGE 167

155 NSA and NRO should be explicitly required to submit annual reports to Congress. The current wording of the law may allow these agencies to report through the Defense De partment, making discrete monitoring of these components difficult. This research shows that in some ways these agencies are less transparent than their non intelligence community peers and may warrant additional attention. Additionally, these agencies sho uld be required to produce and publish a Chief FOIA Officer report as directed by EO 13392. Of the six primary intelligence agencies, only the CIA, DIA and ODNI make any of their Chief FOIA Officer reports available online and only the ODNI posts the most recent report released. Add Or Else clauses into FOIA As analysis using the institutional grammar tool revealed, FOIA, as amended, contains no explicit penalties for agency non performance or non would create enforceable rules and might improve transparency outcomes. Recommendations for FOIA Administrato rs Improve electronic response capabilities. T he experimental round showed that several agencies, including all of the intelligence agencies, do not typically respond electronically. FOIA administrators should improve the use of electronic response technol ogies, including email to communicate with requesters. Besides helping reduce delays in receipt of records and potentially reducing overall time to respond, adoption of electronic response methods would help agencies comply with the letter and intent of th e 1996 E FOIA amendments and EO13556. Expand use of OGIS. resource for agencies and requesters alike. The experiment resul ts show that out of 16 agencies, o nly the USGS referenced OGIS in re sponses to requests Agencies consider OGIS best practices

PAGE 168

156 and should adopt language in their response letters that makes it clear to requesters that OGIS is available to assist in resolving issues. Establish a s tronger connection between records manageme nt and transpare ncy. Round 1 of the experiment shows that f ew agencies associated records retention and management functions with the need to provide information or records to the public. This could impact the way that employees perceive records management and results in records retention processes that make transparency more difficult. Agencies should revise and update their records management policies and directives to clearly indicate how records management supports transparency initiatives. Emphasize C ontrolled U nclassified I nformation implementation The lack of returned records in round 2 of the experiment suggests that CUI policy has not propagated yet through the federal government as anticipated. The sole relevant record returned, from the DCMA, pr ovides questionable guidance to employees on using CUI as a way to prevent information disclosure. This area deserves closer examination by transparency scholars to determine if these experimental results are indicative of broader trends in the federal gov ernment and if agencies ar e using CUI policies to restrict information disclosure contrary to the spirit and direction of Executive Order 13556. Additionally, administrators and policy makers should ensure that agency CUI policy and implementation document s are releasable themselves and promote transparency. The Chief FOIA Officer report should be amended to include questions related to agency management of CUI in relationship to FOIA, requiring agencies to explicitly define steps taken to train employees o n CUI management in light of FOIA requirements and open government initiatives.

PAGE 169

157 Expand log information to provide greater transparency. Round 3 of the experiment indicated differences in what information agencies store and release in their logs. The intelligence agencies fell into the most restrictive category, providing logs with only a few fields. Including disposition outcome and noting which exemptions were used would allow for greater transparency of the FOIA process in intelligence agencies and make future FOIA research easier. Adopting foiaonline.gov as a FOIA management tool would provide the additional benefit of allow requesters to see other records release, potentially reducing duplicative requests and augmenting agency release of frequently requested records. Summary Collectively the results imply that secrecy impacts the instrumental nature of transparency, affecting the how and when of transparency actions providing support to Fung, The experiment suggests that the secrecy by delay present in 1970s (Archibald, 1979) persists over five decades later. Though differences in providing FOIA as a service might be attributable to a number of causes, the possibility exists that intelligence agencies practic e a form of procedural secrecy that works against the spirit of transparency laws such as a FOIA. This research confirms the presence of a meaningful difference in execution of FOIA services between intelligence agencies and non intelligence agencies. Thi s finding provides focus for recommendations for program administrators and policy makers. Intelligence agencies can improve their effective transparency by increasing their ability to respond electronically, offering additional information with responses, and providing points of contact for requester inquiries. Policy makers may need to provide special attention to intelligence agency transparency issues, adopting modifications to reporting requirements and establishing effective monitoring regimes and enf orcement penalties to improve FOIA service performance.

PAGE 170

158 The following concluding chapter addresses the contributions of this research to theory and practice. It also provides a summary of limitations and recommendations for future research.

PAGE 171

159 CHAPTER IX CONCLUSION Introduction Thi s research effort makes an important contribution to knowledge of transparency in theory and practice, particularly in highlighting differences between intelligence agencies and non intelligence agencies This chapter begins by a ddressing the contribution this research makes to theory, methodology and practice. The chapter continues by discussing the limitations of this research and resulting recommendations for future research. The chapter concludes with a brief summary of the di ssertation. Contribution to Theory This dissertation enhances the transparency literature by adding to the fie knowledge of transparency in a number of ways. First, t he key findings outlined in Chapter VIII confirm that intelligence agencies are in som e ways less transparent than other agencies, even on the same topic. The findings also clearly indicate that intelligence agencies do worse at providing transparency as a service than other agencies. Collectively, these findings lead to the inference that secrecy may exert a suppressing effect on transparency, even on non secret topics and information, though other influences cannot be ruled out. Weber (1920/2009) anticipated this influence, but few research efforts have directly addressed the effect of sec recy on transparency. This suggests that models of transparency need to consider the institutional influences of secret keeping and that organization variance due to secret keeping needs to be accounted for in transparency research. The presence of a diffe rential between intelligence agencies and non intelligence agencies bears monitoring and may suggest that these agencies represent a special case of transparency.

PAGE 172

160 Second, this dissertation makes a contribution to theory by proposing an elaborated model of transparency that describes the element of medium using three aspects: content, mechanism and timing. The findings highlight the need to consider these aspects in understanding transparency in practice. The findings suggest that secret keeping organization s may differ considerably from their peer agencies in each of these aspects, particularly in mechanism and timing. Additional work to test performance differences using these aspects may help shed additional light on transparency initiatives and programs. Third t his dissertation makes a unique contribution to understanding transparency as a service by providing knowledge about intelligence agency FOIA program performance. (2013) instrumental definition of transparency supported by the work of Daw es (2010) and Hood (2006) highlights the need to consider not only the what of transparency, but also the how Piotrowski (2007) found that NPM reforms had unintended consequences on FOIA service and performance with the potential to negatively impact effective transparency. Roberts (2006) found transparency in national security networks to be a topic of concerned and advocated additional study. Building off of the conclusions of Meijer (2013), Piotrowski (2007) and Roberts (2006), t his dissertation dir ectly evaluate s several aspects of transparency as a service by examining FOIA performance indicators This research shows that the timing of transparency is an important element and immediacy may distinguish transparency of current workings and activities from historical hindsight. This is an important distinction as the former may be used by principals to actively influence the actions of the actor while the latter only provides after the face awareness. Finally, t his research also sheds light on the conc ept of mediated transparency advanced by Meijer (2009) and Fung, Weil and Graham (2007). In particular, the findings highlight two

PAGE 173

161 areas of consideration. The first is e transparency, transparency mediated through electronic means. The experiment shows tha t nearly twenty years after the modification of FOIA to include electronic dissemination tasks many agencies make use of e transparency tools, but not all. Notably, the intelligence community lags significantly behind peers in adopting or using e transpar the question of the role of brokers in models of transparency. One can argue that paper record transparency lends itself to having a few organizations (journalists, ac ademic institutions) monitor the performance and workings of government agencies on behalf of citizens. While that model is still viable with e transparency, the use of digital media has the potential to lower reproduction, dissemination and discovery cost s if adopted. This suggests that e transparency can support a greater number of principals engaged in direct monitoring of agencies tha n i s viable using paper records. The idea that the choice of medium (paper or electronic) supports different models of tr ansparency (brokered versus direct) is a theoretical consi deration worthy of further examination. Contribution to Methodology comparative conclusions about FOIA program performanc e. The annual report data set provides a large number of observations and, as this research demonstrates, can be used to draw conclusions about FOIA program level outcomes that contribute to an overall assessment of agency and governmental transparency. Ad ditionally, a s this data set grows, it will prove increasingly valuable to researchers looking to examine changes over time. The experimental methodology adopted for this research provides a proof of concept for similar research efforts. Controlling for i dentically worded requests allowed the researcher

PAGE 174

162 to eliminate one source of variance and more directly compare disposition results and service performance This approach can conceivably be used to compare transparency outcomes from different levels of gov ernment (e.g. federal vs. state agencies). Of particular interest given the findings of this research is the ability to examine the mechanisms and service aspects of transparency. Another major contribution of this research is use of a scale of FOIA servi ce indicators that can be used to compare the relative transparency performance of agencies. The six indicators identified in Table 5.6 and used to evaluate the relative performance aspects of agencies in Chapter VII include measures for timing, electronic responsiveness and willingness to provide supplementary information. This service scale can be used to operationalize the availability aspect of definition of transparency by expands the concept of availability from a purely d ichotomous definition (information or no information) to a In other words, agencies that perform worse on the service scale may be considered providing less timely or useful availability than agencies who score higher. Additionally this scale, with modifications, can be extended beyond this scope of this research to look at transparency performance aspects for other government organizations such as local and state government agencies. Contri bution to Practice This dissertation proposes a number of recommendations aimed at different levels of implementation. To recap, the process of using annual report data to evaluate program outcomes resulted in recommendations to enhance the value of this d ata by requiring modifications to reporting criteria and a recommendation to explicitly require reporting from

PAGE 175

163 plicitly defining penalties for non performance or non compliance. Addition of these clauses could strengthen the law and allow parties advocating for increased transparency additional recourse when dealing with federal agencies. The experimental results s uggest that e transparency for intelligence agencies can be substantially improved. Additionally, FOIA administrators may want to consider expanded use of the Office of Government Information Services and adoption of OGIS best practices to improve service to requesters. The experiment also resulted in the recommendation to enhance the connection between records management and transparency by explicitly linking records management practices to transparency outcomes. Limitations Though presumed accurate, the annual report data used in this research has some requests made for legitimately classified information, in the reported analysis. The FOIA logs collected from intelligence agencies do not specify which requests were for classified material, making it difficult to isolate the impact of administrative secrecy from official secrecy. This is a difficult barrier for researchers to overcome. The experimental approach adopted controls for one possible source of variance between agencies by holding the content of the request constant. In order to attempt to get valid responses from agencies with very different types of missions, the requests were for administrative reco rds. Use of mission related records might generate different results, although selecting mission records common to different agencies would likely prove difficult.

PAGE 176

164 The comparative approach used during the experiment assumes that agencies are roughly similar in terms of capacity to research and respond to requests. Control agencies were selected for their similarity to intelligence agencies in terms of FOIA program resources to attempt to reduce potential capacity var iances. However, unobservable differences in information technology capabilities or personnel skill or competence levels may be present that impacted experimental results. These differences are difficult to observe from outside the organization and may req uire that research be performed with the consent and cooperation of the agencies under study. Additionally, the researcher was unable to control for unusual or unplanned circumstances that might have skewed results As an example, the CIA and NRO reported that a technological problem prevented them from receiving round 3 requests in a circumstances was impossible. The FOIA service scale (table 5.6) used for the experi mental rounds allows for comparison of performance between agencies. The scale is subject to construct validity issues, however. Selecting performance indicators that were common to both intelligence and control agencies and valid regardless of disposition status led to weighting practical considerations heavily. The selected indicators were drawn from several sources, including analysis of the FOIA itself and represent several common service dimensions such as timeliness and responsiveness. However, these indicators only partially assess service quality and leave open the possibility of alternate service measures. In addition to the limitations described above, this research rests on two significant presumptions. The first presumption is that the US Intell igence Community is sufficiently

PAGE 177

165 the Intelligence Community as a whole. This is a defensible presumption, but not completely unassailable. Many other agencies c ertainly keep secrets of different sorts, even if they are not designated as such using classification or dissemination control markings. The unique nature of the intelligence agencies leads to the assumption that secrecy is stronger in these agencies than in non intelligence agencies, but this assumption is difficult, if not impossible, to validate. The cultural secrecy of other agencies, including some of those selected as control agencies, may be sufficiently strong enough to skew results and challenge t he generalizability of the research. The second key presumption is that FOIA is sufficiently related to transparency to be and may not be the most significant mea sure. FOIA performance, at the program and individual agencies also practice proactive transparency, publishing key reports and documents online before they a re requested. Measuring proactive transparency proved difficult in practice and was subject to construct validity problem. A combined assessment of reactive and proactive transparency would better mitigate construct validity issues associated with capturin g only a single dimension or aspect of transparency. Recommendations for Future Research Two directions are suggested to continue this stream of research. The first direction deepens our understanding of transparency mechanisms in the federal government i n particular and can help add to the knowledge of FOIA implementation. The second direction seeks to broaden our knowledge of transparency in other contexts and situations.

PAGE 178

166 Deepening Research in Secrecy and Transparency The research topics posed in the fo llowing paragraphs will help improve our knowledge of the specific issues and challenges faced by secret organizations Research in this area will be valuable to understanding special transparency circumstances that merit unique or special policy considera tions. Lasswell (1971) described the need for perspective is needed to understand both secrecy and transparency in practice The use of annual report data provides greater knowledge of transparency programs, particularly as it relates to the intelligence community, and the experimental methodology offers a little e internal dynamics of agencies providing FOIA as a service. The responses received during the experiment hint at legal, technical, procedural and cultural considerations that influence effective transparency. Additional research is needed to provide bette r knowledge in transparency programs and discover how these different considerations impact the end result. As an example, the administrative discretion literature suggests that employees experience a number of competing influences. Case studies are needed to explore how these influences affect transparency outcomes and how influences differ between agencies. Similarly, better insight into comparative secrecy is needed. Weber (1920/2009) states that all bureaucracies keep secrets, but determining how much m ore secret one organization is from another is challenging. For example, should the NSA or CIA be treated has having the same degree of secrecy as other intelligence agencies ? Similarly, should the TSA or NASA be viewed as

PAGE 179

167 just as secret (or not secret) as the USGS or USDA? Developing testable measures of secrecy could go a long way towards building understanding of secrecy in theory and practice. Investigate E Transparency. The experiment found a strong differential between intelligence and control agenci es with respect to electronic responsiveness and e transparency. Additional research is needed to discover the causes contributing to this variance. Of particular interest is whether the difference is attributable to technical capacity, programmatic focus (such as funding for e transparency efforts within agencies) agency level policies or social/institutional reasons. This differential also raises interesting questions about how researchers and citizens should treat the media of transparency. How does the choice of medium affect transparency outcomes? Is the gap between agencies that use electronic means to communicate with citizens and thos e that do not growing or shrinking over time? Should electronic transparency be considered qualitatively different from non electronic transparency? These questions merit additional investigation and discussion. The use of electronic means to disseminate i nformation is directly related to proactive transparency. As mentioned in the limitations section, assessing differences in proactive transparency proved challenging. Additional work is needed in this area to develop a rubric or model that can be used to c ompare proactive transparency efforts. Investigate Records Management and Contr olled Unclassified Information i mplementation The experiment revealed differences between agencies in records management goals. A deeper look at records management practices an d policies is needed to assess if these differences result in appreciable variances in effective transparency. The

PAGE 180

168 experiment also showed that few agencies have yet to adopt controlled unclassified information policies. This should be a source of concern f or open government proponents. As federal agencies develop and release CUI policies, research is needed to evaluate how these policies impact the releasability of information and what oversight is needed to ensure this category of information marking is no t used inappropriately. An Expanded Agenda for Secrecy and Transparency Research In addition to deepening understanding of federal government transparency, a broader research agenda is needed to explore different aspects of the secrecy and transparency dyn amic. This research expands beyond the intelligence community and FOIA implementation to build knowledge in other areas. Exploring directional transparency. The principal (2006) work in particular suggests that transp arency is directional and may be experienced differently by different principals. This is an interesting phenomenon to explore as it could reveal institutional biases that advantage or disadvantage certain classes or categories. The social construction fra mework (Schneider, 1993) might be applied here with intriguing results. Deeper look at transparency service measures needed. This research suggests that there are potentially many different ways of examining transparency as a service. In this research, temporal, electronic responsiveness, and information content indicators were used to compare relative performance of FOIA servi ce. These service performance indicators used in the experiment are only a few of the possible ways to evaluate FOIA as a service or transparency as a service Additional research is needed to develop appropriate measures of transparency and information se rvices provided by agencies. A similar effort is needed to develop measures for evaluating proactive transparency; the information openly provided by agencies.

PAGE 181

169 Brokered versus direct transparency. Another potentially rich area for future research is explo ring the presence of transparency brokers such as journalists or academics. Digital records can be disseminated or re disseminated at low cost, potentially leading to broader consumption of transparency artifacts than paper records permit. This difference in medium has the potential to create a difference in transparency consumption models, a direct to consumer model supported by electronic records and a brokered model supported by select distribution of paper records. The former model needs additional rese arch to understand and compare to the latter. Additionally, measures of transparency are needed to understand if one model is preferable to another and under what circumstances. Comparing federal, state and local transparency. This research was limited to federal agencies to eliminate potential confounding influences. However, the basic experimental methodology can be applied to compare public organizations both within category and between categories. Though FOIA is specific to federal agencies, the service i ndicators, with modifications, can be used to create a baseline of transparency performance to explore differences in performance between organizations. Comparing the grammar of transparency. Using the institutional grammar of transparency to compare tr ansparency related rule sets might reveal interesting institutional differences. The analysis of FOIA contained in Chapter III provides a baseline for comparing state, local or international laws and identifying the actors, aims and conditions of transpare ncy. Exploring other typological sources of variance. This research was limited to exploring the difference between intelligence agencies and other agencies. However, other differences between agencies might result in different transparency outcomes or ser v ice variances. Organizational typology such as mission or service oriented agencies might affect transparency

PAGE 182

170 outcomes. Additionally, the location of responding organizations or the difference between field and headquarters organizations might provide use ful to understanding transparency Conclusion In sum, this dissertation makes an important contribution to our understanding of the relationship between secrecy and transparency at both a theoretical and practical level. Analysis of annual report data shows that intelligence agencies grant fewer and d eny more requests and take longer to render final dispositions than other agencies. The experiment shows that while intelligence agencies make substantially similar disposition decisions to control agencies given identical requests, they take longer to do so and are less likely to respond electronically or s upplement the disposition with additional information than control agencies. These findings lead to the conclusion that aspects of transparency, such as content, means and timing, are affected by secrecy The research resulted in r ecommendations for policy makers and FOIA administrators as well as suggestions for future secrecy and transparency scholarship that can build off the contribution made by this research.

PAGE 183

17 1 REFERENCES Aftergood, S. (1999) Governm ent secrecy and knowledge production: A survey of some general issues. In S. Maret & J. Goldman (Eds.), Government secrecy: Classic and contemporary readings. (2009) Westport CT: Libraries Unlimited. Anand, V., and Rosen, C. C. (2008). The Ethics of Organi zational Secrets. Journal of Management Inquiry 17 (2), 97 101. doi:10.1177/1056492607312785 Archibald, S. (1979). The Freedom of Information Act Revisited. Public Administration Review 39 (4), 311 318. Retrieved from http://www.jstor.org/stable/10.2307/976206 Bannister, F. and Connolly, R. (2011). The trouble with transparency: A critical review of openness in e Government. Policy & Internet, 3 (1), 1 30. Basurto, X., and Kingsley, G., McQueen, K., Smith, M., Weible, C. (2010). A systematic approach Quarterly, 63(3), 523 537. I. Galnoor (Ed.) Government secrecy in democracies (pp. 3 21). New York: New York University Press. Bay, C. (1977). Access to political knowledge as a human right. In I. Galnoor (Ed.) Government secrecy in democracies (pp. 22 39). New York: New York Unive rsity Press. Bennett, C. (1985). From the Dark to the Light: The Open Government Debate in Britain. Journal of Public Policy 5 (2), 187 213. Retrieved from http://journals.cambridge.org/production/action/cjoGetFulltext?fulltextid=2692676 Berkowitz, B. (201 1). The National Reconnaissance Office at 50 years: A brief history. Chantilly VA: National Reconnaissance Office. Blank, L. (2008). Two schools for secrecy: Defining secrecy from the works of Max Weber, George Simmel, Edward Shils and Sissela Bok. In S. Maret & J. Goldman (Eds.), Government secrecy: Classic and contemporary readings. (2009) Westport CT: Libraries Unlimited. Bok, S. (1982). Secrets: On the Ethics of Concealment and Revelation. New York: Pantheon Books. Bozeman, B. (1993). A t heory of g Journal of Public Administration Research and Theory 3 (3), 273 303. Retrieved from http://jpart.oxfordjournals.org/content/3/3/273.short Bozeman, B. (2000). Bureaucracy and red tape Upper Saddle River, NJ: Prentice Hall. Christoph, J. (1975). A comparative view: Administrative secrecy in Britain. Public Administrative Review, 35 (1), 23 32.

PAGE 184

172 Commission on Protecting and Reducing Government Secrecy. (1997). Rethinking classification: Better protection and greater openness. In S. Maret & J. Goldman (Eds.), Government secrecy: Classic and contemporary readings. (2009) Westport CT: Libraries Unlimited. Controlled Unclassified Information Task Force. (2009). Reports and recommendations of the Presidential Task Force on controlled unclassified information. Crawford S. and Ostrom, E. (1995). A grammar of institutions. American Political Science Review, 89 (3), 582 600. Druckman, J. N. (2001). The implications of framing effects for citizen competence. Political Behavior 23 (3), 225 256. Retrieved from http://www.springerlink.com/index/TEJ2E7MKY1P171MM.pdf Dufresne, R. L. and Offstein, E. H. (2008). On the virtues of secrecy in organizations. Journal of Management Inquiry, 17 (2), 102 106. Eisenhardt, K. (1989). Agency theory: An asses sment and review. The Academy of Management Review, 14 (1), 57 74. Etzioni, A. (2010). Is transparency the best disinfectant? The Journal of Political Philosophy, 18 (4), 389 404. Executive Order No. 13392, 3 C.F.R. (2005). Improving Agency Disclosure of Information. Executive Order No. 13556, 3 C.F.R. (2010). Controlled Unclassified Information. Finer, H. (1941). Administrative responsibility in democratic government (pp. 176 187). Public. First Amendment Center (2001). Freedom of Information in the Digi tal Age. Reston, VA: American Society of Newspaper Editors Association. The Freedom of Information Act, 5 U.S.C. § 552, As Amended By Public Law No. 110 175 121 Stat. 2524 and Public Law No. 111 83, § 564, 123 Stat. 2142, 2184 (2009). Friedman, D. D., Landes, W. M., & Posner, R. A. (1991). Some economics of trade secret law. The Journal of Economic Perspectives 5 (1), 61 72. JSTOR. Retrieved from http://www.jstor.org/stable/1942702 Fri edrich, C. J. (1972). Nature and function of secrecy and propaganda. In S. Maret & J. Goldman (Eds.), Government secrecy: Classic and contemporary readings. (2009) Westport CT: Libraries Unlimited. Fung, A., Graham, M., and Weil, D. (2007). Full Disclosure : The Perils and Promise of Transparency. Cambridge: Cambridge University Press.

PAGE 185

173 Fung, A., Graham, M., Weil, D., & Fagotto, E. (2004). The Political Economy of Transparency: What makes disclosure policies effective? Available at SSRN 766287 Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=766287 Galnoor, I. (1975). Government secrecy: Exchanges, intermediaries and middlemen. Public Administration Review, 35 (1), 32 42. Galnoor, I. (1977). Government secrecy in democracies New York: New York University Press. Griffith, J. (1989). The Official Secrets Act 1989. Journal of Law and Society, 16 (3), 273 290 Heald, D. (2006). Varieties of transparency. In C. Hood and D. Heald (Eds.), Transparency: The Key to Better Governance? ( 2006 ) Oxford: Oxford University Press Hermalin, B. E., & Katz, M. L. (2006). Privacy, property rights and efficiency: The economics of privacy as secrecy. Quantitative Marketing and Economics 4 (3), 209 239. Hood, C ( 2006 ). Transparency in historical perspective In C. Hood and D. Heald (Eds.), Transparency: The Key to Better Governance? ( 2006 ) Oxford: Oxford University Press Information Security Oversight Office. (2011). Report to the President. Washington DC: National Archives and Records Administration. Jensen, P. H. and Webster, E. (2009). Knowledge management: Does capture impede creation?. Industrial and Corporate Change, 18 (4), 701 727. Kahneman, D. (2003). Maps of bounded rationality: Psychology for behavioral economics. The American economic review 93 (5), 14 49 1475. JSTOR. Retrieved from http://www.jstor.org/stable/3132137 Kaya, C. (2008). Official secrecy in the United Kingdom and the role of the Freedom of Information Act with regard to reducing secrecy. Erzincan University Faculty of Law Review 12 (3 4), 13 32. -Their Use and Abuse in Organizations. Journal of Management Inquiry 17 (2), 107 110. doi:10.1177/1056492607313233 Kim, M. (2007). Numbers tell part of the story: A comparison of FOIA implementation und er the Clinton and Bush administrations. Communication Law and Policy, 12 (3), 313 339. Kultti, K., Takalo, T. and Toikka, J. (2006). Simultaneous model of innovation, secrecy and patent policy. The American Economic Review, 96 (2), 82 86. Lasswell, H. D. (1971). A pre view of policy sciences New York: American Elsevier Publishing Company. Light, P. C. (2006). The tides of reform revisited: Patterns in making government work, 1945 2002. Public Administration Review 66 (1), 6 19. Retrieved from http://wagne r.es.its.nyu.edu/performance/files/Tides of Reform Revisited.pdf

PAGE 186

174 Lipsky, M. (1980). Street level bureaucracy: Dilemmas of the individual in public services. New York, NY: Russell Sage Foundation. Lowi, T. (1977). The Information Revolution, Politics, and the Prospects for an Open Society. In I. Galnoor (Ed.) Government secrecy in democracies (pp. 40 61). New York: New York University Press. Maynard Moody, S., & Musheno, M. (2000). State Agent or Ci tizen Agent: Two Narratives of Discretion. Journal of Public Administration Research and Theory 10 (2), 329 358. doi:10.1093/oxfordjournals.jpart.a024272 March, J. G. (1991). Exploration and exploitation in Organizational Learning. Organization Science 2 ( 1), 71 87. doi:http://dx.doi.org/10.2307/2634940 Meier, K. J., & Bohte, J. (2001). Structure and Discretion: Missing Links in Representative Bureaucracy. Journal of Public Administration Research and Theory 11 (4), 455 470. doi:10.1093/oxfordjournals.jpart .a003511 Meijer, A. (2009). Understanding modern transparency. International Review of Administrative Sciences, 75 (2), 255 269. Meijer, A. (2013). Understanding the complex dynamics of transparency. Public Administration Review, 73 (3), 429 439. Mosher, F. C. (1968). Democracy and the public service (Vol. 53). New York. The National Security Archive. (2003). The National Security Archive. (20 10 ). Sunshine and Shadows: The Clear Message for Freedom of Information meets Mixed Results. Washington D.C.: George Washington University. The National Security Archive. (20 11 ). Glass Half Full. Washington D.C.: George Washington University. Newbold, S. ( 2011). Federalist No. 27: Is transparency essential for public confidence in government. Public Administration Review, S71 S47 S52. Niskanen, W. A. (1968). The peculiar economics of bureaucracy. The American Economic Review, 58 (2), 293 305. 979). Freedom of Information, Privacy, and Information Control: A Contemporary Administrative Dilemma. Public administration review 39 (4), 323 328. Retrieved from http://www.jstor.org/stable/10.2307/976208 Transparency and the ethics of communication. In C. Hood and D. Heald (Eds.), Transparency: The Key to Better Governance? ( 2006) Oxford: Oxford University Press.

PAGE 187

175 Office of Government Informatio n Services. (2014). Best practices. Retrieved from: https://ogis.archives.gov/about foia/best practices.htm The Openness Promotes Effectiveness in our National (OPEN) Government Act, Pub. L. No. 110 175, 5 U.S.C. § 552 (2007). Pasquier, M. and Villeneuve, P. (2007). Organizational barriers to transparency: A typology and analysis of organizational behavior tending to prevent or restrict access to information. International Review of Administrative Sciences, 73 (1), 147 162. Piotrowski, S. (2007). Government Transparency in the Path of Administrative Reform. Albany, NY: State University of New York Press. Piotrowski, S. (2009). Is transparency sustainable? Public Administration Review 69 (2), 359 361. Retrieved from http://onlinelibrary.wiley.com/doi/10.1111/j .1540 6210.2008.01983.x/abstract Piotrowski, S., & Rosenbloom, D. (2003). Nonmission Based Values in Results Oriented Public Management: The Case of Freedom of Information. Public Administration Review 62 (6), 643 657. Retrieved from http://onlinelibrary.wiley.com/doi/10.1111/1540 6210.00247/abstract Prat, A. (2006). The more closely we are watched, the better we behave? In C. Hood and D. Heald (Eds.), Transparency: The Key to Better Governance? ( 2006 ) Oxford: Oxford University Press Relyea, H. (1975). Opening government to public scrutiny: A decade of federal efforts. Public Administration Review 35 (1), 3 10. Retrieved from http://www.jstor.org/stable/10.2307/975195 Relyea, H. (1979). Faithful Execution of the FOI Act: A Legislative Branch Perspective. Public Administration Review 39 (4), 328 332. Retrieved from http://www.jstor.org/stable/10.2307/976209 Riccucci, N. M. (2005). How Management Matters: Street Level Bureaucrats and Welfare Reform. Washington, DC: Georgetown University Press. Roberts, A. (2000). Less government, more secrecy: Reinvention and the weakening of freedom of information law. Public Administration Review 60 (4), 308 320. Retrieved from http://onlinelibrary.wiley.com/doi/10.1111/0033 3352.00093/abstract Roberts, A. (2001). Structural pluralism and the right to information. The University of Toronto Law Journal, 51 (3), 243 271. Canadian Public Administration, 45 (2), 175 194.

PAGE 188

176 Roberts, A. (2006).Blacked out: Government Secrecy in the Information Age. Cambridge: Cambridge University Press. Roberts, R. (1979). Faithful execution of the FOI Act: One executive branch experience. Public administration review 39 (4), 318 323. Retrieved from http://www.jstor.org/stable/10.2307/976207 Rourke, F. (1957). S ecrecy in the American bureaucracy. Political Science Quarterly, 72 (4), 540 564. Rourke, F. (1960). Administrative secrecy: A Congressional dilemma. The American Political Science Review, 54 (3), 684 694. Rowat, D. C. (1965). How much administrative secrecy ?. The Canadian Journal of Economics and Political Science, 31( 4), 479 498. Sandfort, J. R. (2000). Moving Beyond Discretion and Outcomes: Examining Public Management from the Front Lines of the Welfare System. Journal of Public Administration Research and Theory 10 (4), 729 756. doi:10.1093/oxfordjournals.jpart.a024289 Schneider, A. (1993). Social construction of target populations: Implications for politics and policy. American Political Science Review, 87(2), 334 347. Retrieved from http://www.jstor.org/ stable/10.2307/2939044 Scott, P. G. (1997). Assessing determinants of bureaucratic discretion : An experiment in street level decision m aking. Journal of Public Administration Research and Theory 7 (1), 35 58. doi:10.1093/oxfordjournals.jpart.a024341 Seymour Ure, C. (1977). Great Britain. In I. Galnoor (Ed.) Government secrecy in democracies (pp. 157 175). New York: New York University Press. Shils, E. (1956). The Torment of Secrecy: The Background and Consequences of American Security Policies. Chicag o: Dee. Siddki, S., and Weible, C., Basurto, X., Calanni, J. (2011). Dissecting policy designs: An application of Institutional Grammar Tool. Policy Studies Journal, 39 (1), 79 103. Simmel, G. (1906). The Sociology of Secrecy and of Secret Societies. Americ an Journal of Sociology 11 (4), 441 498. Simon, H. A. (19 4 7). Administrative Behavior (2 nd Ed.). New York : Macmillian Company. Sowa, J. E., and Selden, S. C. (2003). Administrative discretion and active representation: An expansion of the theory of representative bureaucracy. Public Administration Review 63 (6), 700 710. Retrieved from http://onlinelibrary.wiley.com/doi/10.1111/1540 6210.00333/ab stract

PAGE 189

177 Stigler, G. J. (1980). An Introduction to Privacy in Economics and Politics. The Journal of Legal Studies 9 (4), 623 644. doi:10.1086/467657 Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychology of choice. Science (New York N.Y.) 211 (4481), 453 8. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/7455683 Tversky, A., & Kahneman, D. (1986). Rational Choice and the Framing of Decisions. The Journal of Business 59 (S4), S251. doi:10.1086/296365 United States Department of Jus tice. (2004, May). Freedom of Information Act Guide. Retrieved from: http://www.justice.gov/oip/exemption2.htm United States Department of Justice. (2009). DOJ Guide to the Freedom of Information Act (2009 edition). Retrieved from: http://www.justice.gov/oip/foia_guide09.htm United States Department of Jus tice. (2011, February). Frequently asked questions. Retrieved from: http://www.foia.gov/faq.html#important United States Department of Justice. (2012). FOIA data at a glance FY2008 FY2011. Retri eved from: http://www.foia.gov/index.html Vaughan, D. (1997). Structural Secrecy. In S. Maret & J. Goldman (Eds.), Government secrecy: Classic and contemporary readings. (2009) Westport CT: Libraries Unlimited Waldo, D. (1980). The Enterprise of Public Administration. Novato, CA: Chandler and Sharp. Weber, M. (1920). Bureaucracy: Characteristics and the power position. In S. Maret & J. Goldman (Eds.), Government secrecy: Classic and contemporary readings. (200 9) Westport CT: Libraries Unlimited. White House. (2009). Memorandum for the Heads of Executive Departments and Agencies. Retrieved from: http://www.whitehouse.gov/the press office/freedom information act

PAGE 190

AMERICAN SECRETS, AMERICAN TRANSPARENCY 178 APPENDIX A TEXT OF EXPERIMENTAL REQUESTS Trial Round 1 25 Sept 2013 Dear Coordinator: Under the Freedom of Information Act, 5 U.S.C. subsection 552, I am requesting : Agency specific instructions or directives for the management and disposition of ssing the retention and/or disposal of records. Posting these records on the agency FOIA website or electronic reading room with notification to the requester will be considered fulfillment of this request. If there are any fees for searching for, reviewi ng, or copying the records, please notify me before processing if the amount exceeds $10. Because this information is in the public interest and will be used for academic purposes, I am requesting a waiver for fees. Information regarding the retention of r ecords is directly applicable to execution of the Freedom of Information Act. If you deny all or any part of this request, please cite each specific exemption you think justifies your refusal to release the information and notify me of appeal procedures av ailable under the law. If you have any questions about handling this request, you may telephone me at xxx xxx xxxx Sincerely, Cameron Ward Hunt School of Public Affairs, University of Colorado Denver Robert.ward hunt@ucdenver.edu

PAGE 191

179 Trial Round 2 12 Nov 2 013 Dear Coordinator: Under the Freedom of Information Act, 5 U.S.C. subsection 552, I am requesting : Agency specific policy guidance, directives, or instructions implementing or documenting a Controlled Unclassified Information program as directed by EO13556. Electronic copies of the records are preferred. Posting these records on the agency FOIA website or electronic reading room with notification to the requester will be considered fulfillment of this request. If there are any fees for searching fo r, reviewing, or copying the records, please notify me before processing if the amount exceeds $10. Because this information is in the public interest and will be used for academic purposes, I am requesting a waiver for fees. If you deny all or any part of this request, please cite each specific exemption you think justifies your refusal to release the information and notify me of appeal procedures available under the law. If you have any questions about handling this request, you may telephone me at xxx xx x xxxx Sincerely, Cameron Ward Hunt School of Public Affairs, University of Colorado Denver Robert.ward hunt@ucdenver.edu

PAGE 192

180 Trial Round 3 8 Jan 2014 Dear Coordinator: Under the Freedom of Information Act, 5 U.S.C. subsection 552, I am requesting the following records: Agency FOIA Logs or request tracking records for FOIA requests for the period 1/1/12 to 12/31/13. Electronic copies (PDF, XLS, DOC, etc.) of the records are requested. Posting these records on the agency FOIA website or electronic reading room with notification to the requester will be considered fulfillment of this request. If there are any fees for searching for, reviewing, or copying the records, please notify me before processing if the amount exceeds $10. Because this informati on is in the public interest and will be used for academic purposes, I am requesting a waiver for fees. If you deny all or any part of this request, please cite each specific exemption you think justifies your refusal to release the information and notify me of appeal procedures available under the law. If you have any questions about handling this request, you may telephone me at xxx xxx xxxx Sincerely, Cameron Ward Hunt School of Public Affairs, University of Colorado Denver Robert.ward hunt@ucdenver.ed u

PAGE 193

181 APPENDIX B THE US INTELLIGENCE COMMUNI TY The United States Intelligence Community consists of the Office of the Director of National Intelligence (ODNI) and sixteen components. These components differ greatly in size and status as indicated by Table A 1 below. While the CIA and the ODNI are indep endent agencies, the rest of the components fall under one of six departments: Defense, State, Justice, Treasury, Homeland Security and Energy. As Table A 1 indicates, most of these components do not issue a separate FOIA report and are instead part of lar ger office or agency reporting mechanisms.

PAGE 194

182 Table A 1 Intelligence Community Members Name Abrv Status Parent Agency or Component Parent Dept FOIA Report Central Intelligence Agency CIA Independent Agency N/A N/A Y Defense Intelligence Agency DIA Component Agency N/A DoD Y National Geospatial Intelligence Agency NGA Component Agency N/A DoD Y National Security Agency NSA Component Agency N/A DoD Y National Reconnaissance Office NRO Component Agency N/A DoD Y Office of the Director of National Intelligence ODNI Independent Agency N/A DoD Y Army Intelligence G2 Staff USA DoD N Navy Intelligence N2 Staff USN DoD N Air Force Intelligence Surveillance and Reconnaissance Agency AF ISR Component Agency USAF DoD N Marine Corps Intelligence Staff USMC DoD N Coast Guard Intelligence Staff USCG DHS N Office of Intelligence and Counterintelligence Office N/A Energy N Office of Intelligence and Analysis I&A Office N/A DHS Y Bureau of Intelligence and Research INR Bureau N/A State N Office of Intelligence and Analysis OIA Office OTFI Office of Terrorism and Financial Intelligence Treas N Office of National Security Intelligence ONSI Office DEA DoJ N National Security Branch NSB Branch FBI DoJ N