Citation
An analysis of election poll worker costs and cost drivers

Material Information

Title:
An analysis of election poll worker costs and cost drivers
Creator:
Hentze, Iris
Place of Publication:
Denver, Colo.
Publisher:
University of Colorado Denver
Publication Date:
Language:
English

Notes

General Note:
This client-based project was completed on behalf of National Conference of State Legislatures (NCSL) and supervised by PUAD 5361 Capstone course instructor Dr. Wendy Bolyard and second faculty reader Dr. Todd Ely. This project does not necessarily reflect the views of the School of Public Affairs or the faculty readers. Raw data were not included in this document, rather relevant materials were provided directly to the client. Permissions to include this project in the Auraria Library Digital Repository are found in the final Appendix. Questions about this capstone project should be directed to the student author.

Record Information

Source Institution:
University of Colorado Denver Collections
Holding Location:
Auraria Library
Rights Management:
Copyright [name of copyright holder or Creator or Publisher as appropriate]. Permission granted to University of Colorado Denver to digitize and display this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.

Downloads

This item has the following downloads:


Full Text
An Analysis of Election Poll Worker Costs and Cost Drivers
Iris Hentze
University of Colorado Denver School of Public Affairs
This client-based project is submitted in partial fulfillment of the requirements for the degree of Master of Public Administration in the School of Public Affairs at the University of Colorado Denver Denver, Colorado
Summer
2017


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
2
Capstone Project Disclosures
This client-based project was completed on behalf of National Conference of State Legislatures (NCSL) and supervised by PUAD 5361 Capstone course instructor Dr. Wendy Bolyard and second faculty reader Dr. Todd Ely. This project does not necessarily reflect the views of the School of Public Affairs or the faculty readers. Raw data were not included in this document, rather relevant materials were provided directly to the client. Permissions to include this project in the Auraria Library Digital Repository are found in the final Appendix. Questions about this capstone project should be directed to the student author.


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 3
Contents
Executive Summary.....................................................................4
Introduction..........................................................................5
Literature Review.....................................................................7
Methodology..........................................................................16
Results..............................................................................22
Discussion & Recommendations.........................................................25
Conclusion...........................................................................29
References...........................................................................31
Appendices...........................................................................33


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 4
Executive Summary
In the field of election administration, scholars and practitioners alike face a lack of available research and data on the human cost component of elections. The National Conference of State Legislatures (NCSL), which studies the costs of elections and election technology among many other election administration topics has requested data gathering and analysis of the cost of election poll workers. This capstone project accomplishes NCSLs request by both amassing poll worker cost data in one place and by performing statistical analysis on the data to test for relationships between election cost drivers. Two key results are reported. First, the results suggest that there is a strong relationship between the number of polling places present in a jurisdiction and the number of poll workers hired. Next, according to this studys modeling, certain measures implemented in recent years to make voting more convenient for citizens potentially impact the number of poll workers a jurisdiction must hire and the cost of an election. One sweeping recommendation is offered. The most important step NCSL can take to better understand the costs and cost drivers associated with poll workers is to continue to study this topic further. With the information contained in this study, NCSL will have basic elections cost data readily available, which can serve as a starting point for future, in-depth analysis on poll
worker costs in elections.


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
5
An Analysis of Election Poll Worker Costs and Cost Drivers
Elections and voting in the United States have undergone a major transformation in the last two decades. As detailed in the literature, the chaos from the presidential election of 2000 and the subsequent implementation of the 2002 Help America Vote Act (HAVA) began a change in the way elections are administered. HAVA greatly increased administrative requirements on election officials across the country as it intended to help states improve their elections and voting processes for the better (Hill, 2011). Before HAVA, the process of conducting elections in the US was so decentralized that few nationwide standards existed and those that did were a result of legislation intended to deal with illegal restrictions on poll access such as those related to race, income or education level. HAVA sought to standardize election administration, modernize election equipment and strengthen the integrity of elections to prevent any future crises like that which occurred in the 2000 general election.
The variety of new voting requirements and measures in HAVA were accompanied by grants to help local governments implement the policies. A large sum in federal grant funding was allocated to states beginning in 2002 to be used on purchasing new elections technology and meeting the new regulatory requirements. The academic study of election administration and related costs became increasingly popular after the implementation of HAVA. As the regulatory requirements surrounding elections and voting expanded significantly, so did studying the impact of these new regulations on election integrity, voter turnout and the cost of elections.
Given the relative infancy of this field, critical areas have yet to be studied including poll worker costs. The body of research that currently exists on election costs focuses primarily on logistical changes to the way citizens vote including the implementation of vote centers and the expansion of convenience measures such as same day voter registration and early voting. Poll


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
6
worker costs are also important in election administration, but have yet to be focused on with the same rigor due both to the relative infancy of the field and the difficulty in uncovering cost information from overall elections cost data.
Poll workers are an integral part of modern election administration. While voting machines have improved the accuracy and reliability of election results, they have not decreased the need for poll workers. In fact, the switch to electronic voting systems has increased the need for more highly skilled and trained poll workers to field questions and provide assistance to voters in the booth (Montjoy, 2010). Particularly when it comes to serving populations that may not be as technologically literate as others, patient, well-informed poll workers are integral to an accessible voting experience (Classen, Magleby, Monson, & Patterson, 2008).
Funding from HAVA grants has dissipated or is now long gone, but with tightening budgets for many state and local governments, securing new funding sources for election administration proves to be a difficult task. Accessible poll worker cost data are important for local governments to be able to effectively construct budgets for future elections. Additionally, researchers must gain a better understanding of how poll worker costs impact the overall resource burden of elections in order to analyze current trends, make future policy recommendations and help practitioners realize what resources they will need to effectively administer elections in the coming years.
This research aims to add to the literature in the field and provide detail on the topic of poll worker cost data. In aggregating secondary data from the 2012 Election Assistance Commissions Election and Voting Survey (EAVS), state statutes and other sources, the client for this project, the National Conference of State Legislatures (NCSL), will have data from which to build and conduct further research on poll worker costs. In addition to the aggregation


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
7
of data, this research will attempt to provide a basic understanding of some of the cost drivers associated with poll workers. Finally, this research is intended to demonstrate the importance of poll workers and the need for local election officials to track poll worker costs for each election.
This paper is organized into five sections beginning with a review of the literature. A description of the methodology is next, followed by a report of the results. Then, the discussion and recommendations section examines the results in more depth. Finally, this paper ends with a conclusion.
Literature Review
In the aftermath of the 2000 presidential election Congress passed HAVA to modernize and standardize elections across the nation. Since then, state legislatures have continued to add to and modify state election laws. According to Montjoy (2011), referencing NCSLs research, between 2001 and 2009 a total of 1,985 bills tackling election policy became law in the states.
As elections have gained in regulatory complexity so too have they increased in cost. HAVA included substantial grant funding to help states purchase the newest voting equipment, but now states and local governments are struggling to allocate funding to replace these aging systems.
HAVA ushered in two decades of change in election administration. Although HAVA sought to standardize parts of election processes and voting, it also left much of the implementation authority up to states and local election jurisdictions. In their 2010 study examining implementation differences in election laws across the country, Creek and Karnes (2010) state that, As long as localities are compliant with HAVA standards, there is no federal requirement that election districts within a state use the same voting equipment or provide the same training courses for poll workers (p. 276). Inevitably, this dissemination of authority to


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
8
states resulted in a wide variety of state voting laws originating in state legislatures across the nation from 2002 to 2010.
The bulk of existing research on election costs is focused on vote centers, early voting and same day voter registration. While these areas could all benefit from further study, there is a clear gap in the literature on the topic of poll worker costs. The research that has been conducted on the impact of poll workers on elections typically focuses on the behavior of poll workers and how their interactions with voters can impact public perceptions of elections. According to a study conducted by Bowler, Brunell, and Gronke (2014), the voters experience at a polling place impacts their perception of the fairness and integrity of the election process. Bowler et al. (2014) found that positive polling place experiences at the individual level can be aggregated into state level data that presents a picture of the overall public perception of the quality of elections. Similarly, Claassen et al. (2008) argues that the interaction that occurs between poll workers and the voting public at polling places can be compared to other customer service interactions between consumers and retailers.
Poll workers are street-level bureaucrats who use their discretion and decision-making autonomy to implement elections policy. Claassens (2008) study found that better trained poll workers result in better evaluations and recorded experiences from the public. Thus, the individual actions of poll workers impact overall public perception of the quality and integrity of elections.
One study examined the costs associated with poll workers found that personnel/poll worker costs are the single largest expense in election administration (Folz, 2014). Some studies of poll worker behavior do suggest a potential impact on election costs in terms of administrative efficiency and quality, but these studies prove to be preliminary (Bowler, Brunell & Gronke,


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
9
2014; Claassen et al., 2008). Ultimately, it is clear that more research is required to establish what the cost drivers associated with poll workers are and what impact poll workers costs have on election costs.
Vote Centers
The literature on the vote center model is more developed than most of the other components of elections cost. Vote centers are sometimes referred to as non-precinct place voting or NPPV. For the purposes of this review, the term vote center will be used to refer to both. Studies examining the cost-reducing effects of vote centers present varied results that range widely from county to county and depend heavily on contextual factors. For this reason, the literature presents a varied picture with some experts recommending vote centers as a costcutting tool and others expressing concerns over the policy consequences of moving to this new system.
Vote centers have developed as a way for local governments to make election administration more efficient, even in the face of increasing state and federal regulatory requirements and smaller budgets (Folz, 2014). In replacing traditional precinct polling place systems with new vote centers, whereby voters can go to any precinct in a county to vote rather than one assigned precinct, administrators are aiming to decrease costs and increase convenience by centralizing voting activity.
Larimer County, Colorado was one of the first jurisdictions to implement vote centers as a cost-reduction and efficiency-increasing measure in 2003 (Folz, 2014). Geographically, the county proves a difficult case for election administration. It sprawls east-to-west and contains two major population centers, several small towns and rural areas. Before the implementation of the vote center system, there were 143 required polling places throughout the county on election


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
10
day. Due to enabling legislation, beginning in 2003 the county transitioned to the vote center system and consolidated its polling places to a mere 22 vote centers on election day (Folz, 2014). Subsequent studies of the change implemented in Larimer County presented mixed findings on its impact on voter turnout, although gains in administrative efficiency were more prominently demonstrated (Folz, 2014; Montjoy, 2010).
Subsequent research on jurisdictions that have transitioned to vote centers have found relatively consistent cost savings across a number of different cases. One examination of election cost in Weld County, Colorado shows the effect of implementing vote centers in county elections between 2004 and 2008 (Montjoy, 2010). Located adjacent to Larimer County, Weld decreased its number of necessary polling places from 101 to 34 between the two elections and decreased its required number of poll workers from 516 to 415 in 2004 and 2008 respectively (Montjoy, 2010). All else remaining constant, these changes saved the county about $20,000.
Further analysis conducted on counties that had switched to vote center systems in Indiana demonstrated cost savings occurring from the change, principally due to a reduction in poll worker costs (Montjoy, 2010). Finally, a study conducted during the 2008 general election in Knox County, Tennessee resulted in a 9% decrease in overall election costs over 2004 when precinct polling places were still in place (Folz, 2014). The Knox County study found that poll worker costs accounted for the biggest portion of overall election costs at roughly 75% of the total budget. Overall, the vote center pilot program decreased election administration costs by 14.5% in 2008 (Folz, 2014). One key finding from the study includes a reduction in cost due to fewer poll workers overall and, subsequently, less necessary pay for overtime wages.
Vote centers are intended to be conveniently located and to operate during after-work hours and/or on weekends, allowing more voters to make it to the polls at the most convenient


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
11
time and location. However, in a 2011 analysis of vote centers, Chapin argues that vote centers have given voters more choices, which could be making election administration more complex and, as a result, more costly. For example, Chapin (2011) points to the much larger population poll workers must now verify voter eligibility for when they show up to a vote center. Additionally, he discusses the difficulty in the fact that strict geographical requirements are no longer abided by in casting votes, but they still are firmly in place for vote counting (Chapin, 2011). According to Chapin (2011), the way votes are counted has remained the same they must be attributed back to the geographic precinct where the voter resides. This has created an additional administrative burden for poll workers that may result in cost increases rather than decreases.
Vote centers have become a preferred alternative to consolidating polling places since their introduction in 2002 (Montjoy, 2010). Employing this method, based on the available literature, promises to provide counties with cost savings and streamlined elections while simultaneously increasing voter convenience. However, the switch from traditional polling place voting has produced unintended policy outcomes. First, as Montjoy (2010) points out in his analysis, the vote center system pushes election administrators into a resource allocation role that they may be unfamiliar with. With a precinct place system, the details of what resources go to which precincts is largely predetermined by laws that require the same standards of equipment, personnel and accessibility features. Vote centers present election administrators with the opportunity to make decisions about where vote centers should be, how many machines should be in each and how many poll workers should be at each (Montjoy, 2010). State statute still gives counties a baseline requirement to meet, but election administrators now have many more decisions to make in terms of resource allocation to the places in which voters are voting.


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
12
The existing literature presents a mixed narrative on the use of vote centers as a cost saving strategy. Vote centers are an attractive solution to decrease the administrative costs of elections and increase voter accessibility and convenience. The vote center model works by consolidating the many polling places in a county to a typically much smaller number of conveniently-located centers in which any resident of the county can vote. Championed for their cost-saving potential, vote centers have also produced unintended policy consequences that may increase the administrative burdens or costs associated with elections. The success of vote centers in terms of their ability to decrease cost is highly dependent on local conditions and contextual factors.
Voter Convenience Measures
Voter convenience models intended to increase the number of choices voters have in casting ballots have become a popular tool to help administrators modernize elections. Features like same day registration (SDR), election-day registration (EDR) and early voting have become fixtures in some states based on the theory that meeting the voters where they live and accommodating their schedules, will help to encourage turnout in elections. The literature on the impact of voter convenience measures on election costs is presently underdeveloped. More examination of the monetary and administrative impact of voter convenience measures needs to occur for these results to be useful to practitioners.
Early voting. Early voting was one of the first election modernization reforms aimed at providing a more convenient voting experience to the public. As stated by Chapin in his 2011 analysis of United States election reforms, election day used to be the only day available for voters to cast their ballots. Now, election day is simply the last day available to cast a vote in many jurisdictions in the country (Chapin, 2011). The literature on early voting presents an


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
13
incomplete picture of its true impact on elections, but does contain previous examination of its effects on voter turnout and election costs. According to a study conducted on 2008 general election turnout data, early voting decreases voter turnout by reducing, its stimulating effects on marginal voters (Burden, Canon, Mayer, & Moynihan, 2009, p. 22). The theory presented in the study claims that voters who do not vote regularly are more likely to participate on election day when there is excitement and enthusiasm leading up to it. Conversely, they are not as likely to turn out to vote with the anti-climactic effects of spreading the vote out over several weeks or months before election day. Another study also conducted on 2008 general election turnout data concluded that while early voting options do improve voting convenience, they do so for citizens that are already reliable voters, but do not necessarily attract new participation (Giammo & Brox, 2010).
Less conclusive data exist on the effects of early voting on the cost of elections. Early voting was initially implemented in many states because it was thought to decrease election costs by allowing the diffusion of ballot casting, and related administrative duties to occur over a period rather than rushed on a single day (Burden et al., 2009). Some argue that rather than reducing costs, it simply transfers existing costs as counties dedicate a portion of their elections budget away from election-day support to early voting activity. One study found that a number of counties spent nearly half as much on early voting alone as on election day (Giammo & Brox, 2010). Mecklenburg County in North Carolina spent $300,000 on early voting expenses and $650,000 for election day alone. Giammo and Brox (2010) posit that it may be more efficient for counties to avoid transferring funding from election budgets to early voting and instead improve election day services.


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
14
The costs associated with the implementation of early voting can vary by jurisdiction. According to one study, election officials in small municipalities are more likely to perceive the burden of early voting as too much for their jurisdictions to handle, as they may have little state support to rely on (Burden, Canon, Mayer, & Moynihan, 2012). Considering HAVA granted the bulk of election reform implementation authority to states, there is a national patchwork of early voting laws that provide varying degrees of financial support to local jurisdictions. For some small, rural jurisdictions it may be much more onerous a task to implement an early voting system, while larger jurisdictions in the same state may be more well equipped to provide this convenience to voters.
Although originally promised as a voter convenience feature with extensive opportunity to increase voter turnout, early voting has not managed to deliver a significant increase in voter turnout. Instead, early voting has been found to decrease enthusiasm in voters who may occasionally cast ballots, but do not vote regularly. The literature is less clear on the cost impacts associated with early voting. Additionally, the costs of early voting may be more burdensome on the small, rural jurisdictions in a state, a factor state regulations may not take into account. Overall, the literature suggests that if a states goal is to increase turnout, the negligible effects of early voting on turnout may not make it worth the associated costs.
Same day registration and election-day registration. SDR and EDR have also become popular in states that wish to increase the convenience of casting a ballot. SDR is simply the ability to both register and cast a ballot on the same day at the same precinct place or vote center. Similarly, EDR is the ability to do both of those things on election day. The main advantage of conducting elections with SDR and/or EDR in place, based on the available literature, is the reduction in administrative burden this combination of voting convenience provisions is able to


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
15
provide. The ability to register voters up until (and even on) election day has a significant impact on the associated administrative burden. According Burden et al. (2009), particularly when coupled with early voting, SDR can significantly decrease the administrative burden of voter registration by spreading it out over time. Regardless of the monetary cost of SDR, the literature submits it may be worth the expenditure to decrease administrative costs from the perspective of election officials tasked with the work.
Early voting, SDR and EDR based voter convenience models are being implemented nationwide in an effort to increase voter turnout and decrease election costs. Studies on the cost benefits of voter convivence models have yielded results that are diverse and tend to apply differently to different types of jurisdictions. Early voting alone has been largely proven to not increase turnout, although further examination of its cost incidence needs to occur before any conclusions can be made. The literature on SDR and EDR needs to be improved as well before inferences on how these measures impact the cost of elections can be made.
Elections Technology
Voting systems are the machines used to cast and tabulate votes in an election. According to NCSL, at least 25 states have considered or adopted new voting systems in the last three years (The National Conference of State Legislatures, 2017). The systems states bought in the early 2000s with HAVA funds are now reaching their lifespan limits and the replacement of these systems will be a significant cost for states or local governments in the next few years.
In general, uncovering the costs associated with voting technology is more straightforward than uncovering other areas of election costs like poll workers. The cost of elections technology are known by vendors and the purchasers (state procurement officials or individual jurisdictions) who have detailed knowledge of how much an entire system and its


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
16
component parts will cost. Part of this can be attributed to the fact that the systems are large, capital expenditures that governments are not inclined to make without highly detailed information and well-documented purchases.
The difficulty in dealing with voting systems costs often stems from disagreement over which level of government will be facing the bulk of the cost burden for the system purchase. States have differing payment schemes for voting systems, distributing the cost burdens between the state and local jurisdictions. Some states pay for all voting equipment that is needed by local jurisdictions and a few states cover a portion of the cost of equipment, such as Maryland who pays for 50% of the purchase cost of voting equipment and Mississippi that dedicates 50% of its Elections Support Fund to local jurisdictions to assist with the purchase of new equipment (The National Conference of State Legislatures, 2017). Across all 50 states, the payment schemes create various cost burdens on local governments and state governments and are essentially a patchwork of state and local laws.
The costs associated with voting systems purchases and maintenance are some of the clearest components of overall election costs. In contrast with poll worker costs, the costs of a new voting system are the most straightforward costs associated with elections.
Methodology
Research Question and Hypotheses
As evidenced by the literature review, there are a handful of areas in election administration where corresponding cost data has been studied. Poll worker costs are not one of these areas. To help NCSL better understand the available data on poll worker costs, this study examines the following research question:


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
17
What are the costs associated with poll workers in election administration and what are potential cost drivers?
The research question is addressed by first, gathering relevant data on the costs associated poll workers. Then, answering the research question is intended to help practitioners understand potential cost drivers associated with poll workers, its association with the overall costs of election administration and to highlight areas for future study.
The following hypotheses were used to help frame poll worker costs and associated cost drivers in a way that is conducive to future analysis:
Hypothesis 1: States that use the vote-center polling scheme will have fewer poll workers than states that use traditional precinct polling places. This hypothesis is essential to the study of poll worker costs because it can be assumed that the more poll workers there are, the more the total cost of poll workers will be for a jurisdiction.
Null: No difference exists in the number of poll workers hired between states that use vote centers and those that use precinct polling places.
Hypothesis 2: In states that allow early voting, poll worker wages are lower than states with election day voting only.
Null: There is no difference in poll worker wages between states that allow early voting and states that do not.
Measurement and Data Collection
To explore the potential cost drivers associated with poll worker wages, this analysis employs a quantitative research approach to test the hypotheses as detailed above. Specifically, multiple regression was conducted on each dependent variable. The analysis was done entirely with secondary data gathered from a variety of sources (see Appendix A for the measurement


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
18
table). The attributes and source information for all variables included in each model can be found in Appendix B.
Sampling Plan
Poll worker wage (pwwage) serves as a dependent variable in this analysis. The data for poll worker wage is current for 2016 and were obtained from the State Election Directors websites from 44 states. Follow-up calls were placed to State Election Directors offices for those states that did not have the information available online. Data were unavailable for the states of Maine, Mississippi, New Hampshire, Oregon, Washington and Wyoming. The sample size for poll worker wage (pwwage) is n = 39 (see Table 4a in Appendix C).
The data for number of poll workers (pollworkers) and number of polling places (pollingplaces) were obtained from the 2012 U.S. Election Assistance Commissions Election and Voting Survey. The survey, conducted after every general and midterm election, is distributed to all voting jurisdictions in the country. Ideally, this analysis would examine the 2016 version of the survey data, however it has yet to be published so data available from the most recent general election (2012) are used. This analysis uses data from Section D of the 2012 survey (see Appendix E for Section D of the 2012 EAVS). Pollworkers is the second dependent variable in this analysis, while pollingplaces is an independent variable. Alabama, Georgia, New Jersey, Pennsylvania and Utah observations were dropped due to lack of data for the total number of poll workers (pollworkers), polling places (pollingplaces) or both. As a result, the sample size for poll workers (pollworkers) and polling places (pollingplaces) is n = 45 (see Table 1 in Appendix C).


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
19
The data for the use of technology (tech) variable, which serves as an independent variable, were obtained from Verified Voting, a non-partisan, nonprofit organization that studies the reliability and security of voting systems (New Verifier, 2016). Their online tool, The Verifier, presents data on what type of polling place equipment states used in a given election. For this analysis, an accompanying spreadsheet to the online tool was obtained from the website for data from November 2016. The sample size for use of technology (tech) is n = 45 (see Table 3 in Appendix C).
To help control for cost-of-living differences that may account for some of the variation in poll worker wages a cost-of-living index is included as an independent variable (costofliving). The data for this variable comes from the Missouri Economic Research and Information Center, an office in Missouris Department of Economic Development. The data are current for the first quarter of 2017 and presents cost of living data aggregated at the state level on a scale of 85.0 to 187.7 (Missouri Economic Research and Information Center, 2017). The sample size for cost of living (costofliving) is n = 44 (see Table 5a in Appendix C).
Finally, the data for the remaining independent variables were obtained from a variety of NCSL publications. This applies to the following variables: Early Voting (ev), Absentee Voting (absentee), Same day Registration (SDR), Election-Day Registration (EDR), Vote by Mail (VBM), and Polling Scheme (scheme). See Appendix A for detailed data source information and assumptions. The sample size for all of the above variables is n = 45 (see Table 3 in Appendix C).
While there are a few exceptions for each variable, there are data points representing a unique value for each state whenever possible. These samples aggregated at the state-level reflects the desires of the client to collect and analyze data with a nationwide approach, which


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
20
will provide the most relevant information for the population they serve: all 50 state legislatures and U.S. territories. Additionally, the data being examined are from the last general election for which it is available. Therefore, all data reflects the 2016 general election apart from the poll worker wage data which reflects the 2012 general election and the cost of living index which reflects economic data for the first quarter of 2017.
Two units of analyses are used to measure poll worker costs: poll worker election-day wage in dollars and number of poll workers. Due to limitations both in size and scope of this analysis and the availability of data, this analysis aggregates jurisdiction-level data into state samples for two reasons. First, with over 10,000 jurisdictions responsible for election administration in the United States and no resource that currently exists where these data are aggregated, it is beyond the size and scope of this study to examine each jurisdiction. While counties typically are responsible for election administration, other forms of government including cities and townships can be responsible for general election administration instead, depending on the state (The National Conference of State Legislatures, 2016). Additionally, many election laws are codified and enforced at the state level and work as a key part of this analysis.
Validity and Reliability
The nature of elections and voting in the United States federalist system makes it particularly difficult to extrapolate these results, yet it may be possible to generalize the findings of this analysis to future general elections in the U.S., with several qualifiers. According to the State of Colorados Elections Director Judd Choate, the sheer level of decentralization in election administration has produced at least 50 very distinct models for administering elections, with local jurisdictions adding even more variation (J. Choate, phone interview, May 1, 2017). In


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
21
fact, a common language in the field does not exist because most jurisdictions in each state have developed their own individual models with unique definitions and terminology.
The discrepancy in definitions presents a problem with the accuracy of the proposed measures. To minimize this, an effort has been made to ensure the variables included in this analysis are those that have the most wide-reaching, universal definitions possible, regardless of jurisdiction. For example, the Early Voting (ev) variable simplifies the variance in early voting policies across states by only measuring whether or not a state has any early voting measures at all. To improve the internal validity of the model, the variable for poll worker wage (pwwage) was standardized to reflect the total amount poll workers are paid for working on election day, rather than an hourly wage. States varied widely in what they paid poll workers for; some paid extra for training, some gave a meal stipend, some paid extra for early voting, others were all volunteer. This analysis assumes working on election day would be equivalent to working a 15-hour day. Based on this, the poll worker wage data was standardized to reflect total pay for working on election day only. For poll worker wage data and source information, see Appendix D.
The replicability of this analysis fairs well due to the reliance on secondary data. Additionally, the secondary data originates from sources that have reliably and consistently gathered it for years and will continue to do so in the future. The Election Assistance Commissions EAVS has been conducted for every general and midterm election since 2004. Similarly, Verified Voting has produced its Verifier tool every two years since 2012 and has been studying the reliability and security of voting systems since 2003. Thus, despite countless differences between jurisdictions in election administration, there are some resources for


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
22
consistent, reliable data that can be used to study election costs in the future by using data for elections in 2018, 2020 and beyond.
Data Analysis
To test the two hypotheses outlined previously, two multiple regressions were performed on each of the dependent variables (pwwage and pollworkers) to determine what effect, if any, the independent variables were having on each. Before performing the regression, descriptive statistics were run on each variable to gain a better understanding of its characteristics, including the mean, median, mode and range. Then, one regression was performed on the dependent variable pwwage, using all the defined independent variables, keeping all else constant. The process was then repeated on the second dependent variable pollworkers using all the defined independent variables, keeping all else constant. The regressions were performed using the statistical software SPSS. See Appendix B for details on each regression model.
If the first regression yields a high significance level, a beta value close to 1, or an overall r-squared value close to 1.0, these are indicators that a portion of the variance in number of poll workers can be explained by the independent variables in the model and provides supporting evidence for Hypothesis 1. Similarly, if these attributes apply to the results from the regression testing for Hypothesis 2, it may be determined that the variance in poll worker wage is being explained by the independent variables in the second model, which provides supporting evidence for the hypothesis.
Results
Descriptive Statistics
Through examining the descriptive statistics of the variables, it was clear that the analysis would benefit from both dependent variable pollworkers and independent variable pollingplaces


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
23
transforming into more interpretable values. Both variables featured exceedingly large means, ranges and standard deviations values that make sense, given that the data are aggregated at the state level (see Table 1 in Appendix C). The difficulty in interpreting the variables left in this form became even more apparent when including them in the regression model, thus to improve the readability of the results, both were transformed into the following log variables:
Log (pollworkers)
Log (pollingplaces)
Throughout this analysis, the use of any variables in log form refers to log_base-e. See Table 2 in Appendix C for descriptive statistics on transformed variables.
The descriptive statistics show relatively consistent data for the dichotomous independent variables (see Table 3 in Appendix C). Poll worker wages, as adjusted to reflect a days worth of wages on election day, have a normal distribution skewed slightly to the right, with most values falling between $100 and $200, with a median of $127.50 (see Tables 4a, 4b and Figure 1 in Appendix C). Costofliving appears to reflect a relatively normal distribution, although it is skewed further to the right with a median of 97.55 and one extreme outlier at 187.7 (see Tables 5 a, 5b and Figure 2 in Appendix C).
Model 1: Number of Poll Workers
Of the two models tested, the regression examining the number of poll workers, log (pollworkers), as the dependent variable produced more significant results. With an associated p-value of p < 0.001, the model indicates that the impact the number of polling places, log (pollingplaces) has on the number of poll workers is significant (see Table 6d, Appendix C).
The coefficient for log (pollingplaces) is 0.926, which must first be transformed to be interpreted (See Table 6d in Appendix C). According to Benoits (2011) detailed guide on


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
24
logarithmic transformations in regression models, the coefficient for log (pollingplaces) can be interpreted as follows: a 1% increase in log (pollingplaces) results in an eA[(log(1.01))*0.926] = 1.00926 increase in the estimated mean of poll workers. Additionally, the beta coefficient for log (pollingplaces) is 0.980, an indication that the variable is a strong predictor in the model as any beta coefficient greater than 0.50 is said to be a strong measure (Swann, 2017).
The next most significant predictor in the model was the presence of early voting (ev). The p-value for this dichotomous variable is p < 0.10, making it significant at the 90% confidence level (see Table 6d, Appendix C). The coefficient for ev is -0.203, which according to Benoit (2011) can be interpreted in the following way: Each 1 unit increase in ev multiples the expected value of log (pollworkers) by eA(-0.203). Since ev is a binary variable, this can be interpreted to mean that as a state switches from having no early voting (ev = 0) to having early voting provisions (ev = 1) the number of poll workers the state will hire is multiplied by eA(-0.203).
The r-squared of this model is 0.880, indicating that about 88% of the variance in the dependent variable log (pollworkers) is explained by the model (see Table 5b, Appendix C). The adjusted r-squared, which adjusts for the number of predictors in the model, is 0.849 further indicating that the model is explaining the variance in the dependent variable. Since the log (pollingplaces) predictor appeared to be having a significant impact and the model had a high r-squared value, partial and semi-partial correlations were run to see what effect each individual predictor was having on the overall r-squared. The resulting table shows log (pollingplaces) having a much greater impact on the models r-squared value than any other predictors (see table 6e, Appendix C).


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
25
Overall, the model fails to reject the null hypothesis that there is no difference in the number of poll workers states hire and the type of voting scheme they employ. While the model presents interesting results on the relationships of number of polling places and early voting with number of poll workers, it fails to yield any significant results on polling scheme.
Model 2: Poll Worker Wage
The second regression model, which examined poll worker wage (pwwage) as the dependent variable failed to produce statistically significant results. The most significant predictor in this model was absentee voting (absentee) with a p value of p < 0.112 (see Table 7d, Appendix C). The regression coefficient associated with absentee is -61.006, which given the binary nature of the variable can be interpreted simply as if a state has absentee voting (absentee = 1) then there is an associated $61 decrease in pwwage on election day as compared to a state that does not have absentee voting (absentee = 0).
No other predictors in the model are statistically significant. The r-squared for this model is .138, indicating that only 13.8% of the variance in poll worker wage can be explained by the included predictors. The adjusted r-squared is even lower at -0.140 (see Table 7b, Appendix C). Due to inconclusively, no substantive inferences can be made regarding hypothesis two. Therefore, the null hypothesis that there is no difference in poll worker wage between states that do and do not have early voting, fails to be rejected.
Based on the findings from both models there are several implications for the research question: what are the costs associated with poll workers in election administration and what are potential cost drivers? The results address the first part of the research question by indicating that one of the main costs associated with poll workers is simply how many of them are needed to staff the given number of polling places in a jurisdiction. Subsequently, the results from both


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
26
models suggest that the potential cost drivers for poll workers also includes the total number of polling places that need to be staffed in a jurisdiction as well as voter convenience measures that are heavily administrative and potentially create more work for poll workers (i.e. early voting and absentee voting).
Discussion & Recommendations
This analysis aimed to uncover some potential cost drivers associated with poll workers in election administration and to provide a starting point for future research in this area. The associated hypotheses serve as a framework from which to begin interpreting poll worker costs. The findings of this research have several implications for future study of this topic. Implications, Limitations: Model 1
Model 1 tested the impact of a variety of election characteristics on the number of poll workers hired. The rationale behind this model is simple: the more poll workers a jurisdiction must hire, the greater proportion of the elections budget must be devoted to poll worker pay. Therefore, it is valuable to uncover any potentially significant drivers that increase the number of poll workers required in a jurisdiction.
While this model did not provide support for the hypothesis that the voting scheme used by a state impacts the number of poll workers hired, it did illuminate some important relationships between the dependent variable, number of poll workers and other predictors present. First, results from Model 1 show a strong relationship between the number of polling places and number of poll workers in a jurisdiction. This result is intuitive and suggests that if jurisdictions are seeking strategies to reduce spending in election administration, an effective avenue may be to decrease the number of polling places and thereby decrease the number of poll


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
27
workers. These results add to the literature on the topic which includes studies that both support and discredit vote centers as cost saving measures (Folz, 2014; Montjoy, 2010).
The second significant result from Model 1 demonstrates a relationship between early voting provisions and the number of poll workers a jurisdiction hires. The results reveal that the number of poll workers hired decreases when a jurisdiction moves from not having any early voting to adopting early voting provisions. This is a notable result, as the literature that does currently exist on the costs associated with early voting is largely inconclusive. Some scholars have argued that the implementation of early voting is inefficient and forces jurisdictions to bear high administrative costs, particularly in smaller, more rural area, while others have argued for its cost saving potential that spreads administrative burden out over a period of time beyond the single day of election day (Burden et al., 2008). Further research is needed to determine if any cost savings benefits stem from the implementation of early voting laws.
Model 1 is limited in that it fails to consider additional forces that may be impacting the number of poll workers hired. Additional predictors such as the geographic area of a jurisdiction and the number of ballots cast in the last general election would help enrich the model. Thus, to improve the model and gain a more comprehensive understanding of the factors that drive the number of poll workers hired, more independent variables should be considered and the model should be replicated.
Implications, Limitations: Model 2
Model 2 tested for relationships between the wages poll workers are paid and a number of independent variables associated with different types of voting models in the US. The results from the model suggest that only absentee voting is having an impact on the dependent variable. The most statistically significant relationship produced by the model, the results suggest that poll


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
28
workers are paid less in jurisdictions that have in-person absentee voting than those that do not.
A possible explanation for this result could be that jurisdictions with absentee voting can pay their poll workers a lower election-day wage because it may be easier to count ballots ahead of time as they are received rather than all at once on election day.
The predictors in this model are not explaining the variance that exists in poll worker wages. One limitation of the model that could explain this is the exemption of more predictors that are tied to local economic conditions. The results suggest that other factors are contributing more to the variance in poll worker pay across jurisdictions. Some additional predictors that may better explain poll worker pay include local economic conditions such as local minimum wage or local average wage and overall government budget characteristics including total budget amount and proportion of budget allocated to election administration.
Based on the results from this model, practitioners interested in the further study of poll worker costs should expand future research to include other, relevant predictors such as those related to local economic conditions. Additionally, future research may want to discard some of the predictors from this model if they are found to be extraneous.
Overall Recommendation and Limitations
Beyond each individual model, several limitations exist in the data. First, some of the data such as the EAVS survey data which inform the pollworkers and pollingplaces variables are self-reported by local election officials, which could present some errors. Issues with accuracy could arise given the diversity that likely exists between jurisdictions recording methods. While this study attempts to control for as much variation in definitions as possible, some discrepancies may exist that could impact the validity of the data. Additionally, as previously stated the EAVS survey data included in this model was from 2012 as it was the most recent available, while all


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
29
other data are current for the 2016 general election. Another limitation to the overall study is the small sample size; ideally, this study should be replicated in the future with more jurisdictions included. Finally, due to the differences in election administration laws and practices from state to state, certain simplifications of variables were necessary to fit the size and scope of this study. Different conclusions may be reached in studies that do not make these simplifying assumptions (see Appendix A).
This study recommends further research on poll worker cost data to develop an understanding of which costs associated with poll workers are most significant. Specifically, results from this study imply that the number of polling places and types of voter convenience provisions a jurisdiction is using help to determine the number of poll workers a jurisdiction needs during an election. Results also suggest that the type of voter convenience measures a jurisdiction has may impact poll worker wage, however, other factors such as local economic conditions are likely more accurate predictors of poll worker wage and should be included in future research. In conclusion, this study recommends that NCSL continue exploration in this field by studying a broader range of local jurisdictions beyond state-level data and to include more and different predictors in future analysis to better understand the cost drivers associated with poll worker costs.
Conclusion
The cost and cost-drivers associated with poll workers have not been thoroughly studied in the field of election administration. This analysis, through the examination of secondary data, intends to help practitioners make sense of the existing literature and identify a starting point for future research. Results from this analysis demonstrate that factors like voter convenience measures and the number of polling places present in a jurisdiction do have an impact on the


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
30
number of poll workers hired. Results also demonstrate that these factors do little to explain poll worker pay and suggest that future studies should examine other factors, such as local economic conditions, to assess their impact. In conclusion, while the results from this study present a preliminary analysis of factors related to poll worker cost, it provides the building blocks for future study which needs to occur to help practitioners gain a better understanding of the costs associated with poll workers.


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
31
References
Benoit, K (2011). Linear regression models with logarithmic transformations. London School of Economics. Retrieved from http://www.kenbenoit.net/courses/ME104/logmodels2.pdf
Bowler, S., Brunell, T., Donovan, T., Gronke, P. (2014). Election administration and perceptions of fair elections. Electoral Studies, 38(2015), 1-9.
Burden, B., Canon, D., Mayer, K., and Moynihan, D. (2009, October). The effects and costs of early voting, election day registration, and same day registration in the 2008 elections. Paper presented at The Ohio State University Mershon Center for International Security Studies, Columbus, Ohio. Retrieved from https://apw.polisci.wisc.edu/archives/Burden et al.pdf.
Burden, B., Canon, D., Mayer, K., and Moynihan, D. (2012). The effect of administrative burden on bureaucratic perception of policies: evidence from election administration. Public Administration Review, September/October 2012, 741-751.
Chapin, D. (2011). Non-precinct place voting and election administration. Election Law Journal, 10(3), 303-305.
Claassen, R., Magleby, D., Monson, J., and Patterson, K. (2008). At your service, voter
evaluations of poll workers performance. American Politics Research, 36(4), 612-634.
Clark, A. (2014). The cost of elections: money well spent? The Journal of Political Insight, December 2014, 16-19.
Choate, J. (2017, May 1). Phone interview with A. Buchanan.
Creek, H., Karnes, K. (2010). Federalism and election law: implementation issues in rural America. Publius, 40(2), 275-295.
Doyle, S. (2008). Practitioner commentary: vote centers, voting equipment, and uncertainty in election administration. Public Administration Review Symposium on Election Administration, September/October 2008.
Epstein, J. (2007). Electronic Voting. Computer, 40(8), 92-95. Retrieved from: http://0-ieeexplore.ieee.org.skyline.ucdenver.edu/document/4292024/
Folz, D. (2014). Vote centers as a strategy to control election administration costs: findings from a pilot project. SAGE Open, January-March 2014, 1-10.
Giammo, J., and Brox, B. (2010). Reducing the costs of participation, are states getting a return


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
32
on early voting? Political Research Quarterly, 63(2), 295-303.
Haspel, M., and Gibbs Knotts, H. (2005). Location, location, location: precinct placements and the costs of voting. The Journal of Politics, 67(2), 560-573.
Hill, S. (2012). Election Administration Finance in California Counties. The American Review of Public Administration, 42(5), 606-628.
Missouri Economic Research and Information Center, Missouri Department of Economic
Development (2017). Cost of Living Data Series First Quarter 2017. Retrieved from https://www.missourieconomy.org/indicators/cost of living/
Montjoy, R. (2010). The changing nature... and costs of election administration. Public Administration Review, November/December 2010, 867-875.
New Verifier. (2016, May 06). Retrieved June 06, 2017, from https://www.verifiedvoting.org/verifier/
Swann, W. (2017). Module 9: Multiple Regression. Personal Collection of W. Swann, University of Colorado Denver, Denver CO.
The National Conference of State Legislatures (2017). Election costs: what states pay. Retrieved from the National Conference of State Legislatures website:
http://www.ncsl.Org/research/elections-and-campaigns/election-costs.aspx#voting equip
The National Conference of State Legislature (2016). Election Administration at State and Local Levels. Retrieved from the National Conference of State Legislatures website: http://www.ncsl.org/research/elections-and-campaigns/election-administration-at-state-and-1 ocal -1 evel s. aspx


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
33
Appendix A
Comprehensive Measurement Table
Variable Variable Type Relevant Values/Types of Data Assumptions Data Source Notes
Wage Data
Poll Worker Wage DV Continuous, Varies Only examining election day wages; hourly wages adjusted for a 15 hour work day; bonus for training included if exists See Appendix D; Election Director's Websites
EAVS Survey Data
Number of Poll Workers DV Continuous, Varies Total across all of a state's jurisdictions for which data exists EAVS Alabama: No Data Georgia: No Data Assumption: All staff paid, working on election day Some jurisdictions partnered with one another, so reported 0 poll workers like Edmunds ME Pennsylvania: No Data


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
34
Total Number of Polling Places IV Continuous, Varies On election day EAVS Some discrepancy between whether or not jurisdictions were counting early voting into the total Ml totals include physical polling places and central count locations MO Gentry County, includes absentee polling places MO- Pulaski County, includes absentee polling places Wherever possible Included absentee, provisional voting centers-tried to incorporate every place a ballot is cast on voting day
Voter Convenience Group
Early Voting IV Binomial (0 = No Early Voting, 1 = Presence of Early Voting) NCSL Data: http://www.ncsl.o rg/research/electi ons-and- campaigns/absent ee-and-earlv- voting.aspx#a; http://www.ncsl.o rg/research/electi ons-and- campaigns/early-


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
35
voting-in-state-elections.aspx
Absentee Voting IV Binomial (0 = No In-Person Absentee Voting, 1 = Presence of In-Person Absentee Voting) NCSL Data: httD://www.ncsl. org/research/elec tions-and- cam oai gns/absen tee-and-earlv- voting.asDx?ct=d fe09000a89a767 b2718bae98f75f 4d2e93fdc7461c e9a2fl72900a56 a9f6859407480d 5d2fl8181ea527 09edcfcaa8be78 3166cbffa993d0 271a8cld589814 8#a
Same- Day Registrati on IV Binomial (0 = No Same-Day Registration, 1 = Presence of Same-Day Registration) If Same-Day Registration exists during a state's voting period (early voting and/or election day) NCSL Data: httD://www.ncsl. org/research/elec tions-and- camnaigns/same- dav- registration.aspx
Election- Day Registrati on IV Binomial (0 = No Election-Day Registration, 1 = Presence of Election-Day Registration Binomial NCSL: httn://www.bren nancenter.org/an alvsis/voter- registration- modernization- states
Vote by Mail IV Binomial ( 0 = no vote by Measures whether or not a state has some VBM provisions NCSL: httn://www.ncsl. org/research/elec


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
36
mail, 1 = some vote by mail) tions-and- campaigns/all- mail- elections.aspx#st ate
Polling Scheme Vote centers vs. Precinct Polling Places IV Binomial (0 = Uses Traditional Precinct Polling Places, 1 = Uses Vote Centers) NCSL: httD://www.ncsl. org/research/elec tions-and-camnaigns/vote-centers, asnx
Technology
Use of Technolo gy IV Binomial ( 0 = low technology use, 1 = high technology use) Values assigned to each type of election technology used from 1 being the low end of technology used in elections and 6 being the most state-of-the-art system used in election administration. Verified Voting -The Verifier: https://www.venf iedvoting.org/veri fier/#
Then, median found (3.09) and used to break the sample into low technology use (0) and high technology use (1)
Cost of Living Index
Cost of Living IV Continuous https://www.miss ourieconomv.org/i ndicators/cost of living/ Values range from 85.0 to 187.7


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
37
Appendix B
Hypothesis 1
Hypothesis 1: States that use a vote center polling scheme that use traditional precinct pol will have fewer poll workers than states ing places.
Variables Variable Name Variable Type Measure Level of Measurement Data Source
Number of poll workers hired pollworke rs DV Number of poll workers hired in a state for the 2016 general election Continuous EAVS 2012

Early Voting ev IV 0 = No Early Voting 1 = Presence of Early voting Binomial NCSL Data
Absentee Voting absentee IV 0 = No Absentee Voting 1 = Presence of Absentee Voting Binomial NCSL Data
Same-Day Regi strati on SDR IV 0 = No Same-Day Registration 1 = Presence of Same-Day Registration Binomial NCSL Data
Election- Day Regi strati on EDR IV 0 = No Election-Day Registration 1 = Presence of Election-Day Registration Binomial NCSL Data
Vote by Mail VBM IV 0 = No Vote by Mail 1 = Some Vote by Mail Provisions Binomial NCSL Data
Else of Technolog y tech IV 1 = Hand Counted Paper Ballots, 2 = Ballot Marking Device or System, 3 = Optical Scan, 4 = DRE-Push Button, 5 = DRE Dial, 6 = DRE Touchscreen Median was found to be 3.09 states were then reclassified to fit Binomial Verified Voting -The Verifier


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
38
into a 0 category (low technology) and a 1 category (high technology) based on their value relative to the median.
Polling Scheme scheme IV 0 = Uses Traditional Precinct Polling Places 1 = Uses Vote Centers Binomial NCSL Data
Number of polling places pollingpla ces IV Varies Continuous EAVS 2012
Cost of Living Index costoflivin g IV Varies Continuous https://www.miss ouri economy. org/ indicators/costof living/
(Table 1)
Hypothesis 2
Hypothesis 2: In states that allow early voting, poll worker wages are lower than states with election day voting only.
Variables Variable Name Variable Type Measure Level of Measureme nt Data Source
Poll Worker Wage Pwwage DV Dollars Continuous Election Directors Websites/Of fices

Early Voting ev IV 0 = No Early Voting 1 = Presence of Early voting Binomial NCSL Data
Absentee Voting absentee IV 0 = No In-Person Absentee Voting 1 = Presence of In-Person Absentee Voting Binomial NCSL Data
Same-Day Registration SDR IV 0 = No Same-Day Registration 1 = Presence of Same-Day Registration Binomial NCSL Data


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
39
Election-Day Registration EDR IV 0 = No Election-Day Registration 1 = Presence of Election-Day Registration Binomial NCSL Data
Vote by Mail VBM IV 0 = No Vote by Mail 1 = Some Vote by Mail Provisions 2 = All Vote by Mail States (CO, OR and WA) Ratio NCSL Data
Use of Technology tech IV 1 = Hand Counted Paper Ballots, 2 = Ballot Marking Device or System, 3 = Optical Scan, 4 = DRE-Push Button, 5 = DRE Dial, 6 = DRE Touchscreen Median was found to be 3.09 states were then reclassified to fit into a 0 category (low technology) and a 1 category (high technology) based on their value in relation to the median. Ratio, assigned values Verified Voting -The Verifier
Polling Scheme scheme IV 0 = Uses Traditional Precinct Polling Places 1 = Uses Vote Centers Binomial NCSL Data
Number of polling places pollingpl aces IV Varies Continuous EAVS 2012
Cost of Living Index COL IV Continuous https ://www .missourieco nomy.org/in dicators/cost of living/
(Table 2)


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
40
Appendix C
Descriptive Statistics
N Minimum Maximum Mean Std. Deviation
pollworkers 45 782 89440 17583.58 20130.789
pollingplaces 45 36 14686 2154.22 2545.257
(Table 1 Pre-log transformed variables)
Descriptive Statistics
Variable N Minimum Maximum Mean Std. Deviation
log(pollworkers) 45 2.893206753 000000 4.951531791 000001 3.990131830 000000 .4982961910 00000
log(pollingplaces) 45 1.556302501 000000 4.166903524 000000 3.079159785 000000 .5256225430 00000
(Table 2 log transformed variables)
Descriptive Statistics
Variable N Minimum Maximum Mean Std. Deviation
ev 45 0 1 .40 .495
absentee 45 0 1 .29 .458
SDR 45 0 1 .33 .477
EDR 45 0 1 .31 .468
VBM 45 0 1 .42 .499
tech 45 0 1 .49 .506
scheme 45 0 1 .49 .506


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
41
(Table 3 dichotomous independent variables)
Descriptive Statistics
Variable N Minimum Maximum Mean Std. Deviation
pwwage 39 .00 380.00 142.4469 65.87242
(Table 4a Poll Worker Wages)
Statistics
Variable: pwwage
N Valid 39
Missing 6
Mean 142.4469
Median 127.5000
Mode 120.00a
Percentiles 25 108.7500
50 127.5000
75 175.0000
(Table 4B Poll Worker Wages)


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
42
Histogram
(Figure 1 Poll Worker Wage Distribution)
Mean = 142.45 Std. Dev. = 65.672 N = 39
Descriptive Statistics
Variable N Minimum Maximum Mean Std. Deviation
costofliving 44 85.0 187.7 104.391 19.0733
(Table 5a Cost of Living Index)
Statistics
Variable: costofliving
N Valid 44
Missing 1
Mean 104.391
Median 97.550
Mode 101.6
Percentiles 25 92.400
50 97.550
75 112.775
(Table 5b Cost of Living Index)


Frequency
AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
43
Histogram
Mean = 104.39 Std. Dev. = 19.073 N = 44
(Figure 2 Cost of Living Index Distribution)


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
44
Model Results
Model 1
Variables Entered/Removed3
Model Variables Entered Variables Removed Method
1 costofliving, tech, VBM, EDR, log(pollingpl aces), scheme, absentee, ev, SDRb Enter
a. Dependent Variable: log(pollworkers)
b. All requested variables entered.
(Table 6a Model 1)
Model Summary
Model R R Square Adjusted R Square Std. Error of the Estimate
1 .938 .880 .849 .1952596810 00000
a. Predictors: (Constant), costofliving, tech, VBM, EDR, log(pollingplaces), scheme, absentee, ev, SDR
(Table 6b Model 1)
ANOVAa
Model Sum of df Mean F Sig.
Squares Square
1 Regression 9.544 9 1.060 27.815 ,000b
Residual 1.296 34 .038
Total 10.841 43
a. Dependent Variable: log(pollworkers)
b. Predictors: (Constant), costofliving, tech, VBM, EDR, log(pollingplaces), scheme, absentee,
ev, SDR
(Table 6c Model 1)


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
45
Coefficients3
Model Unstandardized Coefficients Standardized Coefficients t Sig.
B Std. Error Beta
1 (Constant) 1.076 .351 3.068 .004
ev -.203 .105 -.201 -1.933 .062
absentee -.097 .094 -.087 -1.034 .309
SDR .327 .224 .312 1.461 .153
EDR -.341 .221 -.320 -1.542 .132
VBM .072 .071 .072 1.006 .322
tech -.086 .072 -.087 -1.192 .241
scheme -.009 .089 -.009 -.097 .923
loq(pollinqplaces) .926 .067 .980 13.749 .000
costoflivinq .002 .002 .065 .812 .422
a. Dependent Variable: log(pollworkers)
(Table 6d Model 1)
Partial Semipartial Partial Semipartial Significance
Variable Corr. Corr. Corr.^2 Corr.^2 Value
ev -0.2985 -0.1077 0.0891 0.0116 0.0770
absentee -0.1444 -0.0503 0.0209 0.0025 0.4006
SDR 0.2507 0.0892 0.0628 0.0080 0.1403
EDR -0.2670 -0.0955 0.0713 0.0091 0.1154
VBM 0.1908 0.0669 0.0364 0.0045 0.2650
tech -0.1839 -0.0645 0.0338 0.0042 0.2829
scheme -0.0479 -0.0165 0.0023 0.0003 0.7814
logpollin~s 0.9210 0.8141 0.8482 0.6628 0.0000
costofliv~g 0.1243 0.0432 0.0155 0.0019 0.4701
(Table 6e Model 1)


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
46
Model 2
Variables Entered
Mode 1 Variables Entered Variables Removed Method
1 costofliving, tech, VBM, EDR, log(pollingpla ces), scheme, absentee, ev, SDRb Enter
a. Dependent Variable: pvwvaqe
b. All requested variables entered.
(Table 7a Model 2)
Model Summary
Model R R Square Adjusted R Square Std. Error of the Estimate
1 .371 .138 -.140 71.22665
a. Predictors: (Constant), costofliving, tech, VBM, EDR, log(pollingplaces), scheme, absentee, ev, SDR
(Table 7b - Model 2)
AN OVA3
Model Sum of df Mean F Sig.
Squares Square
1 Regression 22676.380 9 2519.598 .497 ,864b
Residual 142050.588 28 5073.235
Total 164726.968 37
a. Dependent Variable: pwwage
b. Predictors: (Constant), costofliving, tech, VBM, EDR, log(pollingplaces), scheme, absentee,
ev, SDR
(Table 7c Model 2)
Coefficients3
Model Unstandardized Coefficients Standardized Coefficients t Sig.
B Std. Error Beta
1 (Constant) 125.259 151.925 .824 .417
ev -34.737 41.171 -.263 -.844 .406
absentee -61.006 37.222 -.408 -1.639 .112


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
SDR 29.829 83.166 .211 .359 .723
EDR -1.341 82.664 -.009 -.016 .987
VBM -21.058 28.688 -.159 -.734 .469
tech -1.481 29.362 -.011 -.050 .960
scheme 23.382 35.009 .177 .668 .510
loq(pollinqplaces) 8.069 31.069 .052 .260 .797
costoflivinq .119 .807 .035 .148 .884
a. Dependent Variable: pvwvaqe
(Table 7d Model 2)


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
48
Appendix D
Poll Worker Wage Data, Assumptions and Notes
State State Registered Voters 2016 Poll Worker Wage Notes Poll Worker Wage (assumptions: adjusted for a 15 hour day, includes training bonus)
Alabama 2,526,000 Paid ud to $50/dav $50
Alaska 358,000 $12 an hour for 15 davs before and elections day $180
Arizona 3,145,000 $125 $125
Arkansas 1,456,000 $95 $95
California 16,096,000 $100 $100
Colorado 2,893,000 Varies bv county, $13 and hour for early voting, training and election day in Adams County $195
Connecticut 1,763,000 $380 on election day only $380
Delaware 487,000 $190 $190
Florida 9,604,000 $185 $185
Georgia 4,892,000 $175 $175
Hawaii 530,000 $90 $90
Idaho 790,000 $125 $125
Illinois 6,665,000 $190 for those who work election day and complete training $190
Indiana 3,298,000 $145.00 $145
Iowa 1,657,000 $9.50/hour for election day $143
Kansas 1,438,000 $110- 120/election day $115


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
49
Kentucky 2,253,000 $70 with training, $60 without $70
Louisiana 2,446,000 $200 $200
Maine 830,000
Maryland 3,114,000 $162.50 $163
Massachusetts 3,660,000 $155 $155
Michigan 5,434,000 $150 $150
Minnesota 3,055,000 $7.25/hour $109
Mississippi 1,725,000
Missouri 3,333,000 $82 $82
Montana 581,000 $8.50/hour $128
Nebraska 1,008,000 $7.25/hour + $18 for training $127
Nevada 1,371,000 $10/hour for 14 davs of earlv voting + $120 for election dav $120
New Hampshire 763,000
New Jersey 4,165,000 $200 $200
New Mexico 916,000 $7.25/hour $109
New York 9,142,000 $200 on election dav + $100 for attending training $300
North Carolina 5,194,000 $150+ $25 for training $175
North Dakota 424,000 $12.64/hour $190
Ohio 6,128,000 $160/dav $160
Oklahoma 1,861,000 $97 for election dav + $25 for training $122
Oregon 2,147,000
Pennsylvania 6,909,000 $100 $100
Rhode Island 538,000 $0 $0
South Carolina 2,575,000 $60 for election dav + $60 for training $120


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
50
South Dakota 437,000 $125-$200 depending on nosition held assumption: lowest wage to reflect clerk status $125
Tennessee 3,251,000 Nashville: $85+$ 15 for training $100
Texas 11,724,000 $8.50-$8.00 $120
Utah 1,398,000 Clerks: $150 $150
Vermont 351,000 $0 Volunteers are common among the 14 counties in Vermont $0
Virginia 4,399,000 $175 $175
Washington 3,906,000
West Virginia 913,000 $150 $150
Wisconsin 3,323,000 Widest range between counties that I've seen: $130. $127.5. $192.75 $150
Wyoming 304,000


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
51
Appendix E
2012 Election Assistance Commission Election and Voting Survey
SECTION D Election Administration
D1 asks for Information on the number of creeinets in your Jurisdiction
D2 asks for information or the number and type of polling places in your jurisdiction
D3, D4, and D5 ask for information on poll workers utilized in the November 2012 general election.
D1. Enter the total number of precincts in your jurisdictions for the November 2012 general election.
D1a. Total | |.................I I Data not available
D1 Comments
D2. Enter the total number of physical polling places in your jurisdiction for the November 2012 general election.
Please include physical polling places in operation on Election Day and physical polling places in operation before Election Day (such as early vote centers),
D2a. Total | |.................| ] Data not available
Next, divide the total physical polling places in your jurisdiction (as entered in D2a) into the following categories.
The amounts should sum to the total provided in D2a. If you do not include election offices in your count of polling places, enter 0.
Data not available
Election Day voting
D2b. Physical polling places other than election offices....................| |..........I I
D2c. Election offices.......................................................| |..........f~l
D2d. Other comments:.....................................................I
Early voting
D2e. Physical polling places other than election offices ..............| |.........Q
D2f. Election offices..................................................I I ........Q
D2g. Other -> comments:................................................I I
TOTAL......................................................................| D2a
D2 Comments
OMB Control No. 3265-0006
17
Expiration Date 5/31/2013


AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS
52
D3. Enter the total number of poll workers used in your jurisdiction for the November 2012 general election,
Poll workers may include election judges, booth workers, wardens, commissioners, or other similar terms that refer to persons who verify the identity of a voter; assist the voter with signing the register, affidavits or other documents required to cast a ballot; assist the voter by providing the voter with a ballot or setting up the voting machine for the voter: and serving other functions as dictated by State law.
* include all people recruited specifically for the purposes of working at physical polling places in operation on and/or before Election Day, but, do not include observers stationed at the polling places or regular office staff.
D3a. Total
I I Data not available
D3 Comments
D4. If your jurisdiction has data on the ages of its poll workers (for example, from voter registration records, payroll records or from poll worker applications), enter the total number of poll workers in each age category.
D4a. Under 18 years old................................................| |
D4b. 18 to 25..........................................................| ~l
D4c. 26 to 40..........................................................| ~l
D4d. 4) to 60..........................................................| |
D4e. 61 to 70...........................................................| ~|
D4f. 71 years old and over.............................................| |
I I Data not available
D4 Comments
D5, How difficult or easy was it for your jurisdiction to obtain a sufficient number of poll workers for the November 2012 general election?
I~~l...Very difficult
I I...Somewhat difficult
I I...Neither difficult nor easy
I I...Somewhat easy
..... .. Very easy
I I...Not enough information to answer
D5 Comments
OMB Control No. 3265-0006
18
Expiration Date 5/31/2013


GRANT OF PERMISSIONS
In reference to the following title(s):
Hentze, I. An Analysis of Election Poll Worker Costs and Cost Drivers. 2017.
I, ,..11^^ .------------,L. ________________, as copyright holder or licensee
with the authority to grant copyright permissions for the aforementioned title(s), hereby authorize Auraria Library and University of Colorado Denver to digitize, distribute, and archive the title(s) for nonprofit, educational purposes via the Internet or successive technologies.
I, 'fey (jvKir\d/)______________, as client of the copyright holder
affirm that the content submitted is identical to that which was originally supervised and that the content is suitable for publication in the Auraria Library Digital Collections.
This is a non-exclusive grant of permissions for on-line and off-line use for an indefinite term. Off-line uses shall be consistent either for educational uses, with the terms of U.S. copyright legislation's "fair use" provisions or, by the University of Colorado Denver and/or Auraria Library, with the maintenance and preservation of an archival copy. Digitization allows the University of Colorado Denver and/or Auraria Library to generate image- and text-based versions as appropriate and to provide and enhance access using search software.
This grant of permissions prohibits use of the digitized versions for commercial use or profit.
Signature of Copyright Holder Date of Signature
Amanda Buchanan
Printed or Typed Name of Copyright Holders Client
Attention:
Auraria Digital Library Program I Matthew C. Mariner Auraria Library
1100 Lawrence | Denver, CO 80204 P: 303-556-5817 [ matthew.mariner ucdenver.edu


GRANT OF PERMISSIONS
In reference to the following title(s):
Hentze, I. An Analysis of Election Poll Worker Costs and Cost Drivers. 2017.
I, VW(\V _________________________, as copyright holder or licensee
with the authority to grant copyright permissions for the aforementioned title(s), hereby authorize Auraria Library and University of Colorado Denver to digitize, distribute, and archive the title(s) for nonprofit, educational purposes via the Internet or successive technologies.
I,___________________________________________, as client of the copyright holder
affirm that the content submitted is identical to that which was originally supervised and that the content is suitable for publication in the Auraria Library Digital Collections.
This is a non-exclusive grant of permissions for on-line and off-line use for an indefinite term. Off-line uses shall be consistent either for educational uses, with the terms of U.S. copyright legislation's "fair use" provisions or, by the University of Colorado Denver and/or Auraria Library, with the maintenance and preservation of an archival copy. Digitization allows the University of Colorado Denver and/or Auraria Library to generate image- and text-based versions as appropriate and to provide and enhance access using search software.
This grant of permissions prohibits use of the digitized versions for commercial
11 aM / \7
Signature of Copyright Holder Date of Signature
Iris Hentze
Printed or Typed Name of Copyright Holder
Signature of Copyright Holders Client Date of Signature
Amanda Buchanan
Printed or Typed Name of Copyright Holders Client
Attention:
Auraria Digital Library Program / Matthew C. Mariner Auraria Library
1100 Lawrence | Denver, CO 80204 P: 303-556-5817 | matthew.mariner@ucdenver.edu


Full Text

PAGE 2

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 2

PAGE 3

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 3

PAGE 4

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 4

PAGE 5

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 5

PAGE 6

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 6

PAGE 7

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 7

PAGE 8

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 8 (

PAGE 9

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 9

PAGE 10

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 10

PAGE 11

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 11

PAGE 12

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 12

PAGE 13

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 13

PAGE 14

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 14

PAGE 15

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 15

PAGE 16

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 16

PAGE 17

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 17

PAGE 18

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 18

PAGE 19

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 19

PAGE 20

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 20

PAGE 21

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 21

PAGE 22

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 22

PAGE 23

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 23

PAGE 24

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 24

PAGE 25

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 25

PAGE 26

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 26

PAGE 27

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 27

PAGE 28

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 28

PAGE 29

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 29

PAGE 30

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 30

PAGE 31

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 31

PAGE 32

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 32

PAGE 33

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 33 Appendix A Variable Variable Type Relevant Values/Types of Data Assumptions Data Source Notes Wage Data Poll Worker Wage DV Continuous, Varies Only examining election day wages; hourly wages adjusted for a 15 hour work day; bonus for training included if exists See Appendix D; Election Directors Websites EAVS Survey Data Number of Poll Workers DV Continuous, Varies Total across all of a states jurisdictions for which data exists EAVS Data staff paid, working on election day jurisdictions partnered with one another, so reported 0 poll workers like Edmunds ME Data

PAGE 34

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 34 Total Number of Polling Places IV Continuous, Varies On election day EAVS discrepancy between whether or not jurisdictions were counting the total totals include physical polling places and central count locations Gentry County, includes absentee polling places Pulaski County, i ncludes absentee polling places possible Included absentee, centers tried to place a ballot is Voter Convenience Group Early Voting IV Binomial (0 = No Early Voting, 1 = Presence of Early Voting) ons and eeand early ; o ns and

PAGE 35

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 35 in state Absentee Voting IV Binomial (0 = No In Person Absentee Voting, 1 = Presence of In Person Absentee Voting) Same Day Registrati on IV Binomial (0 = No Same Day Registration, 1 = Presence of Same Day Registration) If Same Day Registration exists during a states election day) Election Day Registrati on IV Binomial (0 = No Election Day Registration, 1 = Presence of Election Day Registration Binomial Vote by Mail IV Binomial ( 0 = Measures whether or not a state has some

PAGE 36

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 36 mail, 1 = some Polling Scheme Vote centers Precinct Polling Places IV Binomial (0 = Uses Traditional Precinct Polling Places, 1 = Uses Vote Centers) Technology Use of Technolo gy IV Binomial ( 0 = low technology use, 1 = high technology use) Values assigned to each type of election technology used from 1 being the low end of technology used in elections and 6 being the most stateof the art system used in election Then, median found break the sample into low technology use (0) and high technology use (1) Verified Voting Cost of Living Index Cost of IV Continuous range

PAGE 37

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 37

PAGE 38

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 38 (Table 1)

PAGE 39

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 39

PAGE 40

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 40

PAGE 41

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 41

PAGE 42

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 42

PAGE 43

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 43

PAGE 44

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 44

PAGE 45

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 45

PAGE 46

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 46

PAGE 47

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 47

PAGE 48

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 48

PAGE 49

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 49

PAGE 50

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 50

PAGE 51

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 51

PAGE 52

AN ANALYSIS OF ELECTION POLL WORKER COSTS AND COST DRIVERS 52