Citation
An inquiry into the twin deficits

Material Information

Title:
An inquiry into the twin deficits
Creator:
Crane, Colleen D
Publication Date:
Language:
English
Physical Description:
vii, 115 leaves : ; 29 cm

Subjects

Subjects / Keywords:
Budget deficits -- United States ( lcsh )
Balance of trade -- United States ( lcsh )
Balance of trade ( fast )
Budget deficits ( fast )
Economic history ( fast )
Economic conditions -- United States ( lcsh )
United States ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Bibliography:
Includes bibliographical references.
General Note:
Submitted in partial fulfillment of the requirements for the degree, Master of Arts, Economics.
General Note:
Department of Economics
Statement of Responsibility:
by Colleen D. Crane.

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:
29195126 ( OCLC )
ocm29195126
Classification:
LD1190.L53 1993m .C73 ( lcc )

Full Text
AN INQUIRY INTO THE TWIN DEFICITS
by
Colleen D. Crane
B.S.B.A., University of Denver, 1988
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado at Denver
in partial fulfillment
of the requirements for the degree of
Master of Arts
Economics
1993


This thesis for the Master of Arts
degree by
Colleen D. Crane
has been approved for the
Department of
Economics
by
V//6/93
Date


Crane, Colleen D. (M.A., Economics)
An Inquiry into the Twin Deficits
Thesis directed by Associate Professor Steve Beckman
ABSTRACT
The problem is to identify the effects of a federal
budget deficit and a corresponding trade balance deficit
on the domestic economy in terms of investment, interest
rates and exchange rate correlations.
A literary review was performed to examine empirical
and econometric analysis on both subjects along with
corresponding recommendations for policy initiatives to
relatively determine cause and effect correlations. The
problem then is, given the recommended deficit reduction,
how are investment, interest rates, exchange rates and
the trade balance affected?
To examine this issue and the corresponding effects,
a simplistic open economy model was estimated utilizing
Two Stage Least Squares (TSLS) estimation technique.
Once a model proved to be relatively stable, it was then
shocked for a decrease of $100 billion dollars in
government spending. The results mirrored theory, except
for a rather surprising inflation effect which fed
through the system of simultaneous equations. This
iii


effect was surprising in its magnitude, but hardly
unrealistic given inflation fears and phobias; and
especially the Fed's position on opting for slow monetary
growth in favor of suppressing inflation rather than
encouraging economic growth.
The conclusion derived is that the size and
persistence of the federal budget deficit has seriously
impaired the Fed's abilities and actions, and could
plausibly be responsible for the persistence of the trade
balance deficit in lieu of dollar depreciation.
This abstract accurately represents the content of the
candidate's thesis. I recommend its publication.
Signed
iv


AUTHOR'S NOTE
An inquiry into the twin deficits started out as a
rather lengthy review of literature on both subjects.
However, what evolved was a rather detailed development
of a rather simplistic open economy macroeconomic model.
The importance lies in the understanding and application
of economic theory, as it creates the foundation for
explanation and analysis of complicated behaviorisms.
What prevailed was the notion that econometric modeling
does not contain the capacity to predict, but rather
serves as a catalyst in our understanding of causal
relationships. The theoretical foundation lends
credibility to econometric modeling, and serves as a
basis for analysis of these rather complex behaviorisms.
The problems associated with the predictive capacity
of econometric modeling are underlying forces, such as
the global environment and government interventions
through the ambiguous results of fiscal and monetary
policies. In a closed free market economy, there is no
question of the ability of econometrics to accurately
predict and explain. However, open economy models, such
as the 12 Brookings Institution Models, demonstrate
pronounced ambiguities.
viii


CONTENTS
Figures...............................................vii
Author1 s Note........................................viii
CHAPTER
1. INTRODUCTION .................................... 1
2. LITERATURE REVIEW ............................... 4
The Federal Budget Deficit ..................... 4
Conclusion............................... 26
The Trade Balance Deficit ..................... 27
The Brookings Institution Simulations ... 44
3. THE MODEL: INTRODUCTION......................... 49
Model Development.............................. 50
The Consumption Function ...................... 57
The Investment Function ....................... 62
The Net Export Function........................ 65
The Exchange Rate Function..................... 69
The Short-Term Interest Rate Equation ... 73
The Moving Average Inflation Rate Equation . 75
Identification Process ........................ 78
Order Condition Box.......................... 78
Rank Condition Analysis ....................... 79
Results........................................ 80
Model Files.................................... 83
v


Policy Experiments ............................ 97
4. CONCLUSION......................................102
APPENDIX
1 105
2 .............................................106
BIBLIOGRAPHY .......................................... 115
vi


FIGURES
Figure
1. Predicted and Actual Consumption, Model 2 ... . 82
2. Linking Up the Self-Correcting Mechanism .... 87
3. O.L.S. Versus Two-Stage Least Squares ........... 95
4. Effects of $100 Billion Cut in Government Spending 98
5. Effects of a Balanced Budget.....................101
vii


CHAPTER 1
INTRODUCTION
Originally, my thesis was to discuss the detrimental
effects of perpetuating a large federal budget deficit on
the domestic economy, through the development of a
macroeconomic model. However, upon review of the wealth
of literature that exists on the subject, it became
increasingly clear that empirically it would be extremely
difficult to address all the effects. By originating my
analysis on the subject of the federal budget deficit,
what became evident and expressly stated, is that more
analysis needs to be directed towards the subject of the
twin deficits. Intuitively, it is assumed that a federal
budget deficit is harmful and the trade balance deficit
is harmful, but that the combination is devastating to
the domestic economy in terms of growth and prosperity.
This inquiry has been limited to the relationship
between the twin deficits in terms of economic impact
through economic theory, empirical investigations and
development of my own macroeconomic model. Even with
this limited scope, the macroeconomic model must allow
for exchange rate effects, since this in turn aeffects
1


the trade balance, and must allow for interest rate
effects, since this also affects exchange rates. And it
must allow for the inflation rate effect, since it
affects both exchange rates and interest rates. The
inflation rate effect is an important component of this
dynamic specification, as the results of the solved model
will indicate. The end result was that a reasonably
complete macroeconomic model was needed, even for this
somewhat restricted exercise.
The literature review has been carefully edited and
chosen to objectively present various econometric,
empirical and theoretical schools of thought on the
subjects of the budget deficit and the trade deficit.
The organization of this study is as follows: a
review of literature on the subject of the budget deficit
as it is presented within three perspectives, categorized
by Charles Schultz as the "pussycat, termite and wolf"
categories.1
. Chapter two contains a review of literature on the
subject of the trade deficit followed by a study
conducted by the Brookings Institution on ambiguities of
policy initiatives and their effects on the trade
balance. Chapter three is comprised of a detailed
^James M. Rock, Debt and the Twin Deficits Debate (Boston: Bristlecone
Bb
Dks,
1991): p.
3.
2


account of the methodology and problems associated with
OLS and TSLS estimation techniques. Chapter four
discusses the results and conclusions of simulation
exercises of the estimated model and progression of the
model files.
3


CHAPTER 2
LITERATURE REVIEW
The Federal Budget Deficit
Depending on one's experience and point of view, a
situation can then be perceived in various perspectives
along with various consequences. In regards to the
Federal Budget Deficit, it can be viewed as a necessary
phenomenon in a monetary society or it can be viewed as
something harmful to growth and prosperity. Charles
Schultz characterizes three divergent views on this
subject, the wolves, the termites and the pussycats.2
Those in the wolves category view the deficit as
seriously harmful, whose effects may include national
bankruptcy, wild inflations and even a depression unless
the deficit is balanced.
Those in the termite category view the deficit as a
gradual process of deterioration of our economic
foundations. For example, the debt is viewed as an
enormous obligation transferred to future generations.
The presumption is that the public deficit absorbs
private savings as it is financed through the private and
2Ibid.
4


foreign sector thus robbing us of savings, thereby
reducing current investment and growth. This line of
reasoning includes discussion of the crowding out effect,
the substitution effect and net foreign investment
effects.
Those in the pussycat school view the deficit as
something that won't hurt the economy, and as a necessary
phenomenon in a monetary society. During slow growth,
government spending helps to maintain incomes through
social transfers which allows debt commitments to be
made. The balance sheet effect is analyzed as deficit
spending increases a bank's willingness to loan funds and
the perceived wealth effect is examined.
Those in the pussycat category are now presented,
along with critiques of the wolves category as presented
within the orthodoxy of mainstream economic theory.
Marshall Robinson contends that it is not "how much
we owe-but how we manage what we owe."3 He maintains
that deficit financing is a necessary and manageable part
of our monetary society. He sees no evidence that
government spending crowds out private investment through
the absorption of private savings. Instead, he perceives
certain aspects of government spending responsible for
^Marshall Robinson, "America's Not-So-Troubling Debts and Deficits, Harvard
Bfobiness Review (July-August, 1989): p. 50.
5


the heavy debt. These are social transfers (Medicaid,
Medicare, unemployment, welfare, farm supports) and
interest payments. It is this form of government
expenditure that is responsible for crowding out
government investment in our infrastructures, such as
education, research, mass transit, job training,
environmental quality, highways and energy development.4
According to Robinson, an article titled "A
Keynesian Presentation of the Relations Among Government
Deficits, Investment, Savings and Growth" by L. Randall
Wray critiques the orthodoxy view through empirical and
theoretical analysis. His critique of orthodoxy suggests
a pussycat perspective, as he discusses a Keynesian
alternative and examines empirical evidence of the budget
deficit on the level of savings and interest rate effect.
In his presentation of the Keynesian view, Wray
defines three major propositions from the General Theory:
1. Spending determines income
2. Investment is a primarily a function of profit
expectations
3. The interest rate is determined in the money
markets.
4Ibid., p. 53.
6


These basic propositions are used to analyze deficit
spending, beginning by justifying deficit spending in
terms of future cash commitments.
Deficit spending raises the level of aggregate
spending and allows nominal incomes to grow. In short,
deficit spending generates incomes, and this is a
necessary phenomenon in a monetary economy. Deficit
spending is then viewed as the process of credit
creation. As some firms or individuals borrow, others
receive income in excess of spending, and at any time
aggregate deficits must equal aggregate surpluses.5
This must be so since most of investment is financed
internally from last period's profits, but growth cannot
occur because spending cannot be greater than last
period's income. Only through the process of credit
creation can a monetary society grow. Financial
institutions then capture the generated surplus spending
in the form of savings to finance positions in assets.
One link between deficit spending and investment is then
made from future expectations regarding investment
opportunities. If expectations are low, investment
falls, profits fall through a multiplier effect, and
consumption may fall as well leading to debt repudiation,
5Rock, p. 979.
7


recession and depression. It is in this respect that
government spending helps to compensate for the
unwillingness of firms to seek deficit finance for
investment.
In an economy financed through tax receipts and
counter-cyclical spending, the deficit is used to
compensate private investment as in the case of a
recession, as well as consumption through social
transfers. As government spending generates incomes,
positive expectations of the economy could be
forthcoming. Now the government must recapture this
generated surplus by the sales of bonds. As these bonds
are held by U.S. citizens, their perceived wealth is
affected by the promise of future interest payments. It
is this deficit spending, whether incurred by households,
firms or government, that produces incomes and allows an
economy to grow through the process of credit creation.
It is then deficit spending, not savings that is utilized
to finance investment and growth in a monetary
societies.6
The main critique to orthodoxy is that savings is
not finance, as stated by Keynes. Orthodoxy defines
savings as postponed consumption, and provides the supply
6Ibid., p. 980.
8


of loanable funds, whereas investment and deficit
spending represent demand. In this presentation of the
loanable funds theory, as the deficit grows, interest
rates increase, which in turn causes the guantity of
savings offered to increase. In this respect, savings
must increase by as much as the increase in deficit
spending, or a part of the deficit is funded by the fall
in the level of private investment. The logical argument
is that government spending and investment is not
constrained to the level of savings, and that financial
institutions aren't quantity constrained and will meet
all demand for credit at a set loan rate. This must be
so since a bank considers government bonds a zero default
risk that can easily be liquidated, thus there is no
reason to assume that deficit spending will increase
interest rates.7
Wray then discusses Minsky's view on deficit
spending, as it relates to the pussycat category. The
three effects presented from Minsky's perspective are the
income-expenditure effect, the cash flow effect and the
balance sheet effect.
The income expenditure effect analyzes the effects
of government spending on incomes. Empirical evidence
7Ibid., p. 997.
9


indicates that government spending contributes directly
to GNP when it purchases goods and services, and
indirectly to GNP through transfers and interest
payments. Historically, government spending fell during
the 1970's; the trend was then reversed through the
1980's, especially during the recession of 1982. The two
types of government expenditures that grew significantly
during the 80's were transfers and interest payments, as
a percentage of personal incomes. It was these transfer
payments that maintained incomes during the recession and
helped to offset the gradual decline in incomes and
wages.* 8
A more detailed analysis of selected components of
spending and income reveals the effects of government
spending as a contributing factor in the growth of GNP.
In the period studied, 1981.2 through 1988.3, wages
accounted for only 61% of personal incomes while transfer
payments accounted for 16.6%; interest incomes accounted
for 13.9%, while government interest payments accounted
for the balance of 5.3%.9
The cash flow effect is an analysis of how
government spending allows private debtors to maintain
8Ibid., p. 986.
8Ibid.
10


cash flows to meet debt commitments. If private deficit
spending declines so does aggregate incomes thereby
creating a situation where debt commitment can not be
maintained. In the period from 1981 to 1983, aggregated
deficit spending by the private sector fell
significantly. As investment is a major determinant of
profit flows according to Kalecki profit equation,
government deficit spending grew allowing corporate
profits to recover while net investment fell.10
Government spending "explained" 82% of corporate profits,
while net investment accounted for only 21%. Following
the recession, net investment remained "sluggish," but
corporate profits more than doubled suggesting the
importance of deficit spending in maintaining profit
flows.
The balance sheet effect examines the relationship
between government spending and savings. The balance
sheet is expanded as banks finance short term until the
short-term debt is retired by long-term debt, which is
sold to individuals or institutions. The orthodox view
is that deficit financing will compete and crowd out
private debt in portfolio investment, with the main
10Ibid., p. 987.
11


proposition that private savings will be absorbed by
deficit financing.
To examine this proposition, personal savings was
compared as a percentage of deficit financing along with
gross savings, and personal savings as a percentage of
personal incomes. From the period studied, 1979 through
1982, both personal and gross savings grew as a
percentage of credit creation by non-financial sectors.
During the recession of 1982, both fell dramatically with
personal savings divided by personal incomes decreasing
steadily through 1988. In 1986, individuals held only
7.4% of outstanding government debt. The majority of
government bonds is not held by individuals to postpone
consumption, but by financial institutions willing to
expand their portfolios with risk free securities. Data
analyzed on the banks willingness to extend credit has
revealed that the extension of credit to private markets
has kept pace with bank credit extended to finance the
government deficit. The data does not suggest that
government debt displaces private debt in bank
portfolios.
Empirical evidence is examined on the effects of
interest rates to deficit spending, through government
debt held by commercial banks. The data does not
indicate a clear correlation between borrowing and
12


interest rates. The supply of foreign funds could have
helped to offset real changes in interest rates. The
empirical evidence presented does not seem to support the
orthodox view that savings must be increased to induce
investment, that government deficit "absorb" savings, and
that the theory of the determination of interest rates is
given by the intersection of the savings and deficit
spending curves. Deficit spending creates surplus
incomes to purchase bonds and other deficit created
liabilities. Interest rates are created by the
willingness of financial institutions to increase their
balance sheets and by the liquidity preference. Most
importantly, Wray argues that deficit spending is
essential for economic growth and stability in the
absence of private investment. He has shown that deficit
financing has financed "leakages" such as the trade
deficit, saving and retained corporate profits. Deficit
spending is a necessary component in a monetary society,
its importance lies in its usage to maintain incomes
during recessions to compensate for low levels of private
investment.
Berstein and Heilbroner also addressed the crowding
out effect through empirical analysis to determine the
relationship between the federal budget deficit and
investment or capital formation. Mainstream contends that
13


"crowding out" occurs because of dis-savings that would
have been otherwise allocated for private expenditure on
plant and equipment. The statistical relationship
between government savings as a percentage of GDP less
net foreign income, to gross private capital formation as
a percentage of GDP plus net foreign investment was
examined for the U.S., U.K., Canada, Japan and France.
All countries, with the exception of Japan, clearly show
a deterioration of capital formation from the period
studied, 1980-1986. This does seem to substantiate
Mainstream theory.11
In an attempt to further examine the affects of the
deficit on investment, Berstein and Heilbroner regress
the change in investment to the change in the deficit.
The results proved to be statistically insignificant with
an r2 of only .16. Next, a regression was run on the
change in private savings to the change in deficits to
try to examine the mainstream idea of savings absorption.
Again the results were not significant, even weaker than
the previous regression on investment. This led to the
conclusion that there is no evidence supporting the
l-'-Ibid., p. 126.
14


orthodoxy view that the deficit crowds out private
investment or absorbs savings.12
The Ricardian Model of Budget Deficits is also an
important contributor to the pussycat category as it
emphasizes the neutrality of budget deficits. The
Ricardian Equivalence Theorem examines the relationship
between budget deficits, the current account, the level
of private savings and the interest rate effect, as it is
presented by Robert J. Barro.13
Barro does not adhere to the general consensus that
sustaining a large budget deficit is harmful to an
industrialized nation. He compares the deficit, defined
by private holdings of public debt, to GNP. Statistical
analysis indicates that relatively it has not been
increasing since 1987. The ratio actually fell from 1987
to 1989 and declined over 4% from 1983-1985.
His study begins with the Standard Theory of Budget
Deficits, which explains why interest rates must rise,
and then compares empirical investigations to the
Ricardian Theory of Budget Deficits.
Suppose the government decides to cut taxes and
incurs a budget deficit, the level of public savings
12Ibid.
l^Rock, p. 133.
15


decline by the same amount, according to the propositions
set forth by the Standard Theory. This assumes that the
perceived wealth effect will then increase consumption
and private savings but by only a fraction of the deficit
thereby lowering national savings. In a closed economy,
national savings must equal domestic investment. If S <
I, real interest rates must increase due to competition
for funds at a sustained level of domestic investment.
Equilibrium is then restored through higher interest
rates reducing investment and increasing savings. The
new equilibrium leads to the "crowding out" of private
investment through deficit expansion. One of the main
problems associated with the Standard Theory is the
assumption of a closed economy.14
Barro then includes borrowing and lending across
countries within this theoretical framework. A country's
national savings finances net domestic and net foreign
investment. Net foreign investment equals the current
account because it is equal to the investment that the
country pays for abroad less what is spend in this
country. In this situation, deficit financing still
creates the situation where domestic investment is in
excess of the national savings level. Instead of the
14Ibid., p. 134.
16


real level of interest rates rising, foreign borrowing
compensates for this excess investment demand. In this
manner, a budget deficit leads to a current account
deficit. Real interest rates would rise only to the
extent that the borrowing country influences the world
economy.
The Ricardian Theory of Budget Deficits defines the
neutrality of a budget deficit created by tax cuts,
thereby leading to higher future tax liabilities that
have the same present value as the initial tax cut. This
present value is then considered by consumers and then
deducted from the present value of their current incomes
to determine a net wealth position, which in turn
determines the current level of consumption. This
decrease in public savings is offset by an increase in
private savings leaving national savings levels
unchanged. In this manner, a budget deficit has no
effect on aggregate consumption demand. This assumes
ceritis paribus.15
The Ricardian Analysis dictates that a budget
deficit would not affect real interest rates or the level
of investment in a closed economy since no change has
been observed in the level of national savings.
15Ibid., p. 136.
17


In an open economy, the current account balance
equals the level of foreign investment, which is defined
as above, the excess of desired national savings over
domestic investment demand. It can then be observed that
the current account balance would also remain unchanged.
Budget deficits do not lead to current account deficits.
This is also known as the Ricardian Equivalence Theorem,
and states that a budget deficit implies an increase by
an equal amount in the present value of future taxes.
There are several objections to this theorem. One is
that consumers aren't too concerned with future tax
consequences. Another is that private capital markets
are imperfect where future taxes and incomes are
uncertain, and that there are many different kinds of tax
consequences. This may imply that budget deficits are
not fully neutral as assumed.16
The Ricardian Theorem does offer some unique
framework compared to the Standard Theory. Barro
examines empirical investigation into the economic
relationships and differences between these theories.
Overall, the Ricardian view supports empirical
investigation in the relationship between deficit
spending and interest rates and in regards to the level
16Ibid., p. 138.
18


of domestic investment and savings. The Ricardian Model
predicts that domestic investment and national savings
will not respond to an increase in the budget deficit or
the stock of public debt.
The Standard Theory contends that domestic
investment and national savings will decline in a closed
economy, and national savings would then decline in an
open economy.
Next is an examination of the effects of the budget
deficits on the current account. Ricardian Theory
predicts no close correlation, whereas the Standard
Theory predicts an increase. Empirical analysis reveals
no direct correlation between 1948 and 1983. One would
assume that if a correlation did exit, it would have been
recognizable during that time period. Since 1983, there
does seem to be a coincidence of a large current account
deficit and budget deficit. The level of private
investment has been at an all-time high since WWII in
relation to GNP since the mid 1980's. The Standard
analysis of the current account which highlights budget
deficits would predict low levels of investment, which
empirically has not been the case. A better and more
appropriate explanation, as stated previously, is that
investment opportunities in the U.S. have been more
attractive than overseas opportunities, thus attracting
19


foreign capital and significantly contributing to the
current account deficit.
Paul Evans examined the current account deficits and
budget deficits in the five major industrialized nations
and concluded that the current account is largely
independent of the budget deficit. Once again the
Ricardian Theorem seems to hold, despite its rigid
assumptions.
However, a major problem associated with the size
and persistence of the Federal Budget Deficit is the
increase in borrowing from abroad, leading our nation
from a creditor to a net debtor nation. The
attractiveness, stability and favorable treatment to
foreigners has encouraged their investment. This can
easily be explained by the high interest rates of the
80's and fiscal expansion making foreign products
cheaper, thus providing foreigners with U.S. dollars to
invest. With a cheaper dollar and a continued demand for
foreign products, foreign investment has responded
positively.17
Minsky raises an interesting point in regard to the
relationship between deficit spending and the trade
balance. He insists that the benefits associated with
17Ibid., p. 55
20


Jb
deficit spending are seriously impaired by a trade
imbalance, with gross profits accruing to foreigners.
Empirical investigation reveals that the current account
deficit accounted for about 40% of corporate profits in
1987. Then an amount equal to 78% of the federal deficit
leaked out of the U.S. due to the trade deficit. Another
"termite" perspective is presented by David Alan Aschauer
in a study initiated for the Federal Reserve Bank of
Chicago. It is important in the respect that it
differentiates between the various types of government
spending and their separate affects on the various types
of investment. He examines the substitution effect,
duration of the perceived wealth effect and the
expenditure effect. A critique of the Keynesian
perspective is discussed in terms of various components
of government spending. The analysis discusses in detail
the affects of public expenditure on private investment,
known as the "crowding out effect," and presents a
complimentary "crowding in" effect as well.18
This study begins with a neoclassical approach to
the analysis of fiscal policy that fiscal policy crowds
out dollar for dollar private investment. Consumption
and public military spending account for only a small
1 ft
David Alan Aschauer, "Does Public Capital Crowd Out Private Capital?"
jrnal of Monetary Economics 24 (1989): p. 171.
21


portion of private investment in plant and equipment. He
agrees with mainstream economists that the rate of return
to private capital is highly correlated with non-military
public expenditures, and that military and consumption
expenditure have little effect on private investment and
the rate of return to private capital.
The general equilibrium framework is expressed as
i = i(mpk,pi,g)
where i is investment, pi is private investment, mpk is
the marginal product of private capital, and g is current
account spending or government consumption.19
This relationship indicates that an increase in the
mpk would in turn raise the level of i due to
expectations concerning higher marginal return to future
production, postponing consumption and theoretically
increasing savings. In this scenario, an increase in
public investment would crowd out private investment if
there is no perceived wealth effect involved. Given the
differential between the marginal products of public and
private expenditures, a wealth effect will prevail. If
the level of public capital stock was too low, an
increase in public investment and equal crowding out of
private investment would then tend to raise future
19Ibid., p. 172.
22


crowding out resulting in a positive income effect.
Consumption would then increase and savings decline,
leading to a further decline in private capital
formation.20
Three situations impact the extent in which public
consumption spending would affect private investment,
namely a substitution effect, the duration of the
perceived wealth effect and the duration of the
expenditure effect.
Aschauer then modifies the equilibrium framework to
include the marginal rate of substitution of public for
private goods and services, the marginal product of
public current account spending in private production,
and the marginal propensity to consume in the future
through the wealth effect. Now, a rise in public
consumption expenditures has less of a negative impact on
private capital formation the better the public goods
substitute for private consumption, and given the
prolonged rise in government current expenditures.
He concludes that a rise in public investment
expenditure has a rather ambiguous affect on private
capital formation. In one situation, if the substitution
effect is predominant, then a dollar for dollar crowding
20Ibid., p. 173.
23


out will occur.21 Prolonged government spending can
enhance private capital formation through the wealth
effect by raising the productivity of private factors of
production. Depending on the interaction between these
forces and the degree of potency, a decrease or increase
in private capital expenditures could result.
Anschauer then compares a Keynesian perspective with
previous analysis of the impact of government spending.
He maintains that the major fault with the Keynesian
analysis is the failure to account for the differential
aspects of government spending. More emphasis is placed
on the methodology of financing the deficit rather than
the composition of the expenditure itself. In this
analysis, there is an ex-post crowding out of private
investment via a rise in the real interest rates.
However, within the Keynesian models, crowding out may
not occur if output rises to allow for higher private and
public expenditures.22
A few opposing views are then offered on the subject
of deficit financing and the affects on private
consumption and investment. One is presented by Bailey.23
2^Ibid., p. 174.
22Ibid., p. 175.
23Ibid.
24


He differentiated between public consumption and public
investment through a series of spending multipliers under
various sets of assumptions. In one case, households may
consider public consumption as a perfect substitute for
private consumption. If this is so, any increase in
government consumption would produce an ex ante decrease
in private consumption, thereby rendering the effects of
government spending as nil.
Next, Anschauer develops a parsimonious empirical
model to demonstrate the relationships of interest with
the following results: private capital accumulation
responds positively to an increase in the rate of return
to capital, whereas the rate of return to private capital
responds negatively to the net private capital stock and
positively to net public stock. (See appendix 1.) These
results seem to confirm neoclassical theory in terms of
the crowding out effect.
Another OLS estimation, which included military
spending and the natural logarithm of net military
capital stock, showed that military purchases accounted
for only an 8 cents on the dollar decrease for private
investment, while more significant was the crowding out
of consumption. The conclusions derived indicate that
public non-military capital has a significant impact on
the level of private investment in plant and equipment,
25


as well as the average return to private capital. The
important underlying assumption is that while higher
investment by the government sector does crowd out
private investment dollar for dollar, given the return to
capital, but it also raises the productivity of capital
thereby crowding in private investment.24
In his conclusion, Anschauer argues that public non-
military expenditures on items that have a complimentary
affect on private inputs such as infrastructure are
expected to have a greater impact on output. He finds
output multipliers for public investment spending
exceeding unity, while output multipliers for public
consumption have less than a unitary impact on output.
The results strongly suggest the critical importance of
distinguishing between the various components of public
expenditure.
Conclusion
The majority of information researched on the
subject of the budget deficit emphasized the various
critiques of orthodoxy, mainly that deficit financing
absorbs savings, interest rates must rise, and deficit
financing "crowds out" private investment. What has been
suggested throughout the literature review on the subject
24Ibid., p. 183.
26


of the budget deficit, are various ambiguities associated
with the empirical analysis on the relationships between
the deficit, interest rates, investment and savings.
There seems to be no doubt as to the paramount
concern regarding the size and persistence of the budget
deficit. There is a general agreement that the deficit
must be reduced; the disagreement arises in how spending
cuts will be achieved and then how these proposed cuts
will affect the economy.
The effect of the budget deficit on the trade
balance deficit remains unclear. Econometric analysis
dictates that the trade balance would be seriously
impaired by the budget deficit in terms of the interest
rate effect. However, this has not been the case.
Foreign investment could have helped to offset higher
levels of domestic interest rates according to the
assumptions of interest rate parity. Or it could be that
slow monetary growth and low levels of domestic
investment could also be responsible for lower domestic
interest rates.
The Trade Balance Deficit
An in-depth study by Peter Hooper and Catherine L.
Mann titled The U.S. External Deficit: Its Causes and
Persistence examines the persistence and unprecedented
27


large trade imbalance. They surmise that the trade
deficit has not responded to macroeconomic variables as
it has in the past, namely relative income growth and the
exchange rate correlations.25
Three distinct but related approaches are presented:
two are macroeconomic in form and one is microeconomic.
All approaches are based on the open economy IS-LM
framework, and all lead to separate policy initiatives.26
The macroeconomic partial equilibrium approach to
the trade deficit approach is based on the reduced form
of a partial equilibrium model of the U.S. external
balance. The model is provided by Helkie-Hooper (1987),
and expressed as X M = f(Y,Y*,EP/P*,Z), where X M is
net exports, Y and Y* are domestic and foreign incomes,
EP/P* is the exchange rate and Z is a vector of other
variables affecting the value of trades in goods and
services.27 In examining the causes of the trade
deficit, this approach involves quantifying the degree of
change in the right side variables based on underlying
structural relationships.
Pr
2^Albert Burger,
aceedings of the
olishers, 1989): p.
U.S. Trade Deficit: Causes, Consequences.
Twelfth Annual Economic Policy Conference
3.
and Cures.
(Academic
2SIbid., p. 7.
2^Ibid., p. 13.
28


Given this expression, the major determinants of
changes in real net exports are relative growth, Y, Y*,
and the relative prices, EP/P* (nominal rate times the
ratio of domestic to foreign prices). Real net exports
are then compared with measures of relative growth and
prices for the past two decades. During the early 1970's
and late 1970's, significant increases in real net
exports corresponded to significant increases in foreign
activity relative to U.S. activity. Since 1980, the
increase in the level of U.S. activity relative to
foreign activity contributed significantly to the decline
of real net exports over that period.
Real net exports were then empirically compared to
relative prices given by the ratio of foreign consumer
price index to U.S. price index in dollars. Increases of
real net exports in the early 1970's and late 1970's
correspond to increases in U.S. price competitiveness.
The decline of real net exports after the 80's followed a
dramatic decrease in price competitiveness after a lag.
This qualitative analysis outlines the reason for
the widening of the trade deficit after the 1980's in
terms of price and national income. A difficulty
encountered indicating problems with partial equilibrium
analysis and simultaneous bias occurred in comparing real
net exports to relative real activity, GNP. Real net
29


exports are correlated to domestic expenditures and GNP
inversely. A decrease in real net exports induced by a
decline in U.S. competitiveness will tend to increase
U.S. domestic expenditures relative to foreign
expenditures while simultaneously reducing U.S. GNP to
foreign GNP.28
In their conclusion of this partial equilibrium
analysis of the trade deficit, Hooper and Mann agree that
between 1980 and 1986, this approach could reasonably
explain the widening trade deficit through dollar
appreciation and relative price changes. However, this
approach does not reasonably explain problems associated
with the adjustment process, including J-curve effects
that cannot be explained by this partial equilibrium
analysis.
Various policy shift and other factors affecting the
worsening of the trade balance were then examined. The
study begins by focusing on factors underlying movements
of the dollar. The analysis consists of the model of
exchange rate determination and the long run interest
rate parity relationship. The basic assumptions of the
model are:
2Ibid., p. 6.
30



1. perfect substitutability of assets denominated
in different currencies;
2. absence of foreign exchange risk; and
3. a constant expected long run equilibrium level
of real exchange rate.29
The essence of the model is that the dollar will
move to equate the expected rate of return on assets
dominated in different currencies. The real dollar was
compared against G-10 currencies and a measure of the
difference between U.S. and foreign long-term real
government bond yields was examined. The study indicates
a strong correlation between the dollar real exchange
rate and long term real interest rate differentials.
Through 1983, there was a strong appreciation that
corresponded to a sharp increase in U.S. real bonds rates
relative to foreign rates. Beginning in early 1984, a
further rapid appreciation occurred despite a decline in
the U.S. real interest rates to foreign rates. Then in
March 1984, rapid depreciation occurred that coincided
with a continual decline in interest rate differentials.
To explain this phenomenon, they turn to a discussion
regarding fiscal and monetary policy shift variables
within the guidelines set forth in the interest rate
^Ibid., p. 16.
31


parity assumptions. The shift variables were obtained
from the 12 multi-country models from the March 1986
Brookings conference. The simulations examined the
affects of sustained exogenous shifts in government
spending equal to 1% of GNP, both in the U.S. and other
OECD countries, holding monetary aggregates exogenous.
Also examined was the effect of an exogenous 4% increase
in the U.S. Ml money stock.
Tight monetary policy of the early 1980's led to
dollar appreciation and rise in U.S. interest rate
differentials and fiscal expansion, following the initial
tax cuts of 1981. The more recent decline in U.S. real
interest rates have been linked to a more accommodative
monetary policies and adoption of the Gramm-Rudman Act in
1985. However, even with the explanation of fiscal and
monetary policy shift variables fail to explain the
widening trade deficit in light of dollar depreciation.30
The microeconomic factors affecting pricing behavior
and protectionist measures reveal that import prices are
adjusting more slowly to the changes in the exchange rate
than they have traditionally, based on an over prediction
of a non-oil import deflator equation.31 In a dis-
30Ibid., p. 54.
^Ibid., p. 61.
32


aggregation study of indexes of industry specific profit
margins revealed that during dollar depreciation, foreign
importers lost profits. It appeared that profit margins
bear the burden of changes in the exchange rate and
foreign costs, leaving the U.S. price of imports
unchanged. This delay and burden on foreign importers,
as they cut their profit margin, has contributed to the
delayed effects on the exchange rate to dollar import
prices. It could be that the foreign importers are
trying to maintain U.S. market share, and are making
profits in other markets.
This could explain the persistence of the trade
deficit. Through the foreigners' willingness to reduce
profit margins to maintain market shares, the U.S.
exporters cannot compete. There is no strong indication
the U.S. exporters absorbed any loss through dollar
appreciation as domestic prices increased as well. What
seems to be evident is that foreign exporters are willing
to absorb losses associated with U.S. dollar appreciation
and use their profit margins to absorb changes in the
exchange rate. The fact that U.S. prices on exports have
changed very little seems to support purchase power
parity in regards to international pricing
competitiveness.
33


Statistical data was then examined to determine the
U.S. and foreign pricing strategy. Common sense indicated
that for homogenous products, U.S. prices adjusted
globally, especially in response to foreign price cuts.
What was revealed is interesting in the fact that the
majority of U.S. exports were relatively inherently
heterogenous, especially in the manufacturing sector, and
could indeed explain why certain pricing was resilient to
exchange rate changes.
In summary, the microeconomic factors responsible
for the widening and persistence of the trade deficit
were attributed to foreign and domestic pricing strategy
and protectionist measures. The macroeconomic factors
which contributed to the trade deficit were excess
domestic expenditures and GNP, compared to the rest of
the world. This accounted for about 1/3 of the deficit.
The rest was attributed to the decline in U.S.
competitiveness associated with dollar appreciation
during 1985.
The remedy to close this gap suggests a budget
deficit reduction. Mann and Hooper recommend
contractionary domestic fiscal policy and expansionary
foreign fiscal policy, which would result in a decline of
relative growth of U.S. domestic spending to that abroad.
They also suggest a policy mix of easing of monetary
34


policy, to keep GNP from falling, to lower U.S. interest
rates to those abroad which would lead to the necessary
dollar depreciation.32
John Taylor analyzed the savings-investment link to
the trade deficit with a seven country econometric
model.33 His analysis began with the gap between savings
and investment in late 1982. The inexplicable result is
that investment has recovered rapidly and is concurrent
with the trade deficit. Simultaneously, trade surpluses
and saving-investment surpluses were experienced in
Germany and Japan. Taylor focuses his analysis on the
saving-investment link in the absence of a trade deficit.
His identity and the basis for his analysis is that X-M =
S-I. This identity intuitively leads to the consensus
that if the current account decreased, savings would
increase compared with investment changes. It is
important to note the source of change in the deficit as
well as interest rate elasticities on investment,
exchange rate elasticities of exports and imports and the
affects of a change in US interest rates relative to
world rates. It is important to note here that the
significance of this analysis relies on capital mobility,
32Ibid., p. 95.
33Ibid., p. 133.
35


elasticities and lags. Empirical evidence in a rational
expectations model permits real interest rates to differ
among countries, and the amount of the expected changes
is the real exchange rates. Also permitted is long term
interest rates and exchange rates to move in anticipation
of future changes in monetary and fiscal policy shifts.34
Taylor focuses on the fundamental factors
influencing the savings-investment link. He prefers not
to use the trade balance as exogenous, but focus on U.S.
Fiscal Policy as the exogenous variable in his analysis.
He also relies on the fact that GNP and the exchange rate
are endogenous to the saving and investment behavior,
too. The model will demonstrate that government spending
in real terms has a much larger impact on the trade
deficit via the savings-investment link than the current
dollar deficit on the current account.
Debate has arisen over trade deficit/savings-
investment identity. The question is whether or not
exchange rate adjustments are necessary to affect the
trade deficit or whether the Gramm-Rudman-Hollings
gradual budget cuts leading to shifts in the savings-
investment balance can affect the trade deficit without
exchange rate adjustments. Taylor will demonstrate
34Ibid., p. 147.
36


empirical evidence indicating that, as people become
aware of decreases in government spending, their
expectations have an immediate effect on interest rates
and exchange rates.
The results of the simulations show that expected
cuts in government spending generate smaller effects on
investment and consumption compared to unanticipated
cuts. Long-term interest rates fall more than short-term
rates due to the forward looking term structure
assumption built into the model. The dollar exchange
rate depreciates by a large amount, creating a
differential to exist between the U.S. interest rates and
foreign rates. Prices fall, causing nominal interest
rates to fall and lead to increased investment given no
change in the money supply; however, given slow
adjustment on wages, prices do not fall immediately.35
Results in Germany and Japan indicate larger effects on
output. The exchange rates appreciated by a larger
amount with a reduction in exports and an increase in
imports. In addition, with future expectations of the
dollar appreciation, interest rates do not fall as much
as they do in the U.S.
35Ibid., p.'147.
37


The simulation shows that a reduction in government
spending has led to an increase in real net exports by
2.1 percentage points as a fraction of real GNP. This
resulted in an increase of savings (Y-C-G) of 2.9% and an
increase in real investment of 0.8 percentage points.
Taylor points out that government spending cuts crowds in
more net exports than investment.
In another scenario, Taylor examines the same
simulation if the money supply had also been increased
during the government spending cuts. The dollar
depreciates more, given the same decline in prices.
Expansionary fiscal policy leads to higher prices where
the dollar must depreciate. In this situation, dollar
depreciation actually makes the trade deficit grow. In
the short run, the increase in net exports is much
smaller because the increase in the money supply leads to
a larger increase in imports than what is offset by the
depreciation of the dollar. The importance lies on the
affect of the nominal exchange rate differentials. The
dollar depreciated more with the increase in money supply
and prices rose slower and by a smaller percentage. In
short, the nominal exchange rate fluctuated and absorbed
the changes in net exports more than domestic prices did.
In a third scenario, given a decrease in government
spending, Taylor examines the role of exchange rates in
38


the change of net exports through the changes in
investment and savings. He demonstrates that it is the
foreign banks that increase money expansion to offset
contractionary fiscal policy to offset the downward
pressure on domestic prices.36 Foreign countries absorb
much of the changes in prices, as was previously noted by
Hooper and Mann.
Taylor has reached three conclusions in his analysis
for exchange rate policy and a reduction in the trade
deficit. The first is to maintain a reduction of
government spending by 3% of real GNP, which would result
in an increase of real net exports of about 2% points, an
increase in real investment of about 0.8% points and an
increase in real savings of about 2.8% points. The real
impact would be on net exports rather that real domestic
investment and output. Given the same reduction in
government spending and an increase in the money supply,
the current dollar net exports and investment are
affected in reverse; nominal exports rise by a small
amount, whereas nominal investment rises by a larger
amount. He attributes this empirical observation to the
fact that the investment effect is larger than the
36Ibid., p. 149.
39


savings effect and the current account is hardly
affected.

Peter L. Berstein and Robert L. Heilbroner initiated
a study that focuses on the relationship between the
savings-investment imbalance, the current trade balance
and the budget deficit.37 Their analysis focuses on the
possibility that the budget deficit is not a cause of the
trade deficit, but a consequence of it. They state the
problem of the federal budget deficit as a global
phenomenon, with many forces beyond our control. This is
another way of stating that monetary and fiscal policy
fail to impact the economy as they have in the past.
The analysis initiates with the definition of the
trade balance. Imports create national incomes for
foreign economies and tax revenues for these economies.
Competition forces down prices and wages in the importing
countries, depressing the flows of national incomes. In
this manner, imports have also tended to reduce our
national income and tax revenues, creating a larger than
normal budget deficit. In addition, slow economic
growth, inflation phobias and reduced exports have all
tended to reduce our national incomes.38 Mainstreamers
AQ.
37Peter Berstein and Robert Heilbroner,
irms/Real Possibilities, p. 109.
38Ibid., p. 115.
The Debt and the Deficit False
40


argue that unless the federal budget deficit is cut, the
current account deficit will remain. They seem to assume
that the budget deficit is an independent variable, and
that causes consumers to purchase more than is
domestically produced. However, we can't be entirely
certain that the current account deficit is entirely
responsible for the budget deficit due to the
complexities of the global forces. The mainstream idea
is that the budget deficit causes dis-savings and
consequently leads to higher real interest rates
domestically, thereby attracting foreign capital that
bids up the exchange rate of the dollar. This, in turn,
makes our exports more costly and encourages imports.
The main proposition is that the budget deficit raises
the real interest rate, but empirical evidence does not
support this.
If mainstream economists are correct, Berstein and
Heilbroner are correct to point out that what is true for
our economy should also be true for other countries as
well. Their empirical evidence does not seem to indicate
any correlation between debt and interest rates for the
foreign countries studied, nor does there seem to be a
41


relationship between budget deficits and current
account.39
Next studied was the relationship between real
interest rates and the flow of savings and investment.
Mainstreamers assert that the budget deficit has reduced
national savings below the rate of national investment,
causing real interest rates to rise in the U.S.;
empirical evidence does not seem to support this. The
history of real interest rates fails to explain how the
budget deficit affects the current account deficit.
The next issue at question here is whether or not
the budget deficit has contributed to dis-savings and
over-consumption, where consumers are buying more than is
produced domestically, leading to a current account
imbalance. The proposition presented here is whether or
not the budget deficit leads to an increase in the
perceived "wealth effect."
The empirical evidence demonstrated by Berstein and
Heilbroner indicates that we are not overspending. Their
analysis computed the long term trend of consumption over
the period of 1954:1-1988:4 compared to real disposable
income. Compared to the established trend, consumers are
actually spending less than in the 1980*s. Empirical
39Ibid., p. 118.
42


evidence indicates that the problems with the economy can
chiefly be attributed to low productivity, competition
and a monetary bias in favor of suppressing inflation
instead of supporting real growth.40
In conclusion, Berstein and Heilbroner concede that
the world economy is far too "noisy" to permit
unambiguous analysis on the popular hypothesis that the
budget deficit has a negative impact on the economy
through the savings-investment link and trade patterns,
and that "other forces" have an impact in the analysis.
They state that historical and empirical evidence lacks
support of mainstream theory, that budget deficits affect
real interest rates or the status of the current account.
They do concur that there is a negative correlation
between the budget deficit and private capital formation,
but data suggests that the budget deficit is not solely
responsible. Their data also suggests that the economy
is not faced by over-consumption compared to historical
values, but rather that slow growth in real incomes has
led to the decline in personal savings.41
40Ibid., p. 122.
41Ibid., p. 125.
43


The Brookings Institution Simulations
Presented is a study of ambiguous policy multipliers
in theory and empirical models. The importance of this
study is the results and comparison of 12 macroeconomic
models simulated for an increase in government spending
and a change in the money supply. The Brookings project,
summarized by Jeffrey A. Frankel, examines whether fiscal
expansion causes domestic currency to appreciate or
depreciate, and examines other ambiguity associated with
the effects on the exchange rate and the current account.
The emphasis is placed on the effects in the second year
of a policy change, just long enough (he contends) for
the trade balance to get past the negative part of the J-
curve.42
Whether the dollar depreciates or appreciates given
by fiscal expansion is theoretically ambiguous. It may
be that higher interest rates caused by fiscal expansion
cause currency appreciation, or it may be that the higher
income increases the demand for foreign goods leading to
an appreciation of the foreign currency and a
depreciation of domestic currency. At least for the
U.S., the 12 models concur that the interest rate affect
dominates and the dollar appreciates following a fiscal
AO
Jeffrey Frankel, Ambiguous Macroeconomic Policy Multipliers. (Washington,
Brookings Institution, 1988): p. 17.
44


expansion.43 The disagreement arises in the non-
ambiguity negative affect of a domestic monetary
expansion on the current account of trading partners, and
through the transmission mechanism of the exchange rates
on foreign output.
The basic model has the following "well known"
effects, fiscal expansion increases domestic output and
the domestic interest rate. This interest rate
differential then attracts foreign capital inflows, ex
post, translates to a trade deficit. If we assume
perfect capital mobility, the balance of payments will
then improve at an unchanged exchange rate. (We also
assume prices are fixed in the short run.) This implies
a domestic currency appreciation, which may be a cause of
the worsening trade balance as well as the increase in
foreign income. A surplus is then generated in the
foreign sector trade balance which increases foreign
output. The ambiguity that occurs is whether capital
mobility is high enough (or the LM curve steep enough)
for fiscal expansion to appreciate the currency.44
Monetary expansion in the Mundell-Fleming model has
a theoretically unambiguous effect. Domestic interest
43Ibid., p. 18.
44Ibid., p. 18.
45


rates lower and domestic income increases, the interest
rate differentials encourage capital outflows, thereby
depreciating domestic currency and the trade balance
improves. This in turn would tend to worsen the foreign
trade balance, but not foreign incomes due to capital
outflows. Frankel describes this as an inverse
transmission; expansionary monetary policy is
expansionary for the foreign trade partners through the
trade balance transmission. Other theoretical literature
features ways in which this inverse transmission can be
reversed. The exchange rate can affect the savings-
expenditure level through the terms of trade equation,
the money demand function through a change in prices,
changes in real wealth through the saving-expenditure
function, a change in supply the price of foreign inputs,
and enter supply through a change in the nominal wage.45
Frankel then proceeds to examine the 12 Brookings
models for a monetary expansion equal to a 4% increase in
the money supply. The simulations showed more conflict
among the 12 Brookings models than did fiscal expansion.
All models depicted a clear depreciation of the domestic
currency, and either a positive or negative transmission
to the foreign sectors. The typical Mundell-Fleming
45Ibid., p. 19.
46


result, an appreciation of the foreign currency and
foreign output declines, were exhibited by the MC, EPA,
LINK, LIVERPOOL, MINIMOD, TAYLOR and the DRI models. The
EEC, VAR, MSG,OECD and the WHARTON models showed a
positive transmission on the foreign sector. This
ambiguity is revealed as a transmission reversal of the
typical Mundell-Fleming result. The explanation is that,
as foreign currency appreciates, it has one of four
expansionary effects: an increase in the real money
supply or real wealth, or a decrease in real wages or
imported input costs, all of which operate through the
aggregate supply function. (Of course, the Mundell-
Fleming model doesn't consider aggregate supply and
therefore could not predict these results.) If any of
these effects are influential enough, a positive
transmission could result.46
The most important aspect of these simulations is
that with a monetary expansion, a depreciation of the
domestic currency and a worsening of the trade balance,
there is a non-interest rate capital inflow, contrary to
basic and theoretical economic theories. A fiscal
expansion can also lead to a worsening of the trade
4Ibid., p. 23.
47


balance through the interest rate effect, leading to
dollar appreciation.
48


CHAPTER 3
THE MODEL: INTRODUCTION
The task was to develop and estimate a small
macroeconomic model, parsimonious in nature using TSLS
estimation methodology. The model is a fairly
conventional open economy model, designed to allow for
the analysis of the effects of the twin deficits on the
domestic economy. The analysis will lend itself to
examination of budget deficit correlation to the trade
balance, through exchange rates, interest rate and
inflation rate effects. The purpose of the model is to
track and examine the progressions from the first
initiative to the policy experiments in regards to
multiplier, accelerator, and self-correcting mechanisms,
which all affect the stability and feed through this
system of simultaneous equations. Finally, the results
of the policy simulations of the model are discussed in
terms of the hypothetical values.
The organization of the development of the
macroeconomic model is as follows: the structural
equations are derived from economic foundations and
theory of aggregate demand functions. Empirical issues
49


and problems associated with estimation and simulation
techniques are then discussed, proceeded by a brief
discussion of the source of the data bases used in this
study. Results of the regression estimates of the
behavioral equations are discussed, followed by the
identification process. The model is then checked for
stability using historical simulations practices. Once
the model is deemed to be stable, it is then shocked to
analyze the correlations between the twin deficits and
the relative effects on the economy. Finally, the
results are summarized and discussed along with some
further areas of research.
Model Development
The structural equations were derived from the
standard aggregate demand and supply functions within the
Keynesian IS-LM framework. Developed by Hicks, the IS-LM
relationships forms the basis for macroeconomic analysis,
by analyzing the relationships between the goods or real
sector and the money market.
The IS curve is determined by the level of aggregate
demand given by the identity ADorY=C+I+G+ X-M.
This identity states that the demand for domestic output
is given by the level of consumption, investment,
government spending and net exports. Consumption (C) is
50


spending by households for goods and services; Investment
(I) is defined by the spending by firms for plant and
equipment; Government spending for goods and services (G)
is an exogenous expenditure component; Net Exports (X-M)
is defined by demand of goods from foreigners. These
relationships define the real sector of the economy, and
define the IS schedule which shows the relationship
between the interest rate and the level of output.
The behavioral equations in the goods market begin
with the consumption function, C = C(YD). This
relationship states that consumption depends positively
upon disposable income. I = I(r) is the investment
function that states that investment depends negatively
on the rate of interest. G is given as an autonomous
component of real government spending. Net exports, X-M,
are a function of exchange rates.
The IS curve is downward sloping and to the right,
showing that lower interest rates lead to increases in
output. If investment was not responsive to changes in
the interest rates, the IS curve would be vertical or
perfectly inelastic. Any increase in an autonomous
variable like government spending or exports will tend to
shift the IS curve to the right for a given interest
rate, and will cause output to increase.
51


The LM curve depicts all possible combinations of
the interest rate and the level of output that yield
equilibrium in the money market, where money demand
equals money supplied. Money demand is inversely related
to the real rate of interest and positively correlated
with the level of real incomes.
The money demand function relates the level of money
demanded by economic agents for speculative, transaction
and precautionary purposes to variables such as income
and interest rates. The supply of money is assumed to be
set by the monetary authorities. An increase in the
demand for money will shift the LM curve upward to the
left, as is the case for rising prices in the transaction
demand function. Alternately, an increase in the real
interest rates will shift the LM curve downward and to
the right. Theoretically, it is the intersection of the
IS-LM curves which then defines the level of aggregate
demand at a given interest rate and level of output.
The methodology I utilized in my regression analysis
was TSLS to obtain consistent parameter estimates for
solving simultaneous equations. Attempting to solely
utilize OLS (Ordinary Least Squares) for estimation of
coefficients for the simultaneous equations would result
in biased coefficients, due to the Classical Assumption
52


that error terms must be uncorrelated to the explanatory
variables being estimated.
TSLS avoids this Classical Assumption by estimating
coefficients in two stages, where proxy variables are
substituted for the endogenous variables as they appear
as explanatory variables in an equation being estimated.
Consistent parameter variables can then be achieved
through this substitution process, as now the "proxy
variables" will be uncorrelated to the endogenous
explanatory variables.
TSLS estimates can only be performed on models that
are "just" or "overly" identified. The criteria for the
identification process are the Rank and Order condition.
Both criteria are based upon the theory of matrix
identification, where Order refers to the number of rows
and columns, and Rank refers to the linearly independent
rows or columns. The basic premise is that there must be
enough exogenous variables, given the number of
endogenous variables, to solve the simultaneous system of
equations.
The Rank condition is sufficient for identification
where the Order condition, is necessary, but not
sufficient. The exogenous variables must be independent
of each other to obtain relevant parameter estimates. To
determine if the Rank condition is satisfied, an array
53


must be constructed to determine if each variable is
contained in each of the separate equations. The Order
condition is then satisfied if the number of exogenous or
predetermined variables in the system are greater than
(over-identified) or equal (just identified) to the
number of endogenous or slope coefficient variables.
Specification and identification are the most
difficult and important parts of the estimation process.
The important criteria is to utilize economic theory to
define behavioral relationships between various economic
agents in a simultaneous set of equations. It is critical
to avoid omitted variables that can lead to specification
bias. This can primarily be determined if the estimated
coefficients have opposite signs than we intuitively
expect. Another important criterion is to determine
statistical significance through the value of the t-stat.
Another concern is with an irrelevant variable appearing
in the estimated equations. Usually an irrelevant
variable will tend to increase the variance of the
estimated coefficients and will almost always reduce the
r2 and reduces the predictive capacity of the estimated
equation.
The overall fit, r, is examined to determine if the
explanatory variables adequately define the behavior of
dependant variable. The bias is examined to determine if
54


the other variable coefficients change significantly with
the inclusion of the variable in question.
Empirically, we need to address several econometric
problems to determine the reliability and stability of
our estimated equations. Specifically, we need to examine
the estimated relationships for problems associated with
autocorrelation, heteroscedasticity, multicollinearity
and simultaneous-equation bias.
Autocorrelation problems can be impure, pure,
positive or negative, and are a violation of the
classical assumption that different observations of the
error terms are independent of each other. The problems
associated with autocorrelation are increased variances
with the estimated coefficients and underestimation of
the standard errors. Heteroscedasticity is the violation
of Classical Assumption V, which states that the error
terms are drawn from a distribution that has a constant
variance. Heteroscedasticity can take the form of pure
or impure, and has the same consequences as problems
associated with autocorrelation, and can be caused by an
omitted variable. Multicollinearity is the violation of
the assumption that no independent variable is a perfect
linear function of one or more other independent
variables (Classical Assumption VI), and can take the
form of perfect or imperfect. Problems are that
55


variances of the estimates will increase and t-scores
will fall or become insignificant. As such problems are
encountered, a respecification of the regression will be
necessary.
The Durbin-Watson test is widely used to determine
if first order serial correlation (autocorrelation)
exists in the error terms of the estimated coefficients
by examination of the residuals. If DW is close to zero,
extreme positive serial correlation exists. If the DW
value is close to four, extreme negative serial
correlation exists, and if the value is close to two, no
serial correlation exists. In the case of a dynamic
equation, where a lagged dependant variable is found in
the right hand side of the equation, to avoid bias it is
recommended to use the Durbin-h statistic to detect the
problem. The White test is used to detect the existence
of a heteroscedasticity problem, or TSP has a function
that will correct for this problem when a regression is
run. The existence of multicollinearity is examined by
the simple correlation of the coefficients between the
independent variables, and TSP will print these values
after the regression is run.
The use of this estimated model in policy analysis
is justifiable if, in fact, the model proves to be stable
over the estimation sample period. To verify, the model
56


is dynamically simulated over the estimation sample
period and then compared with the actual series. The
model is then deemed to be stable in terms of the
coefficient estimates over the sample observations, if
the simulated series fluctuates closely around the actual
series with relatively low errors.
The data was retrieved from the Citibase statistical
series and is presented in quarterly, real 1982 constant
dollars. Before I had access to the statistical data
base, my data was laboriously retrieved from annual
journals of the Survey of Current Business quarterly
statistical series. The data covers the period from
1973.1 to 1989.3, providing 67 quarterly observations.
The reason the data series was initiated with 1973.1 was
the floating exchange rate regime. This is important in
the analysis of the trade and budget deficit correlation,
because of the influence of floating exchange rates on
the level of real interest rates.
The Consumption Function
To begin with, I specified and estimated the
consumption function. The theory of consumption function
was first discovered by Keynes. Given the Keynesian
macroeconomic framework, the behavioral equations in the
goods market begin with consumption as a function of
57


disposable income. As a preliminary regression, I
examined C = f(YD), and ran a regression of c = a + bYD +
e, where c = the desired level of real consumption
a = autonomous consumption
b = the marginal propensity to consume
e = the error term
The results were clearly unsatisfactory as indicated
by the DW statistic of .490, indicating a strong positive
autocorrelation which led me to suspect the coefficient
estimates and the t-values.
I then respecified the regression incorporating
Freidman's Permanent Income Hypothesis, which is most
commonly used to avoid the error of strong positive
autocorrelation error. The theory is that consumption is
not an instantaneous function of disposable income, but
that consumption is influenced from past levels of
disposable income as well as current levels of disposable
income where C = f[(YD, YD(-l), YD(-2) ..)]. The main
proposition is that less weight should be allocated to
past values of YD, as the length of the lag increases.
Freidman suggests that consumption is not based on
current income, but on some perception of lifetime
income, thus changes in transitory income would not
affect current consumption. Then permanent income
58


perceptions can be hypothesized from past levels of
incomes.
If I attempted to model consumption as a function of
the past values of YD, severe multicollinearity would
result and the degrees of freedom would be compromised.
To correct for this, it is standard to use the Koyck
distributed lag model, which assumes that the value of
the coefficients decrease geometrically over time. This
specification includes the lagged values of consumption
in the list of independent variables, and allows for the
impact of the independent variable to decline as the
length of the lag increases.
The respecification and the estimation of the
consumption is as follows: C = f( C(-l), YD ), and a
regression was run on the equation:
c = a + bYD + dC(-l) + e
The results were better in terms of the issue of the
problem of autocorrelation, given the DW statistic of
1.66. Now that the regression includes a lagged value,
the DW statistic is no longer reliable. A more accurate
test is the Pierce Q statistic, which is computed by TSP
in the identification of the residuals for problems
associated with autocorrelation. Under the hypothesis of
no autocorrelation, the Q statistic follows a Chi Square
distribution with the degrees of freedom equal to the
59


number of lags. The Q-stat for 20 lags was 39.9, which
exceeds the critical value of 31.414, and respecification
was again necessary.
Since C(-1) and YD both appear to be significant in
explaining the behavior of consumption, I left them in
the regression and assumed that I might have omitted a
variable.
I felt that I had to include some form of interest
rate, as intuition suggests that consumption responds to
changes in the interest rates through IS framework. Low
interest rates would tend to encourage consumption
through an increase in aggregate demand, leading to an
increase in output.
I tried the nominal short term interest rate, which
includes the inflation premium. The results: C = f(C-l),
YD, RShort, and I ran the regression:
c = a + BYD + Dc(-l) + fRS + e.
All explanatory variables exhibited the appropriate
sign and were significant according to the value of the
t-Stat. The DW was 1.9 and r2 was .99, indicating that
autocorrelation was not a problem, and that the
regression was a good estimation for the actual data. In
examining the covariance matrix, high multicollinearity
was not a problem, given the low values of the
correlation coefficients between the independent
60


variables
. Heteroscedasticity was not a problem because
the program allowed for a consistent covariance matrix.
In examining the residuals, it confirmed my analysis no
autocorrelation problem exists, with a Q-stat of 29.414 <
31.414. I stopped here with the specification process.
However, in order to address the issue of simultaneous
bias, given a dynamic model and the fact that an
endogenous variable, YD, appears in the right-hand side
of the equation, I will rerun using TSLS, after the
specification process is completed for all behavioral
equations in this dynamic model. Results:
LS // Dependent Variable is C82
Date: 1-01-1980 / Time: 0:34
SMPL range: 1973.1 1989.3
Number of observations: 67
VARIABLE COEFFICIENT STD. ERROR T-STAT. 2-TAIL SIG.
C
YD82
C82(-1)
RSHORT
66.647636
0.2760198
0.7428955
2.2190564
22.334387
0.0772125
0.0760372
0.6401824
-2.9840818
3.5748062
9.7701537
-3.4662876
0.004
0.001
0.000
0.001
R-squared
Adjusted R-squared
S.E. of regression
Durbin-Watson stat
Log likelihood
0.998234
0.998150
13.20032
1.955580
-265.8828
Mean of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
2096.027
306.9128
10977.65
11871.80
Coefficient Covariance Matrix
C,C
C,C82(-1)
YD82,YD82
YD82,RSHORT
C82(-1),RSHORT
498.8248
1.350565
0.005962
-0.010836
0.010647
C,YD82
C,RSHORT
YD82,C82(-1)
C82(-l),C82(-1)
RSHORT,RSHORT
-1.436093
-0.526318
-0.005856
0.005782
0.409834
61


The Investment Function
The investment function is derived from the
investment demand schedule for goods and services
purchased by firms. It has already been acknowledged
that investment decisions respond inversely to the rate
of interest. Also acknowledged and derived through the
IS framework is that investment is positively correlated
to national income or output, Y.
As a preliminary regression for the investment
function, I examined: I = f (Y, r) and ran a regression:
i = a + By + cRS + e.
The results indicated extreme positive serial
correlation of the error terms by the DW = .33. Since
both of the explanatory variables were highly
significant, and theory indicates that both interest
rates and output belong in this behavioral equation, I
assumed that I have omitted one or more variables. In my
respecification of the investment function I incorporated
the Koyck distributed lag model demonstrated by Y(-l), to
account for the geometrically decreasing influence of
output on investment. To allow for the planning horizon
of the investment decision-making process, I lagged RS by
four periods.
Respecification: i = a + bY(-l) + cRS(-4) + e.
62


The results indicate that all variables have
significant explanatory power indicated by the t-
Statistic, but a problem with first order autocorrelation
still exists with DW = .59.
In my next respecification, I added GSURP as a
measure of deficit spending to analyze the crowding in or
out effect of government spending on investment, and the
difference of Y to demonstrate the accelerator principle
of investment given a change in output.
The accelerator principle was discovered by John
Maurice Clark in a well-known article, "Business
Acceleration and the Law of Demand: A Technical Factor in
Economic Cycles."47 This concept explains why the demand
for capital fluctuates much more violently than the
demand for the finished product. It positively asserts
that net investment is a function of the rate of change
in the final output rather than of the absolute level of
output. This phenomenon demonstrates the statistical
relationship between the level of investment and the
change in the level of output, given by d(y). The
equation:
i = a + bY(-l) + cRS(-4) + dGSURP + D(Y) + e.
No:
^Wallace C. Peterson, Income Employment and Economic Growth (New York: W.W.
tfton & Company, 1984): pp. 189-194.
63


The results were obviously unsatisfactory in terms
of autocorrelation problems and were evident when I ran a
display of the residuals, with an unsatisfactory Q-stat
of 118.48. At this point I ran the equation with an
AR(1) terra and ran it again with I(-l); both equations
were satisfactory in terms of problems associated with
autocorrelation, heteroscedasticity, statistical
significance and multicollinearity. I preferred the
respecification equation that included I(-l) instead of
the AR term. It seemed intuitively correct to use the
Koyck distributed lag form of investment to allow for the
planning horizon for the level of investment. It seems
reasonable to assume that it would take more than one
year for the desired level of investment to adjust to the
actual level. This estimated investment functions falls
within the recursive block and does not suffer from the
problem of simultaneous bias since the explanatory
variables are all exogenous or predetermined.
The investment function is far from stable as
compared to the consumption function through the values
of r2. This is not surprising given the importance of
confidence and expectations emphasized by Keynes. Such
variables are difficult to measure and are, therefore,
only imperfectly reflected in the accelerator form of
this function. Rising income may be both cause and
64


effect of rising confidence and is therefore positively
associated with investment. Results:
LS // Dependent Variable is 182
Date: 1-01-1980 / Time: 0:37
SMPL range: 1973.1 1989.3
Number of observations: 67
VARIABLE COEFFICIENT STD. ERROR T-STAT. 2-TAIL SIG.
C
182(-1)
RSHORT(-4)
GSURP82
DY82
Y82(-1)
-10.173383
0.8461615
-0.0366347
0.0999493
0.7638888
0.0284086
18.306944 -0.5557117 0.580
0.0595325 14.213441 0.000
0.9593288 -0.0381879 0.970
0.0479117 2.0861147 0.041
0.0683532 11.175605 0.000
0.0146376 1.9407933 0.057
R-squared
Adjusted R-squared
S.E. of regression
Durbin-Watson stat
Log likelihood
0.976346
0.974407
15.65700
2.481403
276.2375
Mean of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
560.2731
97.86954
14953.64
503.5650
COEFFICIENT COVARIANCE MATRIX
C,C 335.1442 C,I82(-1) 0.307475
C,RSHORT(-4) -3.972239 C,GSURP82 -0.454136
C,DY82 0.001055 C,Y82(-1) -0.158598
182(-1),182(-1) 0.003544 182(-1),RSHORT(-4) 0.028493
182(-1),GSURP82 -0.001088 182(-1),DY82 0.001307
182(-1),Y82(-1) -0.000810 RSHORT(-4),RSHORT(-4) 0.920312
RSHORT(-4),GSURP82 0.009226 RSHORT(-4),DY82 0.029948
RSHORT(-4),Y82(-1) -0.005770 GSURP82,GSURP82 0.002296
GSURP82,DY82 0.000347 GSURP82,Y82(-1) 0.000371
DY82,DY82 0.004672 DY82,Y82(-1) -0.000316
Y82(-1),Y82(-1) 0.000214
The Net Export Function
Through IS-LM, the relationship between exports and
output has already been established. For a small open
economy, net exports are taken as exogenous or
independent of a nation's income, because the level of
exports is determined not by the exporting nation's
incomes, but on the incomes of the trading partner or the
65


rest of the world. Exports then, similar to investment,
represent injections into the nation's income stream
(imports), like savings represent leakages out of a
nation's income stream. It is in this relationship that
imports respond positively to the level of national
income.
In order to demonstrate the effects of the J-curve
on net exports, a PDL estimation was applied to the
exchange rate (EXR) variable. This computes a polynomial
distributed lag, which allows for the slope of EXR to
change as XM changes, and allows for a geometrically
decreasing reliance on past values of EXR in explaining
the behavior of XM. This polynomial form of estimation
demonstrates the effects and the behavior of the J-curve
quite well.
The foundation for the J-curve effect is underpinned
by the conditions and assumptions set forth in the
Marshall-Leaner (ML) Condition. The basic premise is
when devaluation or depreciation of the currency occurs,
the balance of trade improves via expenditure switching
and only under certain restrictive assumptions. The J-
curve is known as the adjustment path in the balance of
payments account. Initially, the balance of payments
deteriorates in the short run, as import and export
volumes adjust to changes in the relative price level,
66


and only then improve in the long run. In the short run,
contracts may not be renegotiable so that the dollar
value of imports rises temporarily. Therefore, following
a depreciation, the trade balance may get worse before
showing signs of improvement.
The restrictive assumptions are domestic output is
held constant, domestic prices are unaffected by changes
in the exchange rate, export and import volumes are
affected only through the real terms of trade or
competitiveness of the industry and imports are affected
by changes in domestic output. The domestic economy is
"small," thereby limiting its effects on the world
economy, and supply decisions react passively to demand
at the fixed price level. Expenditure switching refers
to a country's ability to alter the composition of the
current account through the degree of expenditure
switching between imports and exports, given a change in
competitiveness. The ML condition establishes the
economic foundation for the behavioral relationship
between the exchange rate and net exports. I then
examined the equation: xm = a + bY + cGSURP + PDL(EXR) + e.
All coefficients demonstrated statistical
significance by a t-value > 2.0, and also demonstrated
the appropriate signs. The Q-stat value was 14.1, which
is acceptable within the 95% confidence interval, r2 was
67


.93, which demonstrated a good fit for estimated values
of XM by the regression. In examining the residuals, an
AR(1) term was included because of problems with the
autocorrelation functions. Since this regression
includes an endogenous variables, EXR, in the right hand
side of the equation, it does not fall within the
recursive block and will have to be re-estimated using
TSLS to avoid simultaneous bias of the estimated
coefficients. Results:
LS // Dependent Variable is XM82
Date: 1-01-1980 / Time: 0:12
SMPL range: 1973.1 1989.3
Number of observations: 67
Convergence achieved after 4 iterations
VARIABLE COEFFICIENT STD. ERROR T-STAT. 2-TAIL SIG.
C 369.64059
Y82 -0.0689946
GSURP82 -0.0052890
PDL1 -0.2391758
PDL2 -0.1028003
PDL3 -0.0348471
120.32251
0.0294242
0.0773176
0.2467724
0.1123249
0.1171618
3.0720817
-2.3448238
-0.0684061
-0.9692160
-0.9152049
-0.2974272
0.003
0.022
0.946
0.336
0.364
0.767
AR(1) 0.9029542 0.0611445 14.767557 0.000
R-squared
Adjusted R-squared
S.E. of regression
Durbin-Watson stat
Log likelihood
0.947382
0.942120
13.52234
1.999146
265.8632
Mean of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
30.93881
56.20648
10971.22
180.0473
Lag Distribution of EXR Lag Coef S.E. T-Stat
*
*
*

0 -0.17296 0.36673 -0.47164
1 -0.17122 0.19319 -0.88631
2 -0.23918 0.24677 -0.96922
3 -0.37682 0.18932 -1.99038
4 -0.58416 0.34961 -1.67089
0
Sum -1.54435 0.60181 -2.56620
68


Coefficient Covariance Matrix
C,C
C,GSURP82
C,PDL2
C,AR(1)
Y82,GSURP82
Y82,PDL2
Y82,AR(1)
GSURP82,PDL1
GSURP82,PDL3
PDL1,PDL1
PDL1,PDL3
PDL2f PDL2
PDL2,AR(1)
PDL3,AR(1)
14477.51 C,Y82
1.653988 C,PDL1
0.546848 C,PDL3
1.777621 Y82,Y82
-0.000322 Y82,PDL1
2.15E-05 Y82,PDL3
0.000186 GSURP82,GSURP82
0.004336 GSURP82,PDL2
-0.001953 GSURP82,AR(1)
0.060897 PDLl,PDL2
-0.025329 PDL1,AR(1)
0.012617 PDL2,PDL3
0.000989 PDL3,PDL3
-0.001148 AR(1),AR(1)
-2.900218
-4.940577
-1.061609
0.000866
-3.05E-05
-0.000111
0.005978
0.000928
0.001045
1.83E-05
-0.001327
-0.000388
0.013727
0.003739
The Exchange Rate Function
Similar to the respect that domestic interest rates
are the transmission mechanism that links the goods and
the money market determinants, the exchange rate links
the goods, money and net exports markets given flexible
exchange rates, EXR = f(RSHORT). The volatile nature of
exchange rates can be explained by its role in interest
rate parity and purchase power parity.
Interest rate parity was first observed by Keynes.
He strongly argued that arbitrage profits derived from
hedging and speculation in foreign currencies would tend
to be self-eliminating, because the forward exchange rate
would adjust so that interest rate differentials would
return to zero. This can easily be demonstrated by the
simple equation, R = Rf +(Ex Ef/Ef), which states the
domestic interest rates equal foreign interest rates plus
69


the expected capital gain from changes in exchange rate.
In this same equation, if we replace the expected
exchange rate with the forward exchange rate, interest
rate differentials would return to zero due to arbitrage.
Purchase power parity (PPP) states that
international prices for similarly traded commodities
will be roughly equated through the exchange rate
mechanism. Floating exchange rates now tend to equalize
interest rate and price differentials.
It can be demonstrated that as the exchange rate
increases, domestic interest rates must remain stable, as
indicated by e* = r rfA. What is also revealed is that
domestic price levels remain stable (prices are assumed
rigid in the short run). What should be evident is that
with the initial differential in foreign and domestic
interest rates, the dollar depreciates and net exports
increase; but given the new level of domestic exchange
rates, the dollar will appreciate, thereby deteriorating
the trade balance, allowing for the initial overshooting
of domestic output.
I picked GSURP to analyze the effects of government
spending on the exchange rate via the interest rate as
the transmission vehicle. I am trying to establish a link
between deficit spending and the trade deficit. That
link could prove to be in the form of the exchange rate,
70


if indeed mainstream theory is correct in stating that
deficit spending increases interest rates, which in turn
would tend to increase the exchange rate through interest
rate parity. I used a PDL estimated on the short term
interest rate to geometrically link its effect on the
exchange rate.
I ran the initial equation as: exr = a + bPDL (rs,4,
2) + cGSURP + e. The results were unsatisfactory in
terms of problems with autocorrelation. In terms of the
spiked array of the residuals, I reran the equation with
an AR(1) term. I also included the ratio of the Japanese
CPI to U.S. CPI, PRAT and lagged the variable by four
quarters to allow for the gradual adjustment of prices to
changes in the exchange rate. I also lagged exr by one-
quarter to maintain consistency and to allow for the
decreasing importance of the past levels of exr in
explaining the current levels.
The results were satisfactory in terms of
statistical significance and expected signs of the
coefficients. No problems associated with
heteroscedasticity and multicollinearity were evident,
and the regression adequately explained the actual values
of exr. However, the problem with simultaneous bias
still exists due to the endogenous variables on the right
71


hand side of the equation. To correct for this, I will
rerun using TSLS. Results:
LS // Dependent Variable is EXR
Date: 1-01-1980 / Time: 0:30
SMPL range: 1974.2 1989.3
Number of observations: 62
Convergence achieved after 4 iterations
VARIABLE COEFFICIENT STD. ERROR T-STAT. 2-TAIL SIG.
C
EXR(-l)
PRAT(-4)
GSURP82
PDL1
PDL2
PDL3
4.2673316
0.9456599
-3.9595122
0.0027050
-0.1318472
-0.1977570
0.1400665
13.422753
0.0739791
8.7045272
0.0262022
0.2695682
0.1550946
0.1339214
0.3179178
12.782802
-0.4548796
0.1032352
-0.4891051
-1.2750735
1.0458864
0.752
0.000
0.651
0.918
0.627
0.208
0.300
AR(1) 0.3754899 0.1490449 2.5193078 0.015
R-gquared
Adjusted R-squared
S.E. of regression
Durbin-Watson stat
Log likelihood
0.948797
0.942160
4.129545
1.964592
171.6179
Mean of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
106.6754
17.17069
920.8695
142.9475
Lag Distribution of RSHORT Lag Coef S.E. T-Stat
0 0.82393 0.43808 1.88076
1 0.20598 0.20506 1.00447
2 -0.13185 0.26957 -0.48911
3 -0.18954 0.21823 -0.86853
4 0.03291 0.40151 0.08195
0
Sum 0.74143 0.34522 2.14771
Coefficient Covariance Matrix
C,C 180.1703
C,PRAT(-4) -100.4950
C,PDL1 0.121898
C,PDL3 -0.169414
EXR(-l),EXR(-l) 0.005473
EXR(-l),GSURP82 0.000962
EXR(-l),PDL2 -7.18E-06
EXR(-l),AR(1) -0.005646
PRAT(-4),GSURP82 -0.126288
PRAT (-4) PDL2 -0.362382
PRAT(4),AR(1) -0.157968
GSURP82,PDL1 0.000468
GSURP82,PDL3 -0.000368
C,EXR(1) -0.456213
C,GSURP82 0.119437
C,PDL2 0.630385
C,AR(1) 0.651775
EXR(-1),PRAT(-4) 0.011439
EXR(-l),PDL1 -0.000802
EXR(-1),PDL3 -4.11E-05
PRAT(-4),PRAT(-4) 75.76879
PRAT(-4),PDL1 -0.092731
PRAT(-4),PDL3 0.087457
GSURP82,GSURP82 0.000687
GSURP82,PDL2 0.002043
GSURP82,AR(1) -0.000574
72


PDL1,PDL1
PDL1,PDL3
PDL2,PDL2
PDL2,AR(1)
PDL3,AR(1)
0.072667 PDL1,PDL2
-0.034910 PDLl,AR(l)
0.024054 PDL2,PDL3
-0.000827 PDL3,PDL3
0.003138
-0.001906
-0.001744
0.017935
0.022214
0.001637 AR(1),AR(1)
The Short-Term Interest Rate Equation
The interest rate, as stated above, is the
transmission mechanism that links the goods to the money
market. In the IS relationship, interest rate depicts
the various combinations of national income at which the
goods market is in equilibrium, RS = f(Y). In the LM
relationship, interest rates depict the various levels in
which the money market is in equilibrium in relation to
national income, RS = f(Md, MS). We assume that the
money supply is fixed and must then direct our attention
to the demand considerations of speculative, transactions
and liquidity preferences.
Given the Keynesian IS-LM framework, the "liquidity
preference theory of interest," the demand for money is
also a function of the rate of interest in regards to
asset investment, 1 = f(i). Interest rates can then be
derived by the intersection between the liquidity
preference (money demanded as an asset) and a schedule
representing that portion of the total money supply which
is available to hold as an asset. Money demand is a
direct function of a nation's level of income, or output
73


Y. Empirical analysis support this analysis, that the
link between the demand for money varies directly with
real incomes and inversely with the rate of interest.
I included the accelerator principle by first
differencing output, Y. I also differenced the money
supply to examine the effects of the growth rate on the
level of the short term interest rate. I also included
GSURP, to examine the effects of the deficit on interest
rates, and included the moving average of inflation to
examine its effects on the short term interest rate. The
eguation was run as:
rs = a + bD(Y) + cD(MS2) + dGSURP + fINFM + e.
Upon examination of the residuals, it became obvious
that an AR(1) term had to be included. I reran the
regression and was satisfied with the results in terms of
autocorrelation, heteroscedasticity, Q-stat, the signs
and statistical significance of the coefficient
estimates. Results:
LS // Dependent Variable is RSHORT
Date: 1-01-1980 / Time: 0:15
SMPL range: 1973.2 1989.3
Number of observations: 66
Convergence achieved after 4 iterations
VARIABLE COEFFICIENT STD. ERROR T-STAT. 2-TAIL SIG.
C 8.3391552
DY82 0.0057984
DMS2 -0.0152638
GSURP82 0.0141684
INFM 0.3984233
2.2417147 3.7199895
0.0035253 1.6447751
0.0082556 -1.8489010
0.0056717 2.4980697
0.2255836 1.7661893
0.000
0.105
0.069
0.015
0.082
74


AR(1) 0.9235336 0.0489350 18.872659 0.000
R-squared
Adjusted R-squared
S.E. of regression
Durbin-Watson stat
Log likelihood
0.862142
0.850654
1.006766
2.071512
90.94977
Mean of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
8.061667
2.605142
60.81470
75.04603
Coefficient Covariance Matrix
c,c 5.025285
C,DMS2 -0.003736
C,INFM -0.279837
DY82,DY82 1.24E-05
DY82,GSURP82 -2.55E-06
DY82,AR(1) 6.39E-06
DMS2,GSURP82 8.74E-06
DMS2,AR(1) 2.35E-05
GSURP82,INFM -6.89E-05
INFM,INFM 0.050888
AR(1),AR(1) 0.002395
C,DY82 -0.000939
C,GSURP82 0.004302
C,AR(1) 0.010718
DY82,DMS2 4.41E-07
DY82,INFM 8.66E-05
DMS2,DMS2 6.82E-05
DMS2,INFM 0.000377
GSURP82,GSURP82 3.22E-05
GSURP82,AR(1) 4.36E-05
INFM,AR(1) 0.003497
The Moving Average Inflation Rate Equation
Due to the highly volatile and erratic nature of the
rate of inflation, a moving average series was generated
and used instead of the actual values. This behavioral
relationship defines the aggregate supply side of the
model. The Phillips curve plays a major role in most
macroeconomic models, linking the real sector with the
rate of inflation. It is based upon the expectations-
augmented Phillips curve relationship which assumes that
prices are fixed in the short run and production is
altered to match demand at each period. Given a
production function there is a particular level of output
associated with full employment. This natural rate is
75


expressed as YNAT, the natural rate of output. Prices
gradually adjust to sustain the natural rate of output.
Consequently, YNAT and Y are inversely correlated, and
eventually approach equilibrium. The regression was run
as infm = a + bYNAT + c2Y + e.
The results were clearly unsatisfactory in terms of
autocorrelation, and the estimation power of the
regression to the actual value. In examining the
residuals, it became necessary to include an AR(1) term
and a Kocyk distributed lag term for INFM. The equation:
infm = a + binfm(-l) + cYNAT + dY82 + AR(1) + e.
The results were satisfactory in terms of expected
coefficient signs and significance. The r2 was .96,
which indicated a good fit of the data for explaining the
actual values of INFM. YNAT and Y82 were highly
significant and inversely correlated as theory dictates.
The most significant value was demonstrated by INFM(-l),
also known as the speed of adjustment, and was between
zero and one as assumed by the partial adjustment model
to be dynamically stable.
There were no obvious problems associated with
multicollinearity and heteroscedasticity. The problem
with simultaneous bias will be corrected by rerunning
with TSLS. Results:
76


LS // Dependent Variable is INFM
Date: 6-16-1992 / Time: 14:26
SMPL range: 1973.1 1989.2
Number of observations: 66
VARIABLE COEFFICIENT STD. ERROR T-STAT. 2-TAIL SIG.
C
INFM(-l)
YNAT
Y82
2.1837829
0.9458796
0.0034505
0.0029457
0.8667779
0.0362060
0.0006584
0.0006525
2.5194262
26.124946
-5.2409541
4.5145688
0.0143
0.0000
0.0000
0.0000
R-squared 0.960517
Adjusted R-squared 0.958606
S.E. of regression 0.497763
Log likelihood -45.54313
Durbin-Watson stat 1.552643
Mean of dependent var
S.D. of dependent var
Sum of squared resid
F-statistic
Prob(F-statistic)
5.945307
2.446560
15.36163
502.7630
0.000000
At this stage, the estimation and specification
process is complete. I am satisfied with the results of
the regressions based upon economic theory, the
foundations set forth in the specification process, and
the corresponding values of the estimated coefficients.
At this point, I do not suspect the estimated
coefficients based on the problems of estimation
associated with multicollinearity, heteroscedasticity,
autocorrelation. I have examined the "goodness of fit,"
checked for MA or AR processes, examined the covariance
matrix, and the Durbin-Watson stat. At this point, these
are the best estimations possible for this set of
simultaneous equations. However, given the properties
set forth in estimation techniques, to avoid errors in
77


simultaneous bias, the equations will be re-estimated
using TSLS.
The next step towards solving this set of equations
is to satisfy the conditions of identification, given the
Rank and Order conditions.
Identification Process
Instruments: C(-l) RL EXR(-l) I(-l) MS G TAX Y(-l)
YNAT PRAT Endogenous Variables: C I XM RS Y YD GSURP EXR
INFM.
Order Condition Box
Summary criteria: The number of predetermined
(exogenous and lagged endogenous) variables in the system
must be greater than or equal to the number of slope
coefficients in the equation being estimated.
H = # of exogenous variables in the system
h = # of exogenous variables in the single equation
G = # of endogenous variables in the system
g = # of endogenous variables in the single
equation.
IF H h > g 1, the equation is over identified. If
H h = g 1, the equation is just identified. If H h
< g 1, the equation is under identified and can't be
estimated. Results:
78


EQ: 1. Consumption Function 10 1 > 2 1 over
identified
EQ: 2. Investment Function 10 3 > 2 1 over
identified
EQ: 3. Net Export Function 10 1 > 2- 1 over
identified
EQ: 4. Interest Rate Eq. 10 2 > 2 1 over
identified
EQ: 5. Exchange Rate Eq. 10 2 > 3 l over
identified
It appears that the equations can be estimated based
upon the satisfaction of the Order condition, which is a
common condition for most macroeconomic models. However,
the Order condition is necessary but not sufficient
condition for identification.
Rank Condition Analysis
First an array must be constructed to determine
which variable is included in each equation. Second, for
every equation, cross out every row and form an array of
every column that zeros in that equations row. If the
number of rows and columns containing non zero elements
is greater than or equal G 1, the rank condition is
then satisfied.
79


Results
The next step is to assemble and develop a model
file. This model file is necessary to complete the
statistical procedures required to solve for a historical
simulation. An important feature of the model file is to
assign a new name for the dependant variables in the set
of behavioral equations. If this were not explicitly
stated, the new estimated values would overwrite the
statistical values, and one would not be able to compare
the estimated values with the actual values to determine
the relative stability or predictive capacity of the
model. Along with the assign commands, the OLS or TSLS
solved behavioral equations are retrieved along with the
identities. Example:
Modi
1: assign c82 pc82 y82 py82 yd82 pyd82 i82 pi82 rshrt prshrt
2: assign gsurp pgsurp infm pinfm gd pgd tas82 ptax82
3: C82 = -66.64763+.2760198*YD82+.7428955*082(-1)-2.2190 56*RSHORT
4: 182 = -13.33755+.8427052*182(-l)+7.508444D-02*RSHORT(-4)+.7737915
*DY82+2.957472D-02*Y82(-1)+.100305*GSURP82
5: RSHORT = -6.620536+.466191*IMFM(-1)-7.277061D-02*MGROW +1.194244D-
02*GSURP82+4.035208D-03*Y82 +{AR (1)=.9193786}
6: INFM = 2.090638+.9480394*INFM(-1)-3.400638D-03* YNAT+ 2.918418D-
03*Y82
7: GSURP = TAX82 G82
8: Y82 = C82 + 182 + G82
9: YD82 = Y82 TAX82
80


TSP is then asked to SOLVE the model file. It is
quite possible that convergence will not be achieved due
to systematic errors. In this example and first
initiative, the set of equations was solved, but the
results were dismal. By dismal, I mean the model failed
to predict in a meaningful capacity. To observe this, a
historical simulation is generated by plotting the actual
series with the estimated series with the names used in
the assign statements, as demonstrated by actual
consumption and predicted or estimated consumption. (See
f igure 1.)
It can be readily observed that the estimated values
have no close correlation with the actual values.
Therefore, this model is not dynamically stable in terms
of the estimated coefficients over the sample
observations and must be re-estimated.
The re-estimation process tediously involves re-
evaluation of each estimated equation in terms of
specification bias for self correcting mechanisms,
omitted variables and any other theoretical problems that
could be suspect in terms of the failure of the solved
model to predict and track.
This process is discussed in detail in regard to
changes initiated in each progressive model file. The
final model file proves to be dynamically stable and is
then shocked for policy experiments.
81


Billions of 1982
Predicted and actual consumption, model 2


The Model files and the results of the estimated
equations are printed in the appendix.
Model Files
This section outlines the progression of models,
from 1 to 10, and the policy experiments conducted.
Given the length of the path, the description of each
individual model concentrates on highlights of the
changes which were initiated to improve the overall
performance of the model, with the end result being a
dynamically stable model to conduct policy experiments.
Given the scope of the changes, it would be impossible to
detail every change initiated with the respecification
process.
As previously observed, the first model failed to
accurately predict and had to be re-estimated. The task
was then to derive a model that would predict reasonably
well in a closed economy. Model 1 begins this arduous
task with an estimation of a consumption function that
would track well in simulations. This rather limited
closed economy model consists of an income identity, a
consumption function that depends on disposable income,
lagged consumption and a short run interest rate. Results
of the historical simulations show that, in this form,
the multiplier effect is too large; fitted income has
more variation than actual income. This led to the
conclusion that some form of a self-correcting mechanism
83


was missing from the income identity. After ponderous
analysis, a tax function was included that depends on
income. This reduced the multiplier and much improved
the tracking of actual to predicted income levels.
MODI
assign xm82 xmf rsreal rsrf c82 cf infm infmf i82 if y82 yf gsurp82
gsf
assign yd82 ydf rshort rf
XM82=349.6379-1.926814D-02*GSURP82-.1010509*Y82-1.230957*RJAP+1.36709
2*RJAP(-1)+2.921676*RJAP(-2)+3.432795*RJAP(-3)+2.900447*RJAP(-4)+1.32
4633*RJAP(-5)-1.294646*RJAP(-6)-4.957392*RJAP(-7)-9.663604*RJAP(-8)+4
.172882*RSREAL+1.948739*RSREAL(-1)+.258465*RSREAL(-2)-.8979416*RSREAL
(-3)-1.52048*RSREAL(-4)-1.609151*RSREAL(-5)-1.163953*RSREAL(-6)-.1848
882*RSREAL(-7)+1.328045*RSREAL(-8)+[AR(l)=.8972247]
RSREAL=-4.682127-2.474371D-02*D(MS2)+3.160313D-03*D(Y82)+2.601799D-03
*Y82+.1219872*D(RSREAL(-1))+7.591134D-03*GSURP82+[AR(1)=.9099202]
C82=-14.47359+.8709172*C82(-1)-2.068937*RSHORT+.137466*YD82
INFM=.1708842+1.002731*INFM(-1)-1.92247D-03*YNAT+1.888027D-03*Y82+[AR
(1)=.2129255,AR(4)=-.4145245]
l82=-57.9709+.2645608*D(Y82)+.2156738*Y82(-1)-1.742228*RL0NG(-4)-6.57
3028*INFM(-4)+.3977222*GSURP82+[AR(1)=.7924696]
y82 = c82 + i82 + g82 + xm82
gsurp82 = tax82 g82
yd82 = y82 tax82
rshort = rsreal + infm
MODI Revised
assign c82 pc82 y82 py82 yd82 pyd82 tax82 ptax82
C82=-66.64763+.2760198*YD82+.7428955*082(-1)-2.219056*RSHORT
y82 = c82 + i82 + g82 + xm82
yd82 = y82 tax82
84


TAX82=-58.83562+.3171215*Y82
The task of model 2 was then to expand on the
foundations of model 1 by including equations for
investment and the short run interest rate. The results
of the simulations showed that, while the accelerator
principle works well in the investment equation in
isolation, it generates a highly volatile predicted
income path when included in a multiple equation
simulation. To adjust for this, the cure adopted was to
i
de-link the accelerator. That is, the investment
equation uses the actual change in income, and not the
change predicted by the model. This is completely ad
hoc, but is rather commonly done. For example, the model
in Chapter 13 of Pindyck and Rubinfeld's econometric text
de-links the accelerator in just this way.
MOD 2
assign c82 pc82 y82 py82 yd82 pyd82 i82 pi82 rshort prshrt gsurp
pgsurp
assign tax82 ptax82
C82=-66.64763+.2760198*YD82+.7428955*082(-1)-2.219056*RSHORT
I82=-13.33755+.8427052*182(-1)+7.508444D-02*RSHORT(-4)+.7737915*DY82+
2.957472D-02*Y82(-1)+.100305*GSURP82
RSHORT=-14.7619-1.259393D-02*MS2+.4292392*GD+2.02857D-03*Y82+1.832628
D-02*GSURP82+.4525977*INFM
y82 = c82 + i82 + g82 + xm82
yd82 = y82 tax82
TAX82=-58.83562+.3171215*Y82
85


Model 3 represents an initial, and failed, attempt
to incorporate money, prices and the self-correcting
mechanism. The short run interest equation in model 2
uses the price level and the rate of inflation. Model 3
estimates inflation from an aggregate supply equation.
Inflation is regressed on actual GNP, potential GNP and
lagged inflation. An identity was added to this model
file to link the price level to inflation as a self-
correcting mechanism. Figure 2 shows how poorly the
model tracked experience. Model 3 is clearly dynamically
unstable and generates cycles that, instead of being
damped, are explosive. The dynamic properties of a model
are generated by the interaction between equations and
are therefore not part of the single-equation estimation
process. This is one of the unavoidable difficulties of
structural model building. However, theory and common
sense can often be used to generate acceptable and stable
models, as was achieved in model 4.
MOD 3
assign c82 pc82 y82 py82 yd82 pyd82 i82 pi82 rshort
prshrt gsurp pgsurp
assign infm pinfm gd pgd
C82=-66.64763+.2760198*YD82+.7428955*082(-1)-2.219056*RSH
ORT
86


.Linking up the self
Real income
4 500
4000 / "Y"
3300
3000
2500- Xy
2000 ' 1 J 1 1 1 1 1 1 1 1 1 1 l 1 1 1 f 1 1 o 1 1 I 1 1 1 1 1 1 1 1 1 i 1 1 i 1 1 1 1 1 I~ 1 1 1 ri 1 l~rT" 1974 1976 1970 1980 1982 1904 1986 1980 |. Fitted Actual
87 Real income
4 300
4 000
3500
3000
2500 L ri 1 1 . ! ! n 1 1 u 1 m Tr- | 1 1 1 1 1 11 1 1 1 1 11 1 1 >rP~rrl 1974 1976 1978 1980 1982 1984 1986 1908
____Fitted ......Actual
correcting mechanism
Moving average inflation
____ Filled
Moving average inflation

Fig .


I82=-13.33755+.8427052*182(-1)+7.508444D-02*RSHORT(-4)+.7737915*DY82+
2.957472D-02*Y82(-1)+.100305*GSURP82
RSHORT=-14.7619-1.259393D-02*MS2+.4292392*GD+2.02857D-03*Y82+1.832628
D-02*GSURP82+.4525977*INFM
INFM=2.090638+.9480394*INFM(-1)-3.400638D-03*YNAT+2.918418D-03*Y82
gsurp = tax82 g82
y82 = c82 + i82 + g82 + xm82
yd82 = y82 tax82
gd = gd(-l)*(1+.01*infm)
TAX82=-58.83562+.3171215*Y82
Careful examination of model 3 reveals the problem
with the dynamic stability of the estimated equations.
It was discovered that the inflation rate was allowed to
affect other variables only through its impact on the
short-term interest rate. The short run interest
equation, in turn, specified a complicated interaction
between price levels inflation and the money supply that
was difficult to interpret in its relationship to the
investment and consumption functions. The basic
procedure adopted by model 4 is to simplify by
respecifying the short term interest equation as a
function of money growth and inflation rather than money
stock, inflation and the price level. This dramatically
improved the dynamic path as the accompanying figure
shows.
88


MOD 4
assign c82 pc82 y82 py82 yd82 pyd82 i82 pi82 rshort prshrt gsurp
pgsurp
assign infm pinfm gd pgd tax82 ptax82
C82=-66.64763+.2760198*YD82+.7428955*082(-1)-2.219056*RSHORT
I82=-13.33755+.8427052*182(-1)+7.508444D-02*RSHORT(-4)+.7737915*DY82+
2.957472D-02*Y82(-1)+.100305*GSURP82
RSHORT=-6.620536+.466191*INFM-7.277061D-02*MGROW+1.194244D-02*GSURP82
+4.035208D-03*Y82+[AR(1)=.9193786]
INFM=2.090638+.9480394*INFM(-1)-3.400638D-03*YNAT+2.918418D-03*Y82
gsurp = tax82 g82
y82 = c82 + i82 + g82 + xm82
yd82 = y82 tax82
TAX82=-58.83562+.3171215*Y82
Model 5 then attempts to open up the economy by
adding in equations for the exchange rate and the trade
balance. The exchange rate is taken to be a function of
current and lagged interest rates and inflation in both
the U.S. and Japan. Polynomial distributed lags were
used to estimate the lag functions in an attempt to
accurately model the decreasing influences of inflation
and interest as the time path for adjustment increased.
Japan was used because its domination as a trading
partner in the domestic economy thereby affecting
principle capital flows in response to interest rate
differentials. The trade balance is then modeled as a
function of lagged income and a polynomial distributed
89


lag of the exchange rate to allow for J-curve effects.
It became evident that the self-correcting mechanism
needed additional work, as early runs were again
excessively volatile. After several attempts at
respecification of the net export equation, what seemed
to remedy this situation was to incorporate polynomial
distributed lags for the effects of inflation and money
growth on the short-term interest rate, which then
created more channels for the self-correcting mechanism
to operate through and helped stabilize the model.
mod 5
assign c82 pc82 y82 py82 yd82 pyd82 i82 pi82 rshort prshrt
assign infm pinfm gd pgd tax82 ptax82 exr pexr xm82 pxm82
C82=-66.64763+.2760198*YD82+.7428955*082(-1)-2.219056*RSHORT
I82=-13.33755+.8427052*182(-1)+7.508444D-02*RSHORT(-4)+.7737915*DY82+
2.957472D-02*Y82(-1)+.100305*GSURP82
INFM=2.090638+.9480394*INFM(-1)-3.400638D-03*YNAT+2.918418D-03*Y82
y82 = c82 + i82 + g82 + xm82
yd82 = y82 tax82
TAX82=-58.83562+.3171215*Y82
XM82=409.7845-2.111281D-02*GSURP82-8.173769D-02*Y82(-l)-9.000476D-02*
EXR-.1745888*EXR(-1)-.2825665*EXR(-2)-.413938*EXR(-3)-.5687033*EXR(-4
)+[AR(l)=.9108496]
EXR=85.85307+3.847202D-02*GSURP82+.8335189*RSHORT+l.136025*RSHORT{-1)
+1.116647*RSHORT(-2)+.7753842*RSH0RT(-3)+.1122358*RSHORT(-4>+.9184017
*RJAP+5.177969D-02*RJAP(-1)+3.784756D-02*RJAP(-2)+.8766053*RJAP(-3)+2
.568053*RJAP(-4)-4.106912*INFM-1.521336*INFM(-1)-.2812854*INFM(-2)-.3
90


867594*INFM(-3)-1.837758*INFM(-4)+.2369375*INFJPM-.2514952*INFJPM(-1)
297553*INFJPM(-2)+9.876409D-02*INFJPM(-3)+.9374562*INFJPM(-4)+[AR(1
>=.9447558]
RSH0RT=-16.51714+1.263791D-02*GSURP82+5.769123D-03*Y82+.3406153*INFM+
.3306484*INFM(-1)+.2477187*INFM(-2)+9.182609D-02*INFM(-3)-.1370293*IN
FM(-4)-3.2424D-02 *MGR0W+5.446811D-02 *MGROW(1)+8.386596D-02 *MGROW(-2)
+5.576954D-02*MGROW(-3)-2.982116D-02*MGROW(-4)+[AR(1)=.9191322]
Model 6 addresses a problem caused by improvements
in the self-correcting mechanism. While the changes in
models 4 and 5 improved the overall performance, the
performance of the investment equation deteriorated. To
correct for this, it was necessary to re-evaluate the
investment equation, and it was noted that the
distinction between changes in the real and nominal
inflation interest rate had not been made. For this
reason, the lagged value of inflation was included in the
regression. Finally a model that seemed dynamically
stable in terms of the estimated coefficients and
historical simulations over the observation period was
achieved.
MOD 6
assign c82 pc82 y82 py82 yd82 pyd82 i82 pi82 rshort prshrt
assign infm pinfm tax82 ptax82 exr pexr xm82 pxm82
y82 = c82 + i82 + g82 + xm82
yd82 = y82 tax82
C82=-66.64763+.2760198*YD82+.7428955*C82(-1)-2.219056*RSHORT
91


182=29.13634+.7749497*182(-l)+.8991293*RSHORT(-4)+.7343696+DY82+3.388
661D-02*Y82(-1)-3.191656*INFM(-4)+.15215*GSURP82
INFM=2.183782+.9458795*INFM(-1)-3.450473D-03*YNAT+2.94567D-03*Y82
TAX82=-58.83562+.3171215*Y82
XM82=403.4335-1.781171D-02*6SURP82-7.958044D-02*Y82(-1)-.1529224*EXR-
.1736815*EXR(-1)-.2509553*EXR(-2)-.3847439*EXR(-3)-.5750471*EXR(-4)+[
AR(1)=.9069217]
EXR=83.18916+3.997372D-02*GSURP82+.672585l*RSHORT+l.036852*RSHORT(-l)
+1.068232*RSHORT(-2)+.7667263*RSHORT(-3)+.1323326*RSHORT(-4)+.9611668
*RJAP+.1978867*RJAP(-1)+.2202659*RJAP(-2)+1.028304*RJAP(-3)+2.622001*
RJAP(-4)-4.432264*INFM-1.329161*INFM(-1)+.1112609*INFM(-2)-.110997*IN
FM(-3)-1.995935*INFM(-4)+2.254357D-02*INFJPM-.1994674*INFJPM(-l)-.152
9637*INFJPM(-2)+.1620548*INFJPM(-3)+.7455881*INFJPM(-4)+[AR(1)=.92552
53]
RSHORT=16.51714+1.263791D-02*GSURP82+5.769123D-03*Y82+.3406153*INFM+
.3306484*INFM(-1)+.2477187*INFM(-2)+9.182609D-02*INFM(-3).1370293*IN
FM(-4)-3.2424D-02*MGROW+5.446811D-02 *MGROW(-1)+8.386596D-02 *MGROW(-2)
+5.576954D-02*MGROW(-3)-2.982116D-02*MGROW(-4)+[AR(1)=.9191322]
Model 7 addressed simultaneous equation bias through
TSLS. An initial strategy of always using TSLS methods
was abandoned early, because the sensitivity of the
estimated coefficients were impacted by the instrument
list, and the instrument list kept changing as the model
evolved. The decision was made to develop the model and
then correct for simultaneous equation bias. Figure 3
compares the fitted values for ordinary and TSLS. The
92