PLACE IN RETURN BOX to remove this checkout from you! recon.
TO AVOID FINES return on or More due due.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU Is An Affirmdive Action/Equal Opponunlly Inflation
Walla-9.1

 

 

 

THE IMPACTS OF 0.8. FISCAL POLICIES ON

AGRICULTURE

BY

Young Chan Choe

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements

for the degree of

DOCTOR OF PHILOSOPHY

Department of Agricultural Economics

1991

ABSTRACT
THE IHPACTS O! 0.8. I'ISCAL POLICIES ON AGRICULTURE
BY
Young Chan Choe

Macroeconomists have different opinions on how fiscal
policy affects the economy in general. Not surprisingly, these
different views have also resulted in wide disagreements on
how fiscal.policies affect agriculture. Schuh (1981, 1983) and
Barclay and Tweeten (1988) , following the Keynesian
hypothesis, have argued that an increase in the federal
deficit causes unfavorable conditions for the farm economy by
decreasing farm prices. On the other hand, Belongia and Stone
(1985) and Batten and Belongia (1986) have rejected any
possible connection between the federal deficit and farm
prices, based on a New Classical macroeconomic model. Applying
the neo-Keynesian differential price adjustment, Rausser
(1985) and Rausser, Chalfant, Love and Stamoulis (1986) have
argued that the federal deficit, due to sticky industrial
prices, has the same short run impact on the farm economy as
tight monetary policy, which implies a decrease in farm
prices. Just and. Chambers (1987) also applied. the neo-
Keynesian hypothesis, but considered farm prices as sticky as
a result of price supports.

This study attempts to resolve these differing views with
a detailed empirical analysis of the effects of U.S. fiscal

policies on agriculture. Minimally restricted time series

Young Chen Choe
models, in the form of vector autoregressions and error
correction models, are used so that alternative theories of
how agriculture responds to fiscal policies can be tested
rather than imposed a priori. Results support the arguments
of Rausser (1985) and others that the federal deficit
decreases farm prices in the short run without affecting
industrial.prices. Thus, the farm economy suffers,a cost.price
squeeze in the short run. However, farm prices move back to
their long run equilibrium price level after an initial fiscal
shock, reaching equilibrium after about two or three years.
Thus, no long run changes in the relative position between
farm prices and industrial prices are detected. The short run
impact of the federal deficit occurs mainly through its
effects on interest rates and the exchange rate.

Results from simulating the model over a five year period
suggest that spending reductions are the most desirable form
of deficit reduction from the general macroeconomic
perspective, as spending reductions have little impact on
total output or the general price level. A tax increase
results in a slump in both the macroeconomy and the farm
sector. Monetization of the deficit favors the farm sector
initially because there is a short run increase in farm
prices. However, monetization does not affect the relative
price of farm and industrial goods in the long run, and
induces inflation and a decrease in real output after its

initial expansionary effects.

ACKNOWLEDGEMENTS

I wish to express my sincere appreciation to my major
professor, Dr. Robert J. Myers, for his valuable guidance and
counseling throuout author's graduate program and thesis
writing. My special thanks also go to Professors Glenn L.
Johnson, Lindon J. Robison, James F. Oehmke, and Richard T.
Baillie, for their interest, ideas, and comments on the
research. I would like to extend my gratitude to the
Department of Agricultural Economics for continued financial
support. Extra special thanks go to my parents for providing

this opportunity and my wife, Hye Sun Lee, for her love.

iv

TABLE OF CONTENTS

Chapter
I INTRODUCTION
II ISSUES AND CONTROVERSIES

III

IV

1. Fiscal Policies in the Keynesian Paradigm

2. Fiscal Policies in the New Classical Paradigm
3. Fiscal Policies in the Neo-Keynesian Paradigm
4. Empirical Tests of the Effects of Fiscal Policy
5. Fiscal Policies and Agriculture

MODELS AND RESEARCH METHODS

1. Models Used in Empirical Macroeconomics

Page

12
18
24
28

33

33

1.1. The Traditional Simultaneous Equations Model 33

1.2. The Rational Expectations Model
1.3. Unconstrained Vector Autoregressions

2. Extensions of the VAR Approach
2.1. Identifying Contemporaneous Correlatons
2.2. Considerations of Structural Change
2.3. Unit Roots
2.4. Cointegration and Error Correction Models
2.5. Restrictions on Cointegrating Vectors
2.6. Error Regressive Models

3. Methods Used in This Study

PRELIMINARY ANALYSIS

1. Variables and Data

2. Unit Root Tests

3. Cointegration Tests

4. Money Market Equilibrium
5. Exchange Rate Equilibrium

A VAR MODEL OF FISCAL POLICY IMPACTS
ON AGRICULTURE

1. Reduced Form Specification
2 Structural Form Identification

36
38
42
42
45
47
53
58
6O
63

65

65
66
76
80
88

98

98
108

vi

3. Dynamic Responses and

Fore
4. Sens
4.1.
4.2.
4.3.
5. Impl

cast Error Variance Decomposition
itivity Analysis

Structural Identification

Lag Order

Model Specification

ications from the Analysis

VI COMPARISON OF DEFICIT REDUCTION POLICIES

1. Impact of Deficit Reduction Options
on Agriculture

2 P011

cy Simulations

VII SUMMARY AND CONCLUSION

APPENDIX A.
B.

C.
D.
E.

BIBLIOGRAPHY

Data Sources

Definition and Critical Values for
Phillips-Perron Test Statistics
Critical Values for Schmidt-Phillips
Test Statistics

Critical Values for Johansen-Juselius
Cointegration Test Statistics

LM Test of Contemporaneous Correlation

112
116
116
120
120
126
128
128
132

137

141

143

146

147
148

149

LIST OF TABLES

TABLES
1 Results of DF Test for One Unit Root
2 Results of DF Test for Two Unit Roots
3 Results of DP Test for Three Unit Roots
4 Results of DP Test for Two Unit Roots
5 Results of PP Test for One Unit Root
6 Results of SP Test for One Unit Root
7 Autocorrelations of Level Series
8 Autocorrelations of First Differenced Series
9 Lag Selection for AM, AR, AY, AP, AX, and AF
10 JJ Test Results for M, R, Y, P, X, and F
11 Lag Selection for AM, AY, and AR
12 JJ Test Results for M, Y, and R
13 Eigenvalues and Eigenvectors for M, Y, and R
14 Test Results for the Velocity Restriction
15 Lag Selection for AX, AP, and AF
16 JJ Test Results for X, P, and F
17 Eigenvalues and Eigenvectors for X, P, and F
18 Test Results for the Exchange Rate Restriction
19 Lag Selection for AD, AZ“, and AZx
20 Test Results for Structural Change

vii

Page
68
69
71
72
73
74
75
77
78
79
82
83
84
86
92
93
94
95
99

100

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

viii
Historical Decomposition of D
Historical Decomposition of ZM
Historical Decomposition of zx
JJ Test Results for D, ZM' X, P, and F

Eigenvalues and Eigenvectors for D, ZM' X, P,
and F

Lag Selection for AD, AZM, AX, AP, and AF
Summary Statistics for Five Variable ERM

Estimates of Cotemporaneous Parameters in
D-P-X-F-zM Recursive Structure

Decomposition of Forecast Error Variance

Estimates of Cotemporaneous Parameters in
a Simultaneous Structure

JJ Test Results for D, M, R, Y, P, X, and F

Eigenvalues and Eigenvectors for D, M, R, Y, P, X,
and F

JJ Test Results for G, T, M, R, Y, P, X, and F

Eigenvalues and Eigenvectors for G, T, M, R, Y, P,
X, and F

Data Sources
Critical Values for PP Test Statistics
Critical Values for SP Test Statistics

Critical Values for JJ Test Statistics

102

103

104

105

106

107

109

111

115

119

121

122

129

130

141

145

146

147

LIST OF FIGURES

Deficit and Money Supply

Farm Price, Export, and Income

Fiscal Policy in the Keynesian Paradigm

Fiscal Policy in the New Classical Paradigm

Interest Rate, Exchange Rate, and Inflation rate

Plots of Each Series

Actual and Equilibrium Real Money Balances

Actual and Equilibrium Farm Prices

FIGURES

1 Federal
2

3

4

5

6

7

8

9 Impulse
10 Impulse
11 Impulse
12 Impulse
13 Effects
14 Effects
15 Effects

Response Function by ERM with D-P-X-F-ZM
Response Function by ERM with F-D-P-X-ZM
Response Function by Seven Variable ECM

Response Function by Eight Variable ECM

of Spending Reduction

of Tax Increase

of Monetization

ix

Page

10
14
25
67
87
97
113
117
124
131
133
134

135

CHAPTER I

INTRODUCTION

Agricultural economists.havejpaid.considerable attention
to the macroeconomics of agriculture since Schuh (1974) first
considered the exchange rate as an important factor affecting
the farm.economyu Most attention has focused on.exchange rates
and the effects of monetary policy on agriculture (Shei and
Thompson, 1981; Chambers and Just, 1981; Belongia and King,
1983; Rausser, 1985; Orden, 1983, 1986). Recently, however,
the U.S. economy has experienced large federal budget deficits
and a number of agricultural economists have tried to relate
the depressed farm economy in the early 19805 to these
deficits (Schuh, 1981, 1983, 1984b; Rausser, 1985; Rausser,
Chalfant, Love, and Stamoulis, 1986; Belongia.and Stone, 1985;
Batten and Belongia, 1986; Just and Chambers, 1987).

Schuh (1981, 1983, 1984a, 1984b) discussed some potential
effects of fiscal policy on agriculture. The federal deficit,
in his view, tends to increase interest rates and hence the
exchange rate. Agriculture, an export oriented sector, will
suffer from the resulting reduction in exports, prices, and
income. Schuh's conjecture brought immediate responses from
the agricultural economics profession. Barclay and TWeeten

(1988) supported Schuh's conjecture. Their simulation analysis

2

resulted in a negative impact of the federal deficit on farm
exports and. prices ‘through increased. interest. rates and
exchange rates. Rausser, Chalfant, Love, and Stamoulis (1986)
supported the conjecture only in the short run, with no long
run impact from fiscal policy. On the other hand, Just and
Chambers (1987) claimed in their theoretical work that the
federal deficit stimulates the farm economy. However, Belongia
and Stone (1985) , and Batten and Belongia ( 1986) found no
evidence of causality from federal deficit changes to
agriculture in their empirical work.

Controversy over the effects of federal budget deficits
on agriculture implies varying policy recommendations under
the current huge deficit regime. Just and Chambers (1987)
argued.for’a reduction of government spending in other sectors
as the most favorable approach to agriculture for reducing the
federal deficit. Alternatives were a monetary expansion and
a tax increase. Belongia and Stone (1986) argued that focusing
attention on deficit reduction measures diverts attention from
more fundmental changes required in farm commodity programs.
However, Rausser, Chalfant, Love, and Stamoulis (1986) argued
for the dominance of macroeconomic policies over farm policies
in influencing the farm.economy in short run. The.role of farm
policies is confined only to reducing instability in farm
prices and not to providing incentives for overallocation of
resources to agricultural production. They suggest that

frequent use of sectoral policies only brings more instability

3
in the farm economy. Barclay and Tweeten (1988) defined
optimal policy as a balanced budget and payment position,
keeping no interest rate differentials between foreign and
domestic economies.

This study considers some issues surrounding the impact
of fiscal policy on agriculture. The primary objective is to
test empirically the different explanations of how the federal
deficit affects agriculture. Attention will be given to the
significance and the persistence of federal deficit changes
on farm prices. The mechanisms through which the impact of
federal deficit changes get transferred to agriculture will
also be considered.

The second objective of the study is to provide broad
guidelines for policy under current economic conditions. The
effects of changes in government spending, taxation, and
monetization of the deficit on agriculture are examined. The
performance of different deficit reduction policies will be
considered against the alternative of maintaining the status
quo.

Three different models, the simultaneous equations model
(SEM), the rational expectations model (REM), and the vector
autoregression model (VAR), have been used in empirical
macroeconomic policy analysis. In this paper, the VAR model
pioneered by Sims (1980) is used rather than a SEM or a REM
model. VAR models employ only minimal restrictions on the

dynamics of the variables being investigated, where other

4
models incorporate large numbers of overidentifying
restrictions on the model structure. Tests of stationarity,
cointegration, and structural changes will be applied to
selected data to assist in specification of the reduced form
VAR model. Then, the relationS‘ between contemporaneous
variables will be used to identify the structural form VAR.

Given certain identification restrictions, impulse
response functions and decompositions of forecast error
variance can be used to identify how fiscal policy affects
agriculture. Forecasts under the current economic structure
will provide a base projection for the time path of each
variable in the system. Finally, alternative time paths for
each variable under different policy scenarios will be
simulated and compared to the base projection.

In the next chapter, the role of fiscal policies in
agriculture ‘will be stressed by looking at. key summary
statistics. Then, the major issues and controversies
surrounding fiscal policies and agriculture will be discussed
by reviewing the current literature. Chapter III provides
details of the :methods employed. herein. In chapter IV,
variables are defined and stationarity of data. will be
checked. In chapter V, an empirical model is fitted to data
for the selected variables and the impact of fiscal policy on
agriculture will be traced out. Given these estimation
results, chapter VI identifies alternative policy measures for

reducing the current federal budget deficit and compares them

5
through simulation analysis. Chapter VII will provide a brief
summary of findings and conclude the study’ with a few

suggestions for future research.

CHAPTER II

ISSUES AND CONTROVERSIES

Budget deficits in the United States became a major issue
for economists when they rose to an average of $206.7 billion
per year between 1982 and 1986. Deficits, which fluctuated
through the 19705, suddenly grew to alarming levels in the
19805 (see Figure 1). At the same time, agriculture
experienced a prolonged recession. The price of farm products
decreased by an average of 2% per year in the early 19805 and
real net farm income declined by 15% per year. Furthermore,
the value of farm exports fell by an average 8% per year
between 1980 and 1986 (see Figure 2).

Recently, a number of agricultural economists have tried
to relate the depressed farm economy in the early 19805 to the
growing federal budget deficits (Rausser, 1985; Belongia and
Stone, 1985; Just and Chambers, 1987). However, their
explanations are controversial and little has been done to
test alternative theories empirically. Before turning to
empirical tests, however, it is important to get a better
perspective on the theoretical relationship between fiscal
policies and the farm economy. To’this end, the current status
of macroeconomic theories on fiscal policy is first
summarized. Then, the impact of fiscal policies on agriculture
will be discussed within the context of these macroeconomic

theories.

 

7005-. ‘ /

eoo-l- /

ZOO '1

\.
\
\
\
3

 

   
 

 

4» ;—5T —. 1t ; ; .

1970 1975 1980 1985

 

‘ _A A A

A ‘ A... A 9‘—
::}:"_r T_ '

Federal Deficits (Bil. of Dollars)
_ _ Money Supply (Bil. of Dollars)

 

Figure 1. Federal Deficit and Money Supply

 

 

700-1-

e004-

100-1)-

 

    

. U

1980

Federal Deficits (Bil. of Dollars)
_ _ _ Money Supply (Bil. of Dollars)

 

Figure 1. Federal Deficit and Money Supply

1985

 

 

70

so"

401-

301?

204-

ioir

 

 

 

 

L_ l L L k A L L __L__ .1 J L. .1_ L L P
r

L
V '— V r U 1 U iii

1975 1980 1985

1970
Farm Prices (Index of Prices Received by
Farmers, 1977=10)

__ __ Agricultural Exports (Mil. of Dollars)
_{_ Real Farm Income (Bil. of 1982 Dollars)

Figure 2. Farm Price, Export, and Income

9

1. Fiscal Policy in the Keynesian Paradigm

In a Keynesian economy, an expansionary fiscal policy
(increase in government expenditure or reduction in taxes)
shifts the IS curve out from ISO to IS1 (Figure 3). This
causes equilibrium output to increase from Yo to Y1 and the
price level to increase from pc to p1. The real money stock,
Ms/P, decreases because of higher prices causing the LM curve
to move up to LMl. The interest rate increases from ro to r2
and investment drops, pulling the output level back to Y2.
Overall, the output, price, and interest rate of the economy
are all increased in both nominal and real terms. Tight fiscal
policy (decrease in government expenditure or increase in
taxes) will move each of these variables in the opposite
direction.

While increasing government expenditure can help to
produce an initial economic boom, different ways of financing
the expenditure have different effects in a Keynesian economy.
Following Branson (1979) and Canto and Rapp (1982), the
federal government has three distinct alternatives to finance
increased expenditure. First, the government can increase
taxes. A tax increase obviously will offset the initial
stimulative effect of government spending by shifting the IS
curve back towards the origin.

Second, the government expenditure can be financed by

selling bonds to the public. The resulting increase in federal

10

 

 

 

 

I I I
Y0 Y2 Y1

Goods and Money Markets

 

 

\I I l/ L‘.
Li
“a
w. "”L:
we
L1
I
S
Pea
. --a-o
is D.
De

 

 

 

Goods Market

Figure 3. Fiscal Policy in the Keynesian Paradigm

11

debt bids up interest rates, thus choking private investment
and reducing income. The high interest rates also will induce
inflows of foreign capital which will bid up the value of
dollar. The strong dollar makes exports more expensive and
imports cheaper. Exports will fall and imports will expand.
High interest rates also induce a release of commodity stocks
to the market, because interest costs are an important
component of the total costs of carrying stocks. Thus, the
supply of commodities shifts to the right. All of these
effects partially offset the stimulative effect of the IS
shift caused by increased government spending.

Third, the government can finance the deficit by selling
bonds to the monetary authority. The deficit in this case is
financed by additional money creation which shifts the LM
curve out. This causes a decrease in interest rates, an
increase in income, and an increase in the price level. In
this case, secondary effects reinforce the initial
expansionary effect on the economy. This method is frequently
referred to as "monetization" of the deficit.

Although the effect of increased government expenditure
with monetization is always expansionary in a Keynesian
economy, it is not always true that increased government
expenditure will be expansionary without monetization. Tobin
(1969) argued that non-monetized deficits are still
expansionary, because the magnitude of the initial

expansionary effect is greater than the magnitude of the

12

offsetting secondary contractionary effect. Brunner and
Meltzer (1972) and Blinder and Solow (1973) argued not only
that debt financed government expenditure is expansionary, but
also that it is more expansionary than monetization of the
expenditure. They show that higher incomes offset higher
interest payments and hence stabilize the economy. However,
Silber (1970) argues for the reverse case that non-monetized
deficits are not expansionary.

The Keynesian paradigm is based on the disequilibrium
assumption that markets do not always clear immediately, due
to stickiness or slow adjustment processes in the labor
market. Both the size of the budget and the method of
financing the expenditure affect real output.and prices in the
economy (Tobin, 1969; Brunner and Meltzer, 1972; Tobin and
Buiter, 1974; Blinder and Solow, 1973; CEASM, 1978; Branson,

1979; Feldstein, 1982).

2. Fiscal Policies in the New Classical Paradigm

Tatom (1985) described ex-ante crowding out and the
permanent income hypothesis as two important theoretical
considerations from classical economics. Carlson and Spencer
(1975) defined crowding out as a steady state government
spending multiplier (changes of nominal income by a unit
change in government spending given a constant money supply)

of near zero. Canto and Lapp(1982) defined crowding out as a

13
government expenditure multiplier of less than one and full
crowding out as a multiplier of zero. If a fiscal policy
action is largely offset by direct private sector responses,
it fails to stimulate total economic activity. In this case,
the private sector is said to have been "crowded out" by the
government action.

Three different explanations for crowding out phenomena
are distinguished by Blinder and Solow (1973) . First, crowding
out occurs as the LM curve moves back toward the origin after
an expansionary fiscal policy. As shown in Figure 4, the IS
curve shifts up to IS1 from ISo with an expansionary fiscal
policy; At.the initial price level PO' demand for output rises
to Y1. This is shown as a shift in the demand curve to D1 in
the goods market, generating an excess demand gap of Y1 - Yo
which forces price to rise. However the price increase is
fully anticipated by agents and output stays at the natural
rate, Yo. In the financial market, the price increase reduces
the real level of money stock (Ms/P) and, hence, moves the LM
curve up to LMl. This raises interest rates further to r2 and
reduces investment and consumption, which result in a decline
in income to the original level. Overall, the expansionary
fiscal policy actions are offset or crowded out. The policy
affects.nominal variables but not real variables, thus leaving
the equilibrium of the economy unaffected. The general price
level and interest rates are increased by the same proportion.

But real income and real interest rates remain at the original

14

 

 

 

I I
'0 '1

Goods and Money Markets
L:
I l

l /l

 

 

 

Labor Market

 

 

 

g I

P3 \‘\\ '

\\
_ I ‘IDI

R

\r‘ p.

I

YO Y1

Goods Market

Figure 4. Fiscal Policy in the New Classical Paradigm

15
level. By the same reasoning, tight fiscal policy has no
effect on real variables either.

Second, crowding out occurs as the IS curve moves back
toward the origin. As shown in Figure 4, the shift in the IS
curve from ISo to 181 with an expansionary fiscal policy
raises the interest rate. Private investment will be decreased
until there is no more upward pressure on the interest rate.
The stimulative effect of the expansionary fiscal action is
exactly offset by the decreased investment and the IS curve
moves back to the original level, ISO.

Finally, crowding out occurs as government policy actions
are largely offset by direct private sector responses before
they can affect the economy. Tatom (1985) believed that this
type of crowding out can occur regardless of the methods of
financing the government expenditure. A debt financed
government expenditure induces an offsetting change in private
investment, and a tax financed expenditure has a displacement
effect on private consumption. Therefore, fiscal policy
doesn't change the path of the economy. Aggregate demand
(income), interest rate, and the price level are not affected
by the fiscal action (p.10, Tatom, 1985).

The permanent income hypothesis defines consumption
expenditures to be a function of permanent income, which is
a constant fraction of current assets and expected total
future earnings discounted back to the current time (Friedman,

1957). Under the permanent income hypothesis, variations in

16

personal saving' have a large cyclical component. due to
transitory income changes which don't have any effect on an
agent's consumption.plan. The permanent income hypothesis can
also be applied to the government budget constraint, which
indicates that the jpresent value of current and future
government expenditures must equal the present value of taxes.
Debt financed government expenditure must be paid, if not at
present, then sometime in the future. Households perceive and
discount the increased government borrowing as a future tax
liability. Any transitory increase in income caused by a tax
cut will be saved to pay future taxes. Thus, any change in the
tax scheme or the government debt is offset by an equal change
in private saving, leaving agents' consumption plans
unchanged. The shifts between taxation and government
borrowing affect the timing of the tax collection and the
components of personal income, but not aggregate wealth. The
method of financing government expenditure is irrelevant to
the real economy in the case of lump sum taxes. This
hypothesis, known as the Ricardian proposition, is concerned
with the ineffectiveness of shifts between taxes and
government borrowing. (Ricardo, 1951; Crouch, 1972; Barro,
1974, 1978a; Boothe and Reid, 1989).

However, there is also another explanation for the effect
of financing government expenditure on the economy: debt
financed government expenditure doesn't affect the real

economy but causes a monetary expansion. This hypothesis is

17

sometimes called "the monetarist paradigm" because of the
emphasis on monetary factors. In a monetarist framework, there
is a tendency for the Federal Reserve to control interest.rate
movements by conducting monetary policy. When government
deficits place upward pressure on interest rates, the Federal
Reserve tries to reduce the effect of the deficit on interest
rates by printing money. The monetary expansion will cause
inflation but not affect the real economy, by the same
arguments discussed in the case of fiscal expansion. The
monetarist paradigm implies that the shifts between the
monetization and nonmonetization of government deficit do not
affect the real economy, but cause a monetary expansion.
(Hamburger and Zwick, 1981; Fusfeld, 1982; Protopapadakis and
Siegel, 1984).

Although the Ricardian hypothesis is only concerned with
the financing decision of the government, some scholars have
related it to the size of the government budget. For example,
Feldstein (1982) argued that the Ricardian equivalence theorem
implies irrelevance of not only the method of financing, but
also the size of government expenditures.

In contrast to the Keynesian paradigm, the classical
paradigm is based on a market clearing (or equilibrium)
assumption with price and wage flexibility. Neither the size
of the budget nor the method of financing government

expenditures affect the real economy.

18

3. Fiscal Policies in the Nee-Keynesian Paradigm

Andrews and Rausser (1986) described the evolution of the
Neo-Keynesian paradigm as follows;
Traditional Keynesians in a quandary to develop a
rival to the natural rate hypothesis, turned to the
fixed-flex price model first proposed by Means and
expanded it to explain how stagflation can be
generated from exogenous supply shocks. These
modifications of the traditional Keynesian sticky

price model have converged into a competing
paradigm known as the Neo-Keynesian school. (p.414)

The main characteristic of the Neo-Keynesian paradigm is the
heterogeneity in the economy. It contains both Walrasian
auction markets (flexible price sector) and nonclearing
customer markets (fixed price sector). Although the reasons
for sticky' prices in. the short run are not completely
understood, some justifications have been proposed based on
the optimizing behavior of agents. Search costs (information
costs) due to imperfect information (Okun, 1975); transaction
costs (management costs) due to price setting and delivery
lags (Blinder, 1982; Carlton, 1978, 1979, 1980); implicit wage
contracts due to the uncertain environment (Taylor, 1979,
1980); anduasymmetric information (Stiglitz, 1984) are‘various
candidates for Neo-Keynesian microfoundations.

The Neo-Keynesian view does not deny money neutrality and
the natural rate:of employment in the long run, but emphasizes

the short run responses to a shock to the economy. Due to the

19

heterogeneity of markets, fiscal policy as well as monetary
policy leads to changes in the relative price between auction
and customer markets, even under rational expectations. The
price of the flexible sector overshoots its long run
equilibrium level while the price of the other sector change
little during the transition period. The rate of temporary
overshooting depends on the size of the auction sector and.the
speed of the adjustment. Real output, employment, and.the rate
of interest will also be affected by the differential price
movement. After the adjustment period, the price of inflexible
sector responds and the price of flexible sector moves back
to its long run equilibrium level (Chambers, 1984; Frankel,
1984; Rausser, 1985; Stamoulis, Chalfant and Rausser, 1985,
1987; Andrews and.Rausser, 1986; Rausser, Chalfant, Love, and
Stamoulis, 1986).

The impact of fiscal policy in the Neo-Keynesian paradigm
can.be contrasted to the results from other paradigms by using

a macroeconomic model. To measure the fiscal policy effect,

let

(1) M-p=¢y-6r (¢,5>0)

and

(2) y =a-Btr-(pe-p)1+uc (B.u>0)

be the equations for LM and IS curves respectively, where M

is the log of the nominal money supply, p is the log of the

20
overall price level, y is the log of total output, r is the
short term nominal interest rate, pe is the log of the
expected price, and G is the fiscal policy action.
Expectations are formed based on the long run equilibrium
paths of economy.
Both monetary and fiscal actions are governed by feed

back rules

(3) G = f(fl) + CG
and
(4) M = 9(9) + 5M:

where n is an information set available at the previous time
period and 6G and EM represent the random part of G and M,
respectively.

There are two different goods in the model, flexible
price goods with the price pf in log form, and fixed price
goods with the price pn in log form. The flexible price goods
are homogenous and storable. Their expected earnings from
speculative storage are assumed to be equal to storage costs

5 plus the interest cost r;

(5) pg-pf=s+rl

and the overall price level is an average of fixed sector

price with weight 1 and flexible sector price with weight

21

1-1;
(6) p = rpn + <1-r>pf.

Substituting equation (2) into equation (1) yields
(7) H ' p = ¢a - ¢B[r - (pe - p)] + ¢uG - 6r.

By substituting in equations (5) and (6) and rearranging,

equation (7) becomes

(8) M - (1-¢B)[Tpn + <1-r)pf1
= ¢a + ¢epe + ¢uc - (¢B+6)<p§ - pf - s).

Taking expectations will generate the long run version

of equation (8):

(9) Me - (1—¢B)[rp§ + (1-r)p§1
= ¢a + ¢Bpe + ¢uce - (¢3 + 6)<p§ - p? - s)

and subtraction of equation (9) from equation (8) yields

(10) (M - Me) - (1-¢B)r<pn - pi) - (1-r+r¢e+6)<pf - p?)
= ¢u(G - Ge).

Rearranging terms in equation (10) gives

22

 

 

 

(11) pt = p: - 1(1-¢B) (pn - pi)
(1-r+r¢8+6)
on
- (G - 6")
(1-r+r¢3+6)
1
+ (M - Me).

(1-r+r¢8+6)

By taking expectation of equations (3) and (4), Ge=f(fl)

and Me=g(n) are obtained. Thus, equation (11) becomes

 

 

 

_ e - r(1-¢B) - e
(1-1+r¢8+6)
¢u
(1-1+r¢8+6)
1
+ EM.

(1-1+1¢B+6)

The derivative of (12) with respect to G8 is then

 

 

 

 

dpf dp? 1(1-¢B) dpn d9:
(13’ “; = -‘a’- "; ' e )
dG dG (1-r+r¢B+6) dG dG
ou des 1 d6M
- + ’
(1-r+r¢8+6) ace (1-r+1¢8+6) dGe

where den/d6e = d¢s(;/dGe = 0 with the rational expectation

23
assumption.
If the fixed sector’prices do not respond to the expected

fiscal shock (dpn/dGe = 0), then equation (13) reduces to

dpf dp? r(l-¢B) 6p?
(14) — (1 'I' > ——e
dGe dGe 1-r+1¢8+6 dG

 

I

assuming both prices go back to their long run equilibrium
path after the short period of adjustment (dp‘E/dGe = dpfi/dGe
= dpe/dGe). Therefore, flexible sector prices overshoot their
long run equilibrium path during the adjustment period.
Equation (13) shows that if one price does not deviate from
its long run equilibrium path, then the other price would keep
its long run equilibrium path as well. In this case, fiscal
policy should be neutral to all sectors of the economy. The
more flexible sectors (the smaller is r) the economy has, the
less overshooting occurs. With 1 = 0, no overshooting occurs
and the prices are always in equilibrium(the Neoclassical
economy). The more fixed sectors (the bigger is r) the economy
has, the more overshooting occurs and.the longer it lasts. The
effect of an expected government policy shock on any
particular sector will depend on the flexibility of the
economy (the value of r) and will be left as an empirical
question. Rausser (1985) views the agricultural sector as

flexible, and Just and Chambers (1987) view it as fixed.

24

4. Empirical Tests of the Effects of Fiscal Policy

The contrast between different schools of thought in
explaining the effects of fiscal policy on the economy can be
addressed by looking at the experience during the 19705 and
19805. From 1981 to 1986, the U.S. inflation rate declined by
22.1% per year while the rate of interest remained very high.
The U.S. exchange rate rose consistently by an average of
10.6% per year between 1980 and 1985 (see Figure 5). U.S.
federal budget deficits grew over 28.3% per year on average
between 1979 and 1986 (see Figure 1). If we look only at
annual statistics for those years, the association between
deficits and macroeconomic variables appears to strongly
support the Keynesian model.

Yet, when the 19605 and the 19705 are examined, a
different picture emerges. Between 1969 and 1972, U.S. federal
deficits increased continuously, from a $3.2 billion surplus
to $23.4 billion deficit. During that time, U.S. exchange
rates and interest rates fell consistently. The interest rate
also remained quite low averaging 5.39% per year. The
inflation rate remained stable over the period except for a
short fall in 1972. Moreover, no particular pattern is found
in the relationship between the federal deficit and other
macroeconomic variables between 1972 and 1979. Thus, it is
difficult to draw conclusions based on the annual statistics.

Such a narrow focus necessarily raises questions about the

25

 

20
181-
18*)-
14*

’\ / \
12pr

10+

 

 

 

L L l l A L l L
7*1— I r '71

1975

 

1970

Interest Rate (3 Month T-Bill Rate)
__ __ Exchange Rate (Multilateral Trade Weighted Value
of the U.S. Dollars, 1973:10)

_ _ Inflation Rate (Percentage Change from Preceding
GNP Implicit Price Deflator)

Figure 5. Interest Rate, Exchange Rate, and Inflation rate

26
generality of the presumed relationships.

Numerous scholars have tried to test empirically the
different propositions. Gramlich (1971) and Framm and Klein
(1973), in support of the Keynesian view, found a significant
impact of government expenditure increases on real income. On
the other hand, Keran (1969) and Batten and Thoronton (1984)
supported the classical view of crowding out and found no
impact of government spending changes on real income. Carlson
(1982) supported the Ricardian hypothesis and found neither
government expenditures nor deficits affect income, even in
nominal terms

Similar controversies were found in the relationship
between the federal budget deficit and financial variables
(interest rates, the exchange rate, and inflation). Feldstein
and Eckstein (1970), Makin (1983), and Cohen and Clark (1984)
supported the Keynesian view that the budget deficit has a
positive impact on real interest rates. Frankel (1984), in
support of Neo-Keynesian view, found a positive impact in the
short run. However, Belongia and Stone (1985) didn't find any
relationship between real interest rates and the federal
deficit, supporting the classical view. Canto and Rapp (1982)
didn't find any relationship even with the nominal interest
rate. Carlson (1982) and Evans (1986, 1987) supported the
Ricardian hypothesis and found neither government expenditures
nor budget deficits cause changes in interest rates, either

in nominal or' real terms. Plosser (1982) supported the

27
Ricardian view but rejected ex-ante crowding out. He found
that federal budget deficits have no impact on nominal
interest rates, but that a balanced budget increase has
impacts on interests rate in both nominal and real terms.

In the case of exchange rates and the deficit, Hutchinson
and Throop (1985) found a positive relationship and Cohen and
Clark (1984) found a negative relationship. However, Belongia
and Stone (1985) and Batten and Belongia (1986) didn't find
any relationship.

In the case of inflation (or money supply) and the
deficit, Rausser (1985) estimated a temporary price decrease
in the flexible good sector caused by a non-monetized deficit.
Niskanen (1978), Dornbush and Fisher (1981), McMillin and
Beard (1982), and Protopapadakis and Siegel (1984) all
supported the Ricardian equivalence theorem. The government
deficit appeared not to have any impact on inflation (or money
supply) in their estimation. Barro (1977, 1978a, 1978b) also
supports the Ricardian proposition that the budget deficit has
no impact on the real economy, but found that government
expenditure increases stimulate money growth and inflation.
Barr (1979) supported the monetarist view by finding a
positive relationship between the general price level and
budget deficits. Hamburger and Zwick (1981) also found that
both government expenditure and the deficit are responsible
for monetary expansion. Hamburger and Zwick (1982) and Allen

and Smith (1983) supported the monetarist hypothesis by

28
finding an impact of the budget deficit on money supply, but
didn't find any causality from government spending to the
money supply.

The conflicting empirical evidence makes the issue of
fiscal policy impacts on macroeconomic variables an unresolved
puzzle. This tends to place macroeconomics in a state of
disarray (Grossman, 1980; Fusfeld, 1982; Barro, 1984). Bell
and Kristol (1981) refer to this disarray as a "crisis in
economic theory". Not surprisingly, these different
macroeconomic theories have resulted in an wide disagreement

in studies on how macroeconomic policies affect agriculture.

5. Fiscal Policies and Agriculture

Traditionally, agricultural economists.have.devoted most
of their attention to microeconomic issues because the
classical economic paradigm applies to agricultural markets
better than anywhere else (Frankel, 1984). However, attention
has gradually turned to macroeconomic issues after Schuh
( 1974) argued for the important role of a macroeconomic
variable, the exchange rate, in economic fluctuations in
agriculture. Most of the attention so far has focused on
monetary policy impacts on agriculture through macroeconomic
variables, such as the inflation rate, interest.rates, and.the
exchange rate (Shei and Thompson, 1981; Chambers and Just,

1981; Rausser, 1985; Orden, 1986).

29

Schuh (1981) turned his attention to another dimension
of macroeconomic policy, namely fiscal policy. Schuh (1983)
in his testimony to the U.S. Congress argued that the federal
budget deficit, as well as the tight monetary policy, causes
unfavorable conditions for the farm economy. The government
deficit, in his view, tends to increase real interest rates
and hence the exchange rate. Decreases in agricultural
exports, prices, and incomes follow because agriculture is an
export oriented sector. Schuh (1984a) later emphasized this
view by stating that "a more nearly balanced federal budget
probably would do as much as anything to improve our
agricultural export performance" (p.246).

Schuh's initial work was nothing more than an extension
of the Keynesian paradigm to the farm economy and it received
immediate response from a number of researchers. Belongia and
Stone (1985) and Batten and Belongia (1986) criticized the
Keynesian view of fiscal policy impacts on agriculture. Though
a negative relationship between the real exchange rate and
agricultural exports was found in their empirical analysis,
neither money nor the federal deficit caused changes in real
interest rates and exchange rates. They concluded that
"attributions of the decline in farm exports to monetary
policy or the deficit are difficult to support empirically and
still may be regarded, at this late date, only as conjecture"
(p.427, Batten and Belongia).

Barclay and Tweeten (1988) supported Schuh's conjecture

30
by finding a negative impact of federal deficit increases on
agricultural exports and prices. An increase in interest rates
and an appreciation of the U.S. dollar caused by an increase
in the federal deficit is found to be a major mechanism for
the impact.

Rausser (1985) looked at the issue differently. He found
that the speed of price adjustment to any shock in the
monetary variables (money supply, interest rate, and exchange
rate) is much faster in the case of agricultural goods
compared to industrial goods. Chambers (1985) and Bredahl
(1985) related the differential price adjustment to the
stylized facts that‘U.S. agriculture has; (a) highly inelastic
demand and supply, (b) low income elasticities of demand, (c)
high competition, (d) rapid technological change, (e) asset
fixity, (f) variability in supply due to weather, and (g)
foreign agricultural. policy. Rausser found that
nonmonetization of the federal deficit has the same effect on
the economy as tight monetary policy does. It depresses farm
prices through its deflationary impact on the general price
level. Rausser, Chalfant, Love, and Stamoulis (1986) also
supported the short run responses of agricultural prices due
to fiscal deficit changes. However, the neutrality of the
economy is supported in the long run. They argued that
agricultural prices follow a new long run equilibrium path
after a short adjustment period. Thus, the relative price of

agricultural goods to industrial goods remains stable in the

31
long run.

Recently, Just and Chambers (1987) developed a
theoretical model to explain the relationship between the farm
economy and budget deficits. They compared the performance of
three alternative ways to reduce the current budget deficit:
expenditure reduction, monetary expansion, and a tax increase.
Their model appears to be the first theoretical model dealing
directly with fiscal policy impacts on agriculture. Again,
the differential price adjustment scheme is used in their
model, but the direction is just the opposite to Rausser and
others. Farm prices are believed to be fixed due to government
intervention, and industrial prices are allowed to be
flexible. Therefore, any inflationary policy causes a "cost-
price squeeze" in agriculture by increasing industrial prices
relative to farm prices. Expansionary fiscal policy hurts the
farm economy as much as expansionary monetary policy, where
financing the expenditure by borrowing (or a tax) stimulates
it. The results are derived from a comparative static analysis
with a multi-period equilibrium condition in the government
budget. However, the model has many weaknesses, and so far it
lacks empirical support to validate its results.

Thus far, the current literature dealing with fiscal
policy impacts on agriculture have been discussed. The
empirical model by Rausser (1985) and the theoretical model
by Just and Chambers (1987) have been described as Neo-

Keynesian models. The empirical works by Belongia and Stone

32
(1985) and Batten and Belongia (1986) fit the classical
paradigm. Papers by Schuh (1981, 1983) and Barclay and Tweeten
(1988) fit the Keynesian paradigm.

To reduce the current huge federal deficit, Just and
Chambers argued that a government expenditure reduction would
have the most beneficial effects for agriculture. Belongia and
others argued that focusing attention on deficit reduction
measures diverts attention from more fundamental changes
required in farm commodity programs since budget deficits do
not have impacts on agriculture. However, Rausser and others
argued for the dominance of macroeconomic policies over farm
policies in affecting the farm economy in the short run. They
confined the role of farm policies to reducing instability of
farm prices and not providing incentives for over-allocation
of resources to agricultural production. Frequent use of farm
policy would hurt the farm economy by causing more
overshooting to macroeconomic policy shocks later. Barclay and
Tweeten defined optimal policy as a balanced budget and
international account with zero differential between domestic
and foreign interest rates. No jparticular solutions for
reducing the current budget deficit are described in their

simulation study.

CHAPTER III

MODELS AND RESEARCH METHODS

As discussed in the previous section, wide disagreement
exists regarding the effects of fiscal policy on agriculture.
Three different approaches to empirical macroeconomic modeling
can be distinguished: the simultaneous equations model (SEM);
the rational expectation model (REM); and the vector
autoregression model (VAR). In this section, comparisons of
these models will be made and the selection of the VAR
approach for this analysis will be justified. Some recent
developments in time series analysis also will be discussed
and taken into account to establish an improved VAR procedure.
The methods applied in this paper will be introduced at the

end of the section.

1. Models Used in Empirical Macroeconomics

1.1. The Traditional Simultaneous Equations Model

The SEM often has been referred as "Keynesian
macroeconometrics" because it is widely used in the empirical
macroeconomic analysis of Keynesian Models (Cooley and Leroy,
1985). The SEM tends to be large scale, taking account of many
behavioral relations between macroeconomic variables.

A system of g stochastically dependent equations can be

33

34

represented generally as
(15) A(L)yt = B(L)et.

where yt is a (gxl) vector of g macroeconomic variables at
time t and et is a (gxl) vector of disturbance terms. It is
assumed that E(et)=0 and E(ete;)=n for t=s, and O for t#s.
n is a diagonal matrix, implying no contemporaneous
correlation among the error terms across the equations.

Assuming B(L)=I for simplicity gives

(15) A(L)yt = et

p .
with A(L) = E Aij where the Ajs are (gxg) matrices of
i=0

autoregressive parameters and L is the lag operator. The model
is assumed to be stable and all the roots of the
characteristic equation |A(L)|=0 lie outside the unit circle.
The SEM usually distinguishes exogenous and endogenous
variables based on economic theory; By redefining A(L) and.yt,

the structural form of SEM is represented as

(17) [ A11(L’ A12(L) ][ wt ] = [ elt ]

° A22(L) xt e2t '

where A11(L) is a [(g-k)x(g-k)] matrix, A12(L) is a [(g-k)xk]

matrix, and A22(L) is a (kxk) matrix of A(L) elements. The O

35

is an [kx(g-k)] matrix of zeros. wt is a [(g-k)x1] vector of
observations on the endogenous elements of yt and xt is a
(kxl) vector of observations on the exogenous part of the yt
variables. e1t is a [(g-k)x1] vector of disturbance terms for
the wt equations and e2t is a (kxl) vector of disturbance
terms for the xt equations. A11(L), A12(L), and A22(L) are
assumed to have the orders p, q, and r, respectively, which
are not necessarily the same.

The corresponding reduced form is
_ —1 -1

Numerous a priori restrictions are used to identify the
parameters of the behavioral equations. Zero or equality
restrictions are.often.applied.to)A11(L) and A12(L) to exclude
variables from a specific equation. Restrictions on the lag
structure and the error structure are also used. The
predictive power of the model depends on the credibility of
the restrictions.

The problems associated with the SEMs are now well known.
The traditional model uses many restrictions which often cause
over-identification. Some restrictions are based on
controversial aspects of economic theory’ and. not 'tested
empirically. The SEM has a weakness for policy analysis
because its structure may not be invariant to policy changes.

The parameters of the behavioral equations usually do not

36
account for any policy caused structural changes. Such a
change is likely to occur since any change in policy affects
the agent's decision rules by changing their views of the
future. Finally, the errors across the equations are likely
to be related since they are produced by the same decision

making process.
1.2. The Rational Expectations Model

The REM considers the agent's views on. the future
seriously since these views affect the optimizing behavior of
economic agents. With inclusion of the expected values of
endogenous variables, the structural form of REM is

represented as

(19) A11(L)wt + ¢w§ + A12(L)xt = e1t
A22(let = eZt'
where ¢ is an (gxg) matrix of parameters and. w: is
expectation of wt formed in period t-1; that is, w: = E[wt|

wT, x r<t]. By taking conditional expectations,

T,

1 p q

B
£21A11,in-i + A12,0xt +.2 A12,ixt-i]

(20) w€=-(A11’o+¢)-
l=1

and

37

e —1 r

(21) xt = ’A22,oi§1A22,ixt-i'

Substituting equations (20) and (21) into (19) yields

(22) A11(L)wt + 512(L)xt = et

and the reduced form equation is

(23) wt = - 3:1(L)XIZ(L)xt + X;}(L)e1t.

The REM produces the same type of reduced form as the SEM
(compare equation (18) with (23)), but the rational
expectation hypothesis implies restrictions on the parameters
which can be tested. To identify the system from the final
form estimation, the REM uses cross equation restrictions on

parameters and the orthogonality between et and x ('r<t)

1
assuming the error terms are not correlated with past
information. The restrictions representing the optimal
decision rules of agents are also used.

The advantage of the REM, compared to the SEM, is that
the structure of the REM is invariant to policy changes. The
agent's views of the future and his decision.rules are already

taken account in the parameters of structural form and final

form equations. Therefore, the parameters do not depend on

38
changes in policy rules, unless the policy makers have more
information than agents. However, REMs have not been
successful for macroeconomic forecasting. The restrictions on

parameters often fail to be consistent with empirical data.
1.3. "Unconstrained" Vector Autoregressions

Resistance to the use of a priori restriction used in
the SEM, and the poor empirical performance of the REM, have
led some economists to turn their attention to unconstrained
VARs. Sims (1980, 1986) originally applied this method to
identify the effects of government policy on macroeconomic
variables. The structural form VAR can be represented by
equation (16), the same form used in the SEM. However, the
block exogeneity restrictions are removed from A(L) and all
variables are treated as endogenous.

The reduced form equations for the VAR are

P
(24) yt = zlcsyt-s + ut,
3=

-1

—1
where C8 = A0

__1 I _ I-
As and ut = A0 et. B(“tut) - Ao nAO - M.
VAR models tend to be small scale. Restrictions used to
specify VARs are based on statistical properties of the data.

Economic theory has a role in variable selection, but not in

restricting the structure of the system. To recover the system

39

from the reduced form equation, the VAR utilizes the relation
between the structural disturbance et and the reduced form
disturbance ut. The VAR generally assumes that the behavioral
shocks are mutually orthogonal (i.e., n is diagonal). Hence,
if the estimated reduced form covariance matrix M is diagonal,
then the reduced form equation is identified (i.e., A0=I). In
any other case, a (gxg) matrix R, subject to R'R = M51, is
selected to remove the contemporaneous correlation between the
error terms. Then, E(RutuéR') = RE(utu;)R. = MM.1 = I.

Identifying A0 with R and premultiplying the reduced form
equation by it recovers the system. Though various matrices
could satisfy the conditions for R, a unique lower triangular
matrix with positive elements exists for a given variable
order in the system. This method, known as the Choleski
factorization, or triangularization, restricts the system to
be recursive. The present values of variables appear
recursively in the left hand side of equation (16), following
the same order given in n. A particular orthogonal order must
have a sound theoretical and statistical justification.

With no exogenous variables in the system, a policy shock
to the economy cannot be identified by traditional methods.
Various nontraditional methods are applied. First, the moving
average parameters are often used to identify the effects of
a policy shock on the system. The estimated reduced form
equation (24) is transformed to its moving average

representation

40

Q
(25) yt = z Dsut_s,

by either inversion or successive substitution. The moving
average parameters represent the net effect of a particular
shock to economy, assuming no additional shocks occur. The
moving average parameters, D8 are often called the impulse
response weight or dynamic response weight.

Second, a decomposition of variance for the forecast of
future values of variables into components due to particular
shocks provides information on the sources of unpredictable
fluctuations.

Finally, the significance of a variable in a forecast of
the other variables provides a criteria for causality
(Granger, 1969; Sims, 1972). If all the lags of a variable are
significant in a forecast of the current values of the other
variable, but not vice versa, causality runs from the first
variable to the second variable. Insignificance of lags of
both variables in the forecast of a 'variable implies no
relation between two variables. Hsiao (1981), using the same
idea, used Akaike's (1969) final prediction error (FPE) for
the significance criteria.

The VAR approach has been criticized on a number of
grounds. First, the selection of the variables are arbitrary,

depending on the researcher's interests. Therefore, the VAR

approach may not be consistent with the optimal behavior of

41
agents. Rather they reveal the statistical correlations
between the selected variables. Omitted variables could be a
problem of VAR model (Mount, 1989) . Second, policy changes are
sometimes not random as the VAR approach assumes. Third, many
VAR studies specify the model arbitrarily. Different lag
length and. different orderings of ‘variables might cause
different results (Saunders, 1988; Orden and Fackler, 1989).
Fourth, U.S. macroeconomic data usually fail to satisfy the
stationarity condition. The VAR model with levels, in this
case, might not be appropriate for empirical tests (Engle and
Granger, 1987; Mount, 1989). Fifth, U.S. macroeconomic data
often show nonconstant variances conditional on past
information. The existence of conditional heteroscadasticity
doesn't affect consistency'of the estimator, but causes a loss
of efficiency (Engle, 1982; Engle and Bollerslev, 1986).
Sixth, VAR. models are often criticized for their high
parameter to observation ratio (Zellner, 1981; Hsiao, 1979).
With all variables treated endogenously, the number of lags
are heavily constrained by the number of observations.
Nickelsburg (1985) found poor performance and underfitting of
the lag selection criteria in the case of small samples for
VARs. Finally, in the case of unprecedented large movements
in policy instruments, it is doubtful that the linear
structure of the model would remain fixed (Mazon, 1985;

Bessler and Kling, 1989; Todd, 1989).

42

2. Extensions of the VAR Approach

Numerous studies have adopted the VAR approach in
empirical macroeconomic analysis after the mid 19705.
Recently, several papers have considered the critics of the
VAR approach seriously and tried to improve performance in

various ways.

2.1. Identifying Contemporaneous Correlations

As discussed, identification of the VAR system and
causality tests rely on contemporaneous relations between
variables (i.e., the structure of A0) . Though a recursive
order for A0 is often used, many researchers have criticized
this approach. Cooley and Leroy (1985) argued the restriction
was nothing more than imposing the same prior exclusion
restrictions used in the SEM approach. Leamer (1985) didn't
find any reason why AG has to be triangular. They also argued
against the VAR type causality test for policy analysis.
Cooley and Leroy discredited the VAR approach for policy
analysis by saying that

"In the absence of prior predeterminedness

restrictions, such interpretations are completely

unjustified , for an intervention in policy
variables to have clear meaning, an exogeneity
assumption is needed. Such an assumption is, as we
have noted, not testable in the absence of prior
restrictions. We conclude that VAR models are not

useful for analyzing interventions either in
parameters or in variables." (p.301-303)

43

Several recent papers responded to this argument.
Blanchard and watson (1984), Sims (1986), Bernanke (1986),
and Orden and Fackler (1989) introduced a two step procedure
to estimate the contemporaneous correlation matrix, A0, rather
than using the arbitrary Choleski factorization. The relations
between the reduced form and structural form covariances are
used in estimation. Applying the variance operator to “t =

Aalet, the relationship can be represented as
(26) n = AOMAS

The reduced form equation (24) and it's covariance M are
estimated first. Then, the structural form covariance is
recovered from equation (26). With 9 variables in the system,
the
number of distinct covariances in M is g(g+1)/2. Since n has
only 9 components (recall that n is diagonal), the system is
just identified if AC, has g(g-1)/2 unknown parameters. A lower
triangular matrix A0, hence, satisfies the condition for the
just identified system. However, neither simultaneity nor over
(under) identification restrictions have to be precluded.

Though Orden and Fackler (1989) justify the recursive
structure by delays in agent's reaction to the policy shock,
due to information lags and adjustment costs, they argue that
simultaneity is a more likely choice. Blanchard and Watson

(1984), Bernanke (1986), and Orden and Fackler have all

44

emphasized the role of theory in pinning down the initial
structure for the system. Blanchard and Watson had eight
contemporaneous parameters in their four variable VAR. They
used outside information to estimate the parameters and the
variances of the first equation. The findings are used to
estimate the parameters and variances for the other equations
sequentially. Bernanke called this model "quasi triangular"
since the equations are ordered in such a way that the i-th
equation has exactly i-1 unknown parameters and the estimated
residuals from equation 1 through i-1 is used as the
instruments to solve the i-th equation.

Bernanke, in his over-identified model, found a way to
avoid the quasi triangularity assumption. To use up a degree
of freedom from over identification, he allowed an error term
to appear in other equations as well as the equation for its
own variable. Orden and Fackler, however, used the FIML
estimation procedure. Recently, Blanchard and Quah (1989) used
a block exogeneity restriction on the parameters of a moving
average representation to identify a system.

Considering the factors above, it can be said that the
VAR approach tries to place minimal identification
restrictions for any given analysis. Indeed, the restrictions
are based solely on the contemporaneous interactions among the

variables.

45

2.2. Considerations of Structural Change

Schuh (1976) highlighted a change in the macroeconomic
environment in 1973 when ‘the ‘U.S. government adopted a
flexible exchange rate system. Many economists have also been
concerned with the effect of 1980 Monetary Decontrol Act on
the economy (Saunders, 1988). Chambers and Just (1986) have
expressed concern over models which do not take account of
structural changes. Though ‘the ‘VAR. approach. catches the
dynamic response of the economy without a priori restrictions,
an unprecedented large movement in the economy may not allow
the linear structure of the VAR model to remain fixed.

Separation of the sample might be a solution to this
problem. Saunders (1988) used this approach and ran two
different VAR models to check the consistency between data
before and after the 1980 monetary regime change. However,
division of the sample period reduces the number of
observations significantly, and hence causes a degrees of
freedom problem.

Mazon (1985) introduced a dummy variable into a VAR model
to account for a structural changes in the economy. Loss in
the degrees of freedom is negligible in this case. If the
dummy variables solely reflect the structural change, their
effect on the economy can also be traced through a variance
decomposition procedure. To show this, consider a dummy

variable in the reduced form equation (24) . The equation

46

becomes

P
(27) yt = E fsyt-s + aDVt + ut,
s:

where DVt = a (gx1) vector of dummy variables which has 0
values for the period before and 1 after the structural change

has occured. The corresponding moving average representation

will be
oo
(28) yt = z Déut_s + govt.
s=0

The decomposition of variance in forecasts of future expected
values of yt can be used to examine the effect of a structural
change in the economy. The appearance of the dummy variable
in the system will likely affect the parameters of both the
autoregressive and the moving average parameters if it is
significant.

Although. the ‘magnitude of the effect of the dummy
variable is well represented by the differences in base
projection, its statistical significance is not easily derived
since there is no asymptotic distribution theory available for
the decomposition. Mazon used the Monte Carlo Integration
method developed.by Kloek.and.Dijk (1978) to compute the first

and the second moment for the posterior distribution of the

47

decompositionl.

2.3. Unit roots

Thus far, the time series model is assumed to be stable
(i.e., all the roots of the characteristic equation |A(L)|=0
lie outside unit circle). Recently, numerous scholars have
investigated the stationarity properties of macroeconomic data
(Nelson and Plosser, 1982; Schwert, 1987). When the data are
generated by nonstationary processes, the autoregressive
polynomial A(L) is not invertible. Therefore, the moving
average representation (25) cannot be obtained and causality
tests based on impulse response weights cannot be justified.
Phillips and Durlauf (1986) found that OLS produces
consistent estimates but that they are not asymptotically
normally distributed. 'Therefore, approximate maximum
likelihood estimators for the VAR parameters will be difficult
to obtain. Furthermore, test statistics are not easily found
(Sims, Stock and Watson, 1990).

Several techniques have been developed to test for
stationarity. Most commonly, the likelihood ratio test by

Dickey and Fuller (Fuller, 1976; Dickey and Fuller, 1979,

 

1One hundred random samples are drawn from the posterior
distribution of the VAR coefficients, assumed to be Normal-
Inverse Wishart (Zellner, 1971) . The historical decompositions
are generated for each draw and the first moment and standard
errors are computed with the one hundred sample historical
decompositions.

48
1981) and Stock and Watson (1986, 1988) are used. Both tests
are based on the significance of the parameter do in the OLS

regressions

P
(29) Ayt = u + Bt + aoyt_1 +iilaiAYt-i + et

where Ayt = yt - Yt-l' Critical values are reported in Fuller
(1976) for the DF test and Stock and Watson (1988) for the SW
test. The same methods are used for testing larger numbers of
unit roots in economic time series by replacing Ayt with
appropriately differenced series in the test equation.
However, Sen (1986) found that these methods lack power.
Dickey and Pantula (1987) provide a test for multiple unit
roots, which starts with a large number of unit roots and

tests downwards. To test three unit roots in the AR(3) model

3
(30) yt = z ¢jyt-j + et.

J=1
Dickey and Pantula reparameterized (30) to get
3 _ 2
(31) A yt — "1Yt-1 + "ZAYt-l + n3A yt_1 + et.

The appropriate null hypothesis for testing the i—th unit will
be Hi where

49

H3: «1 = n2 = n3 = 0
H2: «1 = n2 = 0, ”3 < 0

H1: "1 a 0, n2 < 0

and alternative hypotheses are H2, H1, and

H0: 111 < 0,

respectively. Test statistics used in the DP test will be same
as the DF test statistics and the equations also can be
augmented by introducing lagged dependent variables, as in the
DF test when auto-correlation appears in the error terms.
Though the DF test is widely used for economic time
series, weaknesses of the test are also well recognized.
First, the test ignores the possibility that the true data
generating process may have MA terms as well as AR terms.
Schwert (1987, 1988) and Lee and Schmidt (1990) found low
power for the DF test in the presence of MA errors in a
process. Too:many rejections in.the case of negative MA.errors
and too few rejections in case of positive MA terms are
experienced. Second, critical values of the DF test are not
valid when the nuisance parameters )1 and 8 appear in a process
(Schmidt, 1988; Nankervis and Savin, 1985; Evans and Savin,
1984) . The meaning of the nuisance parameters depends on

whether or not the unit root is true. Under the null

50
hypothesis, u represents a deterministic time trend and 8
represents a quadratic time trend. Under the alternative
hypothesis, u represents deterministic level and 8 represents
a deterministic time trend.

Phillips and Perron (PP) developed a test to solve the
first problem by allowing a wide variety of weakly dependent
and heterogeneously distributed time series (Perron, 1986;
Phillips, 1987; Phillips and Perron, 1988). The test
statistics are corrected by using Newey and West (1987) type
error covariance corrections and are valid under
autocorrelation or conditional heteroscadasticity of the
errors. The PP unit root test considers following three OLS

regression equation.

(32) yt = E + B(t - n/2) + Eyt_1 + ut

* * t
(33) yt u + a Yt-1 + ut

I
a Yt-l + ut.

(34) Yt

The test procedure starts with the null hypothesis

1, — - —
noon—B

H
O
0
5
D:
9
fl
(.3

with the test statistic 2(42). If the first null hypothesis

is rejected, then test the second null hypothesis

51

with the test statistic 2(a3). If Hg is also rejected, then

test the third null hypothesis

with test statistic 2(ta). Rejection of H3 implies no unit

root and acceptance of Hg implies a unit root in the series.

The trend and drift terms are also important. If H% and

3 cannot be rejected, then.the procedure develops into a more

H
powerful test with.equation (33). The new null hypothesis will

be
Hg: u = o, a = 1

and the test statistic will be 2(41). Rejection of Hg will

lead one to the individual unit root test of the hypothesis

with the test statistic Z(ta*). Acceptance.ofi 3 will lead one

to estimate equation (34) and the individual unit root test

will be done with the hypothesis

52

The test statistic will be Z(ta,). No trend and no)drift terms
will be included in the series. The test statistics Z(t5),
Z(ta*) and Z(ta,) are defined in Appendix B and have the
Dickey Fuller 1, 1“, and. Tr distributions respectively.
Critical values for PP test are also included in Appendix B.

Schmidt and Phillips (1989) developed a test valid with
the nuisance parameters. They modified a unit root test based
on Bhargava's (1986) parameterization. The Schmidt-Phillips

(SP) unit root test starts with OLS regression equation

(35) Ayt = I + ¢§t_1 + 5t (t = 2,....n),

where

§t_1 = yt_1 - u + B(t-l)
n = mean Ay = (yn - 1)/(n-1)
B = y1 - u.

The nuisance parameter p and B can easily be estimated by
(36) Ayt = u + at + et.

The test statistic is the usual t statistic for ¢ and its

critical values appear in Appendix C. The test equation can

53
be augmented with lagged dependent variables when
autocorrelation is present. In that case, the test statistic

will be the ratio of the t statistic to a discount factor p,

where
p2 = 02/02
2 _ -l 2
a6 - lim n E(Eet)
02 = lim n"1 E(S§)
and
2 _ 2
Sn - (233') e

The meaning of the nuisance parameters u and B in the SP test
does not depend on whether the unit root hypothesis is true

or not.

2.4. Cointegration and Error Correction Models

While individual macroeconomic time series data have
often shown stationarity after first differencing (Stock and
Watson, 1986, 1989; Engle and Granger, 1987; Kunst, 1989;
Haslag and Slottje, 1989), it is also believed that many
nonstationary series have a tendency to move together in the
long run. If economic variables are linked by a long run
equilibrium relationship, a linear combination of the
variables should not drift too far apart over time and need

not be differenced. This relationship, termed as cointegration

54

by Granger (1981) , reduces the number of unit roots to be
fewer than the number of variables.

As an example, consider two nonstationary time series Y1t
and y2t in yt, which are stationary after differencing d
times. If there exists a linear combination zt = a'yt which
reveals stationarity after differencing b(<d) times, the two
series are said to be cointegrated of order d, b, denoted
CI(d,b). The (gxh) coefficient matrix a is called as
cointegrating vector and the combination zt is called as error
correction term (Engle and Granger, 1987). zt has expected
value of zero and any nonzero value represents a deviation
from the long run equilibrium relationship.

The existence of cointegration implies a restriction on
multivariate time series models. To show this, consider d=b=1
for the reduced form VAR equation (24) . Taking the first

difference, equation (24) becomes

p-l
(37) Ayt = E“flush-s + I'py'c-p + “t.

where rs = - I + c1 + c2 + ... + c . Therefore, the VAR with

levels will only be appropriate when P has full rank 9. If

P

Pp has zero rank, Ayt is stationary. In any other case, Pp has

rank h with 0<h<g and can be represented as

(38) - Pp = na',

55
where n is dimensioned as (gxh) with rank h. Equation (24) now

becomes

p-1

s=1 '

or equivalently

p-1
(40) Ayt = 5:1Arsyt-s - "zt-p + ut.

These equations are called an error correction model
(ECM) since they contain the error correction term, zt. The
error correction term represents adjustment of the economy to
a disequilibrium in one period (zt¢0) . n represents the speed
of adjustment and (1 represents the long run relationship
between the variables (Engle and Granger, 1987) . The VAR model
with levels, in this case, is inefficient since it ignores the
restrictions on PP. The VAR model with first differences will
be over differenced and misspecified by falsely restricting
Pp = 0. The parameters, 11 and (1, also provide a basis for
identifying the contemporaneous correlations where long run
interpretations of VARs are varying with different
specifications.

An ECM is consistent with economic theory satisfying long

run equilibrium as well as allowing for short run dynamic

adjustments. Thus, ECMs produce accurate forecasts of

56
responses to disequilibrium shocks over reasonable horizons
(Hendry, 1986; Nickell, 1985).

Several measures are available to test for cointegration
and to estimate the cointegrating vectors. The two step method
of Engle and Granger (1987) is based on the autocorrelation
function) of OLS :residuals and focuses exclusively' on a
bivariate.system. Engle and Yoo (1987) found.that the critical
values of the augmented Dickey-Fuller test used by Engle and
Granger increase with the number of variables in the system.
Although Engle and Granger argued for the efficiency of the
two step estimation method in the two variable cointegration
case, the properties of the least square estimators are
dubious when there are more than two variables. Stock (1987)
showed that the least square estimators are biased and not
asymptotically normally distributed.

Stock and Watson (1986, 1989) developed a test based on
the principal component estimators. Though the test is
applicable to a system with more than two variables, Engle
(1987) discredited the test since it is not based on
likelihood theory. Hoffman and Rasche (1989a) are also
skeptical of the method. They argued that inference from the
SW test is highly sensitive to arbitrary lag length
specifications in the structure used to estimate the principal
components. Also, the available critical values generated by
numerical simulation may not be applicable to a wide range of

variables.

57
A consistent maximum likelihood estimation procedure has
been established by Jehansen (1988, 1989) and Johansen and
Juselius (1988, 1989) by using moments and cross moment
matrices of the residuals from auxiliary regressions. In

equation (38), the rank of the matrix P h<g, determines the

pl
number of cointegrating vectors in the system yt. To test H0:

at most h cointegration vectors, a maximum likelihood test

statistic

9 9
(41) LR = -21nQ = -n E ln(1-Ai) z E nxi(n)
i=h+1 i=h+1

is used, where ii are the g-h smallest eigenvalues of
_1 _
(42) IASpp - SpOSOOSOpl - o,
where

n

I , U . .
(43) Sij = i/nt: ritrjt' (i,j=0,p)

1

rot and r t are obtained by OLS regressions:

P

p-1
(44) Ayt =.21aiAyt-i + rot

and

58

p-1
(45) Yt-p =121aiAyt_i + rpt.

The log likelihood test statistic is asymptotically
distributed as c-x(f) where c=.85-.58/f and f=2(g-h)2.
Selected fractiles are given in Appendix D.

The cointegrating vector a and parameter for the error

correction term n are obtained as

n = TSOpa
and
a = eigenvectors of (42) corresponding to the largest h

eigen values.

Johansen's estimator has better properties than least
squares, since it takes accounts the error structure of the

underlying process.
2.5. Restrictions on Cointegrating Vectors

Though the interpretation for the error correction term
is forward straight for a two variable system, it is not as
easy for systems with more than two variables. When there
exist more than two variables in a cointegration equation,

certain restrictions on n and a might be needed for

59
interpretations to make sense. Johansen (1989), Johansen and
Juselius (1988, 1989) introduced a linear restriction
procedure to simplify the parameters to find useful inference
from the cointegration relations. They also established a full
information maximum likelihood method to estimate the
simplified inferable parameters and to test the linear
restrictions on parameters. To test Ho: a = So where e is a

(gxs) and ¢ is a (sxh) matrix, the likelihood statistic

h
(46) LR = -21nQ = n z 1n(1-A;)/(l-Ai)
i=1

is used, where A: are the h largest eigenvalues obtained by

solving

I I _1 _
(47) |xe sppe - e SposooSOp°| - o.

The restricted parameters, a* and n* are obtained as

eigenvectors of (47) corresponding to the largest

9
II

h eigenvalues
and

e
n = -SOpa .

The restriction, HO:K'a = 0, can also be tested by a Wald test

statistic

60

1 1

mo (K'uu'xf 1

(48) w = ntr[(x'u'(soo-aa')‘
where u is the eigenvectors of (42) corresponding to smallest

(g-h) eigenvalues. The test statistic is distributed as X(g-

s)h°

2.6. Error Regressive Models

When more than two components are introduced into the
system Yt' there exists a possibility of more than one
cointegration relationship, each representing an equilibrium
for different variables. For example, a cointegration equation
representing the money market equilibrium condition may not
include all the goods market variables, and vice versa. In
this case cointegration between variables can be used to
reduce the dimension of the system yt. Consider two error
correction terms 21 and 22 in equation (40), each representing
a deviation from equilibrium. in imarket 1 and. market 2
respectively; The ECM: is still applicable 'with. the two

different error correction terms as follows

p-l
(49) Ayt = =21 APSYt-S + nlzlpt-p + "ZZZ't-p + ut.

Alternatively, it is also possible to formulate the time

61
series regression by treating the error correction terms 21
and 22 as a reconstructed variable. 21 will reflect any
deviation of the variables of market 1 andz2 will reflect any
deviation of the variables of market 2 from their long run
equilibrium. The VAR with two stationary variables 21 and 22

can be represented as

pi
(50) 2t =s:1 szt-s + at,

where zt' = [21,t 7‘2,t]' and is referred to an error
regressive model (ERM) in this study. Though the dimension of
the system yt is reduced significantly, the ERM is still
capable of capturing any movement of a variable in yt which
causes deviations from the equilibrium relationship.

The ERM has several benefits over the ECM. First, the ERM
saves the degrees of freedom by reducing the size of system.
Thus, improvement in model fitting and performance is
anticipated, especially in cases when only a small sample is
available. Second, the structural form can be easily
identified in the case of an ERM due to reduced
dimensionality. For the previous example, there are only two
just identified contemporaneous structures possible, either
z1,t before 22,t or zz’t before 21't in the orthogonal order.

The ERM also has disadvantages. There is a loss of

information and it is not easy to detect causes of the

62

disequilibrium movement in zt. Therefore, it is also difficult
to find the impact of any individual series in zt on other
variables, as well as short.run interactions among the zt. For
this reason, the VAR or the ECM would be natural choices if
one is interested in dynamic interactions between individual
series. However, the result from the ERM can serve as a
preliminary analysis. With the improved statistical
performance of the ERM, a VAR analysis consistent with the ERM
will provide confidence for researchers, especially when there
are many possible choices in the model specification.

With cointegration in time series data, the ECM and the
ERM generate stationary models. The parameter a provides an
ordering of contemporaneous variables in structural
identification. VARs in differenced variables suffer
misspecification and VARs in levels underestimate the
parameters near the unit circle, though they are consistent
(Engle and Yoo, 1987) . With the ECM, improved asymptotic
properties are achieved for estimators and tests (Sims, Stock,
and Watson, 1990). Although cointegration between
macroeconomic variables has been found in some foreign data
(Robertson and Orden, 1989; Kunst, 1989), it is not pervasive
in U.S. data. Stock and Watson (1989) found no cointegration
between money, income, price, and.the interest.rate. Engle and
Granger (1987) found cointegration between consumption and
income. But they rejected cointegration between wages and

prices, and between GNP and money.

63

3. Methods Used in This Study

This study adopts the VAR approach developed by Sims.
Although the VAR approach has some disadvantages, as discussed
above, its strengths for empirical analysis are well known.
Sims (1980) justifies use of the VAR approach in cases of no
sound theoretical model by saying that ". . .when adequate
structural models do not exist; as is often the case in
aggregate macroeconomic work, VAR. modeling is a useful
strategy for obtaining empirical evidence on competing
theories." Myers, Piggott, and Tomek (1991) had more reasons

to use the VAR approach; that is,

"... VAR models are relatively simple to specify
and estimate. Only a minimal set of just-
identifying restrictions is employed and no
restrictions are placed on the parameters of the
reduced form. ... given the.uncertainty surrounding
the underlying economic structure of the market,
the unrestricted reduced form VAR provides
flexibility which allows the model to be consistent
with a wide range of alternative economic
structures."

Analysis will begin with unit root tests. If unit roots
are found, further tests will be conducted to find whether
cointegration exists. The maximum likelihood methods developed
by Johansen will be applied to test and estimate the
cointegration relations. Estimates of a cointegrating vector
a will indicate long run relationships between variables and

generate error correction terms zt. A reduced form ERM model

64

will be estimated based on the findings, and several
contemporaneous structures will be imposed to identify the
structural form. Statistical performance and theoretical
consistency will be considered to choose a final model for the
analysis. Then, the impact of fiscal policy on farm prices
relative to industrial prices will be traced out based on
impulse response weights. Decomposition of forecast error
variances will provide a further measure of the impact.

Sensitivity of the ERM interpretations will be analyzed
with respect to different model specifications. Sensitivity
against different lag orders and different contemporaneous
structures will be checked first. Sensitivity analysis will
also be performed against different types of model, ECM and
VAR in both levels and first differences. For the ECM
estimation, a zt series representing the long run equilibrium
relationships between variables will be unrestricted and its
component variables will be included individually. The
parameters for the error correction term will then serve as
another criteria for identification. With all monetary
variables included, the impulse response weights and the error
decomposition will provide a measure of the impact of fiscal
policies on agriculture.

Finally, a policy option to tackle the current budget
deficit will be drawn based on the findings. The policy option
will be compared to other policy scenarios in a simulation

study.

CHAPTER IV

PRELIMINARY ANALYSIS

1. variables and Data

The Federal deficit (D) and farm prices (F) are natural
choices for variables considered here to pursue the study's
objectives. The Federal deficit is also broken down into its
two component parts, federal government spending (G) and tax
revenue (T). Real output of the economy (Y) and industrial
prices (P) are included to represent quantity and price
variables for domestic goods markets; real money stock (M)
and interest rate (R) represent. quantity and. price for
domestic financial markets. The exchange rate variable (X)
enters into the model to represent the international sector.

D is measured by the federal budget deficit ($ bil.)
divided by nominal GNP ($ bil.) to remove any impact of
inflation, size of economy, and business cycle fluctuations
(Belongia and Stone, 1985). Similaereasures are used in other
empirical studies (Barro, 1978a, 1978b; Hamburger and Zwick,
1981; McMillin and Beard, 1982; Protopapadakis and Siegel,
1984). Similarly, G is measured by the ratio of the federal
government expenditures ($ bil.) over nominal GNP and T is
measured by the ratio of the federal government receipts ($
bil.) over the nominal GNP. Nominal M1 ($ bil.) is deflated

by the GNP implicit price deflator (1982=100) to define M and

65

66
the three month treasury bill rate is used for R. X is
measured by the multilateral trade weighted value of the U.S.
dollar (march 1973=100). Y is measured by nominal GNP divided
by ‘the GNP implicit. price deflator. The implicit price
deflators for GNP by farm and nonfarm sectors are used for F
and P respectively.

Quarterly data from 1948:3 to 1989:3 are used for the
analysis (see Appendix .A for sources). The indexes are
seasonally adjusted except for the interest rate and the
exchange rate. All variables, except D, are expressed in
natural logarithms to stabilize their variances (Kunst, 1989) .

Plots of each series are provided in Figure 6.

2. Unit Root Tests

Several measures are used to test for stationarity in
each series. Dickey-Fuller (DF) test and Stock-Watson (SW)
tests are applied and the results are reported in Table 1 for
the first unit root and Table 2 for the second unit root. Lag
lengths for the test equations for each series are selected
based on several criteria (FPE, AIC, SC and LR)2 and appear
in the last row of the tables. The first unit root is rejected

in series G, T and D at the .05 level in both tests. The

 

2FPE: Final Prediction Error Criteria by Hsiao (1981).

AIC: Akaike Information Criteria by Akaike (1969)
SC : Schwartz Criteria by Schwartz (1978)
LR : Likelihood Ratio Test by Sims (1980)

67

.3...

USe‘e

DUN?

In. It...

 

Casi/7?.

OUSeu .

 

 

a... r... t. s... as a... es .s.

 

 

 

 

 

“he

has.

 

Du)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

on.) on...
I. e R,
0’3 Eks a. I 3 O (s
A 4r\
I..C < O. D ‘5
u. e
o...- f o I
0....
lo...
sue.
on“ 14 In.-
..8 PO 8! 89 c. d... Pl ’0. DO 0’ «I d? 00 I...
a. 0
He. «9.
"N. ’8
Ice 0.3
a.) a...
o... a.‘
8? e... g
8.. e .
8? e a
'8 J .... 8. .8 do. «.8 a... ... s «I 1... II 8..
w x
a: ,8:
e L
..e o? ’
a. . .... :- OLA
.0
e a
e...
e I.
U.) do w L
.....a
..5.
a... onus
r8 o 0.. I u.- q.. «.8 I II I... v! e I! I.

 

 

 

Figure 6. Plots of Each Series

68

Amado a. womcdnm om om 4mm" mos cam csfiw zoo"

 

zkconsmmmm mmsfimm

 

can" ”assass.sca

on -~.oa -.os -s... -~.sm -o.se. -~.ss -...s... -..sm... -m..~...
m: -a.ao -.om -.ss -ss.eo -Am.so -s.a~ -~5.s... -oa.sm... -....o...
. . . a
wmwmm43mwwwhnSHnmwas.o~. -.Nm s.am o.s~... ~.ms.. -_.m~ o..s... ~.ms.. o.ss...
. e
Ia.o.cmwuws ss.ov N Na.. .15 s.ms ~.ss... .em ~.sm... -..oo... -..so... -~._s...
zssams om roam o o u o u H _ ~ N

 

wagon" "so as.. exconsmmmm om can" soon on .~o mos 4. .om mos 4» use .o~ mos 44».
DImnsflcsnIos om soon mnoa1ma1n «oddoz OInxok use ms_Los mos m use n mos a.

69

quads N. ”oceans om am down use are csfin woonm

 

 

:«uonsmmmm meadow
< z e m a x s q s
cam" moonaauuovm
on -m.ms... -ss.s~... -o.oo. -s~.ms... -a.a~... -a.ss... -m.~s... -e..s... -o.so...
m: -ss_.m_...-~ss.oa... -ss.Na...-_as.os...-_.o.ao...-_oo.~m...-so~.so...-_mm.8....-s~s.co...
oonmwsmsfimnmn Nassau
Issue. are»: s~.o. -.as .om _.oe .os -.GN -.s_ ...5 -.ms .so
unifies A ...
Lane. areas a .s. o.oa -.sa _.as -.so .so .sm .5» .ms .05
zescos om roam o o . N o N a N N N

 

wagons "so sad. :«coazmmwm om :31" soon on .No «on 4. .om Nos 44 use .oN mos 44:.

amusemesNIos om Home «wonflmnfln moddo: amorok use madame mos o mac

n «cs 3.

70

second unit root is rejected.in all series at any significance
level by the SW test. It is also rejected by the DF test at
any level of significance for all but the P series where it
is rejected only at the .10 level. The results suggest that
all the series except the fiscal measures contain a unit root.
And the fiscal variables, G, T and D, are stationary in
levels.

Dickey-Pantula (DP) test result appears in Table 3 for
the third unit root and Table 4 for the second unit root. As
expected, both units roots are rejected in all series at any
significance level, except for the P series. The second unit
root for P is rejected only at the .10 level. The results from
the DP test affirm the results by the DF and SW tests.

Phillips-Perron (PP) unit root test also provide results
consistent with the DF and SW tests, except for the D series,
as shown in Table 5. A structural change in the mid 1970's for
the D series, as evident in Figure 1, might be responsible for
this anomalous result.

As shown in Table 6, the Schmidt-Phillips (SP) test had
similar results to the DF and SW tests. Except for the fiscal
variables, all series contain a unit root at any level of
significance. However, the unit root is rejected in G only at
the .10 level and in T at the .05 level.

Overall, it is reasonable to say a unit root is evident
in all series except the fiscal variables. Inspection of the

autocorrelations for the levelslgiven.in.Table 7 and.the first

71

asoLo u. zomsdam om as some mos assoc csfin moons

 

 

zxcoaammmu mosimm
e z a m a x a e s
Asian :51» zoos
as -sa.oo... -.o.os... -sa.os.». -No.aa... -_~.ss... -so.ss... -e..ss... -s~.ss... -.o.ss...
zcsuns ow room o o N o N N N N N

 

mInsflm‘nosn on .—c mos 4. .om mos as use .oN mos 44».

72

«made a. wmmsdam om at some mos are coin woonm

 

 

zwcoaymmfim meadow
< 3 v m x x m a o
monosa cswn moon
4: -a.so... -o.~a..* -o.-. -a.ss... -ss.so... -m.ms... -o.os... -o.sm... -o.m~..
zcsams om room o o L o o H H H w

 

mfimsImmnosn on .Ho mos 4. .om mos 4* was .oN mos 444.

73

quads m. womcdnm om we Nome No1 can :51" noon

 

 

«mun maondmnwnm mosdom
< 3 s m a x s 4 o
NIeNL so.om... s.so o_.ss... o.o~ o.ss ~.ms m.ss.. m.as.. ~.ms
Naooc o.os _.ms m.~s.. m.s~ m.~s o.om s..~... s.~s... o.ss
Naemc -~.sm -.Nm -_.m_ -~.as -o.sa. -~.ss -o.sa... -o.so... -~.a~
Nflc_v ~_.am... s.ss s~.os... .ms ~.so ~.os ..mo. m.os.. s.oo
NIo.V -H.os .mo ~.a_ -.wm -~.oo -~.ss -~.m.. -o.oo.. -s.so
Nxo.c a.- s.mo s.os .ss -.Ns .o_ -_.as -.mo -.mo
4 zmzse-=ame pea. . N A o s o s _ o N

 

qusimdnosn on .No «on 4. .om mos 44 use .op mos 44».

74

Amado m. zmmedsm ow we done mos cam :31" woo"

 

 

 

:kcofizmmflm mosflom
< z c m x x a q o
came moon 4 we 444
4 -_.oo -.NN -H.NH -H.oa -N.om -N.m_ -N.Nm -u.uo -a.mm
zcacms om room H H N o w H N o H
TIE

mmmsImlnosn me ._o mos 4. .om wow 4» one .oH mos

'75

quads N. >seonossodun.osm om ro<o. mo1Iom.

 

 

wo51om

< 3 t m a x a 4 o
.omo .omo .omu .oou .ua— .oum .ouc .amo .uoN
.ooo .oNN .cmo .oma .ouu .mmo .mom .NN. .NNw
.ouo .mNm .umu .oam .aNN .mom .Nom .aam .mao
.o—N .m—m .ouN .oNu .auo .NNn .NuN .mNo .mNo
.mom .Nm— .mNo .ooo .NoN .mNm .mmo .uum .amm
.oNs .mmN .oos .mou .Nmu .uuN .mam .Nou ..No
.omu .m—N .maN .omo .NNo .sum .mN_ .Nuo .SON
.muu .muo .mNo .mmo .oma .uw. .mNN .Nao .umN
.mus .smN .muN .mmw .oaN .NaN .m—w .Num .umN
.Nom .uou .oum .msN .oum .Nmu .amN .Nua .umN
.NNo .uNo .mNm .cuN .m—u .ou— .soN .NNo .Nom
.NmN .Nmu .mo~ .mNN .mmm .oom .uoo .Noa .NNN
.Nam .Nao .Nau .mus .mam -.omu .uom .ooN .Nm_
.NNo .Nwa .Nmm .moa .mNN -.ooN .uNs .NNu .NN~
.Npu .omm .NSN .NoN .mm— -.Nuo .wmm .Nus .Noa
.ooo .owo .NNo .Nmo .mwu -.Nos .590 .Noo .uwo
.mmo -.oou .Nuo .NNo .uNu -.NNo ..NH .NNN .umN
.mms -.oso .ocN .Nmm .mpu -.NmN .auN .Nmo .uNw
.aam -.ONN .mNN .NSN ..o— -._mo ..ao .NNo .uNo
.muw -.Noo .oms .NNm ..Nu -.Nmu .amm .Noo .umm

 

76
differenced series in Table 8 does not contradict these
results. The results are also consistent with other empirical

studies3.

3. Cointegration Tests

Though a unit root is found in each component variable
of Yt' except fiscal variables, the number of unit roots in
the system can be reduced if the nonstationary variables are
cointegrated. The Johansen and Juselius method is applied to
test the existence of long run equilibrium relationships among
the six nonstationary variables, M, R, Y, P, X, and F.

Test equations (44) and (45) are estimated first. The
lag p-l is selected based on statistical criteria appearing
in Table 9. Among the information criteria, Lutkepohl (1982)
supported the use of the Schwartz Criterion with his
simulation study. This criterion was minimized for a lag
length of zero. The zero lag model seems underfitted using
Sim's (1980) modified likelihood ratio test. This test
suggests two or seven lags.

The test statistics for cointegration are given in Table
10. Three cointegrating vectors are found with seven lags and

four cointegrating vectors with two lags. Three cointegrating

 

3Nelson and Plosser (1982) found a unit root in output,
price, interest rate, and money series with annual data. Stock
and Watson (1986, 1989) found a unit root in the series with
quarterly data. Similar results are obtained by Hoffman and
Rasche (1989a, 1989b) and Plosser (1986).

77

quads m. >=nono11odon1osm om «Isms oimmoxosnoa mos.om

 

 

 

moonm

< z t m a x a a a
«H .um_ .uNs .mmm .owo .Num .uNN .NNm -.ooN .Noo
1N .NuN .Nuu .mum -.oow -.Nmo .NaN .ouu .omo .05.
«N .oNA -.HNN .mNm .ONm .oNu .NNo -.oom -.omN -.ooN
1. -.oom -.ONN .uoo -.o—o .ooN .Hu— -.NNN -.on -.NNN
1m -.Noo ._.o ...N -.NNN .oNN .NNN -.Nu— -.NmN -.NNN
1m -.owm .NoN .me -.omo -.omm .osm -.omm -.oSN -.Nom
«N -.on -.ouo .uNo .omo -.uNo .oma -.om— -.ooo -.~_w
1m -.omm -.oNN .uma -.ONu -.oom -.omu -.omc -.Nuo -.Nom
so -.oom .ouo .um_ -._~m .Nuo -.omu .ouN .o—_ .oo.
«No . -.oom .Noo .uwm .omN -.oau -.omm .NNN .oN_ .Nom
sNH -.ooo .NNw .Num -.ON_ -.Nsm -.NNN -.Nuo .oom -.cmm
1~N -.NoN .NNu .uam -._.~ -.osm -.NNN -.omm -.oou -.ouo
«Na -.coN -.on .wa .oNo .ONo -.Nmm -.Nom -.__o -.ooo
1N. -.NNm .ONN .Nou .NNN .omm -.os~ -.uou -.ou— -.oms
1pm -.omu .oau .Nuo -.Nus -.ous -.Nos .o—m -.co— -.ow—
sum .ouu .oso .Nsw -.OON -.~N~ -.Nuo -.oo— .omo .omo
sNN .oNo -.osm .uN_ .ONN .o—o -.Num -.ooN .oms .ONN
sum .omN -.oms .Nos -.ouo .Nun -.NNN .omN .ONu .LNS
sue .oNo -.NNN .NmN -.owN -.oao -.Noc .omo .ouN .oNu

«No .ONN -.omm .NNm .oNo -.ouw -.co— .cmw .oua .-m

 

78

Table 9. Lag Selection for AM, AR, AY, AP, AX, and AF

 

 

Lag Criteria 2 1 Sig.
Length x (36) Level
p-1 (Xioggf AIC sc

0 .7830 -46.3ov —46.30v

1 .8421 -45.86 -45.86 570.72 .0000***

2 .5697 -45.89 -44.53 72.76 .0000***

3 .4485 -45.76 -43.73 46.24 .1179

4 .3500 -45.65 -42.94 47.45 .0960*

5 .2115 -45.79 -42.40 55.36 .0206**

6 .1471 -45.79 -41.72 51.44 .0459**

7 .0974 -45.84 -41.09 56.64 .0156**

8 .0646 -45.88 -40.46 46.20 .1188

9 .0476 -45.83 -39.73 41.80 .2333

10 .0253 -46.09 -39.31 62.26 .0042***
11 .0165v -46.16 -38.70 47.63 .0930*

 

1.Likelihood ratio statistics to test Ho: Lag length of p-2
vs. Ha: Lag length of p-l.
2.H9 is rejected at 10% level for *, at 5% for ** and at 1%
or 4*4,

3.v indicates minimum value for each information criteria.

79

Table 10. JJ Test Results for M, R, Y, P, X, and F

 

H0

-21nQ

 

4

5

2

1

Two lags
124.10***
76.73***
52.52***
30.04***
11.91

1.48

Seven lags
154.52**
100.93**

58.64**
22.81
5.95

.43

*

*

*

 

***: Reject the null hypothesis at .01 level.

80
vectors are also found with five or more lags and four
cointegrating vectors with other lags. Thus, at least three
long run equilibrium relationships are evident among the six
nonstationary variables.

Although a cointegrating vector may describe an
equilibrium among all variables in a system, it is also
plausible that only a subset of the variables are significant
in the cointegration. In this case, the dimension of the
cointegrating vector must be reduced. The bigger the system
is, the more possibilty of reduction in the dimension of
cointegrating vector. To simplify the cointegrating vector,
the Johanson and Juselius test is applied to all possible
combinations of two variables among the six nonstationary
variables. However, no cointegration is found in any two
variable combination. This results in moving to a test on
three variable combinations.

Among the various possible three variable combinations,
Hoffman and.Rasche (1989a, 1989b) have shown that the long run
money demand function, a linear combination of M, Y and R, is

a strong candidate.

4 . Money Market Equilibrium

The long run money demand function, known as the LM curve

or portfolio balance schedule, has received much attention

from macroeconomists (Meltzer, 1963; Chow, 1966; Poole, 1970,

81
1988; Goldfeld, 1973). The portfolio balance schedule is

represented as

(51) aMMt + aYYt + aRRt = eMt’

The Johansen and Juselius method is applied to test the
existence of a long run equilibrimm relationship among the
three variables. Two, four or seven lags are adopted for the
cointegration test equations based on lag selection criteria
reported in Table 11. The test statistics for cointegration
are given in Table 12. The estimated eigenvalues and
eigenvectors are presented in Table 13 with parameter
matrices. One cointegration is found at the .05 significance
level and the long run money demand equation is established
as equation (51). The coefficients, “M' any, and “R are
obtained from.the first column in a vector in Table 13 and are
significant by the Wald test. The LM equation is more familiar

after normalization as

N

N
(52) Mt = aYYt + “RRt'

Estimated values for cg, ranging from -.71 to -1.29, suggest
that the equilibrium real income elasticity of money demand
with respect to real balances is unity as many macroeconomists
conjecture. To test.HO: “MI: -aY or 0%.: 1, a likelihood ratio

test statistic, equation (46), and.‘Wald. test statistic,

82

Table 11. Lag Selection for AM, AY, and AR

 

 

Lag Criteria 2 1 Sig.
Length x (9) Level
838 AIC sc
(x10 )
0 .1129 -22.90 —22.90v
***
1 .1171 -22.79 -22.62 28542.00 .0000
2 .0912 -22.97 -22.63 43.46 .0000***
3 .0854 -22.96 -22.45 16.21 .0626*
4 .0783 -22.98 -22.30 19.98 .0180**
5 .0736 -22.97 -22.12 14.29 .1125
6 .0675 -23.98v -21.96 7.29 .6074*
7 .0603 -23.02 -21.83 17.67 .0392**
8 .0556 -23.03 -21.67 6.39 .7002
9 .0554 -22.96 -21.43 6.36 .7035
10 .0503 -22.98 -21.28 14.44 .1074
11 .0469v -22.98 -21.11 13.51 .1409

 

1.Likelihood ratio statistics to test Ho: Lag length of p-1
vs. Ha: Lag length of p.
2.H is rejected at 10% level for *, at 5% for ** and at 1%
or ***,

3.v indicates minimum value for each information criteria.

83

Table 12. JJ Test Results for M, Y, and R

 

 

h m Two lags Four lags Seven lags
o 3 29.85*** 32.58*** 28.35**
1 2 6.55 9.52 12.50*
2 1 1.44 2.68 2.55

 

Reject the null hypothesis at .01 level for ***, at .05 level
for **, and at .10 level for *.

84

Table 13. Eigenvalues and Eigenvectors for M, Y, and R

 

Two lags
Eigenvalues A
.13 .03 .01
Eigenvectors V (=a)
M 14.15 -10.87 9.51
Y -10.10 4.58 2.37
R 5.38 -.75 -.94
-sopv x 1000 (=n)
M 1.64 -1.18 .16
Y 3.10 .72 .21
R 21.99 1.67 -9.91
Four lags
Eigenvalues A
.13 .04 .02
Eigenvectors V (=a)
M 14.01 13.83 11.01
Y -1l.54 -5.30 1.65
R 6.23 1.15 -.28
"SOpV x 1000 (=n)
M .94 1.35 .50
Y 2.37 -.78 .69
R 37.59 .49 -7.44
Seven lags
Eigenvalues A
.10 .06 .02
Eigenvectors V (=a)
M 9.00 24.77 6.43
Y -11.59 -9.23 4.11
R 6.79 3.74 -.84
-sopv x 1000 (=n)
M -.83 1.76 .02
Y 1.55 .63 .77
R 30.01 9.41 -6.13

 

85
equation (48), are used. Results are provided in.Table 14 with
corresponding eigenvalues and eigenvectors. Both tests failed
to reject the unitary income elasticity hypothesis. The long

run LM equation (52) with the restriction is reestimated as

*

with a; having values -.55 for two and four lags and -.58 for
seven lags.

The results are quite consistent with Hoffman and Rasche
(1989a) who found the interest rate elasticity is -.53 for
four lags and -.56 for seven lags with monthly data from 1953
to 1987. However, Stock and Watson (1989) didn't find
cointegration between these monetary variables. The results
are not sensitive to other lag specifications. One
cointegration with the unitary income elasticity is also
accepted with three lags, five, six and eight lags. Estimates
of the interest rate elasticity in the restricted money demand
equation.were -.56 for three, six and eight lags and -.58 for
five lags.

Based on the cointegration results, three variables in
the LM equation can be merged into a synthetic time series
ZMt = Mt - Yt - aERt, which represents deviations from the
long run money market equilibrium. In Figure 7, the actual

balance and the equilibrium balance of money are plotted with

a; = -.55. th is represented by the vertical differences

86

Table 14. Test Results for the Velocity Restriction

 

Two lags

Eigenvalues 1* .12 .03
Eigenvectors V (M & Y) 9.51 5.42
( R ) 5.28 1.24
Test for the Velocity Restriction: x2(1) 1.81
N 1.52
Implied Interest Elasticity of Velocity -.55
(.03)

Four lags
Eigenvalues 1* .13 .03
Eigenvectors V (M & Y) 11.03 5.53
( R ) 6.06 1.23
Test for the Velocity Restriction: x2(1) .53
N .82
Implied Interest Elasticity of Velocity —.55
(.02)

Seven lags
Eigenvalues 1* .10 .03
Eigenvectors V (M & Y) 12.58 8.03
( R ) 7.20 2.48
Test for the Velocity Restriction: x2(1) .14
N .58

Implied Interest Elasticity of Velocity

-.58
(.05)

 

87

 

 

5.01” -’ —-’ — .- _ -

2.5 T

 

2.0 4- 1‘

 

MONEY BALANCE

1.03”

 

.ST

 

v-._.—-

01
48:3 58:3 68:3 78:3 88:3
__ Actual
__ __ Equilibrium

Figure 7. Actual and Equilibrium Real Money Balances

88
between them. Any changes in M, Y and R will be directly

reflected in ZM'
5. Exchange Rate Equilibrium

Unlike the long run money demand function, equilibrium
among P, F and X, has not been studied yet. The rationale for
the long run equilibrium relationship can be derived from the
PPP (Purchasing Power Parity) equation which can be written

as
(54) Xt = a0 + a1(Qt-Q;) + ut.

Qt' and 0; denote, respectively, the domestic and foreign
aggregate price indices. To maintain the terms of trade the
constant, 01 must be unity.

Because PPP tends to be rejected empirically when applied
to aggregate price indices (Frenkel, 1981; Branson, 1981;
Batten Belongia, 1984) , many scholars have replaced the
aggregate price indices with prices of specific commodity
groups. Protopapadakis and Stall (1983) have termed this the
LOP (Law of One Price). Frenkel (1981) disaggregated the
general price indice into prices of traded goods and non-
traded goods, and Isard (1977) disaggregated into
manufacturing goods and primary goods.

In its strictest form, the LOP implies ao=0 and al=1 in

89

(54) where Qt denotes disaggregated prices. The result can be
obtained with competitive behavior in the market;
instantaneous adjustment of prices and the exchange rate; a
high degree of homogeneity between products; and no trade
barriers and transportation costs. Though it is difficult to
find a commodity to fit all of these conditions, Isard (1977),
Protopapadakis and Stall (1983), Jabara and Schwartz (1987),
and Ardeni (1989) considered agricultural goods as the best
possible candidate. Empirical tests of the LOP for
agricultural goods have produced mixed results, depending on
the.commodity, time periods and countrie5‘used.in‘thelanalysis
(Jabara and Schwartz, 1987; Ardeni, 1989).

To test the LOP in agricultural goods, the PPP equation

(54) is expressed as
The aggregate price level is assumed to be an index of

industrial prices with weight 8 and farm prices with weight

1-8 (domestic), and 8* and 1-8* (foreign):

(56) Q = 8P + (l-B)F
and
(57) 0* = 6*P* + (1-B*)F*.

90

Subtracting (57) from (56) and rearranging terms yields

(58) F - F* = (Q - 0*) + B(F - P) - B*(F* - P*).

Substituting equation (58) into equation (55) yields
* 4 * *

When the aggregate price indices in the domestic and foreign
countries have a long run equilibrium relationship (ie., F;-
P; = zit) and the relative price index between sectors abroad
remain stable in the long run (ie., Qt-Q; = 22t)' the third
and the fourth terms in right hand side of (59) can be
replaced.with stationary error processes. Thus, equation (59)

becomes

(60) Xt = a0,t + a18(Ft—Pt) + u;
or, more generally,

(61) xt = as't + aFBFt - apspt+ ué.

For the U.S. economy, the weight of farm prices in the

aggregate price index is considered to be near zero4. With

 

4For the last three decades, the weight of farm prices
on aggregate price index (GNP implicit price deflator)
remained less than .045.

91

B=1, equation (61) becomes

(62) axXt + aPPt - aFFt = eXt'

where ax=1, aP=aF=al, and eXt=°0,t=ut‘ Therefore, the LOP
holds if ap=aF=ax=1.

The Johansen and Juselius method is applied to test the
existence of a long run equilibrium relationship among the
three variables. A dummy variable is introduced into the test
equation to take account of differences in the volatility of
the exchange rate before and after the 1973 exchange rate
system change. Two lags are adopted for the cointegration test
based on lag selection criteria reported in.Table 15. The test
statistics for cointegration are given in Table 16. The
estimated eigenvalues and eigenvectors are presented in Table
17 with parameter matrices. One cointegration is found at the
.05 level and. the long’ run equilibrium :relationship is
established as equation (62).

The coefficients, ax, up, and “F are obtained from the
first column in a vector in Table 17 and found significant by
the Wald test. To test the LOP hypothesis, Ho: ox = up = aF,
the likelihood ratio and Wald tests are used. The test
results, provided in Table 18, fail to reject the equality
restriction. Thus, the long run equilibrium equation (62) is

written as

92

Table 15. Lag Selection for AX, AP, and AF

 

 

 

Lag Criteria 2 1 Sig.
Length x (9) Level

p FP AIC 50

(x10 )

o .1283 -22.78 —22.77v

1 .1331 -22.67 ~22.50 56.91 .ooo***

2 .1202 -22.70 -22.36 23.54 .005***

3 .1167 -22.65 —22.14 9.75 .371

4 .1101 -22.64 ~21.96 15.58 .076*

5 .0879 -22.79 ~21.94 13.42 .144

6 .0840 -22.76 -21.74 12.42 .191

7 .0860 -22.67 -21.47 4.15 .901

8 .0645 -22.88 -21.52 26.90 .001***

9 .0615 -22.85 -21.32 11.46 .246

10 .0512 -22.97 -21.26 17.43 .042**
11 .0452V -23.02v -21.14 19.60 .021**

1.Likelihood ratio statistics to test HO: Lag length of p-1
vs. H : Lag length of p.
2.H is rejected at 10% level for *, at 5% for ** and at 1%
or ***.

3.v indicates minimum value for each information criteria.

93

Table 16. JJ Test Results for X, P, and F

 

 

HO -2an
h m Two lags Four lags Eight lags
o 3 34.71*** 42.52*** 28.63**
**
1 2 10.34 15.10 10.80
2 1 .03 .52 .28

 

Reject the null hypothesis at .01 level for ***, at .05 level
for **, and at .01 level for *.

94

 

Table 17. Eigenvalues and Eigenvectors for X, P, and F
Two lags
Eigenvalues A
.14 .06 .00
Eigenvectors V (=a)
X -5.79 -10.22 1.09
P -4.47 5.57 .90
F 5.61 -5.01 5.87
-sopv x 1000 (=n)
X -6.80 -3.84 .22
P -1.70 .39 -.05
F 3.31 -13.76 -.44
Four lags
Eigenvalues A
.16 .09 .00
Eigenvectors V (=a)
X 10.33 -8.05 .88
P 2.91 7.14 -1.00
F -4.02 -8.33 -6.41
-sopv x 1000 (=n)
X 7.24 -3.82 -.91
P 1.73 .91 .11
F 4.67 -14.74 2.13
Eight lags
Eigenvalues A
.11 .07 .00
Eigenvectors V (=a)
X 17.70 7.84 2.99
P -6.67 7.26 1.62
F 6.82 -8.26 -10.22
~50 v x 1000 (=n)
X 5.22 3.16 -.54
P -.28 1.05 .11
F 15.45 -4.72 1.35

 

95

Table 18. Test Results for the Exchange Rate Restriction

 

Two lags

Eigenvalues A* .14

Eigenvectors V 5.01

Test for the LOP Restriction: x2(1) .78
(a = a ) N .43
(a§=-a;) N .06

(aP=-aF) N .87

 

96

which is called the exchange rate equilibrium equation.

However, the credibility of the LOP hypothesis relies on
the stability of relative prices between sectors in other
countries and the stability of the aggregate price ratio
between countries. The results are also quite sensitive to
lag lengtht As shown in'Table 16, two cointegrations are found
with four lags at the .05 significance level and different
signs for the cointegrating vector are found with eight lags.
However, the four and eight lag models are not considered for
the cointegration test because lags from three to seven are
rejected strongly by the likelihood ratio test.

Based on the cointegrating relationship, three variables
in the exchange rate equilibrium equation can be merged into
a synthetic time series th = Ft - Pt - Xt, which represents
deviations from the long run exchange rate equilibrium. In
Figure 8, the actual and the equilibrium farm prices are
plotted. zxt, is represented. by ‘the ‘vertical differences
between them. Any changes in X, P and F will be directly

reflected in zx.

97

 

43' ,4 ,1“

FARM PRICES

 

 

2+
:1-
0.1
48:3 58:3 68:3 78:3
Actual
- _ _ Equilibrium

Figure 8. Actual and Equilibrium Farm Prices

 

138:3

CHAPTER V

A VAR MODEL OF FISCAL POLICY IMPACTS ON AGRICULTURE

1. Reduced Form Specification

As discussed in chapter III, cointegration between
variables can be used to restrict the VAR in the form of an
ECM, or to reduce number of variables in the form of ERM. The
appropriate form for an ERM will be equation (50) with three
stationary variables Dt' th, and th. Three lags were chosen
for the reduced form on the grounds of statistical tests
reported in Table 19.

Possible structural changes during the estimation period
are considered following Chambers and Just (1986) and Saunders
(1988). To test the significance of a structural change in
1973 due to the exchange rate regime change, and in 1979 due
to the Monetary Decontrol Act, the reduced form ERM is
estimated for two separate sample periods, before and after
the structural changes. Chow tests of the structural changes
are then applied. As shown in Table 20, the structural change
in 1973 has important effects on D and 2),. The structural
change in 1979 is only evident in the D equation. When the
impact of the 1973 structural change is removed by introducing
a dummy variable, the 1979 structural change is no longer
visible. Therefore, it is believed that a structural change

occurred in 1973, but not in 1979. The results are not

98

99

Table 19. Lag Selection for AD, AZ“, and AZx

 

 

Lag Criteria 2 1 Sig.
Length x (9) Level
p FP AIC sc
(x10 )
o .1941 -20.06 -20.06v
***
1 .2012 -19.95 -19.78 1856.12 .000
2 .1583 -20.12 -19.78 44.24 .ooo***
3 .1405 -2o.17 -19.66 26.37 .002***
4 .1303 -2o.17 -19.49 16.90 .050*
5 .1285 -2o.11 -19.26 8.03 .531
6 .1166 -2o.13 -19.12 17.05 .048**
7 .1125 -20.10 -18.91 10.40 .319
8 .0844 -20.31 -18.96 26.89 .001***
9 .0737 -2o.38V -18.85 15.48 .078*
10 .0738 -20.30 -l8.6l 5.41 .797
11 .0647v -20.36 -18.50 21.12 .012**

 

1.Likelihood ratio statistics to test Ho: Lag length of p-1
vs. Ha: Lag length of p.
2.H is rejected at 10% level for *, at 5% for ** and at 1%
or 44*,

3.v indicates minimum value for each information criteria.

100

Table 20. Test Results for Structural Change

 

 

 

Statistics Distribution Equation
1973 D zM zX
H51 F(S3,89) .746 .563 2.167***
cnow F(l3,135) 3.665*** .566 2.472***
1979
as1 F(29,113) .419 .501 1.888***
caow F(10,l42) 2.230** 1.402 1.287

 

1. Test of homoscadasticity is performed by comparing the
variances between two different sample periods.
2. The null hypothesis of homoscadasticity or no structural

change is rejected

e**,

at .05 level for ** and at .01 level for

101
sensitive to the number of lags.

The 1973 structural change is also identified in the
historical decomposition of D as shown in Table 21. The
structural change had positive impact on D implying an
increased federal budget deficit was created by the change.
No evidence of the structural changes was found in ZM and zx
using historical decomposition of these variables (Tables 22
and 23).

Though the three variable ERM has advantages over the
seven variable ECM in.model fitting and identification, it is
difficult to investigate the impact of fiscal policies on farm
and nonfarm prices using the ERM. The responses of P and F to
a shock in D will be reflected in movement of th and is not
easily identified. Replacing zXt with its component variables
turns the system into a five variable ERM and makes it
possible to detect any impact of fiscal variables on the farm
and nonfarm sectors.

With two stationary variables, D and ZM' and one
cointegrating vector among the nonstationary variables, X, P
and F, the number of unit roots contained in the system should
be no more than two. As shown in Table 24, two unit roots are
found.by the Johansen and.Juselius test. Two lags are selected
for estimation of the test equation based on likelihood ratio
tests criteria shown in Table 26. The results remained the
same with three lags through eight lags. The reduced form ERM

is estimated based on these results. The fiscal variable Dt

1172

Table 21. Historical Decoeposition of D

 

TIHE ACTUAL PROJECT PROJECT
ul 0973

1974:1
1974:2
1974:3
1974:4
1975:l
1975:2
1975:3
1975:4
1976:1
1976:2
l976:3
1976:4
1977:l
1977:2
1977:3
l977:4
1978:l
197B:2
1978:3
1978:4
1979:1
1979:2
1979:3
l979:4
1980:1
1980:2
1980:3
1980:4
1981:1
1981:2

(1)

0.00309
0.00726
0.00531
0.01543
0.03083
0.06402
0.03976
0.03921
0.03108
0.02759
0.03059
0.03081
0.02064
0.02133
0.02555
0.02473
0.02250
0.01134
0.01057
0.00862
0.00405
0.00242
0.00786
0.01107
0.01414
0.02414
0.02743
0.02387
0.01591
0.01630

12)

0.01328
0.01853
0.02403

ulo 0V73 D

13) (4) (5) 16)

2e Zx
0.00487 -0.01020 0.00000 0.00000
0.00230 -0.00965 -0.00112 -0.00049
0.00074 -0.01924 -0.00146 0.00197

0.02602 -0.00242 -0.00867 -0.00026 -0.00166

0.02713 -0.00472
0.02787 -0.00567
0.02847 -0.00537
0.02878 -0.00446
0.02895 -0.00330
0.02917 -0.00207 -0.00212

0.00241 -0.00032
0.00517 0.00119
0.00565 -0.00030
0.00242 -0.00062
0.00051 0.00041
0.00023 0.00031

0.00161
0.02979
0.00594
0.00863
0.00121

0.02949 -0.00088 0.00363 -0.00079 -0.00174

0.02989
0.03030
0.03071
0.03110
0.03144
0.03174
0.03199
0.03219
0.03235
0.03248
0.03259
0.03268
0.03275
0.03282
0.03288
0.03294
0.03298
0.03303
0.03306

0.00015 0.00496 -0.00257 -0.00147
0.00095 -0.00557 -0.00379 -0.00030
0.00152 -0.00202 -0.00508 '0.00228
0.00192 0.00284 -0.00669 -0.00170
0.00219 0.00287 -0.00894 -0.00065
0.00238 0.00154 -0.01109 0.00030
0.00252 -0.00576 -0.01196 -0.00292
0.00264 *0.00913 -0.01095 -0.00153
0.00275 -0.01342 -0.00999 -0.00033
0.00286 -0.01696 -0.00942 -0.00205
0.00298 -0.01874 -0.00855 -0.00287
0.00309 -0.01508 -0.00680 -0.00294
0.00319 -0.01143 -0.00521 -0.00503
0.00329 -0.00817 -0.00526 '0.00526
0.00337 0.00257 -0.00507 -0.00624
0.00345 0.00129 -0.00131 -0.00548
0.00351 -0.00128 0.00120 -0.00903
0.00356 -0.01070 -0.00066 -0.00575
0.00361 -0.01230 -0.00103 -0.00343

17)

FORECAST ERROR DUE TO

0973

0.00841
0.01623
0.02329
0.02844
0.03186
0.03354
0.03384
0.03324
0.03226
0.03124
0.03037
0.02974
0.02936
0.02919
0.02918
0.02925
0.02936
0.02947
0.02955
0.02960
0.02962

.0.02961

0.02959
0.02956
0.02953
0.02951
0.02949
0.02947
0.02946
0.02946

(8)

S.E. for

0973

0.00196 4
0.00377 4
0.00518 i
0.00586 !
0.00601 4
0.00589 9
0.00577 9
0.00574 4
0.00580 !
0.00593 4
0.00605 *
0.00615 4
0.00622 4
0.00628 4
0.00636 3
0.00644 4
0.00653 5
0.00662 4
0.00670 4
0.00679 4
0.00687 I
0.00695 4

'0.00703 5

0.00709 4
0.00715 5
0.00718 4
0.00721 4
0.00724 4
0.00726 *
0.00727 4

 

I :significant at
(1) is for actual values for D. (2) is for base projections and (3) for projections
set with DV73=0. Thus, the difference, 121-(31, produces (71, forecast errors of D
due to 0973. 14), (5), and (6) are forecast errors of 0 due to 0, 2e, and Zx,

respectively. The actual value 111 can be recovered by the sue of projected value
and forecast errors, that is, (3)+(4)+(5)+16l+(71. 18) is for the standard errors of
forecast error due to DV73.

.05 level.

1173

Table 22. Historical Decoeposition of Zn

 

(1)

121

(31

TIME ACTUAL PROJECT PROJECT

1974:l
1974:2
1974:3
1974:4
1975:1
1975:2
1975:3
1975:4
1976:1
l976:2
1976:3
1976:4
1977:1
1977:2
1977:3
l977:4
1978:1
197B:2
197B:3
197B:4
1979:1
1979:2
1979:3
1979:4
1980:1
1980:2
1980:3
19BO:4
1981:l
1981:2

-0.5659
-0.5335
-0.5435
-0.6l69
-0.7383
-0.7950
-0.7282
-0.8105
-0.9039
-0.8824
-0.8901
-O.9504
-0.9655
-0.9605
-0.9059
-0.8363
-0.8170
-0.8438
-0.7832
-0.7046
-0.6738
-0.6768
-0.6567
-0.5515
-0.4997
-0.6753
-0.6970
-0.4919
-0.5053
-0.4840

ul 0973

-0.6527
-0.6610
-0.6510
-0.6579
-0.6772
-0.6926
-0.7004
-0.7054
-0.7112
-0.7l71
-0.7220
-0.7257
-0.7290
-0.7322
-0.7353
-0.7380
-0.7405
-0.7427
-0.7447
-0.7465
-0.7480
-0.7494
-0.7505
-O.7516
-0.7524
-0.7532
-0.7539
-0.7545
-0.7550
-0.7554

u/o DV73

-0.6688
-0.6896
-0.6764
-0.6706
-0.6769
-0.6822
-0.6818
-0.6801
-0.6810
-0.6841
-0.6877
-0.6911
-0.6945
-0.6979
-0.7012
-0.7040
-0.7065
-0.7086
-0.7104
-0.7119
-0.7133
-0.7145
-0.7156
-O.7165
-0.7174
-0.7182
-0.7189
-0.7l95
-0.7201
-0.7206

(4)

15)

(6)

(7)

FORECAST ERROR DUE TO

0

0.0000
0.0090
0.0199
0.0335
0.0366
0.0230
-0.0148
-0.0290
-0.0204
-0.0100
-0.0041
-0.0074
-0.0142
-0.0080
0.0010
-0.0010
-0.0074
-0.0088
'0.0003
0.0121
0.0234
0.0339
0.0437
0.0481
0.0462
0.0414
0.0292
0.0182
0.0160
0.0252

ls

0.0868
0.1172
0.0891
0.0147
-0.0967
-0.1302
-0.0160
-0.0803
~0.1804
-0.1604
-0.1646
'0.2206
-0.2271
-0.2223
-0.1764
-0.1059
-0.0736
-0.0941
-0.0523
0.0049
0.0255
0.0130
0.0229
0.1255
0.1811
0.0133
0.0086
0.2223
0.1951
0.2014

lx

0.0000
0.0014
-0.0015
-0.0072
-0.0010
0.0049
0.0029
0.0043
0.0082
0.0052
0.0006
0.0033
0.0048
0.0020
0.0048
0.0087
0.0044
0.0019
0.0141
0.0249
0.0253
0.0258
0.0272
0.0264
0.0254
0.0233
0.0191
0.0221
0.0387
0.0449

0973

0.0161

0.0286

0.0254

0.0127
-0.0003
-0.0104
-0.0185
-0.0253
-0.0302
-0.0331
-0.0343
-0.0346
-0.0345
-0.0343
-0.0341
-0.0340
-0.0340
-0.0342
-0.0343
-0.0345
-0.0347
-0.0349
-0.0350
-0.0350
-0.0350
-0.0350

_-o.0350

-0.0350
-0.0349
-0.0349

(8)
S.E. for
0973

0.0197
0.0426
0.0563
0.0603
0.0606
0.0603
0.0598
0.0588
0.0579
0.0578
0.0582
0.0589
0.0596
0.0602
0.0606
0.0609
0.0611
0.0612
0.0613
0.0613
0.0614
0.0613
0.0613
0.0613
0.0613
0.0612
0.0612
0.0612
0.0611
0.0611

 

(l) is for actual values for Zn. (2) is for base projections and (3) for projections
set with 0973=0. Thus, the difference, 121-(3), produces (71, forecast errors of la
due to 0V73. 14), (5), and (6) are forecast errors of la due to 0, 2s, and Zx,
respectively. The Actual value 111 can be recovered by the sue of projected value
and forecast errors, that is, 131+(4)+(51+(6)+(71. (B) is for the standard errors of
forecast error due to 0973.

1174

Table 23. Historical Deco-position of 2x

 

(8)
S.E. for
0973

15) (6) (7)
FORECAST ERROR DUE TO

0973

(11 (2) 131 (4)

ACTUAL PROJECT PROJECT
w/ 0973 w/o 0973 0 ll Zx

TIHE

1974:1
1974:2
l974:3
l974:4
1975:l
1975:2
1975:3
1975:4
1976:1
1976:2
1976:3
1976:4
1977:1
1977:2
1977:3
1977:4
197B:1
l978:2
1978:3
197B:4
1979:1
1979:2
1979:3
1979:4
1980:1
l980:2
1980:3
1980:4
1981:l
1981:2

-4.0387
-4.2910
-4.2567
-4.2347
-4.3351
-4.3212
-4.2743
-4.3471
4.4419
-4.4170
-4.4179
-4.5143
-4.4549
-4.4018
-4.4140
-4.5136
-4.3388
-4.1927
-4.1786
-4.1168
-4.0598
-4.1107
'4.1206
-4.2057
-4.2979
-4.4539
-4.2440
-4.1904
-4.3345
-4.4590

-4.0827
-4.1519
-4.2164
-4.2706
-4.3163
-4.3587
-4.3985
-4.4338
-4.4640
-4.4900
-4.5125
-4.5323
-4.5495
-4.5644
-4.5774
-4.5889
-4.5990
-4.6079
-4.6158
-4.6226
-4.6286
-4.6338
-4.6384
-4.6423
-4.6457
-4.6487
-4.6513
-4.6535
-4.6554
-4.6571

-4.0906
-4.1662
-4.2365
-4.2924
-4.3347
-4.3698
-4.4010
-4.4283
-4.4518
'4.4725
-4.4916
-4.5093
-4.5256
-4.5406
-4.5541
“4.5662
-4.5771
-4.5866
-4.5950
-4.6023
-4.6086
-4.6141
-4.6189
-4.6231
-4.6268
-4.6300
-4.6328
-4.6353
-4.6374
-4.6393

0.0000
-0.0002
0.0004
0.0036
0.0078
0.0127
0.0135
0.0070
-0.0062
-0.0122
-0.0129
-0.0099
-0.0067
-0.0062
-0.0061
-0.0032
-0.0006
-0.0005
-0.0015
-0.0015
0.0014
0.0057
0.0104
0.0150
0.0184
0.0193
0.0176
0.0137
0.0070
0.0010

0.0000
-0.0142
-0.0157
-0.0190
-0.0213
-0.0168
-0.0181
-0.0276

0.0033

0.0212

0.0235

0.0462

0.0756

0.0944

0.1187

0.1372

0.1518

0.1676

0.1812

0.1767

0.1729

0.1705

0.1644

0.1497

0.1214

0.1038

0.1101

0.0739

0.0263

0.0308

0.0440
-0.1247
-0.0250

0.0513
-0.0053

0.0416

0.1288

0.1074

0.0251

0.0640

0.0841
-0.0183

0.0256

0.0743

0.0510
'0.0587

0.1090

0.2481

0.2575

0.3305

0.3945

0.3469

0.3429

0.2720

0.2080

0.0718

0.2795

0.3755

0.2877 .

0.1663

0.0080

0.0144

0.0201

0.0218

0.0184

0.0111

0.0026
-0.0054
-0.0122
-0.0174
-0.0210
-0.0230
-0.0238
-0.0238
-0.0233
-0.0227
-0.0220
-0.0213
-0.0208
-0.0203
-0.0200
-0.0197
-0.0194
-0.0192
-0.0189
-0.0187
-0.0185
-0.0182
-0.0180
-0.0178

0.0181
0.0377
0.0564
0.0712
0.0829
0.0920
0.0994
0.1057
0.1116
0.1176
0.1237
0.1299
0.1361
0.1422
0.1482
0.1538
0.1591
0.1640
0.1685
0.1725
0.1762
0.1796
0.5775
0.1854
0.1878
0.1901
0.1922
0.1941
0.1958
0.1974

 

(1) is for actual values for 2x. (2) is for base projections and (3) for projections
set with 0973=0. Thus, the difference, 121-(31, produces 17), forecast errors of 2x,
due to 0973. (4), (51, and (6) are forecast errors of 2x due to 0, la, and 1x,
respectively. The actual value (1) can be recovered by the sun of projected value
and forecast errors, that is, (3)+(4)+(5)+161+171. (B) is for the standard errors of
forecast error due to 0973.

105

Table 24. JJ Test Results for D, ZM' X, P, and F

 

 

Ho -21nQ
h m
o 5 117.46***
1 4 65.77***
2 3 25.93**
3 2 8.82
4 1 .10

 

**: Reject the null hypothesis at .05 level.
***: Reject the null hypothesis at .01 level.

106

Table 25. Eigenvalues and Eigenvectors for D, ZM' X, P,
and F

 

Two lags

Eigenvalues A

.27 .22 .10
Eigenvectors V (=a)
D -.60 -.56 -.00
Z .57 -.50 .52
XM -.31 -.14 .65
P -.17 .65 .55
F .43 .05 -.01
-SOPV x 1000 (=n)
D -2.79 -1.88 -1.11
2 2.51 -15.03 18.42
X” -2.39 6.59 3.48
P -1.34 1.48 .58
F 6.72 -4.45 5.30

 

Three largest eigenvalues and corresponding eigenvectors and
n vectors are appeared.

107

Table 26. Lag Selection for AD, AZM, AX, AP, and AF

 

 

Lag Criteria 2 1 Sig.
Length x (25) Level
FEE AIC sc
(x10 )
o .3843 -37.80 -37.80v
1 .4084 -37.49 -37.02 371.35 .ooo***
2 .2600 -37.70 -36.76 82.12 .000***
3 .2176 -37.64 -36.22 34.52 .097*
4 .1882 -37.54 -35.65 32.37 .147
5 .1180 -37.76 -35.40 49.59 .002***
6 .0990 -37.69 -34.86 32.48 .145
7 .0645 -37.88 -34.57 51.42 .001***
8 .0398 -38.12 -34.34 55.13 .000***
9 .0322 -38.08 -33.44 30.02 .223
10 .0231 -38.17 -33.44 39.68 .031**
11 .0178v -38.19v -32.99 29.64 .238

 

1.Likelihood ratio statistics to test H0: Lag length of p-1
vs. H : Lag length of p.
2.H is rejected at 10% level for *, at 5% for ** and at 1%
or ***.

3.v indicates minimum value for each information criteria.

108
will be replaced by its component variables, Gt and Tt in a
following section.
Table 27 provides summary statistics for the two lag

reduced form ERM. R2

s for the equations indicate a significant
proportion of the variation in dependent variables is
explained by the model. The statistics are obtained after the
model has been reparameterized to get an equivalent VAR in

levels. The ZM equation had the lowest R2

, as anticipated,
since it includes movement of three variables. The D and P
equations showed weak serial correlations based on Ljung-Box
Q statistics. The Ljung-Box Q statistics on squared error
terms, which are asymptotically equivalent to a LM test for
conditional heteroscedasticity, could not reject ARCH type
errors, except in the X series. However, it is not clear how
VAR methods, such as impulse response analysis, can be applied
with ARCH errors. No theoretical or empirical work has been
done in this area. Instead, OLS is applied and is consistent
(Engle, 1982). Overall, the summary statistics for the OLS
estimator imply that the ERM provides reasonably robust

statistical results.
2. Structural Form Identification
The reduced form is identified as the structural form if

the off-diagonal elements of the covariance matrix are zero.

However, the LM test rejected the null hypothesis of no

109

Table 27. Summary Statistics for Five Variable ERM

 

 

Stat's Dist'n Dependent Variable
D ZM X P F
R2 .884 .781 .941 .999 .983
DW 2.09 1.87 2.01 2.15 2.02

AC x2(20) 36.12** 27.71 17.90 37.03** 18.74

* * *

* 61.92** 28.38 89.02**

ARCH x2(20) 97.88** 38.30**

 

**: significant at .05 level.
***:significant at .01 level.

110
contemporaneous correlation at the .05 level (see Appendix E) .
Therefore, Fackler's (1988) maximum likelihood estimation
method is applied to identify the structural form.

A recursive order of D-P-X-F-ZM is considered and
estimated as shown.in.Table 28. This order allows for the most
possible influence of the federal deficit on other variables.
D is placed first in the order because the budget is set in
a long term perspective. Furthermore, fiscal policy affects
goods markets and money markets within a quarter because
agents adjust to a perceived policy changes quickly. This
order also allows effects from the goods market to the money
market within a quarter, assuming a more flexible money market
than goods market. Farm prices are ordered after industrial
prices, thus indicating farm prices are more flexible, as
Rausser (1985) has argued. ZM is ordered last in the order
because the interest rate in ZM is much more sensitive than
goods market prices. X is placed between P and ZM, since the
exchange rate equation reflects conditions of both goods and
financial markets. A recursive order of D-ZM-P-X-Fcan also
be used if the money supply in ZM is considered as a policy
variable, as discussed in the next section.

Estimates of n in Table 25 support the recursive order,
except for D and P. An order of P-D-X-F-ZM is suggested by the
speed.of adjustment parameter and this order’will be.discussed
in a following chapter.

A positive contemporaneous coefficient for D in the P,

Table 28. Estimates of Contemporaneous Parameters in

111

D-P-X-F-zM Recursive Structure

 

 

Dep. Explanatory Variables S.E.
Var.
D P X F ZM
D .0071
P .279 .0049
(.054)
X .462 .579 .0267
(.319) (.432)
F 2.687 1.591 -.247 .0598
(.718) (.971) (.176)
ZM 1.799 1.263 -.290 -.007 .0622

(.779)

(1.018) (.184) (.082)

 

Standard errors for the parameters

are in parenthesis.

112
X and F equations is expected because excess demand in the
goods market, caused by a federal deficit, induces
inflationary pressure. A positive coefficient for P in the F
equation is also expected because inflation in the industrial
sector induces inflationary pressure in the farm sector. The
coefficient for P in the X equation is not significant. The
negative contemporaneous coefficient of X in the F equation
is expected, as indicated by Schuh (1974, 1976). Direct
interpretations of the other parameters is difficult since ZM
consists of three different variables. The coefficient of D
in the F equation is much bigger than in the P equation,
supporting the assumption of flexible farm prices and fixed

industrial prices.

3. Dynamic Responses and Forecast Error Variance Decomposition

To detect dynamic responses of variables to the fiscal
policy shock, the reduced form ERM is reparameterized to its
equivalent VAR in levels. The response of each variable over
six year periods to a one standard deviation shock to each of
the variables are shown in Figure 9.

A positive shock to the D results in a disturbance in the
money market equilibrium as well as an increase in X. A sharp
decrease in F follows, causing a temporary cost price squeeze
in agriculture. F then increases back towards its long run

equilibrium level as P starts to decrease towards the new

1113

newcosmo om mun: <os.ucdo 96 o c msonx

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

o o nouuosuo N: zomuosmo x moueosuo t assesses
.2. _ .3 ...fi .3 .
.3. L ... .... .2:
w L / A .8 ....
7 .... .... .... _ .... p
o 3.2.. (1 ...: (\11 ....- ..SH .
n ...... ...; f... ....
r a...o-:.I.I.I.I.I¢l a...%.I-I.ol.l|.I.L a...“...:.1...a...1......cL ...-.1 .I.I_..I .I..I. I;
...-...: sac-...: oa....o: oao....:
possesses om u you n no a msonx .: mun: <u1.oo.o
a mean: N: meonx x meonw v macaw
v ... . .3 ...d .3
.04 .04 3‘4 . ..J;
r ... .3. .3 ..L
m .... “uni o... o... 14./III .... .111
m ...L ...? ...-L ...-H.—
u ...; ...-L ...: ......P L.
.... ...JJJ.I...I...L ....Arl “13.3431 -... ”1. u. ... ... ... .... ”In...n............:u..
... ... ... .3
m. .3 . .... 1 .....fl 3:
...+ on .0. .0.
. A A I \11111
a .... .... ...-q .... x
m ...... ll..\ a... HIIII .... ....
m a... o.... ...3 ...a
a I I I I1.L . II II I I c.1l c on I .I .4 . .L— 0.. elf-loo... oar-4.6;;
u 3L. n u bl: 2 .... .1» u ... ... 3 ...-#1” u ... ... ... a o a o 8 8 o.

swasuo m. HscsHmo womcosno mssonwos 0% mm: ten: Uimixiuinz

m zomcosoo
... J
.0; 1

..3
o .3

...: 1.1\

....A

 

 

 

 

.11..1..1..l . 1..-.L

a msonx

.ee ..IIIII .II ...... .I#
.04.—

... ..

o .8 II
...-

......

 

U... 45" 1-.-]‘01
. J o p. .- .-

   

 

.3 .. 1.1.1....) 1.93
.04“ ’
.8...

o .3 . .

....1

4......

.
...-F1191 .1.. 1 . . :1.
O J O a. u. D.

114
equilibrium. It takes about 2-3 years for the exchange rate
and both prices to reach equilibrium. Thus, the long run
equilibrium relationship among the three variables are
obtained.

A positive shock to P induces an increase in P and a
decrease in F initially. F then increases rapidly for a year
and reaches a new equilibrium with P after 2-3 years. A
positive shock to the money market equilibrium reduces F and
a positive shock to F increases F without affecting P very
mush. A positive shock in X decreases F initially and then
decreases P later. Responses of F are quicker and larger than
responses of P to any shock in the economy.

The impulse responses support the fixed industrial
prices-flexible agricultural prices paradigm discussed by
Rausser. The federal deficit brings a temporary cost price
squeeze in the farm economy which remains about 2-3 years. No
long run price squeeze is found. The exchange rate as well as
money market variables seem to be responsible for transferring
the policy shock to farm prices. A

The fixed industrial.prices-flexible.agricultural prices
paradigm is also supported by the decomposition of forecast
error variance. As shown in Table 29, The forecast error
variances in F are bigger than those in P. Own shocks explain
a relatively large proportion of the forecast error variance
for each series. The effect of D on F is immediate and steady

whereas the effect of ZM on F takes time.

12155

Table 29. Deco-position of Forecast Error Variance

 

Quarters
Ahead

Forecast
Error
Variance
in

12

16

12

16

20

Shocks to

 

0

0.757
(31.981

2.546
(31.05)

4.813
(27.04)

7.493
(23.86)

10.253
(21.31)

11.633
(7.66)

21.771
17.26)

25.187
15.93)

27.161
(5.13)

29.149
14.70)

ls

0.094
(3.96)

o. 392
(4.78)

0.603
13.391

0.669
12.13)

0.674
11.401

1.048
1.69)
10.736
(3.58)

24.550
15.78)

36.956
(6.981

46.328
(7.47)

0.059
12.501

0.702
18.561

2.869
116.12)

7.016
(22.34)

12.962
(26.94)

6.029
(3.97)

18.443
16.15)

27.736
(6.53)

32.667
16.171

34.669
(5.59)

P

1.380
(58.27)

3.894
147.49)

7.360
(41.35)

11.673
137.171

16.624
134.55)

1.625
(1.07)

6.268
(2.09)

10.534
(2.481

13.713
12.59)

15.939
(2.571

F

0.078
(3.29)

0.666
(8.12)

2.152
(12.09)

4.550
(14.49)

7.607
(15.81)
131.544
(86.62)

242.664
(80.92)

336.736
(79.28)

419.060
179.151

494.105
(79.67)

 

Nuebers in parenthesis are proportions of variance in Z teres.
Actual Variances x 10,000 are used for easy coeparison.

116

4. Sensitivity Analysis

4.1. Structural Identification

To investigate the question of whether the results are
robust to alternate identifications, other structural forms
are estimated. A recursive order of D-F-P-X-zM is considered
first. This order allows the most possible influence of the
agricultural sector on other sectors of the economy.
Industrial prices are believed to adjust faster than farm
prices, following' Just and Chambers (1987). The impulse
response functions remained virtually the same as before.
Thus, overshooting of F and sluggish adjustment of P are
supported by these results.

Other recursive orders tried include D-ZM-P-X-F and P-
D-X-F-ZM. The first order allows the money market to
contemporaneously affect the goods market, as well as more
flexible F than.P and the second order is implied by the speed
of adjustment parameter. Various other recursive orders, which
often don't have any natural economic justification, were also
tried. The results remained virtually the same unless D is
ordered after F.

With D ordered after F, the impulse response function
displayed a positive response of F after about two years to
a positive shock in D, as illustrateded in Figure 10. P

doesn't respond to the shock. The result is not consistent

1:17

momoosmo om moo: <o1Iocdo so u o msonx

rnozrm C

zomuosmom om c use a no a mass: 1: men: <~sisa3o

‘omrozi V

we

‘0me

c womcosmm

.... .

 

.......

.1.

 

/I|\\\I

-....4.

....H....._..............--..1

. a ops—au-

o m:onx

... .11-11.11.11. ..... 1.1L
..ag
....

 

......
...:

 

 

 

...-F1111. 111.1...L

0 ...—upon.

 

3
.UH”Wruuunuinnnnununuunnu;

 

 

...-LVNIJI 4|..ICI90‘L
. J 0 an .0 u.

N: xmmuosmm

 

j

 

”a

 

/r\\\\(

 

— I. I all. I.-.Ia-.I_

I J .-.o~o

N3 msonx

 

...:
.3 .
o .3.
o... .

 

”I;

 

.I.I.Il.|l..l...lL
o J a an .- no

 

...
.... .._

3&4

 

 

SeCJo

.... .C/fllillilil

. ...—up.»-

.0_ C—n—l

x ”success

:1 _

..u
. ... I

 

 

 

 

a ..u .4—

..u...— I—

::c.-;s.a.).z.
o a ops-cu.

x «eon:

 

...A
.3.—

HM“”A_......-.....-L

o a ..uu so".

 

 

a
. L

 

.... up

.;7_/II)\\\\\)))

 

”Mu.-.........-.-...._

. AOpUNCHC

 

t momuosoo
..e fl 11...- .. .11
L .
....
0.:
...-L
-1

 

 

o nope-on.

t msonx

 

..21
1.1
....

a
O‘IIII

 

 

 

......

..._ _
a.o..—..I.II.I..11......L
o a e u. .e u-

...
......
.3;

a... \IIII

-..;

...JL

 

 

 

 

A... 4 c u d e 1
c 4 u .u .e no

 

weasno Ho. Hemopmo womuosmo masonwo: 0% mm: ten: mioiulxlaz

m momcosmm

... .11.).-.11. ..... 11.1
..2
...

o .0. —
o..-

.s.‘ L
a... 4.1 .1 ......lo ... .n

OJO—U-ou.

 

 

 

 

 

m meonr

...

..s _

...“
.... _ (xiii-......IIIuiiiuu.
....
...a
0.: ..IJI .10-41114.1

 

. a Damp--

-.'e' e...sOO.-e ‘.'ed

  

 

I....F.I ..I.I. . .4312...
C 4 a an a. -

118

with the discussions of chapter II. Furthermore, it is hard
to believe that the federal government takes an expansionary
fiscal policy within a quarter whenever there is inflationary
pressure in farm prices, as implied by the positive
contemporaneous parameter estimates of F in the D equation.
Hence these structures are not considered appropriate.

Various simultaneous structural forms are also
considered. simultaneity between farm prices and industrial
prices is considered first. The maximum likelihood estimates
appearing in Table 30, gave similar results as the recursive
structures. Both price variables have positive contemporaneous
impacts on the other price, but the impact of F on P is much
larger than the impact of P on F. The impulse responses under
this structure remained the same as in Figure 9. Allowing for
the simultaneity between the prices in other structures
produced similar impulse responses. .

simultaneity between P and ZM is also considered, but the
contemporaneous parameter estimates have different signs from
the previous section and unrealistically large values for some
values as well as explosive impulse responses.

In summary, the implications drawn by the recursive
structure, D-P-X-F-ZM, remain quite stable over various

structural forms.

119

Table 30. Estimates of Contemporaneous Parameters in
a Simultaneous Structure

 

 

Dep. Explanatory Variables S.E.
Var.
D ZM X P F
D .0071
ZM 1.952 1.442 .0627
(-749) (1.015)
X .358 -.053 .501 .0265
(.330) (.033) (.491)
P .279 .000 .0049
(.092) (.032)
F 2.683 -.247 1.576 .0598
(1.525) (.176) (4.934)

 

Standard errors for the parameters are in parenthesis.

120

4.2. Lag Order

Three, four and.eight lags are considered for the reduced
form ERM estimation. As discussed before, the number of
cointegrating vectors among the variables remained the same
as in the two lag model. Although more cyclical movements are
added with more lags, the impulse response functions are quite
similar to Figure 9. Thus, implications drawn by the two lag

ERM remained stable.

4.3. Model Specification

Replacing ZM with its component variables turns the
system in a seven variable ECM. As shown in Table 31, four
cointegrating relationships at the .01 level and five
cointegrating relationships at the .05 level are found by the
Johansen and Juselius test. Though the results are not very
sensitive to the number of lags and cointegrating
relationships, a three lag ECM ‘with four cointegrating
relationships provides the most consistent impulse response
function.

Keeping the recursive structure D-P-X-F-ZM, as used in
the ERM, an order of D-M-P-Y-X-F-R is recommended to identify
the structural form ECM. According to estimates of the 1:
matrix shown in Table 32, R.has bigger values in the speed of

adjustment parameter than any other, followed by F and X.

Table 31. JJ Test Results for D, M, R, Y, P, X, and F

 

 

Ho -21nQ
h m
0 7 206.54***
1 5 145.79***
2 5 92.53***
3 4 51.83***
4 3 28.37**
5 2 11.28
6 1 .01

 

Reject null hypothesis at .05 level for

** and .01 for ***.

122

Table 32. Eigenvalues and Eigenvectors for D, M, R, Y, P, X,

 

 

and F
Eigenvalues A
.31 .28 .22 .14
Eigenvectors V (=0)
D -10.60 -97.62 -45.12 56.31
M 9.39 -6.42 10.64 -4.17
R 4.58 -2.41 -4.44 -2.03
Y -4.57 5.79 12.66 7.42
P -7.86 -1.35 -3.34 -7.56
X -5.15 10.45 -3.57 5.96
F 4.86 4.99 -.96 5.75
-sopv x 1000 (=n)
D .29 -2.93 .71 1.67
M 3.06 -1.31 1.16 -.97
R 15.22 4.88 -39.43 5.40
Y 2.39 .49 -2.11 -.69
P -2.56 -.25 -.79 -.84
X -3.90 4.56 3.34 3.50
F 2.90 15.12 -3.67 5.98
Four largest eigenvalues and corresponding eigenvectors and

n vectors are appeared.

123

However, it is not obvious how to order the other variables
based on the adjustment parameters. The policy variables M are
assumed to have a slower response than Y and P on the grounds
that agents adjust to a perceived policy change quickly where
as the monetary authority has a more complicated decision
' process. Among the policy variables, M is believed to be more
flexible than D since the Federal Reserve Board sets monetary
targets more frequently than the annually set government
budget.

To detect the dynamic responses of variable to the fiscal
policy shock, the reduced form ECM is reparameterized to its
equivalent VAR in levels of the variables. The impulse
response functions from the recursive ECM provide very similar
implications to the ERM analysis. As shown in Figure 11, a
positive shock in D leads to a sharp decrease in F, leaving
the farm economy in a temporary cost price squeeze. F then
starts to increases towards its long run equilibrium level.
Both prices reached the same equilibrium level after about 2
years. The response of F is much bigger than the response of
P to any shock in the economy. Again Rausser's fixed
industrial prices-flexible agricultural prices paradigm is
supported.

As expected, an increase in R and X has strong and
permanent negative impacts on F. An increase in Y and M
induces a temporary overshooting of farm prices which vanishes

after about two years. A positive shock to P causes a decrease

1J24

xmmcosmm on man: <~s1~a1m no u o mgonx

 

 

 

 

  

 

 

   

 

 

 

 

 

 

 

 

o xmmucsmm : ammuosmm x «mmuoamm x mamuoawm < ammoosmm v ammuoamm m ammuoamm
o .... ...-II...- ... . .. .... ...III..-....-. 11--.5..- l ... I
.1 1.1 ...1 .111 1.1.
W 3”“ ...: o. .. /\ III .... 1i _ .. ..._ _ .. ...“ .l\ ._
o ..... “...; J Hun" .1 .uQ . ...... _. . fun“ _ . I”... _
n ...: ...-1.1:...L .... -1.-11......- ........-..-...1..-.. .... .......-.......L. ....._.I....-......-... .... ..--.11....- ....1.-.......!-.L
r c u a. c a .u o o .o o o .n o o .o o o .o o o .-

xmmuoamam om v mag m no a mscnr 1: man: <~11~o1m

 

 

 

 

 

 

 

 

 

 

 

 

o msonx : mgonx a msonx x macaw < manor 6 wronr m msonx

.0 ...I ........ I. .. .. ...-n.3,..- .... . Isl-.... . 1.1.-1...... . .
x ..J ... ...»... ...." .... .... ...
m . ..IflJ.... 1| ... 11...“- ....M ....1}_ 33th

.. .. .... . .. .. .... .. . 3
. .111 .. ..1 .1 -_...1 4.1.1.1 1 _
U o .- ..l'.l¢l...l..l— c...»...!t.u.1l.llo.l1 o... 1.0.4.11... I. I .- ...Cl...l.l|oolto a... 3"...lalch a... onlol...lcca.xl-oo ...—Halo...11||11.l.—

. . . 1. . . .. c o .. o . .. o a .. o . .. - u _.

i
!

 

1.1

1111-...-II1 ....1. _ 1.1.1.- ...... . 111-1.1-4 .111. 1 1.11.1.1 It. 1.1.1.4121!!!“
.11W-\IIJH111J1M111/ .111jn..Hi...1.1..1-. {—1.151. 1

 

 

 

....h — 0.. c C.— . ...”
...b.a.1 «3 4|.0L I cub.-- a... ..Ilataololld I... 1.31.1.l0
. a .o o a. O o u. c 0 a. o .. .... a. o o. -.

UMCDFU

muocno Hp. Hsvswmm womvosmo masonwos UK mm<o= <mnHmUHm mnz

125
in F for very short periods. The cost price squeeze is removed
when P quickly overshoots and then returns to its long run
equilibrium level. A positive shock in F resulted in a
favorable situation for the farm economy.

Though many scholars (Schuh, 1981, 1983, 1984a, 1984b;
Rausser, 1985; and Barclay and Tweeten, 1988) have argued that
interest rates are a major transmital mechanism of
macroeconomic policy shocks to the economy, the interest rate
responds to the policy shock much differently than expected.
R falls as the federal deficit increases. Increased money
supply after the fiscal shock may be the main cause. As the
‘monetarists (Hamburger and.Zwick, 1981, 1982; Allen.and.Smith,
1983) have argued, monetization of the deficit is evident in
the impulse response function. When increased federal deficit
puts pressure on the interest rate and hence decreases output
and farm prices, the monetary authority tries to remove the
pressure by increasing the money supply. Interest rates, which
have the largest speed of adjustment among any of the
variables, immediately fall in response to a monetary shock.
Farm prices start to increase right after the shock. After
about 2-4 quarters, the exchange rate decreases and output
increases. Thus, farm prices would have fallen further, as a
result of the higher interest rates and exchange rates,
without the monetization.

Shocks to the financial variables (R and X) dominate

shocks to the policy variables (D and M) in their effects on

126
farm prices. However, there was no single dominant factor
affecting farm prices in the ECM, besides its own shock.

A‘VAR in levels produced impulse responses quite similar
to the ECM, but the VAR with first differences had different
impulse responses.

The impulse responses with seven variable VAR models were
quite sensitive to the lags selected. The ECM with lags two,
four, six, seven or eight were tried, but failed to generate
a consistent response function. Thus, it is verified that VAR
analysis becomes more sensitive to the number of lags selected

as the number of variables increases.
5. Implications from the Analysis

A number of implications can be drawn regardless of the
models, structural forms, lag orders and specifications that
have been considered.

First, a consistent result on the impact of the federal
deficit on prices has been found. Farm prices are reduced
sharply due to an unexpected increase in the federal deficit,
while industrial prices respond only gradually. This results
in a cost price squeeze on the farm economy. However, both
prices moved toward and reach a long run equilibrium level
after about 2-3 years.

Second, farm prices are found to be more responsive to

compared to industrial prices. The results support the fixed

127
industrial prices-flexible agricultural prices paradigm
suggested by Rausser and others.

Third, the interest rate and the exchange rate are major
transmital mechanisms for fiscal shocks to the farm economy.
Expansionary monetary policy after a federal deficit increase
causes interest rates to fall. However, the interest rate and
the exchange rate have a substantial negative impact on the
farm prices and respond significantly to policy shocks.

Thus, the continuous increases in federal deficits in the
1980's appear to have a significant negative impact on
agriculture through their impacts on the macroeconomic

variables.

VI. COMPARISON O? DEFICIT REDUCTION POLICIES

Three broad options are considered for reducing the deficit.
As discussed in Just and Chambers (1987), these are reduction in
government spending, a tax increase, and monetization. Recently,
the U.S. government has attempted to undertake tight monetary
policy with spending reductions to cut the federal deficit. The
policy is likely to be changed in the near future as the Bush
administration is considering a tax increase. In this chapter, some
of the effects of each option will be examined using the

econometric model developed earlier.

1. Impact of Deficit Reduction Options on Agriculture

To compare the impact of the three policy alternatives on the
economy, fiscal variables G and T are introduced into the ECM. By
replacing D with G and T, the model now consists of eight
variables. As list in Table 33, five cointegrating vectors are
detected at the .01 level. The reduced form ECM is estimated and
reparameterized to its equivalent VAR in levels. A recursive order
G-T-M—P-Y—X-F-R is established to identify the structural form. G
is placed before T since T is more variant to economic conditions.
This is also supported by the error correction parameter matrix H,
appearing in Table 34.

The ECM produced impulse responses that are consistent with

models in the previous section. As shown in Figure 12, all three

128

129

Table 33. JJ Test Results for G, T, H, R, Y, P, X, and F

 

 

HO -21nQ
h m
o 8 245.08***
1 7 176.39***
2 6 121.10***
3 5 81.48***
4 4 52.64***
5 3 24.76**
6 2 7.66
7 1 .08

 

Reject null hypothesis at .05 level for ** and .01 for ***.

130

 

Table 34. Eigenvalues and Eigenvectors for G, T, M, R, Y, P,
X, and F
Eigenvalues A
.35 .29 .22 .16 .16
Eigenvectors V (=a)
G -5.60 18.16 -ll.23 9.88 -3.00
T 5.29 -21.02 8.94 -8.28 -23.77
M 9.08 6.57 -7.10 -6.18 -3.52
R 4.02 2.07 -5.29 -1.91 .73
Y -3.71 -4.46 14.55 8.05 -3.19
P -7.84 .55 -3.22 -7.55 2.78
X -4.86 -10.25 -1.54 5.46 1.70
F 5.26 -4.03 -.04 6.19 3.46
'SOpV x 1000 (=n)

G .02 12.32 -.14 6.69 -1.17
T -1067 -3049 -2057 025 -7063
M 3.25 1.32 .61 -l.23 -.75
R 12.89 -12.00 -37.98 5.32 2.44
Y 2.14 -.97 -2.32 -.39 -1.83
P -2.70 .08 -.68 -.83 -.30
X -3.75 -3.80 3.93 3.49 -3.83
F 2.18 -12.09 -3.31 7.71 -2.59

 

Five largest eigenvalues and corresponding eigenvectors and
n vectors are appeared.

131

C)

RODS“!

7:00:20! 3 35:19:90 *3

: ammwozum

 

.ou._.1 1

.3.”

.3.+Lnuuuillllur

I o.”H-I.Il.l1~. all. clean-L
. O a.

 

 

3

o .9. .

 

 

...”:‘1‘44 ‘ ‘ q ‘

 

a” I I? E: b
o

O.”
. o.-

 

 

-3sl

 

 

 

|.ou ‘1 < ‘1}

1
o o u-

..u g =32.sz

.._r

...F
O 111‘

 

 

 

-3.~ _

....» he}... 1.1.1.1....
.

 

 

 

 

- p-
..udllspr r
....“
...-whiA M
.-3.u .
a..u.—.1¢11<1111I‘L
- u no
.311--- -i} p -
_
3.. m
...-“hurlnnnnuuunnuw
I.-.._ a

[

w ammwozmm
a. , - -

a».
.3.
-3».
-33.

-... Lr1... O ‘0' ‘9... ‘0‘

 

 

L.m_.“-_4

....
.3.
1.4

 

In. ...- ... ...—-

 

 

 

3
M
a . Loo...

 

mHQsHo Hm. Havcwmo womvosmo masonwo: 6% award <mnwmowm mo:

1 ammwozmm

 

 

 

o a no

 

 

 

 

 

..uA
0.0»;
0.0;;
.3. - - . - .

 

 

 

132

policy measures have similar impacts on output, increasing it
initially but decreasing it after about a year. Among the policy
alternatives, a spending reduction has the least impact and tax
increase has the most impact on output. All three options have a
similar impact on farm prices, increasing initially but eventually
decreasing to their long run equilibrium level. Spending reductions
and monetization induce farm prices to be stabilized at the
equilibrium after about two years. All the policy measures,
especially monetization, have little impact on industrial prices.

As Tatom (1985) and Just and Chambers (1987) suggest, the
spending reduction seems the most favored deficit reduction policy.
It provides the most stable environment for both the macroeconomy
and agriculture, while tax increases contribute to unstable

economic conditions.

2. Policy simulations

The implications drawn in previous section rely on economic
conditions at hand. A simulation study bearing current economic
conditions is set to compare how the economy will evolve under
different policy scenarios. A base projection is set for 1991—1995
periods using the eight variable ECM. Simulated projections are,
then, drawn under each different policy option and compared to the
base projection. The reduced form ECM is used for projections.

In Figures 13-15, base projections for output, nonfarm prices

and farm prices were compared with the projected time paths of the

133

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

L
0
0
B
E .
fl
1. a.
1 a.
N
c 3.
g ..
amusfif, fi,rflfi-rfi 1,.r-
3 3 ‘l 1 3 3 '1 1 3 3 'l 1 3 3 'l 1 3 3 ‘l
1330 1331 1332 1333 133%
L 3.33 o . .
0
6 3.01— . /
,1, ......
0
. 1.33-—
E 1.33-
I
I: '1."—
E
‘.I§—--—-c*v ‘ —r v~ #— T v fl 4 r—‘v w +fi 1 . 1
3 3 1- 3 3 'l 1 3 3 1 1 3 3 'l 1 3 3 ‘l
1330 1331 1332 1333 133%
'0‘ 1.33 g
. ......J
F
g ......
11
P 1.31-
H
g s...—
E
"-" 'fi I‘fi'_l'**1*"'1_'fiT
3 'I I 3 3 ‘l 1 3 3 ‘1 I 3 3 'l I 3 3 '1
1330 1331 1332 1333 1333

Figure 13. Effects of Spending Reduction

KM

 

3.33-—

 

MZODZN ram: 003"
3.
u
M

 

 

 

333%

!

a."—

3.v—l

373':

° "fi I r ' T ' r f r * 'fif r w
33113311331133‘11331
1380 1331 1332 1333 ISSH

 

..‘c—u

0 DZ." not.
.

”0"."-

‘0’:—

‘o" —'

"no-‘3'!

 

'IJ‘I

 

 

 

‘0’:-

1.31%

manaw 3231 nor
.8

 

 

 

‘*————

 

I
3 1 1

a a ;
1332

Figure 14. Effects of Tax Increase

3133'!
1330 1331

r i

a : §
1333

135

 

 

”SDHZH ram: CIDI"
I.
I
N

 

 

3 3 1 a a 1 x a a 1 a a a 1 1 a a 1
1330 1331 1332 1333 1333

 

I.” _

'.'.-d

o 33:" DD!“
3.

‘0’.—

1.33-—

 

1.33-:

 

flint-1:!"

 

1.11 , . .7 , ‘_ 1
a a 1 t a
15330

a 1
13531

 

1...-..7..-

 

mnnaw 333'" OD!“
.3

 

 

 

I —r v v— r V v v j V V W
3

a 1 a a ‘1 x a a 1 1 a s 1 a 1
1330 1331 1332 1333 1333

Figure 15. Effects of Monetization

3 3

136

variables under different policy options. Three options are
considered: a 5% increase in the money supply, a 5% reduction in
spending, and a 5% increase in taxes, all occuring in 1991. As
shown in Figure 13, no significant impacts are found in each of the
projected time paths as a result of a spending reduction. A 5% tax
increase resulted.in a‘depressed farm sector as shown in Figure 14.
Though its impact on industrial prices are minimal, a large
decrease in output and farm prices occur. A 5% increase in the
money supply increases industrial prices little, as shown in Figure
15. However, the policy is favorable to the farm sector since it
initially increases farm prices considerably. Farm prices then move
back to their normal level, after the first quarter of 1993. The
policy also lifts the output of the economy a small amount in the
second and third quarter of 1991, but decreases output for the rest
of the period.

The simulation study thus suggests that a spending reduction
is the most favored option for tackling the current deficit. A
monetary expansion would mostly favor the farm sector. A tax
increase is the least favored policy. A recession in both the farm

and nonfarm economy could be expected with this option.

VII. SUMMARY AND CONCLUSION

Macroeconomists have different opinions on how fiscal policy
affects the economy. The different macroeconomic views have also
resulted in wide disagreements on how fiscal policies affect
agriculture. Schuh (1981, 1983) and Barclay and TWeeten (1988),
following the Keynesian hypothesis, argued that an increase in the
federal deficit causes unfavorable conditions for the farm economy
by increasing interest rates and the exchange rate. This leads to
decreases in agricultural exports, prices, and income. Others,
following the neoclassical model, disagree with this Keynesian
explanation. Belongia and Stone (1985) and Batten and Belongia
(1986) rejected any possible connection between the federal deficit
and the farm economy. Applying the neo-Keynesian differential price
adjustment, Rausser (1985) and Rausser, Chalfant, Love and
Stamoulis (1986) supported Schuh's conjecture partly. They argued
that the federal deficit, due to sticky industrial prices, has the
same short.run impact on the farm economy as tight.monetary policy,
which decreases farm prices. However, they also supported the
neutrality argument in the long run. Just and Chambers (1987) also
used the neo-Keynesian idea, but considered farm prices as sticky,
because of farm programs. Their results were just the opposite to
Rausser's in the short run.

The main objective of this study is to evaluate the different
explanations of deficit effects on agriculture using the VAR

analysis pioneered by Sims (1980). Modifications are made on the

137

138

VAR model because of unit roots and structural changes found. A
unit root is evident in all series except the fiscal variables. The
long run money balance (LM) and exchange rate equilibrium,
established with Johansen and Juselius cointegration tests, play
an important role in the modification. A structural change in the
federal budget deficit variable, found by Chow Test and Mazon's
Simulation method, is also considered in the reduced form ERM
estimation.

The LM test rejected the null hypothesis of no contemporaneous
correlations between reduced form error terms. A sequential
ordering is established based on the priors and the maximum
likelihood method by Fackler is used to identify the structural
form ERM.

The results from the impulse response analysis suggest a
substantial impact of the federal deficit on farm prices in the
short run. A sharp decrease in farm prices occurs as a result of
an increase in the federal deficit, leaving the farm economy in a
cost.price squeeze. However, the price decrease remains only in the
short run. Farm prices move back to the long run equilibrium price
level after the initial shock.and.reach.equilibrium‘after about two
or three years. Thus, no long run changes in the relative position
between farm prices and industrial prices are detected. The short
run impact of the federal deficit occurs mainly through its impact
on interest rates and the exchange rate, as conjectured by Schuh
and others.

The results are quite consistent with the results of Rausser

139
and others using a fixed industrial price and flexible farm price
model. Rausser's explanation is also supported by the quick and
substantial movement in farm prices due to other changes in the
macroeconomic environment, including the money supply.

The decomposition of forecast error variances also supports
the fixed industrial prices and flexible farm prices. The forecast
error variances for farm prices are much bigger than those for
industrial prices, regardless of shocks in the economy.

Results are not very sensitive to the type of structural form
applied. Various recursive and simultaneous structures were
applied, but the impulse responses remained stable. The impulse
responses also remained stable with the ECM and the VAR in levels
when lags for the models are selected based on the ERM. The VAR in
first differences performed poorly.

The impulse responses were not sensitive to different lag
structures in the ERM. Since more variables are included in the ECM
and the VAR, results become more sensitive to the number of lags
in these model. The lag selection criteria also didn't work well
in that case, as Nickelsburg (1985) and Hsiao (1979) have verified.
Thus, the number of lags for large systems are selected based on
consistency in performance compared to the small ERM model. As the
number of variables increases, the results from the VAR with levels
become closer to those of the ECM.

An eight variable ECM is estimated to compare three policy
alternatives for reducing the budget deficit: a spending reduction,

a tax increase and monetization. A tax increase has big negative

140
impacts on real output while a spending reduction has minimal
impacts.

Results from simulation over the next five years confirmed
this conclusion. A spending reduction scheme does not have much
impact on output and prices in the simulation. A tax increase
results in.a big slump in the macroeconomy and the farm sector. The
simulation also identified that monetization is a favorable option
for the farm sector since it initially increases farm prices
considerably. However, the monetization option induces a decrease
in real output after its initial expansionary impact on the economy
has disappeared. It also induces inflation for the economy. Thus,
the simulation.experiment suggests that a spending reduction is the
most favored option for tackling the current budget deficit and a
tax increase is the least favored option.

The implications drawn from the VAR analysis have certain
limitations. Results are only credible where no further structural
changes are expected. With the U.S. economy now experiencing the
unprecedented triple digit federal deficit, the results might be
different from these simulated, since the linear structural
relationship of time series analysis may not catch the structural
changes. Other farm indicators including export and income must be
included to draw a better picture on how fiscal policies affect

farm economy.

Appendices

Appendix A

Data Sources

 

 

Table 35. Data Sources
Variable Period Source
R 1948 Federal Reserve Bulletin
1949 - 1970 Banking and Monetary Statistics
1971 - Treasury Bulletin
Annual Statistical Digest
X 1948 - 1966 International Financial §tatistics1
1967 - 1978 Federal Reserve Bulletin
1979 - Annual Statistical Digest
Federal Reserve Bulletin
GNP 1948 - 1982 The National Income and Product
(nominal) Accounts of the United States
1983 - Survey of Current Business
Implicit 1948 - 1982 The National Income and Product
Price Accounts of the United States
Deflator 1983 - Survey of Current Business
for GNP
P 1948 - 1982 The National Income and Product
Account of the United States
1983 Survey of Current Business
F 1948 - 1982 The National Income and Product
Account of the United States
1983 Survey of Current Business
Federal 1948 - 1982 The National Income and Product
Tax Account of the United States
Receipts 1983 Survey of Current Business
Federal 1948 - 1982 The National Income and Product
spendings Account of the United States
1983 Survey of Current Business
Federal 1948 - 1982 The National Income and Product

(continued)

141

142

Table 35 (Cont'd).

 

 

Variable Period Source

Deficit Account of the United States
1983 Survey of Current Business

M1 1948 - 1960 Federal Reserve Bulletin3
1961 - Jun. 89 Citibank Data Base 1990
Jul. 89 - Federal Reserve Bulletin

 

Data obtained from the sources are most updated version and
selected issues for the source publications are used unless
notified. XR, IR, and M1 data are created by average of
three momthly data.

1.Generated by the formula in Federal Reserve Bulletin Aug.
1978 issue (p. 700)

2.Data from Aug. 1978 issue of the Bulletin.

3.Data from Oct. 1969 issue of the Bulletin.

The test
2(t- ) =
Z(ta*)=
2(tal)=
2(61) =
2(92) =
2(Q3) =

where

ta
ta*
ta'
’1
Q2
’3

*
a

I
a

*
u

Appendix B

Definition and Critical Values for
Phillips-Perron Test Statistics

statistics are defined by

(su/sn1)t;-(1/2sn1)(831-83)[n'22(yt-1-Y-1)21'1’2
- 2 _ 2 -2 _— 2.-1/2
(su/sn1>ta. - (n3/4<3Dy )1/23 n1)<sf,1-s 3)
(83/S§1>¢1 - (1/zsn1)(S§1—s§){n(a* -1)
- 1/4(s§1-su)tn 22(yt-1-Y-1)21'1}
(Si/sn1)92 - (1/3sn1)(sn1-su)[n(a'-1)
- (n6/480y)(s§1-S§)1
(sfi/sfil)¢3 - (1/2sfil)(s§1-sfi)[n(a'-1)

- (n6 /480 y)(sn1- 83)]

= (E - a){2(yt-1-Y-1)2}1/2/§
= (a* - a){z(yt_1-Y-1)2}1’2/s*

= (a' - a) / (s'2c3>1/2

= (25*2)'1{ns§ - ns*2}

= (3s'2)‘1(nsg - ns'z)
= (zs'z)'1(nsg -n(§o - ?_1)2- ns'z)
= ztyt - ?o)yt-1/2(yt-1 - ?_1)2

-1

= (Xvi-1) 2mm

= YO - a v-1

143

144

Y = n"1 E Yt-i (i=0,1)

and §, 8* and S. are the standard errors of regression (32),

(33) and (34), respectively.

|
S0 is 8' when a = 1. Si is a consistent estimator of a

1 2

2

u = 11m

1

n' 2 B(ui) and 8:1 is a consistent estimator of a =lim n-

B(Sfi) under the appropriate null hypothesis, where 3n = 2 ut.

n :represents 'the. number' of observations. The consistent

2

estimation of 0 concerns the appropriate choice of truncation

lag parameter. Though the choice will be an empirical matter,
Perron (1986) recommended to inspection of the sample
autocorrelation of first differenced data. In this paper, the
LR in first differences is used together with the
recommendation. C. is the (i,i) element of the matrix

1

OUT)“1 and D denotes the determinant of the (Y'Y) which is

Y
represented as

_ 2 2_ 2 _ 2
Dy — (n (n 1)/12) 2yt_1 n(2tyt-1) + n(n+1) Eth-1 EYt-1

-(n(n+l)(2n+l)/6) (2yt_1)2.

The critical values for the test statistics are presented as

145

Table 36. Critical values for PP Test Statistics

 

Test Statistics Percentiles

10% 5% 2.5%
Z(§2) 4.03 4.68 5.31
2(63) 5.34 6.25 7.16
2(t5) -3.12 -3.41 -3.66
Z(Q1) 3.78 4.59 5.38
2(a*) -2.57 -2.86 -3.12
Z(a,) -l.62 -1.95 -2.23

 

Appendix c

Critical values for Schmidt-Phillips Test Statistics

Table 37. Critical Values for SP Test Statistics

 

n Percentiles
10% 5% 2.5% 1%
25 -2.85 -3.18 -3.50 -3.90
50 -2.80 -3.11 -3.39 -3.73
100 -2.77 -3.06 -3.32 -3.63
200 -2.76 -3.04 -3.30 -3.61
500 -2.76 -3.04 -3.29 -3.59
1000 -2.75 -3.02 -3.28 -3.58
2000 -2.75 -3.02 -3.27 -3.56

 

146

Table 38. Critical Values for JJ Test Statistics

Appendix D

Cointegration Test Statistics

Critical Values for Johansen-Juselius

 

5%

154.3

103.1

78.1

57.2

38.6

23.8

12.0

percentiles

41.2

26.1

13.9

1%

165.2

112.7

86.6

63.9

44.5

28.5

15.6

5.3

 

147

Appendix 3

Ln Test of contemporaneous Correlation

The LN statistic for testing Ho: n=I against H1: n+1 is

given by
g i-l 2
X = n2 2p..
LN i=2 j=1 13
where
2
_ oij
Pij - -
011011

The test statistic is distributed as x2

with g(g-l)/2 degree
of freedom. In the five variable ERM estimated in Chapter V,
the value of the LM test statistic was 48.55508 which is

significant at the .01.

148

Bibliography

BIBLIOGRAPHY

Akakike, Hizotogu. "Fitting Autoregressive Models for
Prediction." ls o st'tu e f t ' ' a

Marhgmarig_. 1969. pp.243-247.

Allen, Stuart and Michael Smith. "Government Borrowing and
Monetary Accommodation." J; Monetary Econ. 12(1983):605-
616.

Andrews, Margaret S. and Gordon C. Rausser. "Some Political
Economy Aspects of Macroeconomic Linkages with

Agriculture." Amer. J: Agr. Ecgm. 68(1986):4l3-417.

Ardeni, Pier Giorgio. "Does the Law of One Price Really Hold

for Commodity Prices?" Amer. J. Agr. Eggm. 71(1989):661-
669

Baillie, R. T. "Commodity Prices and Aggregate Inflation:
Would A Commodity Price Rule be Worthwhile?"
Econometrics and Economic Theory Pap. No. 8808. Dept.
of Econ. Mich. State Univ. E. Lansing. MI. April. 1989.

Baillie, R.T. and Tim Bollerslev. "The Message in Daily
Exchange Rates: A Conditional Variance Tale." Q, 9:

Business gag Eggn. §tg . 7(1989):295-307.

Baillie, R. T. and Robert J. Myers. "Modeling Commodity Price
Distribution and Estimating the Optimal Future Hedge."

MM- Forthcoming 1991 -

Barclay, Tom and Tweeten, Luther. "Macroeconomic Policy
Impacts on United States Agriculture: A Simulation

Analysis." Agrggm;rmr§; Ecgnomics. 1(1988):29l-307.

Barr, Wallace. "Inflation, Agriculture, and the Food Economy:

Discussion." Amer. J, Agr, Econ. 61(1979):909-911.

Barro, Robert J. "Are Government Bonds Net Wealth?" o i .
£999. 82(1974):1095-1117.

Barro, Robert J. "Unanticipated.Money Growth and Unemployment
in the United States." Amer. Econ, Rev. 67(1977):101-
115.

149

150

Barro, Robert J. "Comment from an Unconstructed Ricardian."

m. Menetary Ecen, 4(1978a):569-581.

Barro, Robert J. "Unanticipated Money, Output, and the Price
Level in the United States." Q. Polit, Eeen. 86(1978b):
549-580.

Barro, Robert J.'Macreeconomies, 2nd. New York: John Wiley &
Sons. 1984.

Barro, Robert J. neereeeenemiee, 3rd. New York: John Wiley &
Sons. 1987. .

Batten, Dallas S. and Michael T. Belongia. "Monetary Policy,
Real Exchange Rate and U.S. Agricultural Exports."

Amer, g. Agr. Econ. 68(1986):422-427.

Batten, Dallas S. and Daniel L. Thoronton. "How Robust Are
the Policy Conclusions of the St. Louis Equation?: Some

Further Evidence." Fed. Bee, flank of St, Lenis Rev.
(1984: June/July):26-32.

Bell, Daniel and Irving Kristol. WW
Theery. New York: Basic Books Inc. 1981.

Belongia, Michael T. "The Impact of Inflation on the Real
Income of U. S. Farmers: Discussion." Amer, J, Agr,
Ee_n. 67(1985):398-399.

Belongia, Michael T. and Richard A. King. "A Monetary Analysis
of Food Price Determination." Amer. J. Agr. -Eeon.
65(1983):131-135.

Belongia, Michael T. and Courtenay C. Stone. "Would Lower
Federal Deficits Increase U. S. Farm Exports?" Fed,
Res, Bang er Sr, Lonis B 1. (Nov. l985):5-19.

Bera, A. K. and C. M. Jarque. "Model Specification Tests: A

Simultaneous Approach." ;. Econometrics. 20(1982):59-
82.

Bernanke, Ben 8. "Alternative Explanations of the Money-

Income Correlation. " Carnegi e-Rocnester Conference Seri ee
22 zublie 261191. 25(1986):49-100.

Bessler, David A. and John L. Kling. "The Forecast and Policy
Analysis." Amer, g. Agr, Egon. 71(1989):503-506.

Bhargava, A. "On the Theory of Testing for Unit Roots in

Observed Time Series." Rev. of Econ, Studiee. 52(1986):
369-384.

151

Bisignanno, Joseph. ”Fiscal Deficits and Exchange Rate: A
Look at Recent Policy Assertions and Their Theoretical
and Empirical Support." Fed. Res. Bank of San
Francisco. Working Pap. No. 85-04, June 1985.

Blanchard, O. J. and M. W. Watson. "Are All Business Cycles
Alike?" Paper Presented at Conference on Business
Cycles. Nat. Bur. Econ. Res. Puerto Rico. Mar. 22-25.
1984.

Blanchard, O. J. and Danny Quah. "The Dynamic Effects of
Aggregate Demand and Supply Disturbances." Amerr_Eeenr
Bey. 79(1989):655-673.

Blinder, Alan 8. "On the Monetization of Deficits." Working
Pap. Series, No. 1052. Nat. Bur. Econ. Res. Dec. 1982.

Blinder, Alan S. and Robert M. Solow. "Does Fiscal Policy
Matter?" 1, Englie Econ. 2(1973):319-337.

Blinder, Alan S. and Robert M. Solow. "Analytical
Foundations of Fiscal Policy." in Alan S. Blinder and
Robert M. Solow, ed., h cono ' s o b ' ce,
pp. 3-15. The Brookings Institution, 1974.

Board of Governors of the Federal Reserve System. Annnel
Emerieriee1_pigeer. Washington, D. C. selected issues.

Board of Governors of the Federal Reserve System. Eenking

WW9- Washington. D- c.
1976.

Board of Governors of the Federal Reserve System. Eegerel
Beeerye_Enllerin. Washington, D.C. selected issues.

Bollerslev, Tim. "Generalized Autoregressive Conditional
Heteroskedasticity." J, Econemerries 31(1986):307-327.

Bollerslev, Tim. "A Conditional Heteroskedastic Time Series
Model for Speculative Prices and Rates of Return." Bey;
W. 69 (1987) =542-547-

Bollerslev, Tim, Robert F. Engle and J. M. Wooldridge. "A
Capital Asset Pricing Model with Time Varying

Covariances." g, 20011;, Ego n, 96(1988): 116-131.

Boothe, Paul M. and Bradford G. Reid. "Asset Returns and
Governments Budgets In a Small Open Economy: Empirical
Evidence for Canada." n, Eonetary Ecen. 23(1989):65-77.

152

Box, George E. P. and Jenkins, Gwilym M. I1me_§er1e__
Analxsis_Egrssasrins_and_£2nrrol Holden-Day. San

FtflflCiSCO e 197 0 .

Box, George E. P. and Pierce, David A. "Distribution of
Residual Autocorrelations in Autoregressive-Integrated

Moving Average Time Series Models." 1, Amer, finer, As .
65(1970):1509-1526.

Box, George. E. P. and G. C. Tiao. "A Canonical Analysis of
Multiple Time Series Models." E1emerr1ke. 64(1977):355-
365.

Bredahl, Maury E. "The Effects of Inflation on the Welfare and
Performance of Agriculture: Discussion." Amer. J, Agr,
Eeen. 67(1985):400-401.

Branson, William H. Meeroeconomic Tneory ang P011ey. 2nd.
Harper & Row. New York. 1979.

Branson, William H. "The Collapse of Purchasing Power Parities
during the 1970's by Frenkel: Comments." Enreneen_Eeen1
Bey. 16(1981):161-171.

Brown, W. W. and G. J. Santoni. "Unreal Estimates of the

Real Rate of Interest." Eed1_Bssl_§ank_2f_§t1_heni§
Eeyr (Jan. 1981):18-26.

Brunner, Karl and Allan H. Meltzer. "Money, Debt, and
Economic Activity." g, 2911;. Econ. 80(1972):951-977.

Buchanan, J. M. and R. Wagner. Qemocracy 1n nef1eit; Tne

Eel1t1ee1 Le geeyo of Lo rg Keynes. Academic Press.
New York. 1978.

Buiter, William. "Deficits, Crowding Out and Inflation: The
Simple Analytic." Working Pap. No. 1078. Nat. Bur. Econ.
Res. Feb. 1983.

Bureau of Economic Analysis. Ens1ness §tet1sr1cs. U.S.
Department of Commerce. Washington, D.C. selected issues.

Bureau of Economic Analysis. Enryey_efi_gnrrenr_Ene1ne_e. U.S.

Department of Commerce. Washington, D.C. selected issues.

Burger, Albert E. "The Implementation Problem of Mbnetary

Policy." Eeg, Ees. Eank n: gt. Leuis Rev,(Mar. 197l):20-
30

Canto, Victor A. and Donald Rapp. "The Crowding Out

Controversy: Arguments and Evidence." Eeg1_Eee1_Eenk_ef
Atlsn§a_Bsx- (Aug 1982)=33-37-

153

Carlson Keith M. ”The Mix of Monetary and Fiscal Policies:
Conventional Wisdom vs. Empirical Reality." Eeg1_3ee1

Bank_2f_§t1_Lguis_Bsx- (Oct- 1982)=7-21-

Carlson, Keith M. and Roger W. Spencer. "Crowding Out and

Its Critics." Eeg, Ees. Eenk ef Er, Leu1s Eev, (Dec.
1975):2-17.

Carlton, Dennis W. "Market Behavior with Demand Uncertainty
and Price Inflexibility." Amer1_Eeen1_Bey. 68(1978):57l-
587.

Carlton, Dennis W. "Contracts, Price Rigidity, and Market
Equilibrium." 1, 2911;, Econ. 87(1979):1034-1063.

Carlton, Dennis W. "Modeling Price Rigidity or Predicting the
Quality of the Good that Clears the Market. " Working Pap.
No. 503. Nat. Bur. Econ. Res. July 1980.

CEASM (Council of Economic Advisers Staff Memorandum).
"Financing a Federal Deficit." Ronald L. Teigen, ed.
{‘od°1°= '1 101‘ .,'-na, Ou‘ an: _ as. ° 1 f
Eeliey. 4th. R.D. Irwin. Homewood, Ill. 1978. pp.277-
279.

Chambers, Robert G. "Interrelationships between. Monetary
Instruments and Agricultural Commodity Trade." Amer1_g1

Agr1_Eeen. 63(1981):934-94l.

Chambers, Robert G. "Agricultural and Financial Market

Interdependence in the Short Run." Amer1_g1_Agr1_Ee_n.
66(1984):12-24.

Chambers, Robert G. "Credit Constraints, Interest Rates, and

Agricultural Prices." Amer1_J1_Agr1_Eeen. 67(1988):12-
24.

Chambers, Robert G. and Richard E. Just. "Effects of Exchange
Rate Changes on U.S. Agriculture: A Dynamic Analysis."

Ansrl_11_Asr1_Esgn- 63(1981)=33-46-

Chambers, Robert G. and Richard E. Just. "A Critique of the
Exchange Rate Treatment in Agricultural Trade Models:

Reply-" Amer1_£1_Agr1_Esen- 68(1986)=994-997-

Chow, G. C. "Tests of Equality Between Sets of Coefficients
in Two Linear Regressions." Economerrice. 28(1960):59l-
605.

Chow, G. C. "On the Long-Run and Short-Run Demand for Money."
11.221131_EQQD- 74(1966)=111-131-

154

Christ, Carl F. "A Short-Run Aggregate-Demand Model of the
Interdependence and Effects of Monetary and Fiscal
with Keynesian and Classical Interest Elasticities."

Amsr1_Eson1_Bex- 57(1967):434-443-

Christ, Carl F. "A Simple Macroeconomic Model with a
Government Budget Restraint." q, P011t, Econ.
76(1968):53-67.

Cohen, Darrel and Peter B. Clark. "The Effects of Fiscal
Policy on the U.S. Economy." Staff Studies No. 136.
Board of Governors of the Fed. Res. Sys., Jan. 1984.

Cooley, Thomas F. and Stephen F. LeRoy. "Atheoretical
Macroeconometrics: A Critique." J, Monetary Eeen. 16
(l985):283-308.

Corrigan, E. Gerald. "The Measurement and Importance of
Fiscal Policy Changes." ed. es. Ba k ew r

nenrn1y_3ey. (June l970):133—145.

Crouch, Robert L. neereeeenem1ee. Harcourt Brace Jovanovich
InC. 1972.

Crouhy-Veyrac, L., M. Crouhy, and J. Melitz. "More about the
Law of One Price." Enrenean Econ, E y. 18(1982):325-44.

Cullison, William E. "Money, the Money Base, and Nominal

GNP." Eed, Bee, Eenk e: R1ehmend Eeonem1e Rey,(May/
June, l982):3-13.

Davis, Richard G. "How Much Does Money Matter? A Look at

Some Recent Evidence." Eed, Ree. Eank ef new Xerk
Menrh1y_E_y. (June 1969):119-131.

Dickey, David A. and Wayne A. Fuller. "Distribution of the
Estimators for Autoregressive Time Series with a Unit

Root." J, Amer. stat. Ass. 74(1979):427-431.

Dickey, David A. and Wayne A. Fuller. "Likelihood Ratio
Statistics for Autoregressive Time Series with a Unit

Root." Eeenemerr1e_. 49(1981):1057-1072.

Dickey, David A. and Sastry'G. Pantula. "Determining the Order
of Differencing in Autoregressive Processes." 11
Busins§§_and_£s2n1__tat 5(1987)= 455- 461-

Diebold, F. X. and M. Nerlove."The Dynamics of Exchange Rate
Volatility: A Multivariate Latent Factor ARCH Model." 11

Aenlisd.£sen2mstris_. 4(1989>=1-21-

155

Dobson, William D. "Effects of the Macroeconomic Environment

on 1985 Farm Legislation.” Annual Ag. Ontleek genfierence.
USDA. Washington, D. C. Dec. 3-5. 1984. pp.48-58.

Dornbush, Rudiger. "Expectations and Exchange Rate Dynamics."

1, 2911;. Been. 84(1976):1161-ll76.

Dornbush, Rudiger and Stanley Fisher. "Budget Deficits and
Inflation." M. June Flanders and Assaf Razin, ed.

Qeye1enmenr 1n en lnflatienary Wor1g. Academic Press.
New York. 1981.

Dornbush, Rudiger and Stanley Fisher. neereeeenemiee(3rd.).
McGraw-Hill Book Co. New York. 1984.

Dwyer, Gerald. "Inflation and Deficits." Eeenem1e_1ngn1ry.
20(1982):315-329.

Engle, Robert F. "Autoregressive Conditional
Heteroskedasticity with Estimates of the Variance of
United Kingdom Inflation." Economerr1ca. 50(1982):987-
1007.

Engle, Robert F. "On the Theory of Cointegrated Economic Time
Series." mimeo. 1987.

Engle, Robert F. and Tim Bollerslev. "Modeling the
Persistence of Conditional Variances." Eeenemerr1e
Bey. 5(1986):l-50.

Engle, Robert F. and C. W. J. Granger. "Cointegration and
Error Correction: Representation, Estimation, and

Testing." Eeenemerr1ee. 55(1987):251-276.

Engle, Robert F., C.W.J. Granger, and Dennis Kraft. "
Combining Competing Forecasts of Inflation Using a

Bivariate ARCH Model." J, Econ. Qynam1cs eng Qentre .
8(1984):151-65.

Engle, Robert F., David M. Lilien, and Russell P. Robins.
"Estimating Time Varying Risk Premia In the Term
Structure: the ARCH-M Model." Ecenemetrica. 55(1987) :391-
407.

Engle, Robert F. and B. S. Yoo. "Forecasting and Testing in

Co-integrated Systems." J, Eeonometr1e . 35(1987):l43-
159.

Evans, G. B. A. and N. E. Savin. "Testing for Unit Roots:
l." Eeenemerriee. 49(1981):753-779.

156

Evans, G. B. A. and N. E. Savin. "Testing for Unit Roots:
2 ” Eeenemerr1en. 52(1984):1241-1270.

Evans, Paul. "Is the Dollar High Because of Large Budget
Deficits?” ;1_nenerery_Eeen. 18(1986):227-249.

Evans, Paul. ”Do Budget Deficits Raise Nominal Interest
Rate: Evidence from Six Countries." Q1_jmnmmgunnjgmnp
20(1987):281-300.

Evans, Paul. "Interest Rates and Expected Future Budget
Deficits in the United States." J1_2911r1_Eeen. 95
(1987): 34- 58.

Fackler, P. F. "Vector Autoregressive Techniques for
Structural Analysis." Working Pap. No. 113. Dept. of
Econ. and Business. N. C. State Univ. Raleigh. N. C.
1988.

Feldstein, Martin S. "Government Deficits and Aggregate

Demand." J, nonerery Ecen, 9(1982):1-20.

Feldstein, Martin S. and Otto Eckstein. "The Fundamental
Determinants of the Interest Rate." Bey1_efi_Eeen1_eng
finer. 52(1970):363-375.

Framm, Gary and Lawrence R. Klein. "A Comparison of Eleven
Econometric Models of the united States." Amer1_Eeen1
Eev. 63(1973):385-393.

Frank, de Leeuw and Thomas M. Holloway. "The High Employment
Budget: Revised Estimates and Automatic Inflation
Effects." u r Bus' . (April 1982):21-
23.

Frankel, Jeffery A. ”Commodity Prices and Money: Lessons

from International Finance." Amer1_qr_Agr1_Eeen.
66(1984):560-566.

Frankel, Jeffery A. "Expectations and Commodity Price

Dynamics: Overshooting' Model." Amer, J, Agr, Eeen.
68(1986):344-348.

French, K.R., G.W. Schwert.and R.E. Stambaugh. "Expected Stock
Returns and‘Volatility." 1, E1nan, E; n. 19(1987):3— 29.

Frenkel, Jacob A. ”The Collapse of Purchasing Power Parities

During the 1970's." W. 16(1981):145-
165.

Friedman. Milton. Waggon
Princeton Univ. Press. 1957.

157

Friedman, Milton. "A Monetary Theory of Nominal Income." J.
Eel1r1_Eeen. 79(1971):323-337.

Friedman, Milton. "Comments on the Critics." g 2911; Econ.
80(1972):906-50.

Fuller, Wayne A. Stat's 'c i 'es.
Wiley. New York. 1976.

Fusfeld, Daniel R. W. 4th. Scott.
Foresman and Co. Glenview. Ill. 1982.

Goldfeld Stephen. "The Demand for Money Revis1ted. " Ereokings
Ban, Eeon. A . (1973):577-638.

Gramlich, Edward M. "The Usefulness of Monetary and Fiscal
Policy as Discretionary Stabilization Tools." . one
Credit_and_nanking 3(1971)= 506- 532-

Granger, C. W. J.

"Investigating Causal Relations by
Econometric Models and Cross-Spectral Methods."
Esonomstrisa- 37(1969)=24-36-

Granger, C. W. J. "Some Properties of Time Series Data and

Their Use in Econometric Model Specificat1on." J.
Eeenemerriee. 16(1981):121-l30.

Grossman, Herschel I. "Rational Expectations, Business Cycles,
and Government Behavior. " at'ona

ta ' ns
edited by Stanley Fisher. Nat.
Econ. Res. Chicago. 1980.

Eoononis.l:fl111-

Bur.
Hafer, R. W.

"Select1ng a Monetary Indicator: A Test of the
New Monetary Aggregates." d. Res. Ba
(Feb. 1981):12-18.

t. o is
Hafer, R. W.

V

"Money, Debt, and Economic Activity." Eeg. Ree,
ui ev. (Jun/July 1984):18-25.
Hamburger, Michael J. and Burton Zwick. "Deficits, Money,
and Inflation." g, Monetery Econ, 7(1981) 141 150.
Hamburger, Michael J. and Burton Zw1ck.

and Inflation-

"Deficits,
. Reply.”
283.

Money,
1, Monerary Eeen 10(1982):279-
Haslag, Joseph H. and Daniel J. Slottje. "Macroeconomic

Policy and Income Inequality: An Error-Correction
Representation." Working Pap. No. 8909. Fed. Res. Bank
of Dallas. June 1989.

158

Hendry, D. F. "Econometric Modelling with Cointegrated

Variables: An Overview." 931, Eull, Eeen, S er.
48(1986): 201—212.

Hoffman, Dennis and Robert H. Rasche. "Long-Run Income and
Interest Elasticities of Money Demand in the United
States." Working Pap. No. 2949. Nat. Bur. Econ. Res.
1989a.

Hoffman, Dennis and Robert H. Rasche. "The Demand for Money
in the U.S. during the Great Depression: Estimates and
Comparisons with the Post War Experience." Unpublished
Working Pap. Nat. Bur. Econ. Res. Dec. 1989b.

Hsiao, Cheng. "Autoregressive Modeling of Canadian Money and
Income Data." 1. Amer. Stat, Ase. 74(1979):353-560.

Hsiao, Cheng. "Autoregressive Modeling and Money-Income
Causality Detection." g, nenerary Ee n. 7(1981):85-106.

Hsieh, D. A. "The Statistical Properties of Daily Foreign

Exchange Rates: 1974-1983." 1, Interner, Ee n. 18
(1988):129-146.

Hutchinson, Michael M. and Adrian W. Throop. "U.S. Budget
Deficits and the Real Value of the Dollar." Fen. Ees.

Bank 9f Een Ereneisce Econ, E y. (Nov. 1985):26-43.
Isard, Peter. "How Far Can We Push the Law of One Price?"
Amsr1_Eson1_Bex. 67(1977)=942-48-

Jabara, C. L., and N. E. Schwartz. "Flexible Exchange Rates
and Commodity Price Changes: The Case of Japan." Amer.

J1_AQIL_EQQQ. 69(1987):580-90.

Johansen, Sdren. "Statistical Analysis of Cointegration

Vectors." Q, Eeen, Qynem1es end Conrr 1. 12(1988):
231-254.

Johansen, Sbren. "Estimation and Hypothesis Testing of
Cointegration Vectors in Gaussian Vector Autoregressive
Models." Preprint 3. Inst. of Math. Stat. Univ. of
Copenhagen. Jan. 1989.

Johansen, Sbren and Katarina Juselius. "Hypothesis Testing
for Cointegration Vectors with an Application to the
Demand for Money in Denmark and Finland." Preprint 2.
Inst. of Math. Stat. Univ. of Copenhagen. Feb. 1988.

159

Johansen, Sbren and Katarina Juselius. "The Full Information
Maximum Likelihood Procedures for Inference on
Cointegration with Applications." Preprint 4. Inst. of
Math. Stat. Univ. of Copenhagen. Jan. 1989.

Just, Richard. E. and. Robert G. Chambers. "Deficits and
Agriculture: An Alternative Parable." Working Pap. No.
87-08. Dept. of Agr. and Resource Econ. Univ. of
Maryland. July. 1987.

Keran, Michael W. "Monetary and Fiscal Influences on
Economic Activity: The Historical Evidence." Eeg. Bee,

Bank_of_§t11221§_zs_- (NOV- 1969)=5-24-

Keran, Michael W. "Monetary and Fiscal Influences on

Economic Activity: The Foreign Experience." R s.
Bank_of_§11Louis_Esx- (Feb- 1970)=16-28-

Kloek, T. and H. K. Van Dijk. "Bayesian Estimates of Equation
System Parameters: An Application of Integration by Monte

Carlo." Eeenemerr1ee. 46(1978):l-20.

Kochin, Levis A. "Are Future Taxes Anticipated by

Consumers?" 11_nonsx_§rsd1t_and_§anking. 6(1974)=385-
394.

Kormend, Roger C. "Government Debt, Government Spending, and
Private Sector Behavior." Amer, Econ. Bey. 73(1983):994-
1010.

Kunst, Robert M. "Cointegration in a Macro Economic System:
Sensitivity and Prediction." Working Pap. Institute for
Advanced Studies. Vienna. Austria. 1989.

Leamer, Edward E. "Vector Autoregressions for Causal

Inference?" Qarnegie-Eeehester Cenference Seriee en
Puh11g_291iey. 22(1985):255-304.

Lee, Junsoo and Peter Schmidt. "Unit Root Tests With ARMA
errors." unpublished Working Pap. Dept. of Econ. Michigan
State Univ. E.Lansing. MI. March. 1990.

Leonall C. Andersen and Jerry L. Jordan. "Monetary and
Fiscal Policy Actions: A Test of Their Relative
Importance in Economic Stabilization." Een1_3ee1_Eenk
e: an, Lon1s Bey. (Nov. 1968):1l-24.

Ljung, G.M. and G.E.P. Box. "On A Measure of Lack of Fit in
Time Series Models." BiQEQSEiKQ- 65(1978):297-303.

160

Lutkepohl, H. "Comparison of Criteria for Estimating the Order
of a vector Autoregressive Process." Working Pap. No.
8207 . Fachbereich Wirtschaft-swissenschaften . Universitat
Osnabruck. 1982.

Makin, John H. "Real Interest, Money Surprises, Anticipated
Inflation and Fiscal Deficits." B§!1_QI_EQQD1_§DQ_§§§§-
65(1983):374-384.

Mazon, Cristina. "The Impact of Government Policy on the
U.S. Steel and Cigarette Industries." Unpublished Ph.D
Dissertation. Dept. of Econ. Univ. of Minn. Minn. 1985.

McCurdy, T. and I. G. Morgan. "Tests of the Martingale
Hypothesis for Foreign Currency Futures with Time
Varying Volatility." Innerner, J. er Eerecesring.
3(1987):131-148.

McMillin, Douglas W. and Thomas R. Beard. "Deficits, Money,
and Inflation: Comment." J, Monetery Eeen. 10(1982):
273-277.

Meltzer, Allan. H. "The Demand for Money: the Evidence from
the Time Series." E, 2011;, Ecen. 71(1963):219-246.

Milhoj, "A Conditional Variance Model for Daily Deviations

of An Exchange Rate-" J1_of.5251ns§§_and_fison1_§tat-
5(1988):99-103.

Mount, Timothy D. "Policy Analysis with Time Series

Econometric Models: Discussion." Am§£1_l1_£EBL_J$EHL
71(1989):507-508.

Myers, Robert J., R. R. Piggott, and William G. Tomek. "
Estimating Sources of Fluctuations in the Australian
Wool Market: An Application of VAR Methods." Amer. J,

Agr1_Eeen. Forthcoming. 1991.

Nankervis, J. C. and N. E. Savin. "Testing the
Autoregressive Parameters with t Statistics." 11

Eeonomerr1e . 27(1985):l43-161.

Nelson, Charles R. and Charles I. Plosser. "Trends and
Random Walks in Macroeconomic Time Series: Some
Evidence and Implications." J, nonetery Egon. 10
(l982):139-162.

Newey, W. K. and K. D. West. "A Simple Positive Definite
Heteroskedasticity and Autocorrelation Consistent
Covariance Matrix." Eeenemerriee. 55(1987):703-708.

161

Nickell, S. "Error Correction, Partial Adjustment and All

That: An Expository Note.“ Qxf, Eull, Eeen, Eta . 47
(1985):ll9-129.

Nickelsburg, Gerald. "Small Sample Properties of
Dimensionality Statistics for Fitting VAR Models to
Aggregate Economic Data: A Monte Carlo Study." J1

Eeenemerr1_§. 28(1985):183-192.

Niskanen, William. "Deficits, Government Spending, and

Inflation." g. genetary Eeen, 4(1978):591-602.

O'Driscall, G. P. "The Ricardian Nonequivalence Theorem." Jr

EQlit: Econ, 85(1979):207-210.

Okun, Arthur. "Inflation: Its Mechanics and Welfare Costs."

Bro281ns§_Ban1_2n_Eson1_A_t- (1975):351—4o1.

Orden, David. "Capitol Flows, the Exchange Rate, and Goods
Markets in a Neoclassical Trade Model: Towards a
Framework for Evaluating Monetary Aspects of Agricultural
Trade. " Paper Presented at Canada/U.S. International
Agricultural Trade Research Consortium. Ontario. Canada.
June 24. 1983.

Orden, David. "Agriculture, Trade, and Macroeconomics: the

U.S. Case." J, Bolicy noge11ng. 8(1986):27-51.

Orden, David. and P. F. Fackler. "Identifying Monetary and
Other Macroeconomic Impacts on Agricultural Prices in

VAR Models." Amer, J. Agr. Econ. 71(1989):495-502.

Ott, Mack. "Money, Credit, and Velocity." . . 0
5r, Len1s Bey. 64(1982):21-34.

Perron, Pierre. "Trends and Random Walks in Macroeconomic
Time Series: Further Evidence from a New Approach."
unpublished. Dept. of Econ. Universite' de Montre'al.
Oct. 1986 .

Phillips, P. C. B. "Time Series Regression with A Unit Root."
Eeenemerr1ee. 55(1987):277-301.

Phillips. P. C. B. and S. N. Durlauf. "Multiple Time Series
Data and Their Use in Econometric Model Specification."

8111.822n1_§tud- 53(1986)=473-495-

Phillips, P. C. B. and Pierre.Perron. "Testing for a‘Unit.Root
‘in Time Series Regression." E1emerr1re. 75(1988):334-
346.

162

Plosser, Charles I. "Government Financing Decisions and

Asset Returns." 21.!QDELQIY_EQQQ- 9(1982):325-352.

Poole, W. "Wither Demand for Money?" Ereekinge_£en1_en_Eeen1
Act. (l970):485-500.

Protopapadakis, Aris A. and Jeremy J. Siegel. "Government
Debt, the Money Supply, and Inflation: Theory and
Evidence for Seven Industrialized Economies." Working
Pap. No. 84-4, Fed- Res. Bank of Philadelphia. Aug. 1984.

Protopapadakis, Aris A. and H. R. Stoll. "Spot and Futures
Prices and the Law of One Price." 11__E1nenee.
38(1983):1431-1455.

Rausser, Gordon C. "Macroeconomics and U.S. Agricultural
Policy." Bruce L. Gardner. ed. Q,S, Agricu1ture1
2211921_Ihe_12§§_Farm_Lesislation. American Enterprise
Institute for Public Policy Research. Washington D. C.
1985.

Rausser, Gordon C., James A. Chalfant, H. Alan Love, and
Kostas G. Stamoulis. "Macroeconomic Linkages, Taxes, and
Subsidies in the U.S. Agricultural Sector." Amer_._J_,_Agr,_
Eeen. 68(1986):399-412.

Ricardo, D. "Funding System." in v.4 of P. Sraffa. ed. Ine
W s orre enc v . Cambridge
Univ. Press. Cambridge. England. 1951.

Robertson, John and David Orden. "A Vector Error-Correction
Model of Money and Price Dynamics in New Zealand."
Working Pap. Dept. Agri. Econ. Virginia Tech. Blacksburg.
Virginia. Sept. 1989.

Runkle, D. E. "Vector Autoregressions and Reality." J, of
Busines__and_fic_n1_§_at 5(1987)= 437-442-

Said, 8. E. and D. A. Dickey. "Testing for Unit Roots in
Autoregressive-Moving Average Models of Unknown Order."

Eieme_r1ke. 71(1984): 599- 608.

Santoni, G. J. and Courtenay C. Stone. "Navigating Through
the Interest Rate Moraces: Some Basic Principles." Eed.

Ees, Eenk ef §t, Lon1s Rev. (Mar. 1981):ll-18.
Sargent, Thomas J. and Neil Wallace. "Some Unpleasant
Monetarist Arithmetic." e . a o ' ne 0 's

QBQIEQIL¥_B§_- 5(1931)=1‘17-

163

Saunders, Peter J. "Causality of U.S. Agricultural Prices and
the Money Supply: Further Empirical Evidence." Amer1_11

Agr1_Eeen. 70(1988):588-596.

Schmidt, Peter. "Dicky-Fuller Tests with Drift."
Econometrics and Economic Theory Pap. No. 8717. Dept.
of Econ. Mich. State Univ. E. Lansing. MI. 1988.

Schmidt, Peter and P. C. B. Phillips. "Testing for a Unit
Root in the Presence of Deterministic Trends."
unpublished. Dept. of Econ. Mich. State Univ. E.
Lansing. MI. Oct. 1989.

Schuh, G. E. ”The Exchange Rate and U.S. Agriculture." AEQL1
l1_AQI1_EQQn- 56(1974)=1-13-

Schuh, G. E. "The New Macroeconomics of Agriculture." Amerr
11_Agr1_reen. 58(1976):802-811.

Schuh, G. E. "The Global Macroeconomic Environment of U. S.

Food and Agricultural Policy-" £22d_§ecuritx_1n_nunsrx
flerlg. 1981.

Schuh, G. E. "U.S. Agricultural Policy in An Open World
Economy." Testimony to Congress. 1983.

Schuh, G. E. "Future Directions for Food and Agricultural
Trade Policy." Amer1_g1_Agr1_Eeen. (l984a):242-247.

Schuh, G. E. "Trade and Capital Market Linkages As Components
of The Domestic Farm Policy Environment." Annne1_Agr1

Qnrleek_§enrerenee. (l984b):59-66. USDA. Washington,
D.C.

Schwert, G. William. "Effects of Model Specification on
Tests for Unit Roots in Macroeconomic Data." 11

Monetarx.£con- 20(1987)=73-103-

Schwert, G. William. "Tests for Unit Roots: A Monte Carlo

Investigation." 1, Business and Econ, E; r. 7(1988):147-
159.

Seater, John J. "Are Future Taxes Discounted." J, Money,
Credit_and_nankins- 14(1982)=377-389o

Sen, D. L. "Robustness of Single Unit Root Test Statistics
in the Presence of Multiple Unit Roots. " unpublished Ph.D
Dissertation. Dept. of Statistics. North Carolina State
Univ. 1986.

164

Sen, D. L. and D. A. Dickey. "Symmetric Test for Second
Differencing in Univariate Time Series." J1_er_Ene1neee

and.Bcon1_§tat 5(1987L 463-473-

Shei, Shun Yi and Robert Thompson. "Inflation and Agriculture:
1A Monetarist Structuralist Synthesis." Mimeo. Dept. of
Agr. Econ. Purdue Univ. West Lafayette, Indiana. sept.

1981.

Silber, William L. "Fiscal Policy in IS-LM Analysis: A
Correction." J1_none21_gredit_and_hankins- 2(1970)=461-
472.

Simon, Herbert A. "Causal Ordering and Identifiability." in
William C. Hood and Tjalling C. Koopmans, ed., §§ud1es
1n_Eeenemerr1e_merne_. Cowles Foundation. 1953.

Sims, C. A. "Money Income and Causality." Amer. Eeon. Rev. 62
(1972):54o-552.

Sims, C. A. "Macroeconomic and Reality." Econometr1ca.
48(1980):l-48.

Sims, C. A. "Are Forecasting Models Usable for Policy

Analysis?" Eeg, Ees. Eenk er m1nneeno1is Qnerrerly Rev.
lO(l986):l-16.

Sims, C. A. "Models and Their Uses." Amer1_11_Agr1_Eeen.
71(1989):489-494.

Sims, C. A., J. H. Stock, and M. W. Watson. "Inference in
Linear Time Series Models with Some Unit Roots."

Eeenemerr1ee. 58(1990):113-144.

Spanos, A. ta 5 ’ oundations of E no t 'c ode 'n .
Cambridge Univ. Press. Cambridge. 1986.

Stamoulis, Kostas G., James A. Chalfant, and Gordon C.
Rausser. "Monetary Policies and Overshooting of Flexible
Prices: Implications for Agricultural Policy." Selected
Pap. Presented at Amer. Agr. Econ. Ass. Annual Meeting.
Ames, Iowa. 1985. '

Stamoulis, Kostas G., James A. Chalfant, and Gordon C.
Rausser. "Monetary Policy and Relative Farm Prices."
Univ. of Ill. at Urbana Champaign and Univ. of Cal. at
Berkeley. unpublished. 1987.

Stiglitz, Joseph. "Theories of Wage Rigidity." Working Paper.
Nat. Bur. Econ. Res. Sept. 1984.

165

Stock, J. H. "Asymptotic Properties of Least Squares
Estimators of Cointegrating Vectors." Eeenemerr1ee. 55
(1987): 1035-1056.

Stock, J. H. and M. W. Watson. "Does Real GNP Have a Unit
Root?" Economic_1etter§ 22(1986L 147-151-

Stock, J. H. and M. W. Watson. "Testing for Common Trends."

£1_Amer1_§tat1_As§- 83(1988)=997-1107-

Stock, J. H. and M. W. Watson. "Interpreting the Evidence on
Money-Income Causality." 1, Eeonomerr1cs. 40(1989):161-
181.

Tatom, John A. "A Perspective on the Federal Deficit

Problems." Eed1_Bes1_Bank_gf_§tl_Louis- (June/July
1984):5-17.

Tatom, John A. "Two Views of the Effects of Government

Budget Deficits in the 1980's." Eeg_,__3ee,__Een]g_efi_§r1
Len1e. (Oct. l985):5-16.

Taylor, John. "Staggered Wage Setting in a Macro Model. " Amer1
Eeen1_3ey. 69(1979):109-ll3.

Taylor, John. "Aggregate Dynamics and Staggered Contracts."
11.221111_Ec_n- 88(1980)=

Teigen, Ronald L. "A Critical Look at Monetarist Economics."

Eeg, Res, Eenk 9f 5r, Len1s Rev. (Jan. 1972):10-25.

Theil, H. and J. C. G. Boot. "The Final Form of Econometric

Equation Systems.” Bs21_2f.1he_1nternatl_51ati_1ns. 30
(1962):l36-152.

Thompson, Robert L. "U.S. Macroeconomic Policy and
Agriculture." Paper Presented at Univ. Consortium for
Agricultural Trade Research Conference. July 24-29.
1986. Tahoe City. California.

Tobin, James. "A General Equilibrium Approach to Monetary
TheorY" W 1(1969)=15-29-

Tobin, James. "Friedman's Theoretical Framework." 1. 2911;,
Econ. 80(1972):852-863.

Tobin, James. "The Wrong Mix for Recovery." gnellenge.(May/
June 1982):25.

Tobin, James and William Buiter. "Long Run Effects of Fiscal
and Monetary Policy on Aggregate Demand." Cowles
Foundation Discussion Pap. No. 384. 1974.

166

Todd, Richard M. "Policy Analysis with Time Series
Econometric Models: Discussion." Amer1_g1_Agr1_Eeen.
71(1989):509-510.

Tweeten, Luther. "Impact of Federal Fiscal-Monetary Policy on

Farm structure." WM- 15(1983)=61-
68.

Tweeten, Luther. Eerm__zel1ey__Ane1ye1e. Westview Press.
Boulder. Colorado. 1989.

Vela, R. P., D. W. Wichern, and G. C. Reinsel. "A Note on
Nonstationarity and Canonical Analysis of Multiple Time

Series Models."11.11me_§er125_8nalxsis 8(1987L 479-
487.

Wallis, Kenneth F. "Multiple Time Series Analysis and the
Final Form of Econometric Models." Eeenemerr1ee. 45
(l977L 1481-1498.

Weintraub, Robert E. "Deficits: Their Impact on Inflation
and Growth." Staff Study for the Subcommittee on
Monetary and Fiscal Policy of the Joint Economic
Committee, GPO, July, 1981.

Yohe, William P. "The 'Crowding Out' of Private Expenditures

by Fiscal Policy Actions." Eed1_Be§1_Bank_of_§t1_Louis
Bey. (Oct. 1970L 12-24.

Zellner, Arnold. "The Current State of Business Econometrics. "
Mimeo. Univ. of Chicago. Ill. 1981.

Zellner, Arnold. uc to es 'n
Eeenemerr1ee. New York: John Wiley & Sons. 1971.

      

“whammy/@111; 71011111171111 “

07645777