Department of Food and Resource Economics 

Is the Export-Lead Growth

Hypothesis Valid for
Canada?
Titus Awokuse

College of Agriculture and Natural Resources  University of Delaware

University of Delaware

) 5 ( & 6 W D I I 3 D S H U

)22'
5(6285&(
(&2120,&6

FREC SP02-01
March 2002

Is the export-led growth hypothesis valid for Canada?

Titus O. Awokuse*
Department of Food and Resource Economics, University of Delaware

Abstract
Empirical evidence linking exports to economic growth has been mixed and
inconclusive. This study re-examine the export-led growth (ELG) hypothesis for Canada
by testing for Granger causality from exports to national output growth using vector error
correction models (VECM) and the augmented vector autoregressive (VAR)
methodology developed in Toda and Yamamoto (1995).

Application of recent

developments in time series modeling and the inclusion of relevant variables omitted in
previous studies help clarify the contradictory results from prior studies on the Canadian
economy. The empirical results suggest that a long-run steady state exists among the
models six variables and that Granger causal flow is unidirectional from real exports to
real GDP.
JEL Classification: F43, C32

_____________________________________________________
* Author can be contacted at the Department of Food and Resource Economics,
University of Delaware, Newark, DE 19716, USA. Email: [email protected].

1. Introduction
The nature of the relationship between exports and national output growth has
been one of the most debated in the recent past, yet with little consensus. Central to this
debate is the question of whether strong economic performance is export-led or growthdriven.

This question is important because the determination of the causal pattern

between export and growth has important implications for policy-makers decisions
about the appropriate growth and development strategies and policies to adopt. The fact
that strong correlation exists between exports and real GDP growth has been well
documented in the literature. But previous empirical studies have produced mixed and
conflicting results on the nature and direction of the causal relationship between export
growth and output growth. Although most studies focus on the causal link between
exports and output growth in developing countries (Michaely 1977; Balassa 1978; Chow
1987), some researchers have examined the export-led hypothesis with emphasis on
industrialized countries (Marin 1992; Serletis 1992; Henriques and Sadorsky 1996;
Yamada 1998). However, very few empirical studies have been done in the recent past to
investigate the export-led growth (ELG) hypothesis for Canada (Serletis 1992; Henriques
and Sadorsky 1996).
Using data from selected industrialized countries, Marin (1992) examines the
causal link between exports and productivity and finds that the ELG hypothesis cannot be
rejected for Germany, Japan, the United Kingdom, and the United States. But Marins
(1992) study did not include Canada. Only two recent studies are found that use
Canadian data. The first study by Serletis (1992) examines the ELG hypothesis by using

single equation techniques to analyze Canadian annual data (exports, imports, and GNP)
from 1870 to 1985. Serletis (1992) finds empirical support for the ELG hypothesis in
Canada. More recently, Henriques and Sadorsky (1996) also focused on the export and
output growth relationship for Canada using just three variables (GDP, exports, and terms
of trade). They employ a multivariate cointegration estimation methodology that
accounted for potential feedback and simultaneity effects between the three variables. In
contrast to Serletis (1992) earlier result, Henriques and Sadorsky (1996) find that
changes in GDP precede changes in exports.
The lack of consistent causal pattern between exports and output growth in the
studies on Canada may be due to one or more of the following modeling issues. The
causal models in these studies may be mis-specified because of: i) the omission of an
important variable such as capital, labor, and foreign output shocks; ii) traditional
Granger causality F-test in a regression context may not be valid if the variables in the
system are integrated since the test statistic does not have a standard distribution (Toda
and Philips 1993); and iii) temporal aggregation issues from use of annual time series
may yield erroneous causation results (Bahmani-Oskooee and Alse (1993).
This paper contributes to the literature on the export-output growth nexus in the
following ways. First, previous studies on the dynamic linkages between exports and
output growth are extended through the application of recent advances in time series
statistical techniques: i) vector error correction modeling (Toda and Philips 1993), and ii)
augmented level VAR modeling with integrated and cointegrated processes of arbitrary
orders (Toda and Yamamoto 1995; and Dolado and Lutkepohl 1996).

These two

methodological procedures are useful because they allow tests of Granger causality
between exports and output growth while accounting for the long-run information often
ignored in systems that requires first differencing and pre-whitening prior to inference. In
contrast to previous studies on Canada that used annual data in their analysis, this study
uses a higher frequency quarterly time series data on Canada from 1961:1 -2000:4.
In addition to employing recently developed time series modeling techniques and
use of higher frequency data, this study also expands on the three variables used by
Henriques and Sadorsky (1996) to include labor, capital, and foreign output shock.
Using VECM and Toda and Yamamotos (1995) augmented VAR procedure, our results
suggest that changes in real exports precede changes in real GDP. Furthermore, our
inclusion of additional variables appear to be justified, as capital, labors, and foreign
output shocks are statistically significant in the models. Our finding of evidence in
support of ELG hypothesis for Canada is consistent with Serletis (1992) results, but
different from Henriques and Sadorsky (1996) findings.1 The rest of this paper is
organized as follows. Section 2 discusses the analytical framework and methodological
issues while section 3 presents empirical findings and section 4 summarizes the papers
findings.
2. Analytical Framework and Methodological Issues
2.1. Exports and Output Growth Nexus
In the literature, causality from exports to real output is denoted as ELG hypothesis while
the reverse causal flow from real output to exports is termed growth-driven exports. ELG
hypothesis reflects the view that export-oriented policies help stimulate economic

growth.

Export expansion can be a catalyst for output growth both directly, as a

component of aggregate output, as well as indirectly through efficient resource

allocation, greater capacity utilization, exploitation of economies of scale, and
stimulation of technological improvement due to foreign market competition. Exports
provide foreign exchange that allows for increasing levels of imports of capital goods and
intermediate goods that in turn raise the growth of capital formation and thus stimulate
output growth (McKinnon 1964; Balassa 1978; Buffie, 1992). Furthermore, export
growth through expanded market base allows for the exploitation of economies of scale
for open economies and promotes the transfer and diffusion of technical knowledge in the
long run (Helpman and Krugman 1985; Grossman and Helpman 1991). Exports can be
viewed as economies of scale that are external to the individual firms in the non-export
sector but internal to the overall economy. The ELG hypothesis implies that the rate of
export growth will lead to economy-wide efficiency and productivity growth.
Early empirical formulations tried to capture the causal link between exports and
real GDP growth by incorporating exports into the aggregate production function
(Balassa 1978; Sheehey 1990). This paper expands on the growth equation by including
other potentially relevant variables such as terms of trade, and foreign output shock. The
terms of trade reflects the possible linkages of the real exchange rates (and the possible
effects of restrictive trade policy such as tariff and non-tariff barriers) and real output.
The foreign output shock variable controls for export growth not influenced by Canadian
price competitiveness or productivity, but by growth in the rest of the world.
Accordingly, the aggregate production function is expressed as:

Y = F[(K, L); X, TT, Y*]

(1)

where Y represents real GDP growth, K, L, X, TT, Y* represent real capital, labor, real
exports, real terms of trade, and a foreign output shock respectively. The export-output
growth causal link is a long-run behavioral relationship whose analysis requires
estimation techniques appropriate for long-run equilibria. Therefore, the variables in the
system must be tested for cointegration, prior to testing for Granger causality. This paper
will apply recent advances in time series techniques by estimating VECM and level VAR
models with integrated and cointegrated processes of arbitrary orders (Toda and
Yamamoto 1995; and Dolado and Lutkepohl 1996).

2.2. Multivariate Cointegration Analysis and Error Correction Modeling

The concept of cointegration is intuitively appealing because it is supported by the

notion of long-run equilibrium in economic theory. While variables in a system may
fluctuate in the short run, they are expected to return to their steady state in the long run.
A common method for testing for cointegration between economic series is the EngelGrangers two-step bivariate residual-based method.

Although this approach is an

improvement over simple correlation coefficient analysis, it has been shown to be weak
in modeling multivariate cases because it: i) is sensitive to the choice of endogenous
variables in the cointegrating regression; ii) makes a priori assumption of a single
cointegrating vector in the system; and iii) tends to yield biased parameter estimates in
small samples (Banerjee, et al 1990). Johansen (1988) and Johansen and Juselius (1990)
maximum likelihood (ML) procedure is a very popular alternative to the Engle-Granger

method. The main attraction of this procedure is that it tests for the possibility of multiple
cointegrating relationships among the variables. Johansen and Juselius (1992) modeled
time series as reduced rank regression in which they computed the ML estimates in the
multivariate cointegration model with Gaussian errors. The model is based on the error
correction representation given by
p 1

X t = + i X t i + X t 1 + t

(2)

i =1

where Xt is an (nx1) column vector of p variables, is an (nx1) vector of constant terms,

and represent coefficient matrices, is a difference operator, k denotes the lag

length, and t is i.i.d. p-dimensional Gaussian error with mean zero and variance matrix

(white noise disturbance term). The coefficient matrix is known as the impact
matrix and it contains information about the long-run relationships.
Equation (2) resembles a VAR model in first differences, except for the inclusion
of the lagged level of Xt-1, an error correction term, which will contain information about
the long run among variables in the vector Xt. This way of specifying the system contains
information on both the short- and long-run adjustment to changes in Xt through the
estimates of and respectively. The VECM equation above allows for three model
specifications. (a) If is of full rank, then Xt is stationary in levels and a VAR in levels
is an appropriate model. (b) If has zero rank, then it contains no long run information,
and the appropriate model is a VAR in first differences (implies variables are not
cointegrated). (c) If the rank of is a positive number, r and is less than p (where p is
the number of variables in the system), there exists matrices and , with dimensions

(p x r), such that = . In this representation contains the coefficients of the r

distinct long run cointegrating vectors that render ' X t stationary, even though Xt is
itself non-stationary, and contains the short run speed of adjustment coefficients for the
equations in the system.
Johansens methodology requires the estimation of the VAR equation (2) and the
residuals are then used to compute two likelihood ratio (LR) test statistics that can be
used in the determination of the unique cointegrating vectors of Xt. The first test which
considers the hypothesis that the rank of is less than or equal to r cointegrating vectors
is given by the trace test below:
Trace = T

ln(1 )

i = r +1

(3)

The second test statistic is known as the maximal eigenvalue test which computes the
null hypothesis that there are exactly r cointegrating vectors in Xt and is given by:

max = -T ln(1-r).

(4)

The distributions for these tests are not given by the usual chi-squared distributions. The
asymptotic critical values for these likelihood ratio tests are calculated via numerical
simulations (see Johansen and Juselius 1990; and Osterwald-Lenum 1992).

2.3 Toda-Yamamoto Levels Vector Autoregression (VAR)

Sims, Stock, and Watson (1990) show that inference based on levels VAR is valid
since the Wald test used in Granger causality restrictions has a limiting chi-square
distribution if the time series are cointegrated and the long run relationship involves the

variable that is excluded under the null hypothesis.

However, this approach has

limitations because of its dependence on pre-tests for cointegration and its inapplicability
to mixed orders of integration processes.

A recent method proposed by Toda and

Yamamoto (1995) is complementary to the Sims, et al. (1990) technique because it

allows for causal inference based on augmented level VAR with integrated and
cointegrated processes.

This method is useful because it bypasses the need for

potentially biased pre-tests for unit roots and cointegration, common to other
formulations.
The Toda and Yamamoto (1995) procedure uses a modified Wald (MWALD) test to
test restrictions on the parameters of the VAR(k) model. This test has an asymptotic chisquared distribution with k degrees of freedom in the limit when a VAR[k+d(max)] is
estimated (where d(max) is the maximal order of integration for the series in the system).
Two steps are involved with implementing the procedure.

The first step includes

determination of the lag length (k) and the maximum order of integration (d) of the
variables in the system. Measures such as the Akaike Information Criterion (AIC) and
Hannan-Quinn (HQ) Information Criterion can be used to determine the appropriate lag
structure of the VAR. Given the VAR(k) selected, and the order of integration d(max) is
determined, a levels VAR can then be estimated with a total of p=[k+d(max)] lags. The
second step is to apply standard Wald tests to the first k VAR coefficient matrix (but not
all lagged coefficients) to conduct inference on Granger causality.

3. Empirical Analysis and Results

3.1. Data and Integration Properties

The data set consists of Canadian observations on real GDP, real exports, real
terms of trade (export unit value divided by import unit value), manufacturing
employment as proxy for labor, gross capital formation as proxy for capital, and
industrial production index for all industrialized nations as the proxy for foreign output
shock. Industrial production for all industrialized countries control for export growth not
influenced by Canadian price competitiveness or productivity, but by growth in the rest
of the world. The data set is quarterly and covers the period 1961:1 to 2000:4. All data
series are in natural logarithms and are obtained from the IMF database. Time series
univariate properties were examined using two unit root tests: augmented Dickey and
Fuller (1979) and Phillip and Perron (1988) tests. Neither of these tests reject the I(1) null
for any of the variables (see Table 1). Both the AIC and HQ information criteria test
results suggest using a lag length of five (which has white noise residuals). Subsequent
analysis therefore proceeds with the use of VAR with lag length k=5.
Results of cointegration rank tests, presented in Table 2, suggest the existence of
at most one cointegrating vector present in the system. This implies the presence of five
independent common stochastic trends in this system of six variables. This finding is
consistent with results of Henriques and Sadorsky (1996) that also found evidence of
cointegration between Canadian real exports, real GDP, and real terms of trade.

3.2. Granger causality tests based on VECM

Since there exist a unique cointegrating vector in the six-variable VAR used in
the cointegration tests, it is best to estimate models with one error correction term
included to capture long run relationships.

Beyond the analysis of the long-run

relationships among the six variables in the system, the short-run dynamics is also
explored by performing multivariate Granger causality tests for the VECM. F-statistics
and probability (in parentheses) for Granger causality tests from the VECM specification
are presented in Table 3. Also included in Table 3 are the t-statistics for the error
correction terms (ECT) from each of the six equations. Emphasis is placed only on the
relationships between Canadian real exports and real GDP.

For each variable in the

system, at least one channel of Granger causality is active: either in the short-run through
the joint tests of lagged-differences or a statistically significant ECT. This latter channel
is provided by the VECM specification and is only significant in the two equations for
real GDP and foreign output shocks. A significant ECT coefficient implies that past
equilibrium errors pay a role in determining current outcomes. The short-run dynamics
are captured by the individual coefficients of the differenced terms. It is noted that the
ECT coefficient for the real GDP equation is statistically significant while the ECT
coefficient for the real exports equation is not significant. This implies that export growth
did Granger cause the growth in real GDP (but not vice versa) in the long-run.
Similarly, in the short-run, the ELG hypothesis is also supported since the test
that real GDP does not Granger-cause real exports could not be rejected at the 5% level.

10

The empirical evidence strongly suggests that real exports Granger-causes real GDP at
the 1% level of significance. In support of our inclusion of other variables omitted in
other studies, changes in capital and labor do influence real GDP growth. This finding
for labor and capital, which was omitted in previous studies on Canada, is consistent with
the augmented aggregate production function assumed earlier.

This implies that in

addition to the influence of exports, other variables (such as capital/investment, terms of

trade, employment level, and foreign output shocks) also matter for the growth of
national output. In contrast results from the output equation, real exports seem to be
exogenous at the 5% level of significance. Only foreign output shocks have a notable
influence on the growth of real exports. These results provide strong empirical evidence
in support of the ELG hypothesis.
3.3. Granger causality tests based on Toda-Yamamoto Levels VAR
Results from a VAR estimated using the procedure developed by Toda and
Yamamoto (1995) are presents in Table 4. Although the optimal lag order chosen by the
AIC, FPE, HQ, and likelihood ratio criteria was five, a (k+1=6) order VAR was
estimated with restrictions placed on lagged terms up to the kth lag. Since all the
variables are in levels, no short-run causality flows exist as was the case with the VECM.
Rather the results provide information about the long-run causal relationships among the
variables in the system. Consistent with the VECM approach, this model specification
tends to further confirm the ELG hypothesis for Canada in the long-run.

11

Furthermore, real exports growth is Granger-caused by foreign output shocks

measured by real industrial production index for all industrialized nations.

Also of

importance in Table 4 is the result that real exports does not Granger-cause real terms of
trade. This contrasts with the conclusion from Henriques and Sadorsky (1996) that their
result suggests that historically Canada may not be the small open economy that it is
often though of. The result of the present study is consistent with the ELG hypothesis
and supports the widely held view that Canada is a small open economy.
3.4 Structural Stability Results
Evidence from test plots for the recursive residuals and CUSUMSQ for the six
equations suggest that structural breaks may have occurred at the following points: 1973:4
and 1992:2. Chow test results confirm structural instability for all six equations at 1973:4,
but the hypothesis of no structural break at 1992:2 could not be rejected.3 The break point
at 1973:4 is plausible because it coincides with the OPEC oil price shocks of late 1973 that
had impacts on productivity in the world economies. Given the results from the tests for
structural break, the VAR system was re-estimated over two sub-samples: 1960:1-1973:4
and 1974:1-2000:4. The F-statistics and p-values for the resulting Granger-causality tests
over the two sub-periods are generally consistent with those reported for the full sample.
The notable difference is the finding of a bi-directional causal flow between productivity
and exports in the early sub-sample (1961:4-1973:4). This result supports the view that in
the post-1974 period the Canadian economy has become increasingly dependent on
exports abroad, especially to the United States.

12

4. Concluding Remarks

The objective of this paper is to determine if the export-led growth (ELG)

hypothesis is valid for Canada. Two recent studies on the causal linkage between exports
and output growth in Canada (Serletis 1992; Henriques and Sadorsky 1996) yielded
contradictory findings.

This paper extended these previous studies on Canada by

employing recently developed time series estimation techniques and including previously
omitted relevant variables. To determine whether Canadian data are consistent with
export-led growth or growth driven exports, two alternative methodological procedures
were used to test for Granger causality: VECM and the augmented level VAR model
with integrated and cointegrated processes of arbitrary orders developed by Toda and
Yamamoto (1995). The analysis focused on the dynamic causal relationship between
exports, capital, labor, terms of trade, foreign output shock, and output growth using
quarterly Canadian data (1961:1-200:4).
Empirical evidence from Granger causality tests based on both alternative models
indicate that changes in real exports precede changes in real GDP. Furthermore, our
inclusion of additional variables was justified, as capital, labor, and foreign output shocks
are statistically significant in the models. In addition to finding support for the ELG
hypothesis in the short-run, the results from cointegration analysis and levels VAR also
provide support for ELG in the long-run.

This papers conclusions supporting the

validity of the ELG hypothesis are similar to results from Serletis (1992) for Canada and
Marin (1992) empirical findings for other industrialized nations such as the United

13

States, Japan, United Kingdom, and Germany.

Given recent depreciations of the

Canadian dollar and improved access to U.S. and other foreign markets, this finding is
plausible and consistent with prior expectations that increasing Canadian exports
stimulates economic growth.

One limitation of this study is the absence of an

explanation for the contemporaneous relationship between exports and productivity

growth. A useful extension of this paper would be to complement the Granger causality
analysis with a structural analysis of the contemporaneous error structure. Some recent
studies advocate the use of directed graphs theory as an alternative means of identifying
contemporaneous

VAR

innovations

(see

Bessler,

et

al

2002).

14

References
Akaike, H. (1974) A new look at the statistical model identification. IEEE Transactions
on Automatic Control, AC-19, 716-23
Bessler, D. A., and S. Lee (2002) Money and prices: U.S. Data 1869-1914, A study with
directed graphs. Empirical Economics 27, 427-446
Bessler, D. A., and J. Yang (2003) The structure of interdependence in international
stock markets. Journal of International Money and Finance, forthcoming
Buffie E. (1992) On the condition for export-led growth. Canadian Journal of
Economics 25, 211-225
Bahmani-Oskooee, M., and J. Alse (1993) Export growth and economic growth:
application of cointegration and EC modelling Journal of Developing Area 27, 535-42
Balassa, B. (1978) Exports and economic growth: further evidence. Journal of
Development Economics 5, 181-89
Banerjee, A., J.J. Dolado, J.W. Galbraith and D.F. Hendry (1993) Co-Integration,
Error Correction, and the Econometric Analysis of Non-stationary Data
(Oxford: Oxford University Press)
Chow, P.C.Y. (1987) Causality between exports and industrial performance: evidence
from NICs. Journal of Development Economics 26, 55-63
Dickey, D.A., and W.A. Fuller (1979) Distribution of the estimators for autoregression
time series with a unit root. Journal of the American Statistical Association 74, 427-31
Engle, R.F., and C.W.J. Granger (1987) Cointegration and error correction:
representation, estimation, and testing. Econometrica 55, 251-76
Dolado, J.J., and H. Lutkepohl (1996) Making Wald test work for cointegrated VAR systems.
Econometrics Review 15, 369-86

15

Grossman, G.M., and E. Helpman (1991) Innovation and growth in the global economy.
(Cambridge, MA: MIT Press)
Helpman, E., and P. Krugman (1985) Market structure and foreign trade. (Cambridge, MA:
MIT Press)
Henriques, I., and P. Sadorsky (1996) Export-led growth or growth-driven exports? The
Canadian case. Canadian Journal of Economics 29, 540-55
Johansen, S. (1988) Statistical analysis of cointegrating vectors. Journal of Economic
Dynamics and Control 12, 231-54
Johansen, S., and K. Juselius (1990) Maximum likelihood and inference on cointegration: with
applications to the demand for money. Oxford Bulletin of Economics and Statistic 52,
169-210
Marin, D. (1992) Is the export-led growth hypothesis valid for industrialized countries?
Review of Economics and Statistics 74, 678-88
McKinnon, R. (1964) Foreign exchange constraint in economic development and efficient aid
allocation. Economic Journal 74, 338-409
Michaely, M. (1977) Exports and growth: an empirical investigation. Journal of Development
Economics 40, 49-53
Osterwald-Lenum, M. (1992) A note with quantiles of the asymptotic distribution of the
maximum likelihood cointegration rank test statistics: four cases. Oxford Bulletin of
Economics and Statisitics 54, 461-72
Phillips, P.C.B., and P. Perron (1988) Testing for a unit root in time series regression.
Biometrika 75, 335-46
Serletis, A. (1992) Export, growth, and Canadian economic development. Journal of
Development Economics 38, 133-45

16

Sheehey, E.J. (1990) Exports and growth: a flawed framework. Journal of Development
Economics 27, 111-16
Sims, C.A., J.H. Stock, and M.W. Watson (1990) Inference in linear time series models with
unit roots. Econometrica 58, 113-44
Toda, H.Y., and P.C.B. Phillips (1993) Vector autoregressions and causality. Econometrica 61,
1367-93
Toda, H.Y., and T. Yamamoto (1995) Statistical inference in vector autoregression with
possibly integrated processes. Journal of Econometrics 66, 225-250
Yamada, H. (1998) A note on the causality between export and productivity.Economics
Letters 61, 111-114

17

Table 1. Unit Roots Tests, 1961:1 - 2000:4

ADF (k)
Levels
GDP
EXP
TOT
K
L
IP

-2.52(1)
-2.54(4)
-1.49(1)
-2.56(1)
-2.62(5)
-2.23(1)

First differences
GDP
EXP
TOT
K
L
IP

-7.18(1)
-5.72(3)
-7.55(1)
-6.65(1)
-4.57(4)
-5.87(1)

PP (k)

-2.75(1)
-1.63(3)
-1.37(1)
-2.48(1)
-2.72(3)
-3.04(1)

***
***
***
***
***
***

-9.93(1)
-36.59(3)
-10.62(1)
-8.19(2)
-10.99(3)
-6.60(1)

***
***
***
***
***
***

Notes:
***, **, and * denote that a test statistic is significant
at the 1 %, 5%, and 10% signigificance level, respectively.

The critical values for the tests at the 1%, 5%, and 10%
significance level, respectively, are -3.51, -2.89, and -2.58
respectively. Values in parentheses are lag lengths.

18

Table 2. Johansen Cointegration Test Results, 1961:1 - 2000:4

# of Cointegrating
Vectors
r=0
r1
r2
r3
r4
r5

Trace
Statistics
112.86**
66.09
38.80
21.14
8.47
3.31

-max
C(5%)
94.15
68.52
47.21
29.68
15.41
3.76

Statistics
46.77**
27.29
17.66
12.67
5.16
3.31

C(5%)
39.37
33.46
27.07
20.97
14.07
3.76

Notes:
Critical values used are taken from Osterwald-Lenum (1992)
** indicates rejection at the 95% critical values

19

Table 3. Granger causality test results based on VECM

Short run lagged differences
GDP
Dep.
Variables
GDP
EXP
TOT
K
L
IP

5.0736
(0.407)
3.4490
(0.6311)
7.1604
(0.209)
2.7518
(0.7382)
10.5266
(0.0616)

EXP

TOT

F-Statisitics
24.3867
9.4106
(0.0002) (0.0938)
1.4809
(0.9153)
5.0241
(0.413)
9.5095
7.7207
(0.0904) (0.1723)
6.8499
5.9292
(0.232) (0.3132)
5.8072
8.5824
(0.3254) (0.1269)

16.4925
(0.0056)
1.9249
(0.8594)
1.5497
(0.9073)
14.2087
(0.0143)
9.0227
(0.1082)

14.1324
(0.0148)
2.7519
(0.7382)
2.3077
(0.8051)
8.0635
(0.1528)
11.7277
(0.0387)

IP

6.9763
(0.2224)
10.8223
(0.055)
6.3932
(0.2698)
1.9868
(0.851)
11.6272
(0.0403)
-

Lagged
ECT

t-statistic
[ 4.63125]
[-0.99646]
[ 1.62995]
[ 1.67959]
[ 0.26676]
[ 3.19124]

Notes:
Figures presented in final column are estimated t-statistics testing the null that the lagged error correction
term (ECT) is statistically insignificant for each equation. All other estimates are asymptotic Granger
F-statistics. Values in parentheses are p-values.

20

Table 4. Granger causality test results based on Toda-Yamamoto Procedure

GDP
Dep.
Variables
GDP
EXP
TOT
K
L
IP

5.6633
(0.4619)
6.9814
(0.3226)
11.5921
(0.0717)
2.9371
(0.8167)
23.1542
(0.0007)

EXP

TOT

Modified Wald Statistics

18.8303
12.1429
13.4062
(0.0045)
(0.0589)
(0.037)
3.3359
1.8317
(0.7657)
(0.9345)
5.7835
2.6884
(0.4479)
(0.8468)
7.6059
5.8779
(0.2684)
(0.437)
10.7324
9.7526
14.7767
(0.097)
(0.1355)
(0.0221)
10.9203
8.6471
9.9760
(0.0909)
(0.1944)
(0.1257)

9.3199
(0.1564)
6.4550
(0.3742)
6.7598
(0.3436)
11.6992
(0.069)
12.4330
(0.053)

IP

8.8916
(0.1798)
12.9240
(0.0443)
8.3179
(0.2157)
3.7550
(0.7098)
11.9449
(0.0632)
-

Notes:
The [k+d(max)]th order level VAR was estimated with d(max)=1 since the order of integration is 1.
Lag length selection of k=5 was based on AIC and HQ information criteria test results.
Reported estimates are asymptotic Wald statistics. Values in parentheses are p-values.

21

The Department of Food and Resource Economics

College of Agriculture and Natural Resources
University of Delaware
The Department of Food and Resource Economics carries on an extensive and coordinated
program of teaching, organized research, and public service in a wide variety of the following
professional subject matter areas:
Subject Matter Areas
Agricultural Finance
Agricultural Policy and Public Programs
Environmental and Resource Economics
Food and Agribusiness Management
Food and Fiber Marketing
International Agricultural Trade

Natural Resource Management

Operations Research and Decision Analysis
Price and Demand Analysis
Rural and Community Development
Statistical Analysis and Research Methods

The departments research in these areas is part of the organized research program of the
Delaware Agricultural Experiment Station, College of Agriculture and Natural Resources. Much
of the research is in cooperation with industry partners, other state research stations, the USDA, and
other State and Federal agencies. The combination of teaching, research, and service provides an
efficient, effective, and productive use of resources invested in higher education and service to the
public. Emphasis in research is on solving practical problems important to various segments of the
economy.
The departments coordinated teaching, research, and service program provides professional
training careers in a wide variety of occupations in the food and agribusiness industry, financial
institutions, and government service. Departmental course work is supplemented by courses in other
disciplines, particularly in the College of Agriculture and Natural Resources and the College of
Business and Economics. Academic programs lead to degrees at two levels: Bachelor of Science
and Masters of Science. Course work in all curricula provides knowledge of tools and techniques
useful for decision making. Emphasis in the undergraduate program centers on developing the
students managerial ability through three different areas, Food and Agricultural Business
Management, Natural Resource Management, and Agricultural Economics. The graduate program
builds on the undergraduate background, strengthening basic knowledge and adding more
sophisticated analytical skills and business capabilities. The department also cooperates in the
offering of an MS and Ph.D. degrees in the inter disciplinary Operations Research Program. In
addition, a Ph.D. degree is offered in cooperation with the Department of Economics.
For further information write to:

Dr. Thomas W. Ilvento, Chair

Department of Food and Resource Economics
University of Delaware
Newark, DE 19717-1303

FREC Research Reports

are published as a
service to Delawares
Food and Agribusiness
Community by the
Department of
Food and Resource
Economics, College
of Agriculture and
Natural Resources
of the University of
Delaware.