A time-series investigation of the impact of corporate and personal current taxes on economic growth in the U.S.
Saunders, Peter J.
Abstract
This paper investigates the impact of corporate income taxes and
personal current taxes on economic growth. Time-series tests analyzing
the relationship between corporate taxes and nonresidential investment
are implemented. Also tests of the impact of personal current taxes on
the labor supply, approximated by the average weekly overtime hours in
manufacturing, are undertaken. These tests include cointegration, vector
error correction (VEC) estimation, and Granger causality testing of the
relevant time-series data. Cointegration tests indicate the existence of
a stable long run relationship between corporate taxes and
nonresidential investment. Further investigation of this relationship is
undertaken within the VEC testing framework. VEC test results indicate
that corporation taxes have a negative impact on nonresidential
investment. Therefore, corporate taxes appear to affect negatively the
economic growth in the U.S. Econometric tests of the personal current
taxes data and the average weekly overtime data indicate that these two
time series data are statistically independent. Therefore, under this
study's testing framework, there is no statistical evidence of a
negative impact of personal current taxes on the labor supply and
economic growth.
JEL classification: E62, H20
Keywords: Corporate income taxes, personal current taxes, economic
growth, VEC estimation.
INTRODUCTION
Recent tax cuts enacted in the U.S. in the 2001 and 2003
legislation, have led to renewed interest in analyzing the impact of
taxes in general, and economic policies aimed at reducing taxes in
particular, on economic growth. The current economic decline underlines
further the urgency of understanding the impact of taxes on economic
growth. (1) The proponents of tax reducing policies rely on the
supply-side theory to explain the link between taxes and economic
growth. The key element in the supply-side analysis of tax policies
rests upon the incentive effects of such policies on the supply of labor
and capital in an economy. Since an economy's output (X) is the
function of capital (K), labor (L), and technology (T), {X = x (K, L,
T)}, any tax policy that will lead to an increase in these resources
will bring about economic growth. In theory, reducing personal income
taxes will provide incentives to work harder and longer, thereby
increasing the supply of labor, while reducing corporate taxes will lead
to higher capital investment by business firms. Higher amounts of labor
and capital will lead to economic growth. The objective of this study is
to provide new empirical evidence on the impact of corporate and
personal income taxes on economic growth through the above outlined
channels. Clearly, investigating the theoretical connection between
taxes and economic growth is crucial in view of not only the 2001 and
2003 enacted fiscal legislation, but even more so when the merits of
competing fiscal measures and their effectiveness to achieve economic
recovery are considered. (2) This issue needs immediate empirical
attention as the 2001 and 2003 tax reduction legislation is due to
expire in 2010. Empirical research can provide crucial information on
how this type of fiscal policy can impact the U.S. economy. The
supply-side empirical investigation must focus on examining the
relationship between taxes and economic growth.
There is perhaps no more important issue in the field of
macroeconomics than identifying factors that influence economic growth.
This subject has received extensive attention in economic literature.
Recent contributions in the on-going debate concerning the factors that
affect economic growth include research by Dollar (1992) and Edwards
(1992). These authors analyze the impact of trade on economic growth.
Numerous studies address the impact of various taxes on economic growth.
Helms (1985) found that when state and local taxes are used to fund
transfer payments, then these taxes significantly retarded economic
growth. However, the disincentive effects of these taxes are negated if
tax revenues are used to finance improved public services. Similar
conclusions about the effects of state and local taxes on economic
growth are reached by Mofidi and Stone (1990). The results of their
study also indicate a negative impact of these taxes when tax revenues
are used to fund transfer payments, but a positive effect when tax
revenues are used to finance expenditures on health, education, and
public infrastructure.
Jorgenson and Wilcoxen (1997) analyze the impact of tax reform,
such as the flat-rate consumption tax, on economic growth. Their study
indicates that this tax would increase economic growth as measured by
the GDP. Auerbach and Hines (1988) analyze the impact of taxes on
investment in the U.S. Hulten (1984) also investigates the impact of tax
policy on investment decisions. Dye (1980) investigates the impact of
taxing and spending on economic growth in the U.S. Dye finds that while
tax policies do not have much impact on economic growth, expenditure
policies do impact economic growth significantly. Engen and Skinner
(1996) also examine the impact of taxation on economic growth. Contrary
to Dye's research, the joint authors report a positive impact of
tax policies on economic growth. Blanchard and Perotti (2002) claim that
both increases in taxes and increases in government expenditures affect
investment spending negatively. Zagler and Durnecker (2003) also find
that several tax rates and categories of expenditures have a long run
impact on economic growth.
The impact of corporate taxes on economic growth has also been
investigated extensively. For example, Lee and Gordon (2005) used
cross-country data ranging from 1970 to 1997 to investigate the impact
of tax policies on a country's economic growth. Their study finds
that increases in corporate taxes have a negative impact on economic
growth. In fact, a ten percent reduction in the corporate tax rate will
result in a one to two percent increase in the annual rate of growth.
Similar conclusions about the impact of corporate taxes on economic
growth are reached by Djankov, Ganser, McLiesh, Ramalto, and Shleifer
(2008). Their cross-sectional study of 85 countries in 2004 indicates
that corporate taxes have a large negative impact on aggregate
investment and economic growth in countries under their empirical
investigation.
Clearly, empirical results on the effects of taxation on economic
growth to date, although numerous, are controversial and inconclusive.
Since much of the empirical research is conducted within a
cross-sectional testing framework, the issue of the short and the long
run effects of tax changes on economic growth remains largely
unresolved. Further research of the short and the long run effects of
tax policies is needed to provide additional empirical evidence on how
taxes affect economic growth. The objective of this paper is to provide
such evidence by analyzing the impact of corporate taxes on business
investment, and by investigating the impact of personal income taxes on
the labor supply. The novelty of the present research is twofold. First,
it provides new empirical information on both the short and the long run
effects of taxes on economic growth. Second, the issue of causality in
the taxes and economic growth relationship is investigated. Granger
(1969) causality testing framework is used to accomplish this task.
These above stated objectives can best be accomplished within a reduced
form modeling of time series data. Unlike structural modeling, reduced
form modeling of the data can be used to shed light on the short and the
long run relationship among all test variables. It can also indicate the
existence or the absence of causal flows among the test variables.
Therefore, reduced form modeling is the most appropriate testing
structure to accomplish the above stated research objectives. This
modeling is comprised of unit root and cointegration testing, vector
error correction (VEC) estimation, and Granger (1969) causality testing.
Consequently, the remainder of this paper is organized in the following
way. Initially the time series methodological issues and the data
selections are outlined. Thereafter, test results of all time series
data are reported. Overall conclusions about the impact of corporate and
personal taxes on the U.S. economy are summarized in the final part of
this paper.
DATA AND METHODOLOGY
Taxes impact economic growth through numerous channels. For
example, one obvious way that corporate taxes affect economic growth is
by decreasing the amount of financial capital available for expenditures
on plants and equipment. Investment in new plants and equipment leads to
an increased production of goods and services and, thereby, to economic
growth. Therefore, the most appropriate macroeconomic measure of these
types of investment expenditures is the nonresidential investment (NRI).
Consequently, nonresidential investment is used in the present study to
approximate investment. Tax on corporate income (CTAX) is the best test
variable that measures the impact of taxation on corporations. Personal
current taxes (PCT) are used to investigate the effect of taxes on the
labor supply. In theory, personal income taxes affect the work-leisure
choice. Higher taxes will lead to a reduction in the work effort, and to
an increased consumption of leisure (3) The selection of the variable
that is best suited to capture the impact of income taxes on the work
effort-leisure choice is affected by the constraints faced by most labor
force participants. The majority of workers cannot vary their basic 40
hour work week. However, in most cases, they can choose the number of
overtime hours worked. Therefore, it is possible to capture the impact
of personal current taxes on the labor supply by analyzing the impact of
these taxes on the overtime hours worked. Average weekly overtime hours
in manufacturing (OH) are used in all subsequent tests to approximate
the overtime variable. Time series data on all the above described
variables ranging from the first quarter of 1947 to the third quarter of
2008 are used to analyze the effects of taxes on economic growth in the
U.S. (4)
Econometric modeling of the above described time series data
necessitates undertaking several steps prior to the estimation of any
test equations. These steps include, among others, unit root and
cointegration testing of all time series data. The results of these two
tests determine the appropriate testing framework to be deployed in all
the further data modeling. Initially, each individual time series data
must be subjected to stationarity tests. Time series data often exhibit
a trend. Such data are nonstationary, and as such are not suitable for
any econometric estimation. The objective of stationarity tests is to
determine whether the data are stationary or nonstationary, i.e., to
determine the degree of integration of each individual data series. If
the data are integrated of the same order, then a cointegration testing
framework can be used to determine the existence or the absence of a
long run relationship among the group of the test variables. (5)
Cointegration test results determine the type of estimation framework
that must be deployed in further time series analyses of all
relationships under investigation. If the data are found to be
cointegrated, then vector error correction (VEC) modeling is the
appropriate testing framework to be used in further data analyses. VEC
estimation can be used to investigate the short run dynamics of
relationships under empirical investigation. This modeling can also be
used to determine the direction of causal flows between test variables
in the short run. The absence of cointegration among test variables
necessitates using a different approach in further analyses of the
relationships. A VAR testing framework, such as the Granger (1969)
causality technique, can be used in such a case.
UNIT ROOT AND COINTEGRATION TESTS
As mentioned above, the first step in investigating the
relationship between corporate and personal income taxes and economic
growth is to subject each individual data series to stationarity tests
to determine the degree of integration of each time series. Unit root
testing accomplishes this objective. Unit root tests are numerous and
well known. They are commonly used in research that involves time series
data. Therefore, any further detailed outline of these tests would be
redundant. In the present study, the augmented Dickey-Fuller [Fuller
(1976), Dickey and Fuller (1979)] (ADF) test was deployed initially to
determine the degree of integration of the CTAX, NRI, PCT, and OH
variables. Thereafter the Phillip-Peron (1988) (PP) unit root test was
used to test the robustness of the initially obtained ADF test results.
An intercept and a deterministic time trend variable was included in all
test equations. (6) Test results for all the variables are reported in
Table 1 below. These tests indicate that CTAX, NRI, and PCT variables
are nonstationary in levels. These variables contain unit roots. It is
also clear that they contain only one unit root since their first
differences are stationary. Therefore, CTAX, NRI, and PCT are all
integrated of the first order, I(1). At the same time, both the ADF and
the PP tests indicate that the OH variable is stationary, i.e., it is
I(0).
The above reported unit root test results determine the appropriate
reduced form testing framework of the effect of taxes on economic
growth. CTAX and NRI variables are nonstationary, and integrated of
order one, I(1). Given this fact, cointegration estimation is the most
appropriate testing structure for investigating the impact of corporate
taxes on nonresidential investment. Since CTAX and NRI are
nonstationary, and of the same order of integration, it is possible that
corporate taxes and nonresidential investment may be related in the long
run. Cointegration tests of these two time series can be used to make
this determination. However, a cointegration testing method cannot be
used to examine the effect of personal current taxes on the labor
supply, since the PCT and the OH variables are of different orders of
integration, I(1) and I(0), respectively. Therefore, these variables
cannot be cointegrated. The absence of cointegration requires that an
alternative testing framework, such as the Granger (1969) causality
estimation within a VAR modeling, be used to analyze the effect of
personal current taxes on the labor supply.
One of the key advantages of reduced form modeling of the time
series data is its ability to provide some information on the long run
relationships among test variables, such as the relationship between
corporate taxes and nonresidential investment. If such a relationship is
established, than it would be reasonable to conclude that corporate
taxes may impact the long run economic growth in the U.S. As mentioned
above, a cointegration testing framework can be used to accomplish this
objective. Numerous cointegration tests are available to investigate the
long run relationship among any I(1) time series data, such as the
corporate taxes and nonresidential investment data. The Engle and
Granger (1987) test, the Stock and Watson (1988) estimation, as well the
Johansen (1988) procedure, among others, can be used to determine
whether corporate taxes and investment are related in the long run.
Although all cointegration tests share the same common objective of
finding the most stationary linear combination of the vector time series
under investigation, some important statistical differences exist among
these tests. Gonzalo (1994) investigated relative merits of alternative
cointegration tests. His investigation found Johansen's test
superior to the other cointegration tests. Dickey, Jansen, and Thornton
(1991) reached similar conclusions with respect to the merits of the
three above mentioned cointegration tests. Given the above findings
concerning relative merits of various cointegration tests,
Johansen's (1988) estimation was adopted in analyzing the long run
relationship between corporate taxes and nonresidential investment. In
addition to the above mentioned advantages of Johansen's estimation
method, there is one other compelling reason to deploy this particular
procedure in the present data analyses. There exists an important
econometric connection between Johansen's cointegration tests and
further VEC analyses of the CTAX and the NRI data. Test results
generated by Johansen's cointegration estimation can be used in
subsequent VEC data modeling. (7)
Results of the Johansen's cointegration testing of the CTAX
and NRI data are summarized in the above Table 2. (8) Both the trace and
the eigen value tests reject the null hypothesis of no cointegrating
equation. The likelihood ratio test implies the existence of one
cointegrating equation at the conventional five percent level of
statistical significance, given the value of the trace statistic of
24.630. The eigen statistic of 23.689 confirms this conclusion. These
test results provide new empirical evidence on the relation between
corporate taxes and nonresidential investment in the long run. They
indicate the existence of a stable long run relationship between these
two variables. Consequently, it appears that corporate tax policies may
influence inveStment decisions in the U.S. in the long run. However,
establishing the existence of a stable, long run relationship between
corporate taxes and nonresidential investment gives no indication of the
causal flows in this relationship. It would certainly be of crucial
importance to determine if corporate taxes have a causal impact on
investment. Cointegration tests alone cannot make this determination.
This objective can be accomplished within VEC analyses of the corporate
tax and nonresidential investment data.
VEC ESTIMATION
Given the fact that corporate tax and nonresidential private
investment time series data are cointegrated, it is possible to
investigate the short run dynamics of the relationship between these two
variables within a VEC testing framework. This investigation can provide
key evidence of the short run impact of corporate taxes on investment
and economic growth. In this research, the Engle and Granger (1987) VEC
estimation method is used to provide this vital information. One obvious
advantage of using this particular method for VEC estimation is its
connection with Johansen's (1988) cointegration estimation. Several
steps must be followed in the Engle and Granger VEC testing procedure.
These steps include initial integration and cointegration testing of the
CTAX and the NRI data, and thereafter, VEC estimation. Essentially, if
two time series variables such as CTAX and NRI are both integrated of
the same order I(1) and are also cointegrated, then the residuals from
the Johansen's cointegrating equation can be used in further Engle
and Granger VEC data modeling2 In accordance with the Engle and Granger
methodology, the two following equations are estimated as follows:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)
The Engle and Granger VEC model is estimated in the first
differences of levels. Therefore, [DELTA]CTAX and [DELTA]NRI are changes
in corporate taxes and nonresidential investment. It is assumed that at
least one of the coefficients ([rho] or [[rho].sub.1]l) is nonzero. The
[z.sub.t] terms are the residuals from Johansen's (1988) previously
estimated cointegration equation reported in Table 2. These lagged
[x.sub.t] terms provide key information about the short run relationship
between corporate taxes and investment. Their analysis provides
information about the dynamics of the model under empirical
investigation, as the two rho coefficients in equation (1) and (2) are
the speed of adjustment coefficients. (10) The combined analysis of
these coefficients can also provide information about the causal flows
between corporate taxes and nonresidential investment. In general, a
statistically insignificant [z.sub.t] term implies a state of
equilibrium in the model under investigation. Finding a statistically
significant coefficient of the lagged [z.sub.t] term indicates that
disequilibrium prevails in this model. In such a case, the signs of the
two coefficients determine the direction of causal flows between
corporate taxes and nonresidential investment. VEC results of
estimations of equations (1) and (2) are reported in the following Table
3.
The results reported in the above table provide key empirical
evidence on the short run impact of corporate taxes on nonresidential
investment in the U.S. Focusing on statistical estimates of the
[z.sub.t] terms in equations (1) and (2) individually provides this
information. The lagged [z.sub.t] coefficient in equation (1) is
negative and statistically significant. The coefficient of the same
variable in equation (2) is statistically insignificant. Therefore,
these VEC estimation results indicate that corporate taxes have a
statistically significant impact on nonresidential investment in the
short run. It appears that this impact is negative, as the lagged
[z.sub.t] term is negative in equation (1). VEC estimates reported above
also provide information about causal flows between corporate taxes and
nonresidential investment. This information is obtained by analyzing
jointly the test results of equations (1) and (2). As noted above, the
coefficient of this lagged term in equation (1) is statistically
significant while the same coefficient in equation (2) is statistically
insignificant. One interpretation of these results is to conclude that
changes in corporate taxes have a negative causal impact on
nonresidential investment in the short run. At the same time, there is
no evidence of a causal flow from nonresidential investment to corporate
taxes. Consequently, corporate taxes impede economic growth in the U.S.
GRANGER CAUSALITY TESTS
Cointegration and VEC modeling cannot be used to investigate the
impact of personal current taxes on the labor supply. This testing
structure can only be used when all test variables are integrated of the
same order of integration. In the present case, the PCT data are
nonstationary and I(1), while the OH data are stationary and I(0). This
outcome necessitates adopting a different estimation approach other than
cointegration and VEC modeling. A VAR Granger (1969) causality testing
framework can be used in further analyses of the relationship between
personal current taxes and overtime hours. Standard bivariate Granger
causality testing requires estimating the following equations;
[ILLUSTRATION OMITTED] (3)
[ILLUSTRATION OMITTED] (4)
where [X.sub.t] and [Y.sub.t] are the two time series variables
under investigation. Lags on the two test variables have to be selected
in the two above equations. Lag selection can be either arbitrary, or it
can be based upon a statistical criterion. Arbitrary lag selection in
causality testing suffers by two serious shortcomings. The first
involves the loss of degrees of freedom in cases when relatively short
sample sizes are analyzed and long lag lengths are chosen. The second
problem is perhaps even more serious, as an arbitrary lag selection
itself can influence Granger causality test results. (11) Numerous
statistical criteria, such as the Schwards information criterion (SIC),
the Akaike information criterion (AIC), and Hsiao's (1979 and 1981)
minimum final prediction error (FPE) are available for selecting the
appropriate lag test structure of the time series variables. Using any
one of such statistical criteria can solve the two above mentioned
problems that plague the arbitrary lag selection in causality testing.
In the present study, Hsiao's minimum final prediction criterion
(FPE) was used to analyze the causal flows between personal current
taxes and overtime hours.
The Hsiao's (1979 and 1981) minimum FPE causality testing
method is uniquely suited for bivariate causality tests involving
relatively short sample periods, such as those used in the PCT and the
OH data analyses. (12) It allows an empirical examination of all lags in
a predetermined lag range. The appropriate lag structure is determined
by minimizing the final prediction error (FPE). FPE is calculated as
(SEE) (2). (T + K) / T, where SEE is the standard error of the
regression, T is the number of observations, and K indicates the number
of parameters. Causality implications are determined by analyzing
results of several statistical tests whose main objective is to find a
lag structure that yields minimum FPE in each step of the estimation.
(13) This procedure was used to determine the optimum lag length of X
(PCT) and Y (OH) in equation (3). Lag lengths ranging from one to ten
were examined in each case. The optimum lag length of PCT was determined
to be three. This constituted the first step in Hsiao's causality
testing procedure. The particular lag structure of the PCT variable
obtained in this way was maintained while lagged values of OH were added
ranging from one to ten lags. Minimizing the FPE resulted in a one lag
selection for the OH variable. This estimation completed the second step
in causality testing of the PCT and OH data. The same procedure was used
in estimating equation (4) where the roles of X and Y were reversed.
This led to a selection of five lags for OH and one lag for PCT.
Causality implications were obtained by comparing minimum FPEs of the
two estimation steps.
There are three causality outcomes possible in the test cases
described above. First, there can be a unidirectional causal flow from
PCT to OH, or from OH to PCT. That is to say, PCT can Granger (1969)
cause OH, or OH can Granger cause PCT. Second, there can be a
bi-directional causality between the two test variables. Finally, PCT
and OH variables can be statistically independent of one another. These
inferences are based upon comparisons of the minimum FPEs of the two
above described estimation steps. If the minimum [FPE.sub.OH] without
the lagged values of PCT variable is greater than the minimum
[FPE.sub.OH] with the lagged values of the PCT variable, then causality
flows from personal current taxes to overtime hours. Similarly, if the
minimum [FPE.sub.PCT] without the lagged values of the OH variable is
greater than the minimum FPE with these lagged values, then a
unidirectional causality flows from OH to PCT. It is also possible that
adding lagged values of test variables in bivariate testing of the data
does not decrease the minimum FPEs obtained under univariate data
testing in the two test cases. Such a result would imply that the two
test variables are statistically independent.
The Granger (1969) causality test results reported in the above
table provide important new information on the causal relationship
between personal current taxes and overtime hours. These results
indicate the absence of causal flows between these two variables, as
adding the lagged variables in bivariate test structures in both
equations does not reduce minimum FPEs obtained in the univariate test
setting. In the first test case, the minimum FPE of 747.37 (step one) is
smaller than minimum FPE of 747.64 (step two). The same results are
obtained when the roles of OH and PCT are reversed. Consequently, it
appears that the two test variables are statistically independent.
Therefore, changes in personal current taxes do not appear to impact
negatively the overtime hours worked in the manufacturing sector of the
U.S. economy. Therefore, reducing personal current taxes does not appear
to have any statistically significant impact on the labor supply, under
the present econometric testing structure. Given these results, reducing
these taxes may not have the desired effect on economic growth in this
country.
OVERALL CONCLUSIONS
The rapidly deteriorating economic situation in this country and
throughout the rest of the world places utmost importance on determining
which economic policies are most likely to reverse this universal
economic decline. Fiscal policy can be used to accomplish this crucially
important task. This policy can be used either to stimulate the
aggregate demand, primarily through increases in various types of
government expenditures, or it can be designed to affect positively the
economy's aggregate supply. The latter task can only be achieved by
various tax reductions. In theory, tax decreases will have positive
incentive effects on economic resources. In particular, personal income
tax cuts will increase the labor supply while reducing corporate taxes
may lead to higher capital investment. Reducing taxes will lead to
economic growth through these above described channels. The objective of
this paper is to provide new empirical evidence on this key economic
issue of the impact of taxes on economic growth. This task is
accomplished within a time series reduced form modeling of economic
data. The novelty of the present research is two fold. First, it
analyzes both the short and the long run effects of taxation on economic
growth in the U.S. It also provides new empirical evidence on the
causality in the taxes and economic growth relationship. The focus of
the present paper is on investigating the impact of corporate taxes
(CTAX) on nonresidential private investment (NRI), and on the effect of
personal current taxes (PCT) on the average weekly overtime hours (OH)
in the manufacturing sector of the U.S. economy.
Time series data on these variables ranging from the first quarter
of 1947 to the third quarter of 2008 were used to analyze these
relationships. The initial data analyses were carried out within the
unit root and the cointegration testing structures. Thereafter, vector
error correction (VEC)and Granger (1969) causality estimations were used
to determine the impact of taxes on economic growth in the U. S.
Initially all time series data were subjected to unit root testing. The
augmented Dickey-Fuller (1976, 1979) (ADF) and the Phillips-Perron
(1988) (PP) tests were used to determine the order of integration of
each individual data series. These tests indicated that while the CTAX,
NRI, and PCT data are nonstationary and I(1), the OH data are
stationary, i.e., these data are I(0). These test results implied that a
cointegration and vector error correction (VEC) estimation framework is
appropriate for investigating the impact of corporate taxes on business
investment, while a VAR estimation must be used to investigate the
effects of personal taxes on the labor supply. The impact of personal
current taxes on average weekly overtime hours in the manufacturing
sector was analyzed within the Granger (1969) causality testing
framework. Hsiao's (1979 and 1981) minimum final prediction error
(FPE) estimation was used to determine the causal impact of personal
current taxes (PCT) on the labor supply approximated by the average
weekly overtime hours in manufacturing (OH). Test results indicated the
absence of any causal flows between these two test variables. These
results implied that lowering personal current taxes may not have the
desired positive impact on the labor supply and, thereby, on economic
growth in this country.
The Johansen (1988) cointegration test was used to determine the
long run relationship between CTAX and NRI. Both the trace and the eigen
value tests indicated the existence of a stable long run relationship
between these two variables. One obvious way to interpret these results
is by concluding that corporate taxes may have an impact on
nonresidential investment in the long run. However, cointegration tests
alone cannot be used to make meaningful inferences about the causal
impact of corporate taxes on business investment. The Engle and Granger
(1987) vector error correction (VEC) method was used to accomplish this
objective. VEC test results provided new empirical evidence on the
impact of corporate taxes on nonresidential investment. These results
indicated that increasing corporate taxes will impact negatively
nonresidential investment in the short run. Conversely, lowering
corporate taxes will not only have a positive long run effect on
economic growth, but such a tax policy may have a much desired positive
short run impact on business investment and economic growth. This result
is of crucial importance in the present economic circumstances when the
most important feature of any economic policy must be its ability to
lead to a quick economic recovery. The results of this research indicate
that lowering corporate taxes may accomplish this key economic
objective.
References
Auerback, A. and Hines, J. "Investment Tax Incentives and
Frequent Tax Reforms." American Economic Review, Vol. 78, No. 2,
1988, pp. 211-216.
Blanchard, O. and Perotti, R. "An Empirical Characterization
of the Dynamic Effects of Changes in Government Spending and Taxes on
Output." Quarterly Journal of Economics, Vol. 117, No. 4, 2002, pp.
1329-68.
Dickey, D. A., and W. A. Fuller. "Distribution of the
Estimators for Autoregressive Time Series with a Unit Root."
Journal of American Statistical Association, 74(366), 1979, 427-31.
Dickey, D. A., D. W. Jansen., and D. L. Thornton. "A Primer on
Cointegration with an Application to Money and Income." Review,
Federal Reserve Bank of St. Louis, 73(2), March/April, 1991, 58-78.
Djankov, S., Ganser, T., McLiesh, C., Ramalto, R., and Shleifer, A.
"The Effect of Corporate Taxes on Investment and
Enterpreneurship."Working Paper, American Law & Economic
Association Annual Meeting, 2008.
Dollar, D. "Outward-Oriented Developing Economies Really Do
Grow More Rapidly: Evidence from 95 LDCs, 1976-1985." Economic
Development and Cultural Change, 40, 1992, 523-44.
Dye, T. R. "Taxing, Spending, and Economic Growth in the
American States." Journal of Politics, VoW. 42, No. 4, 1980, pp.
1085-1107.
Edwards, S. "Trade Orientations, Distortions, and Growth in
Developing Countries." Journal of Development Economies, 39, 1992,
31-57.
Enders, W. Applied Econometric Time Series. New York: John Wiley
and Sons, 1995.
Engen, E. and Skinner, J. "Taxation and Economic Growth."
National Tax Journal, Vol. 49, No. 4, 1996, pp. 617-42.
Engle, R. F., and C. W. J. Granger. "Cointegration and Error
Correction: Representation, Estimation, and Testing."
Econometrica,, 55(2), 1987, 251-76.
Fuller, W. A. Introduction to Statistical Time Series. New York:
John Wiley and Sons, 1976.
Gonzalo, J. "Five Alternative Methods of Estimating Long-Run
Equilibrium Relationships." Journal of Econometrics, 60, 1994,
203-33.
Granger, C. W. J. "Investigating Causal Relations by
Econometric Models and Cross-Spectral Methods." Econometrica, 37,
1969, 424-38.
Helms. L. J. "The Effect of State and Local Taxes on Economic
Growth: A Time-Series--Cross Section Approach." Review of Economics
and Statistics, Vol. 67, No. 4, 1985, pp. 574-82.
Holden, K., and J. Thompson. "Co-Integration: An Introductory
Survey." British Review of Economic Issues, 14(33), 1992, pp. 1-55.
Hsiao, C. "Autoregressive Modeling of Canadian Money and
Income Data." Journal of the American Statistical Association, 74,
1979, pp. 553-60.
Hsiao, C. "Autoregressive Modeling and Money-Income Causality
Detection." Journal of Monetary Economics, 7, 1981, pp. 85-106.
Hulten, C. "Tax Policy and the Investment Decision."
American Economic Review, Vol. 74, No. 2, 1984, pp. 236-241.
Johansen, S. "Statistical Analysis of Cointegrating
Vectors." Journal of Economic Dynamics and Control, 12, 1988, pp.
231-54.
Jorgensen, D. W. and Wilcoxen, P. J. "The Long Run Dynamics of
Fundamental Tax Reform." American Economic Review, Vol. 87, No.2,
pp. 126-32.
Lee, Y., and Gordon, R. H. "Tax Structure and Economic
Growth." Journal of Public Economics, Vol. 89, Issues 5-6, 2005,
pp. 1027-43.
McCallum, B. T. "Unit Root in Macroeconomics Time Series: Some
Critical Issues." Economic Quarterly, Federal Reserve Bank of
Richmond, 79, 1993, pp. 13-43.
Mofidi, A. and Stone, J. A. "Do State and Local Taxes Affect
Economic Growth?" The Review of Economics and Statistics, Vol. 72,
No. 4, 1990, pp. 686-91.
Phillips, P. C. B., and P. Perron. "Testing for a Unit Root in
Time Series Regressions." Biometrica, 75, 1988, pp. 335-46.
Saunders, P. J. "Causality of the U.S. Agricultural Prices and
the Money Supply." American Journal of Agricultural Economics, 70,
1988, pp. 588-96.
Stock, J. H., and Watson, M. W. "Variable Trends in Economic
Time Series." Journal of Economic Perspectives, 2(3), 1988, pp.
14-74.
Thornton, D. L., and Batten, D. S. "Lag-Length Selection and
Tests of Granger Causality between Money and Income: Further Empirical
Evidence." Journal of Money, Credit, and Banking, 17, 1985, pp.
164-78.
U.S. Department of Commerce, Bureau of Economic Analysis.
U. S. Department of Labor, Bureau of Labor Statistics.
Zagler, M., and Durnecker, G. "Fiscal Policy and Economic
Growth." Journal of Economic Surveys, Vol. 17, Issue 3, 2003, pp.
397-418.
Notes
(1.) The U.S. economy is currently in the midst of a severe
economic recession, as shown by a rising unemployment rate and a rapidly
falling output, as well as other indicators The negative fourth quarter
2008 growth rate of the real GDP of 3.8 is unprecedented in recent
economic history of this country.
(2.) Fiscal policies can affect economic growth through two
distinct channels: by stimulating the aggregate demand (standard
Keynesian hypothesis), or by increasing an economy's aggregate
supply. Typically the aggregate demand policies favor increasing fiscal
expenditures, while tax policies are most suited to affect the aggregate
supply. Therefore, the key economic issue facing policy makers is
whether to use expenditure or tax policy to achieve economic recovery.
(3.) This theoretical result will depend on the relative strength
of the substitution and income effect.
(4.) The CTAX, NRI, and PCT data were obtained from the U.S.
Department of Commerce, Bureau of Economic Analysis. The OH data came
from the U.S. Department of Labor, Bureau of Labor Statistics. The OH
data are only available from the first quarter of 1956 onwards. The
sample period is always adjusted to accommodate this limitation in all
tests that use the OH data.
(5.) For a further excellent discussion of unit root and
cointegration tests see Holden and Thompson (1992), and McCallum (1993),
among others.
(6.) The SIC method was used to determine the test lag structure in
the ADF tests.
(7.) This connection is explained in a greater detail in the
following section of this paper.
(8.) The assumption of linear deterministic trend in the data was
used in cointegration data testing.
(9.) This estimation procedure was followed in the present paper in
the CTAX and NRI data analyses. Interested readers are referred to
Enders (1995), pages 373-81 for a more detailed description of this
estimation.
(10.) These coefficients describe the dynamics of the short run
relationship between corporate taxes and nonresidential investment. In
general, they outline the short run disequlibrium responses of the model
under investigation. A statistically insignificant coefficient implies
that change in [z.sub.t] does not respond to the deviations from the
long run equilibrium, indicating the state of equilibrium in the system
under investigation. A statistically significant coefficient implies the
existence of a disequlibrium in the model. For a more detailed
explanation of these issues see Enders (1995), among others.
(11.) For a further discussion of this issue see Thornton and
Batten (1985) and Saunders (1988), among others.
(12.) The OH data are only available from the first quarter of
1956. This constraint necessitated using a shortened sample period in
all subsequent causality tests.
(13.) A detailed description of the minimum FPE causality testing
procedure would be redundant, as it is well documented in economic
literature. Interested readers are referred to Hsiao (1979 and 1981) for
a more complete description of this causality testing technique.
PETER J. SAUNDERS, Professor, Department of Economics, Central
Washington University, 400 East University Way, Ellensburg WA
98926-7486, E-mail: saunders@cwu.edu
Table 1
ADF and PP Test Results for CTAX, NRI, PCT, and OH.
Variable ADF test results PP test results
CTAX (1) -2.866 -1.930
CTAX (2) -5.405 * -14.778 *
NRI (1) -0.810 -0.241
NRI (2) -5.764 * -8.181 *
PCT (1) -2.485 -1.494
PCT (2) -4.850 * -18.887 *
OH (1) -12.752 * -12.908 *
OH (2) -6.515 * -78.647 *
(1) ADF and PP test results for the levels of variables.
(2) ADF and PP test results for the first differences of levels.
* Indicates statistical significance at the five-percent level.
Table 2
Johansen Cointegration Test Results for CTAX and NRL Lags 1.2
Variables Trace Test Results Eigen Value Test Results
CTAX and NRI (1,2) 24.630 * 23.689 *
(1) Trace test indicates one cointegrating equation at the five
percent level of significance.
(2) Eigen value test indicates one cointegrating equation at the
five percent level.
* Indicates statistical significance at the five percent level.
Table 3
VEC Estimates of Equations (1) and (2)
Equation Dependent Independent Coefficient "t" Statistic
Variable Variable
(1) [DELTA] constant 2.867 4.0277 *
NRI z(-1) -0.018 -4.120 *
[DELTA] NRI(-1) 0.336 5.100 *
[DELTA] NRI(-2) 0.158 2.500 *
[DELTA] CTAX(-1) 0.232 3.793 *
[DELTA] CTAX(-2) 0.032 0.524
(2) [DELTA] constant 0.405 0.505
CTAX z(-1) 0.007 1.366
[DELTA] CTAX(-1) 0.126 1.830
[DELTA] CTAX(-2) 0.049 0.709
[DELTA] NRI(-1) 0.016 0.220
[DELTA] NRI(-2) 0.108 1.528
* Indicates statistical significance at the five-percent level.
Numbers in parentheses indicate number of lags.
Table 4
Causality Testing by Computing FPEs for OH and PCT *
Dependent Independent FPE Causality
Variable Variable Implications
PCT(3) 747.37
PCT(3) OH(1) 747.64 747.37<747.64
OH # PCT
OH(5) 2.359
OH(5) PCT(1) 2.379 2.359 < 2.379
PCT # OH
* Numbers in parentheses are lags for minimum FPEs.
