Government spending and private consumption among select African countries: a panel data approach.

Anoruo, Emmanuel

Abstract

This paper explores the relationship between government spending and private consumption for 24 African countries using panel data approach. The results indicate that government spending engenders private consumption. This finding supports the complementarity's view relative to government spending-private consumption nexus. From a policy standpoint, the results suggest that fiscal policy could be used to stimulate economic activity by manipulating private consumption.

JEL Classification: C22, E21

Keywords: Crowding-in effect, panel data, random effects, fixed effects, private consumption.

Introduction

The relationship between government spending and private consumption continues to attract the attention of fiscal officials and economists alike. The question is whether government spending crowds in or out private consumption. There are two opposing views with regard to the relationship between government spending and private consumption. These include substitutability and complementarity views. The substitutability argument stipulates that an increase in government spending depresses private consumption. In other words, this view maintains that an increase in government spending "crowds-out" private consumption. The substitutability view was first advanced by Baily (1971) and has been empirically confirmed by a number of subsequent studies, including Barro (1981), Aschauer (1985), Ho (2001a, 2001b, 2001c), Kormendi (1983), Ahmed (1986), Baxter and King (1993), Katsaitis (1987), Leiderman and Razin (1989), Rock, Craigwell and Sealy (1989), Dalamagas (1992a, 1992b), Hatzinikolaou (2000) and Campbell and Mankiw (1989, 1991). Giorgioni and Holden (2003) using data from ten developing countries examined the effects of temporary and permanent government spending on private consumption. They find that both temporary and permanent components of government spending have partial impact on private consumption. Contrary to the substitutability theory, the complementarity view holds that government spending "crowds-in" private consumption, According to this view, an increase in government spending engenders private consumption. Karras (1994) and Devereux et al (1996) have validated the complementarity's proposition. Giavazzi, Jappelli and Pagano (2000), Hoppner and Wesche (2000), Khalid (1996), Kuehlwein (1998) and Giavazzi and Pagano (1996) used regime-switching models to assess the effect of government spending on private consumption. The results from these studies find government spending to be expansionary in the first regime. However, in the second regime, government spending is found to retard private consumption.

The previous studies in the extant literature have mainly focused attention on the impact of government spending on private consumption in the context of OECD countries. Developing African countries have not been accorded adequate attention on this issue. After all, the macroeconomic dynamics that govern the relationship between private consumption and government spending in industrialized nations are different from those in developing countries. For instance, developing countries are often associated with massive external debt, rapid population growth, imperfect capital markets, capital flight and financial repression. In addition, most of the earlier studies applied cross-sectional analyses in examining the relationship between private consumption and government spending.

In light of these drawbacks, the present study uses recent advances in panel data approach to investigate the effect of government spending on private consumption for a panel of 24 African countries. In particular, the fixed--and random-effect models are used to investigate the dynamics between government spending and private consumption for the sample countries. These models are attractive given the diverse nature of the sample countries. Unlike the cross-sectional models, these procedures are able to detect both time--and country-specific effects.

This paper is organized as follows. Following the present introduction, section 2 discusses the data and the descriptive statistics. Section 3 furnishes the methodology. Section 4 presents the empirical results. Section 5 presents the summary and policy implications of the study.

Data and Descriptive Statistics

The data used in this study were collected from the International Monetary Fund's International Financial Statistics (IFS) CDROM data disk, 2001. The data consist of annual observations on government spending, private consumption and disposal income for 24 African countries. The variables are expressed in real terms (i.e. deflated by the CPI). The private consumption, government spending and disposable income series were converted into the US dollar (US$). The data span from 1980 through 1999 for each country. The sample countries include Algeria, Benin, Botswana, Burkina Faso, Burundi, Cameroon, Congo Republic, Cote D Ivoire, Ethiopia, Gabon, Kenya, Lesotho, Madagascar, Malawi, Mali, Morocco, Mozambique, Niger, Rwanda, Senegal, Sierra Leone, Togo, Zambia and Zimbabwe.

Table 1 presents the descriptive statistics for real government spending, (RGS), real private consumption (RPC) and real disposable income (proxied by real GDP) (RY). As can be seen from Table 1, the mean values of RGS, RPC and RY are $0.124, $0.426 and $0.693 billions. The maximum and minimum values show cross-country variability among the variables used in the study. The standard deviation reveals that RPC deviated the most from the sample mean while RGS clustered the most around the mean.

Methodology

This study uses both the fixed-and random-effect models to examine the impact of government spending on private consumption among 24 African countries. The models are based on the following equation:

Yit = [chi]'it [gamma]it + [mu]it ...(1)

where Y represents the dependent variable (real private consumption), c' is a vector of explanatory variables (in our case, real income and real government spending), i stands for the countries in the sample (i= 1, 2, 3, 4 ..., 24), t is the period under investigation (t = 1980, 1989, 1990, 1991, ... 1999) and [mu]it is the error term. From equation (1) we derive the fixed effects model in terms of the notations used in the study as follows:

RPCit = [beta]1 RGSit + [beta]2RYit [alpha]i + [delta]i + [mu]it ... (2)

where RPC represents real private consumption, RGS stands for real government spending, while [mu] is the error term. In equation (2), [a.sub.i] captures unobserved country-specific effects assumed fixed over time. The year-effects represented by [a.sub.i] are included to account for shocks that are common to all countries in the sample, such as rapid population, slow economic growth, and imperfect capital markets.

From equation (1) we again generate the random effects model as follows:

RPCit = [beta]1RGSit[gamma]i + [beta]2RYit[gamma]i + [delta]i + [mu]it, [gamma]i = [bar.[gamma]] + [[bar.h].sub.i] ...(3)

where RPC represents real private consumption, RGS stands for real government spending, [mu] is the error term, [[bar.h].sub.i] stands for random country effect while [bar.[gamma]] represents the mean of the coefficient vector. Under the random effects model, the slope coefficients are allowed to vary randomly across countries. In this study, real income (Y) (proxied by real GDP) serves as a mediating variable between government spending and private consumption. Graham (1993) argues that real income should be included in the examination of the relationship between government spending and private consumption to ensure consistent regression estimates.

Most of the previous country-studies applied the standard OLS procedure to examine the impact of government spending on private consumption. These studies assumed that the omitted variables are independent of the explanatory variables and are independently, identically distributed. This assumption however leads to **biased inferences especially when country-specific features, such as policy changes; political regimes and tax policies that might affect private consumption are not taken into consideration. Hsiao (1986) points out that the OLS procedure yields biased and inconsistent estimates when the omitted country-specific variables are correlated with the explanatory variables.

The panel data approach provides avenues through which the country-specific characteristics (whether observed or unobserved) can be incorporated into cross-country studies to avoid biases resulting from the omission of relevant variables. The fixed-effects procedure yields unbiased and consistent estimates when the omitted country-specific variables are correlated with the explanatory variables. It is important to point out that one of the weaknesses associated with fixed-effects model is that it assumes that differences across countries represent shift in the regression equation. This assumption implies that the fixed-effects model is appropriate when the entire population rather than the sample is investigated. However the random-effects model is applied when a sample rather the population is considered. The random-effects model is not without flaws. It yields biased regression estimates if the omitted country-specific variables are correlated with the explanatory variables. This study considers both the fixed-effect and random-effect procedures given the weaknesses associated with each of the models. Furthermore, our sample (24 countries) is large enough to warrant the application of both techniques. However, the Hausman (1978) test procedure is implemented to gauge the performance of the fixed-effects model against the random-effects approach. Under the Hausman test, the null hypothesis is that the conditional mean of the disturbances' residuals is zero. The fixed-effects approach is preferred over the random-effects model if the null hypothesis is rejected. However, the random-effects procedure is preferred over the fixed-effects method if the null hypothesis is accepted.

Prior to estimating equations (2) and (3), we implement the Im et al. (1997) (IPS) panel unit root procedure to determine the time series properties of real private consumption, government spending and real income. The standard ADF unit root procedure is based on the following equation:

[DELTA]Xt = [alpha]0 + [beta]Xt-1 + [delta]t + [p.summation over (i=1) [theta]i[DELTA]Xt-1 + [epsilon]t ...(4)

where [DELTA] is first-difference operator, t represents time trend, and e stands for stationary random error, and n is the optimal lag length. The null and alternative hypotheses under the ADF unit root test are that [beta] = 0 and [beta] [not equal to] 1, respectively. IPS suggest that the average of the individual ADF t-statistics from independent cross-sections should be used to determine the panel unit root. The average ADF is calculated as follows:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] ...(5)

where tN, T = (1/N)[[summation].sup.N.sub.t=1] ti, represents the t-statistic from the OLS estimate of n in equation (4) for each of the units of the cross-section, and E[[bar.t]N, T([rho],0)] assumes that ni, = 0 for all i and with the selection of n = ([n.sub.1], [n.sub.2] ... [n.sub.N]) for each cross-section. The calculated value of [[theta].sub.t] is compared to critical value for one-sided N(0,1) distribution to determine the order of integration for each of the series in the panel.

Empirical Results

This section discusses the empirical results of the study. Table 2 displays the panel unit root procedures of IPS. The results suggest that real private consumption; government spending and real disposable income are level stationary at the 5 percent significance level for with and without time trend. In other words, the panel unit root test results suggest that real private consumption; government spending and real disposable income have zero order of integration or I(0).

Having ascertained the time-series properties of real private consumption, real government spending and real disposable income, we next estimate the fixed--and random-effect models. The fixed-effects model was estimated using the least squares dummy variable approach (LSDV). However, the generalized least squares (GLS) technique was used to estimate the random-effects model. (1) We apply the Breusch-Pagan (1980) Lagrange-multiplier test to gauge the performance of the random-effects model against the pooled OLS. The Breusch-Pagan test statistic (33.00) presented in Table 3 suggests that the random-effects model should be favored over the pooled OLS. The Hausman test result of 3.27 (with p-value of 0.19) presented in Table 3 suggests that the null hypothesis that the conditional mean of the disturbances is zero should not be rejected. The acceptance of the null hypothesis indicates that the results from the random-effects model are superior to those obtained from the fixed-effects framework. To this effect, we only display and interpret the results from the random-effects model in Table 3.

It can be seen from Table 3 that the regression coefficient on real government spending (RGS) is significantly positive at the 1 percent level. The results reveal that a US$1.00 increase in real government spending causes private consumption to rise by approximately US$0.63. The finding that real government spending and real private consumption are complementary is consistent with Karras (1994) and Devereux et al (1996). We next examine the effect of real disposable income on real private consumption. Consistent with economic theory, the results indicate that real disposable income has significantly positive effect on real private consumption. The t-ratio is significantly different from zero at the 1 percent level. "The results suggest that a US$1.00 increase in real disposable income causes real private consumption to increase by roughly US$0.44".

Summary and Policy Implications

This paper has examined the relationship between private consumption and government spending for a panel of 24 countries of Africa using the fixed- and random-effect frameworks. In particular, this paper examined the issue as to whether government spending crowds out private consumption. The results from the fixed- and random-effects models suggest that government spending has expansionary effect on private consumption. This finding is consistent with the Keynesian proposition, which stipulates that during a recessionary period, government spending should be used to stimulate economic activity and employment. In other words, increase in government spending engenders economic activity through aggregate demand. The results from this study implicate fiscal policy as a potent tool that could be used to stimulate aggregate demand and hence economic activity for the sample countries.

Acknowledgements

The author is grateful to the participants at the 56th Annual Conference of the International Atlantic Economic Society (held in Quebec City Canada) for their helpful comments and suggestions. The usual disclaimers apply.

REFERENCES

Ahmed, S. (1986), Temporary and Permanent Government Spending in an Open Economy: Some Evidence for the United Kingdom. Journal of Monetary Economics 17: 97-224.

Aschauer, D. A. (1985), Fiscal Policy and Aggregate Demand. American Economic Review 75: 117-127.

Bailey, M. J. (1971), National Income and the Price Level, New York, McGraw-Hill.

Barro, R. J. (1981), Output Effects of Government Purchases. Journal of Political Economy 84: 343-350.

Baxter, M., King, R. G. (1993), Fiscal Policy in General Equilibrium" American Economic Review, 83, 315-334.

Breusch, T. S., Pagan, A. R. (1980), Lagrange Multiplier Test and Its Applications to Model Specification in Econometrics. Review of Economic Studies 47: 239-253.

Dalamagas, B. A. (1992a), How Rival are the Ricardian Equivalence Proposition and the Fiscal Policy Potency View? Scottish Journal of Political Economy 39: 457-476.

Dalamagas, B. A. (1992b), Testing Ricardian Equivalence: A Reconsideration. Applied Economics, 24(1): 59-69.

Devereux, M. B., Head, A., Lapham, B. J. (1996), Monopolistic Competition, Increasing Returns and the Effects of Government Spending. Journal of Money, Credit and Banking 28: 233-254.

Giavazzi, F., Jappelli, Pagano, T., M. (2000), Searching for Non-Linear Effects of Fiscal Policy: Evidence from Industrial and Developing Countries. European Economic Review 44(7): 1259-1290.

Giavazzi, F., Pagano, M. (1996), Non-Keynesian Effects of Fiscal Policy Changes International Evidence and the Swedish Experience. Swedish Economic Policy Review 3: 67-103.

Giorgioni, G., Holden, K. (2003), Does the Ricardian Equivalence Proposition Hold in Less Developed Countries? International Review of Applied Economics 17(2): 209-221.

Graham, F.C. (1993), Fiscal Policy and Aggregate Demand: Comment. American Economic Review 83: 695-666.

Hatzinikolaou, D. (2000), Sensitivity of Consumption to Income and Government Purchases: Some Specification and Estimation Issues. Applied Economics 32: 767-775.

Ho, T. (2001a), The Government Spending and Private Consumption: A Panel Cointegration Analysis. International Review of Economics and Finance 10: 95-108.

--. (2001b), Consumption and Government Spending Substitutability Revisited: Evidence from Taiwan. Journal of Political Economy 48(5): 589-604.

--. (2001c), Analyzing the Crowding-Out Problems of Taiwan. Journal of Economic Development, 26(1): 1-16.

Hoppner, F., Wesche, K. (2000), Non-linear Effects of Fiscal Policy in Germany: A Markov-Switching Approach. Discussion Paper 9/2000, Bonn Economic Discussion Papers.

Hsiao, C. (1996), Analysis of Panel Data, Cambridge: Cambridge University Press.

Im, K., Pesaran, H., Shin, Y. (1995), Testing for Stationarity in Heterogeneous Panels Manuscript, University of Cambridge.

Karras, G. (1994) Government Spending and Private Consumption: Some International Evidence. Journal of Money, Credit and Banking 26: 9-22.

Katsaitis, O. (1987) On the Substitutability between Private Consumer Expenditure and Government Spending in Canada. The Canadian Journal of Economics 20(3): 533-543.

Khalid, A. M. (1996) Ricardian Equivalence: Empirical Evidence from Developing Countries. Journal of Development Economics 51: 413-432.

Kuehlwein, M. (1998), Evidence on the Substitutability between Government Purchases and Consumer Spending within Specific Spending Categories. Economics Letters 58(3): 325-329.

Kormendi, R. C. (1983), Government Debt, Government Spending, and Private Sector Behaviour. American Economic Review 73: 994-1010.

Campbell, J. Y., Mankiw, N. G. (1989), Consumption, Income, and Interest Rates: Reinterpreting the Time Series Evidence In Stanley Fischer and Oliver J. Blanchard, (ed.), NBR Macroeconomics Annuals, Cambridge, Mass.: MIT Press.

--. (1991), The Response of Consumption to Income: A Cross-Country Investigation. European Economic Review 35: 723-767.

Leiderman, L., Razin, A. (1988), Testing Ricardian Neutrality with an Intertemporal Stochastic Model. Journal of Money, Credit and Banking 20(1): 1-21.

Rock, L. L., Craigwell, R. C., Sealy, R. C. (1989), Public Deficits and Private Consumption: Empirical Evidence from Small Open Economies. Applied Economics 21(5): 697-710.

NOTE

(1.) We also applied other procedures including the maximum likelihood procedure (ML) and the Instrumental Variables (IV) to equations (2) and (3). These procedures yielded similar results as those reported in Table 3.

Emmanuel Anoruo, Ph.D., Department of Management Science and Economics, Coppin State University, 2500 W. North Avenue, Baltimore, MD 21216, E-mail: [email protected].

Table 1
Descriptive Statistics (Billions of US $)

Series Mean STD Min Max

RGS .124 .538 .001 5.518
RPC .426 2.734 .002 13.000
RY .693 1.586 .002 27.214

RPC = real private consumption, RGS = real government spending,
RY = real income (GDP), Max = maximum, Min = minimum, STD = standard
deviation.

Table 2
Unit Root Tests for Heterogeneous Panel

Series Test Without Trend With Trend

TRGS IPS ADF-stat -2.72 *** -2.87 ***
 (0.00) (0.00)

RPC IPS ADF-stat -2.77 *** -3.10 ***
 (0.00) (0.00)

RY IPS ADF-stat -1.70 ** -1.69 **
 (0.05) (0.05)

*** and ** indicates the rejection of the null hypothesis of
non-stationarity at the 1% and 5% level, respectively. The 1% and 5%
one-sided critical values for IPS are -2.33 and -1.645, respectively.
The null hypothesis is rejected if the test statistic is less than the
critical value (-1.645). The p-values are in parentheses.

Table 3
Random-Effects Estimation Results (Dependent Variable: RPC)

Constant -0.04
 (0.47)
RGS 0.63 ***
 (6.09)
RY 0.44 ***
 (21.72)
Adj. [R.sup.2] 0.92
Hausman's Test Statistic (/ (2)) 3.27
Breusch-Pagan Test Statistic (/ (2)) 33.00
Sample Size 480

*** indicates significance at the 1% level. RPC = real private
consumption, RGS = real government spending, and RY = real income
(GDP). Numbers in parentheses represent tstatistics.