DOES GOVERNMENT SPENDING GROWTH EXCEED ECONOMIC GROWTH IN SAUDI ARABIA?

International Journal of Economics, Commerce and Management United Kingdom Vol. IV, Issue 2, February 2016 http://ijecm.co.uk/ ISSN 2348 0386 DOES GOVERNMENT SPENDING GROWTH EXCEED ECONOMIC GROWTH IN SAUDI ARABIA? Guo Ping College of Economics and Trade, Hunan University, Changsha, China Alotaish Mohammed Saud M. College of Economics and Trade, Hunan University, Changsha, China mohd.alotaish@hotmail.com Ihtisham ul Haq College of Economics and Trade, Hunan University, Changsha, China ihtishamamin_99@yahoo.com Abstract This study was carried out to examine Wagner s Law in case of Saudi Arabia. A time series data covering period from 1969 to 2013 was analyzed for the said purpose. Bivariate cointegration analysis was conducted to see how economic growth influence government spending. Unit root tests identified non-stationarity when variables were taken at levels. These tests verified that order of integration of study variables is one. Cointegration technique was applied to check long run relationship. This technique assured the presence of long run relation between government spending and economic growth. The long run estimates showed that GDP growth as well as per capita growth has positive significant effect on government spending and magnitude of coefficient of both is high than unity confirming existence of Wagner s Law. Causality analysis confirmed that a unidirectional causality is running from economic growth to government spending in long run. Keywords: Wagner s Law, GDP, Government spending, Economic growth, Saudi Arabia Licensed under Creative Common Page 354

International Journal of Economics, Commerce and Management, United Kingdom INTRODUCTION Adolf Wagner theorized that government spending will increase faster than economic growth in an economy. Responsibilities of government increases day by day as complexity of administrative and social related issues has been enhanced. Thus, the portion of gross domestic product (GDP) spare to government spending has been increased as compare to past. This shows that causality will run from economic size to government spending. Wagner (1883) considered government spending as endogenous variable depending on the size of economy (GDP or GDP per capita). For him, increase in government spending is not a cause but a consequence of the increase in size of economy. On other hand, Keynes (1936) postulated that government spending is exogenous and it can boost economic growth in time of economic downturn. According to Keynes economy is operating less than full employment and the effective demand can be increased through government spending. The increase in government spending will utilize idle (human and physical) resources which will help to achieve high economic growth. The papers related to government spending and economic growth nexus in the past for Saudi Arabia restricted to causality analysis and concluded that causality runs from economic growth to government spending, but did not find strong evidence to conclude whether Wagner s Law exists or not. This study is unique and differentiates itself from other studies that besides causal analysis cointegration regressions are carried out to estimates long run magnitude of economic growth. The results obtained from fully modified ordinary least squares, dynamic ordinary least squares and conical cointegration regression suggested that the effect of economic growth is more than unity so Wagner s Law does exist in case of Saudi Arabia. Long run granger causality postulated that economic growth ganger causes government spending. The rest of paper is organized in following manner. Literature review is discussed in section two of the paper. Third section talks about data and research methodology adopted for the current study. Results are interpreted and discussed in fourth section while last section of paper concludes the study. LITERATURE REVIEW Numerous empirical research studies have been devoted to the nexus between government spending and economic growth. However, these studies come up with mix results and concluded that this relationship varies from region to region and from country to country. The studies like Al-Faris (2002); Aregbeyen (2006); and Omoke (2009) among other found support for Wagner s Law. However, empirical studies like Abizadeh and Yousefi (1998); Burney (2002); Huang (2006); and Babatunde (2007) among others either found no evidence to support the Licensed under Creative Common Page 355

Guo, Alotaish & Ihtisham Law or found weak support. Some researchers opined that there did not exist any kind of causal relationship between government spending and economic growth (Dakurah, Davies and Sampath, 2001; Muhlis and Hakan, 2003). Wijeweera and Garis (2009) investigated Wagner s Law for Saudi Arabia by employing Engle and Granger cointegration technique. Although results of their study declared positive impact of economic growth on government spending but the income elasticity was found to be less than unity. They concluded that growth in government spending was less than economic growth hence, their results is inconclusive regarding Wagner s Law. RESEARCH METHODOLOGY Data on gross domestic product (GDP), government spending (GS) and population was obtained from World Bank online database (2014) over period 1969-2013. Wijeweera and Garis (2009) analyzed the data from 1969 to 2007 to test Wagner s Law in Saudi Arabia so we also considered 1969 as a starting point for our study and extended till 2013. However, we applied different techniques than them for same purposes. The GDP per capita (GDPPC) and GS per capita (GSPC) was obtained by dividing GDP and GS by corresponding year population respectively. All variables were then converted to natural log so that the estimated coefficients can be interpreted as elasticities. The natural log of GDP and GDPPC is used as a proxy for economic growth. This study tested two versions of Wagner s Law for Saudi Arabia that are shown respectively in equation 1 and 2. log(gs t ) = b 0+ b 1 log (GDP t ) + t (1) log (GSPC t ) = γ 0 + γ 1 log(gdppc) + u t (2) Unit roots problem of time series has to be tested through augmented Dickey-Fuller, ADF, (1979) and Phillips and Perron, PP, (1988). Once order of integration is decided then this study will apply well-known Johansen cointegration technique (Johansen and Juselius, 1990; Johansen, 1991) to capture long run relationship. The long run and short run causality has to be determined through vector error correction model (VECM). In VECM system of equations each variable is treated as endogenous and there are as many numbers of equations as many variables are included. The VECM also includes error correction term (ECT) that shows disequilibrium adjusted per period. This study used three distinct cointegration regressions to obtained long run estimates. First, fully modified ordinary least squares (FMOLS) developed by Phillips and Hansen (1990) is written in equation 3. Licensed under Creative Common Page 356

International Journal of Economics, Commerce and Management, United Kingdom ˆ β fm T = [ (x t x i ) t ] 1 T [ (x t x i )yˆt + T u ] (3) t=1 t=1 Where yˆt is variable transformed to account for endogeneity, x t is explanatory variable and u error correction term for autocorrelation. This method distinguishes between dependent and independent variables as compare to VECM which takes variables as endogenous. Also, this method gives free estimates from simultaneity bias. Second approach we adopted to capture the long run effect of economic growth is dynamic ordinary least squares that can be attributed to Stock and Watson (1993). The leads and lags of first difference of independent variables correct for autocorrelation, simultaneous bias and small sample bias between regressions. The estimation of DOLS is presented in equation 4. k y t = α + βˊx t + Πˊj x t j + ῡ t (4) j= k The lead-lag truncation parameter is presented by k. Third method we applied is canonical cointegrating regressions (CCR) proposed by Park (1992). It generates asymptotically efficient estimators that have normal distributions, so that their standard errors can be interpreted in the usual way. CCR is presented in equation 5. p s y t = c + ɳ i t i + ɳ i t i + γx t + ε t (5) i=1 i=p+1 Equation 5 comprises superfluous time polynomials up to order s in order to test for stochastic and deterministic cointegration. The CCR advantage is that it does not require the assumption of a Gaussian vector autoregressive (VAR) structure. EMPIRICAL RESULTS AND DISCUSSION ADF and PP test were applied to check non-stationarity problem in time series. Results are depicted in Table 1 below. Non-stationarity problem was detected when variables were taken at levels. This problem was removed by taking study variables at first difference. It is proved by these tests results that time series data of the current study is trended and its mean and variance is not constant over time so we cannot apply ordinary least squares. Licensed under Creative Common Page 357

Guo, Alotaish & Ihtisham Table 1. Results of ADF and PP unit root tests ADF PP Variable Level First Difference Level First Difference Conclusion loggdp 2.33-4.15*** 2.54-4.28*** I(1) logpc 0.90-1.91** 2.08-1.96** I(1) loggs 1.62-2.72** 2.67-2.19** I(1) Note: *** and ** shows significance at 1 and 5 % respectively. Once it is known that data is trended so a methodology is needed to take variables at first difference without losing long run relationship between variables. Cointegration technique developed by Johansen (1988) and Johansen and Juselius (1990) is a technique to figure out long run relationship between variables and this technique takes variables at first difference. However, before moving to perform this technique, we have to find out lag length that is very crucial in determining cointegration vectors between variables. There are basically two VECM systems; first one, between GDP and government spending, and second one between GDP per capita (PC) and government spending so we have to examine lag length separately for these VECM systems. The results of selected criterion are given in Table 2 and Table 3. One lag length has been identified by Akaike information criteria (AIC) and Schwarz-Bayesian criterion (SBC). Table 2. Lag length identification for cointegration technique Lag AIC SBC 0 3.234355 3.316272 1-2.306096* -2.060347* 2-2.287582-1.878000 Note: * presents lag length selected by criterion Table 3. Lag length identification for cointegration technique Lag AIC SC 0 2.255463 2.339052 1-4.328073-4.077307 2-4.707114* -4.289170* 3-4.619762-4.034639 Note: * presents lag length selected by criteria After lag length has been specified for cointegration analysis. The Johansen cointegration results for long run relation between GDP growth and government spending has been given in Licensed under Creative Common Page 358

International Journal of Economics, Commerce and Management, United Kingdom Table 4. There exists a single cointegrating vector which is verifying that a long run relation is existed between GDP growth and government spending. Table 4. Cointegration results for GDP and government spending Series: loggdp loggs Hypothesized No. of CE(s) Trace Statistic 0.05 Critical Value Max-Eigen Statistic 0.05 Critical Value None * 17.57961 12.32090 17.18001 11.22480 At most 1 0.399596 4.129906 0.399596 4.129906 Note: * shows cointegrating vector at 5 % critical value Cointegration results for GDP per capita and government spending per capita is documented in Table 5. Again these results confirmed presence of long run relationship between economic growth and government spending as a cointegration vector was witnessed. Table 5. Cointegration results for GDP per capita and government spending Series: logpc loggs Hypothesized No. of CE(s) Trace Statistic 0.05 Critical Value Max-Eigen Statistic 0.05 Critical Value None * 18.26167 12.32090 14.35274 11.22480 At most 1 0.022236 4.129906 0.201274 4.129906 Note: * shows cointegrating vector at 5 % critical value Bivariate causality analysis based on VECM for GDP growth and government spending is provided in Table 6. Short run causality is captured through Wald F-stat. whereas long run is presented by the significance of ECT (t-stat.). It can be seen that long run as well as short run causality is running from GDP growth to government spending and not vice versa. Thus, these results proved that Wagner s Law is prevailing in Kingdom of Saudi Arabia. Similarly, same results are obtained in case of long run when GDP per capita instead of GDP was taken as economic growth of the economy. However, there is bidirectional causality between GDP per capita and government spending in short run. These results are shown in Table 7. Table 6. Causality results for GDP growth and government spending Short Run Results (F-stats) Long run (t-stats) Variable ΔlogGS ΔlogGDP ECT ΔlogGS --- 3.30** -4.14*** ΔlogGDP 1.13 -- -1.68 Note: *** and ** shows significance at 1 and 5 % respectively Licensed under Creative Common Page 359

Guo, Alotaish & Ihtisham Table 7. Causality results for GDP per capita and government spending Short Run Results (F-stats) Long run (t-stats) Variable ΔlogGSPC ΔlogGDPPC ECT ΔlogGSPC --- 6.59*** -3.68*** ΔlogGDPPC 6.01*** -- 0.108 Note: *** shows significance at 1 %. Three cointegration methods are being employed to check the validity of Wagner s Law in Saudi Arabia. These long run results are presented in Table 8. The results of FMOLS, DOLS, and CCR showed that GDP growth has positive significant effect on government spending and its magnitude is above unity, it can be witnessed in Panel (A) of Table 8. Results estimated through cointegration regressions for GDPPC and government spending is presented in Panel (B) of Table 8. In both cases results confirmed existence of Wagner s Law in case of Saudi Arabia as opposed to what Wijeweera and Garis (2009) claimed that their results neither support nor deny Wagner s Law. Table 8. Long run estimates of cointegration regressions Panel A: Dependent variable: loggs FMOLS DOSL CCR Constant -3.064867*** -1.846215*** -3.504024*** loggdp 1.065197*** 1.020876*** 1.082263*** R-squared 0.941928 0.980706 0.942950 Adj. R-squared 0.940545 0.978620 0.941591 Part B: Dependent variable: loggspc FOMLS DOSL CCR constant -1.934641*** -1.568196*** -2.147612*** loggdppc 1.058860*** 1.022349*** 1.082210*** R-squared 0.862570 0.953082 0.864089 Adj. R-squared 0.859298 0.948009 0.860853 Note: 1 % significance is shown by ***. CONCLUSIONS This study was carried out to examine Wagner s Law in case of Saudi Arabia. A time series data covering period from 1969 to 2013 was analyzed for the said purpose. Bivariate cointegration analysis was conducted to see how economic growth effect government spending. Unit root tests identified non-stationarity when variables were taken at levels. These tests verified that order of integration of study variables is one. Cointegration technique was applied to check long run relationship. This technique affirmed presence of long run relation between government spending and economic growth. The long run estimates obtained through three different Licensed under Creative Common Page 360

International Journal of Economics, Commerce and Management, United Kingdom techniques (FMOLS, DOLS, and CCR) proved that Wagner s Law is prevailing in Saudi Arabia. Causality analysis confirmed that a unidirectional causality is running from economic growth to government spending in long run. Like every empirical study this study also is not exceptional to limitations of the study. This study limits itself to two version of Wagner s Law, however; future study can be carried out for all known versions of this law in literature. Similarly, this study analyzed aggregate government spending and did not extend analysis to disaggregate data for government spending like government expenditure on health, education and infrastructure etc. REFERENCES Abizadeh, S. and Yousefi, M. (1998). An Empirical Analysis of South Korea s Economic Development and Public Expenditure Growth. The Journal of Socio-Economics, 27(6): 687-94. Al-Faris, A. F. (2002). Public Expenditure and Economic Growth in the Gulf Cooperation Council Countries. Applied Economics, 34(9): 1187-95 Aregbeyen, O. (2006). Cointegration, Causality and Wagner s Law: A Test for Nigeria, 1970-2003. Central Bank of Nigeria Economic and Financial Review, Volume 44 Number 2. Babatunde, M.A. (2007). A bound testing analysis of Wagner s law in Nigeria: 1970-2006. Proceedings of Africa Metrics Conference. Burney, Nadeem A. (2002). Wagner s Hypothesis: Evidence from Kuwait Using Cointegration tests. Applied Economics, 34, 49-57. Dakurah H., Davies S. And Sampath R. (2001). Defense Spending and Economic Growth in Developing Countries: A Causality Analysis. Journal of Policy Modeling, 23(6), 651-658. Dickey, D., and Fuller, W. (1979). Distribution Of The Estimators For Autoregressive Time Series With A Unit Root. Journal of the American Statistical Association, 74:427 731. Huang, C-J. (2006). Government Expenditures In China And Taiwan: Do They Follow Wagner s Law? Journal of Economic Development, 31(2) Johansen, S. and Juselius, K., (1990).Maximum likelihood estimation and inference on cointegration with applications to the demand for money. Oxford Bulletin of Economics and Statistics, 52(2), 169-210. Johansen, S., (1991). Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models. Econometrica, 59, 1551-1580. Johansen, S., (1995). Likelihood-Based Inference in Cointegrated Vector Autoregressive Models. Oxford: Oxford University Press. Muhlis, B and Hakan C. (2003). Causality between Public Expenditure and Economic Growth: The Turkish Case. Journal of Economic and Social Research 6 (1): 53-72. Musgrave and A.T. Peacock (eds), Classics in the Theory of Public Finance, London: Macmillan, 1958. Omoke P. (2009). Government Expenditure and National Income: A Causality Test for Nigeria. European Journal of Economic and Political Studies, Vol. 2: 1-11 Park, J. Y. (1992). Canonical Cointegrating Regressions. Econometrica, 60 (1), 119-143. Phillips P.C.B. and Perron P. (1988). Testing for a unit root in a time series regression. Biometrika, 75, 335-346 Phillips, P. C. and Hansen, B. E. (1990). Statistical Inference in Instrumental Variables Regression with I(1) Processes. The Review of Economic Studies, 57 (1), 99-125. Licensed under Creative Common Page 361

Guo, Alotaish & Ihtisham Stock, J. H. and Watson, M. W. (1993). A Simple Estimator of Cointegrating Vectors in Higher Order Integrated Systems. Econometrica, 61 (4), 783-820. Wagner, A., (1883) Three Extracts on Public Finance, translated and reprinted in R.A. Wijeweera, A. and Garis, T. (2009). Wagner s law and social welfare: the case of the kingdom of Saudi Arabia. Applied Econometrics and International Development, 9-2: 199-209 Licensed under Creative Common Page 362