Output Volatility and Government Size in Nigeria

Abstract Research background: Output volatility has potentially adverse consequences on the economy and the stabilizing role of fiscal policy is linked to the share of government size in an economy. Hence, given the relative large share of government in developing countries, government size is expected to play an important role in stabilizing output volatility. Purpose: This study examines the relationship between output volatility and government size in Nigeria. The study seeks to establish if government size mitigates output volatility in Nigeria. Research methodology: The study employs the Autoregressive Distributed Lag (ARDL) technique after conducting stationarity and co-integration tests. Results: The results of the ARDL estimate showed that government size lessens output volatility but the magnitude was insignificant. Further, the study found that volatility in aggregate government spending; international oil price and public debt were significant determinants of output volatility in Nigeria. Novelty: This showed that the automatic stabilization role of government size on output volatility could not be established. The automatic stabilization role of fiscal policy can be improved by increasing social security transfers (pension payment), payments of unemployment benefits and increasing civil servants minimum wage.


Introduction
The automatic stabilization role of fiscal policy has remained a contending issue in the literature. From a conceptual perspective, the stabilizing role of fiscal policy is linked to the share of government size in an economy. Government size in any economy has a potentially important role in stabilizing aggregate demand and hence output for two reasons. First, higher government expenditure may be associated with the larger provision of public goods and services as well as a larger fraction of workers employed in the public sector to the extent that government expenditure is more stable than other components of aggregate demand.
Consequently government expenditure is expected to reduce volatility in aggregate personal disposable income, aggregate personal consumption and hence reduce volatility in aggregate output (Mohanty, Zampoli, 2009). Secondly, a higher share of government expenditure may also reflect the provision of social security schemes or transfers such as unemployment benefits and state pensions to a large number of citizens. Such transfer reduces volatility in disposable income and helps consumption smoothening which is important for reducing output volatility.
Furthermore, tax share could contribute to stabilizing output volatility -a higher tax share among other things remaining equal reduces the volatility of household disposable income and firms' cash flow in the face of fluctuation in their gross income (Mohanty, Zampoli, 2009).
On the other hand, higher tax rates reduce the effort to increase the growth rate, hence the growth perspectives decrease.
With respect to empirical literature, there appears to be ambiguity on the relationship between government size and output volatility. A large number of studies seem to suggest a negative relationship between government size and output volatility (see Gali, 1994;Cohen, Follete, 2000;Fatas, Mihov, 2001;Anres, Domenech, Fatas, 2004). These studies concluded that government size acts as an automatic stabilizer in an economy by reducing output volatility. However, P. Van der Noord (2000) and J-T. Guo and S. Harrison (2006) observed a positive relationship between government size and output volatility (that is government size had a destabilising effect on the economy) while M. Mohanty and F. Zampoli (2009) observed no clear evidence on the relationship between government size and output volatility.
The above mentioned empirical literature focused on OECD countries and are crosssectional in nature (Gali, 1994;Fatas, Mihov, 2001;Anres, Domenech, Fatas, 2004;Kim, Lee, 2007) while the only country specific-time-series study was on the US economy (Cohen, Follete, 2000). Literature on this issue appears scanty with respect to Nigeria. Most indigenous or local studies have focused exclusively on the relationship between government spending and economic growth (Agu, Okwo, Ugwunta, Idike, 2015;Ogbole, Amadi, Essi, 2011;Nurudeen, Usman, 2010;Oyinlola, 1993). This is particularly worrisome because: One, the share of both government size (government expenditure to real GDP) and aggregate government expenditure have been on the increase as observed in figures 1 and 2. Two, the Nigerian government (public sector) is a key employer of labour and largely responsible for the provision of certain goods and services (such as road construction, medical and education, fire, water, waste management and security services among others). The government is also involved in the payment of pensions to retired civil servants. Hence, given the important role of government, it is therefore pertinent to examine if government size has helped in stabilizing the Nigerian economy by reducing volatility in aggregate output. 1963 1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 1961 1964 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997  Drawing from the above, this study seeks to address the following research questions.
(i) Is government size a stabilizer of output volatility in Nigeria? (ii) Is government size a key determinant of output volatility in Nigeria? The objective of this study is to examine the relationship between government size and output volatility in Nigeria over the period of 1961 to 2017. The rest of the paper is as follows: section two presents a review of literature on government size and output volatility, section three discusses the research methods while section four discusses the data analysis and results. Section five presents the summary, conclusion and policy recommendation.

Literature Review
With respect to theoretical literature, classical economists emphasised the absence of government intervention in the belief that government intervention causes instability in the economy and advised that a government should maintain a balanced budget. In contrast, the Keynesian economists emphasised the need for government intervention particularly during recession in order to restore the economy on the path of growth by accelerating aggregate demand.
The Ricardian equivalence proposition holds that the impact of fiscal policy (either through tax The study employed the panel ordinary regression technique and the result of the study showed a modest negative relationship between government size and output volatility. Furthermore, the study observed that variables such as volatility of inflation, exposure to terms of trade shocks and trade openness were important determinants of output volatility. K. Erkki and V. Matti (2004) also examined the relationship between government size and output volatility for the period 1980 to 1999. The study covered a group of ninety-one (91) countries and employed the panel regression technique. The result of the study showed that the negative relationship between government size and output volatility among OECD countries could not be generalized to a wider international data set. Also, the result showed a non-linear relationship between government size and output volatility. The negative impact of government size on output volatility was significantly negative only for countries with a high and low public sector.
J. Gali (1994) examined the relationship between government size and output volatility for a group of twenty-two (22)

Theoretical Framework
The Keynesian theory assumes that the economy is inherently unstable and that output volatility involves significant economic costs. Traditionally, the automatic stabilizing policy reduces the gap between potential output and actual output, reducing the gap implies a reduction in output volatility. Thus, government involvement in the economy would enhance stability due to the proportional link between the magnitude of the impact of an automatic stabilizer and the size of government expenditure (Blinder, Solow, 1974). The central idea of Keynes is that by lessening the impact of liquidity constraints faced by households, government size dampens the impact of exogenous shock to aggregate income, aggregate current consumption and hence output volatility (Xavier, Jean, Andre, 2008).

Model Specification
Based on the theoretical link between the automatic stabilizing role of government size and output volatility, the study specifies the model below.
where Y t is output volatility and GS t is government size. Introducing control variables that is important to the link between output volatility and government size. These variables are classified into two. The first set of control variables are trade openness denoted by OPX; GDP share of Agricultural output denoted by AGR and crude oil price denoted by OIL. These variables have been found to be positively associated with output volatility and government size. Omitting these variables may result in biased estimates (Erkki, Matti, 2004;Persson, Tabellini, 2001;Rodrik, 1998).  (1) gives: Linearing equation (2) Output volatility (Y) is measure the volatility series of real gross domestic product. Government size (GS) is measured by the ratio of total government expenditure to real GDP, VGP is volatility in government spending, openness (OPX) is measured by the ratio of import plus export to real GDP, AGR is measured by share of agricultural output to real

GDP, oil price (OIL) is measured by international crude oil (bonny light) price, debt (DBT)
is measured by total public debt to real GDP and CPI is measured by average consumer price index. The data are sourced from the Central Bank of Nigeria statistical bulletin, 2017 edition.
Volatility series (such as output (real GDP) and government expenditure (VGP)) are computed using the E-GARCH approach.

E-GARCH Model for Volatility Series
The volatility series for this study are generated using the Exponential Generalize Autoregressive Conditional Heteroeskedaticity (EGARCH) [1,1]. The E-GARCH model is employed due to its capturing of asymmetric effects and its non-imposition of non-negative constrain on the parameters (Jamil, Streissler, Kunst, 2012;Kontonikas, 2004). The E-GARCH process is expressed as: Equation (4) describes the conditional variance. The estimates of the conditional variance for the series (that is, output (Y) and government expenditure) are used in equations (3) (Nwosa, Omolade, 2017;Demachi, 2012).

ADF and PP Stationarity Test Model Specification
The ADF test tests the null hypothesis that a time series y t is I(1) against the alternative that it is I(0), assuming that the dynamics in the data have an Autoregressive Moving Average (ARMA) structure. The ADF test is based on estimating the test regression where D t is a vector of the deterministic terms (constant, trend etc.). The p lagged difference terms, t j y − ∆ , are used to approximate the ARMA structure of the errors, and the value of p is set so that the error t ε is serially uncorrelated and also assumed to be homoskedastic (Dickey, Fuller, 1981 where u t is I(0) and may be heteroskedastic. The PP tests correct for any serial correlation and heteroskedasticity in the errors u t of the test regress by directly modifying the test statistics 0 t π = and ˆ. Tπ these modified statistics denoted by t Z and , Z π are given by ( ) Under the null hypothesis that 0 π = , the PP t Z and Z π statistics have the same asymptotic distributions as the ADF t-statistic and normalized bias statistics. One advantage of the PP tests over the ADF tests is that the PP tests are robust to general forms of heteroskedasticity in the error term u t . Another advantage is that the user does not have to specify a lag length for the test regression (Phillips, Perron, 1988).
Sequel to the mix in the unit root results, the co-integration test and the regression estimate were carried out using the Auto-Regressive Distributed Lag (ARDL) Bound Co-integration technique. Unlike the Johansen co-integration technique which requires that the variables must be integrated of order one, the ARDL co-integration approach is applicable irrespective of whether the variables in the estimating model are purely I(0), purely I(1)) or mutually integrated (Oteng-Abayie, Frimpong, 2006). The result of the co-integration estimate showed that the value of the F-statistics (5.76) is higher than the value of the upper bound at all levels of the critical value, suggesting the presence of co-integration among the variables in equation (3). From the ARDL regression estimate reported in Table 3 (see full estimate in the appendix), it was observed that volatility in government expenditure (VGP) had a negative but significant impact on output volatility while international oil price (OIL) and debt (DBT) had a positive and significant impact on output volatility in Nigeria. All other variables including government size had an insignificant impact on output volatility. With respect to objective two, the regression estimate showed that volatility in government spending and international oil price were the significant determinants of output volatility in Nigeria. In the case of oil price, the Nigerian economy is oil driven and movements in international oil price shapes the behaviour of aggregate output. Periods of high oil prices have usually been synonymous with periods of impressive real GDP growth and vice-versa. The positive link between international oil price and output volatility further implies that international oil price contributes to output volatility in Nigeria.
In contrast to a priori expectation, volatility in government spending reducing output volatility in Nigeria. The evidence could reflect discretionary or counter-cyclical fiscal policy whereby the government increases spending during a recession and cut-back spending during periods of impressive economic growth. The importance of the volatility in government spending and international oil price as significant determinants of output volatility is supported by the regression estimate which focused only on the significant variables in Table 3 (see Table A2 in appendix).
The error correction term (ECM-term) from the short run ARDL estimate is expected to be negatively signed and statistically significant. From the estimate, the coefficient of the error correction term was correctly-negatively signed (-0.496) and is statistically significant. The coefficient estimate of the error correction term of -0.496 implied that the model corrects its short-run disequilibrium by about 0.50 per cent speed of adjustment in order to return to the long-run equilibrium.
To evaluate the robustness of the regression estimate, some diagnostic tests were carried out and the results are presented in Table 4. From the table, both the serial correlation LM and heteroskedaticity ARCH estimates showed the absence of serial correlation in the ARDL estimates. This is because the null hypothesis of both tests were accepted as their probability values were greater than 0.05. The results are also supported by the estimate of the Durbin-Watson of 2.03, indications the absence of serial autocorrelation in the regression estimate.     Source: authors' computation using E-views 9, 2019.