Fiscal Policy Shocks and Private Consumption in Nigeria: Blanchard-Perotti (2002) Approach

This paper examines the effects of fiscal policy shocks on private consumption in Nigeria. Albeit, there is a considerable number of works examining the effects of fiscal policy shocks on private consumption globally but in Nigeria, no study has used the structural VAR approach by Blanchard and Perotti (2002) as used in this paper. This approach relies on institutional information about the tax and transfer systems and the timing of tax collection to identify the automatic response of taxes and spending to private consumption as well as to infer fiscal shocks. The key result of this paper is that positive government spending shocks in Nigeria have an instantaneous negative effect on private consumption. The effect becomes significant in the period following the shock. Also, positive tax shocks have a negative effect on private consumption in the period of a shock and the effect becomes statistically insignificant afterwards. On this premises, one-off changes in government spending and taxes in Nigeria are long-lived and short-lived respectively. Thus, the government expenditure changes can be used to support private consumption in the long-run while that of taxes can only be used to support private consumption for a short period.


Introduction
The major challenge in most studies on fiscal policy is the inability to correctly identify changes in current policy variables that are attributable to actual policies rather than to endogenous responses to economic conditions. Possible delay in legislation, the lags in actual implementation of the policies and the time to recognize that there is actually a need for stabilization in the first place are also crucial problems encountered in empirical analysis of fiscal policy. In addressing this issue, studies have examined the responses of government spending shocks and tax revenue shocks on private consumers' behaviour in Nigeria (Onodje, 2009;Sousa, 2009;Favero, Giavazzi & Francesco, 2007;Orisadare, 2012). Previous studies that have attempted to address the above issue only used varieties of Vector Autoregressive (VAR) models such as Vector Error Correction Mechanism (VECM), Narrative, Recursive and Sign Restriction Structural VAR (SVAR). In addition, these models only accounted for either dynamic interactions among the variables or unanticipated shocks associated with ordering and sign restriction behaviours. However, these methods neglected the institutional information about the fiscal policy variables (most especially the tax system) and how this information affects macroeconomic variables (private consumption) upon implementation (Ramey, 2011).
Furthermore, studies that have used Blanchard and Perotti (2002) SVAR approach in developed countries noted that the inability to provide full institutional information on the structure of the tax system and government spending was one of the major reasons for slow recovery from the 2001 recession in the U.S. (House and Shapiro, 2006;Yang & Shu-Chun, 2005). Also, the responses of unanticipated fiscal policy shock from private consumption using Blanchard and Perotti (2002) SVAR approach has not been fully explored among the previous studies most especially in developing countries (House and Shapiro, 2006;Yang & Shu-Chun, 2005;and Den-Haan, & Georg, 2006). In addition, the relationship between fiscal shocks and private consumption in Nigeria using Blanchard and Perrotti (2002) approach of SVAR has not been explored in the literature. Meanwhile, this SVAR approach on like other approaches relies on institutional information about the tax and transfer systems as well as the timing of tax collection to identify the automatic response of taxes and spending to private consumption as well as to infer fiscal shocks (Blanchard & Perrotti, 2002). Whereas, this approach will enhance the Nigerian policy makers on the timing of tax collection and also provide significant and in fiscal stress, spending and tax multipliers became higher and more persistent more in the EMU. Shaheen and Turner (2009) characterized the dynamic effect of fiscal policy shock in Pakistan for the period 1973:1-2008:4 by employing a five variable SVAR model and using Blanchard and Perotti (2002) and the Recursive approaches to identify exogenous fiscal shocks. They found that government expenditure increases output in the short run and decreases it in the medium term. Interest rate and inflation increases following government spending shock. Revenue shocks also increase output, inflation and interest rate. While output decreases in the long run as in the case of government spending shock, interest rate increases at a higher level in the medium and long term. They concluded that in Pakistan, increases in government spending and tax can be used to expand output in the short run at the expense of higher output and inflation and SVAR models are rarely used to study fiscal policy in Africa. Abderrahim et al (2010) studied macroeconomic effects of fiscal shocks in a five variables SVAR mode for Algeria. They found that positive structural shock in government spending has positive effect on output in the short term with very small multiplier. This turns into negative effect in the medium term and in the long run through crowding-out of private investment thereby increasing the average interest rate in the economy. Prices increase persistently following this shock. Public revenue shocks, on the other hand, result in positive impact on government spending in the short term through which the effect on output is channeled i.e., output responds in similar pattern as in the case of government spending shock. The effect on prices and interest rate is persistent and negative in the medium and long term. They used, however, annual data that makes identification of truly exogenous shock rather less reliable.
Mancellari (2011) studied the dynamic effects of change in spending and net revenue on macroeconomic variables in Albania. He used SVAR model and the Blanchard and Perotti identification approach. He estimated multipliers from the model. Accordingly, he found that reduction in tax has the highest impact on output reaching a multiplier of 1.65 after five quarters. Between the two spending components, capital spending has higher role in stimulating the economy than current spending. While interest rate is not responsive for change in any of the spending components, it decreases following a tax cut. Prices; slightly increase following positive shock in current spending, do not respond for a change in capital spending and fall with respect to a cut in taxes. Orisadare (2012) examined the effect of fiscal policy shocks on economic activities in Nigeria from 1970 to 2009 using Recursive SVAR approach. The study found that fiscal policy shock indicated a positive and relatively low significant relationship between government revenue shock and economic activities while a negative significant relationship between government spending shocks and economic activities in Nigeria. Nazir et al (2013) investigated the long and short run effect of fiscal policy on GDP growth of Pakistan. Government consumption expenditure and per capita real revenues are used as fiscal variables while discount rate, trade openness and gross fixed capital formation are treated as control variables to analyze their impact on GDP growth for the economy of Pakistan for the period of 1980-2012. Johansen cointegration and vector error correction model are applied to know the effects of fiscal variables on GDP growth in short and long run. The findings show that fiscal policy has vital role for meaningful economic progress. This study pointed out that government consumption expenditure has negative relation with GDP growth and public revenues have a progressive impact on economic activity of Pakistan. Less consumption expenditure and effective revenue structure is advocated to boost the economic growth of Pakistan.
In summary, only few of the studies reviewed that used Structural VAR technique propounded by Blanchard and Perotti (2002) are either from developed nations or developing nations excluding Nigeria. In Nigeria, this study therefore fills this gap by using Blanchard and Perotti (2002) approach of SVAR to examining the relationship between fiscal policy shocks and private consumption.

Methodology
In order to estimate the empirical relationship between fiscal policy shock and private consumption in Nigeria the paper employed Blanchard and Perotti (2002)  Since this paper used quarterly data, as a result, a lag length of four is supposed to be chosen as suggested by various lag length identifiers ranging from the Akaike information criterion (AIC) to Hannan-Quinn information criterion.

Identification of Fiscal Policy Shocks
The empirical literature has classified the identification of a structural VAR into four approaches. These approaches include: first, the recursive approach introduced by Sims (1980) and applied to study the effects of fiscal shocks by Fatas and Mihov (2001); second, the structural VAR approach proposed by Blanchard and Perotti (2002) and extended in Perotti (2005Perotti ( , 2008; third, the sign-restrictions approach developed by Uhlig (2005) and applied to fiscal policy analysis by Mountford and Uhlig (2005); and, fourth, the event-study approach introduced by Ramey and Shapiro (1998) to study the defence spending of large unexpected increases in government defence spending and also used by Edelberg et al. (1999), Eichenbaum and Fisher (2005), Perotti (2008) and Ramey (2007). This paper used one identification approach i.e. the structural VAR approach proposed by Blanchard and Perotti (2002). The main reason for chosen this approach is that this method relies on institutional information about the fiscal policy variables and the impulse response results are more realistic than other approaches (Cayen & Desgagnes, 2009).
Furthermore, the sign restrictions on impulse responses by Canova and Pappa (2002) as well as Mountford and Uhlig (2002) fail to pin down when the shock occurs and its identification conditions might be too strong. For instance, by identifying revenue shocks through the condition that tax revenues and private consumption do not co-vary positively in response to the shock. This sign restriction approach rules out by the assumption a whole set of "non-Keynesian" output responses to fiscal shocks. Finally, the approach represented by Fatas and Mihov (2001) and Favero (2002) essentially relies on Choleski ordering to identify fiscal shocks. Thus, ordering the fiscal policy variables is equivalent to assuming that all automatic elasticities of fiscal variables to macroeconomic variables are equal to zero.

The Blanchard-Perotti Approach
The identification approach introduced by Blanchard and Perotti (2002) relies on institutional information about tax and transfer systems and about the timing of tax collections in order to identify the automatic response of taxes and government spending to economic activity. This identification scheme relies on a two-step procedure: first, the institutional information is used to estimate cyclically adjusted taxes and government expenditures. In the second step, estimates of fiscal policy shock are obtained. Blanchard and Perotti (2002) and Perotti (2005) applied this approach to estimate the effects of government spending and tax shocks for the United States. This study follows the identification scheme used by Perotti (2005) as he also employs a five-variable VAR model while Blanchard and Perotti's (2002) analysis built on a threevariable system. In addition to the adaptation of the identification scheme used by Perotti (2005), this study follows the four steps approach of Giordano et al (2006) to identify the fiscal shocks since the matrices A and B of the system equation are not identified without constraint. Meanwhile, this study used private consumption in lieu of economic activity that other studies have used including Blanchard and Perotti (2002) and Perotti (2005). Although, economic activity is large in volume than private consumption but it is a superset of private consumption. In addition, private consumption has a larger proportion of economic activity of almost 70% (Mankiw, 1998). Also, since both economic activity and private consumption response to the same shocks from government spending, thus, private consumption is used in lieu of economic activity in this study.
In the first step, the reduced-form VAR is estimated from the reduced-form residuals Perotti, (2005).The reduced form residuals of interest rates; (ii) the systematic discretionary response of policymakers to private consumption, debt and interest rate innovations e.g. increase in government spending implemented systematically to fight insecurity in Nigeria; and (iii) random discretionary shocks to fiscal policies; these are the "structural" fiscal shocks, which unlike the reduced form residual are uncorrelated with all other structural shocks. The study therefore express the reduced form residuals of government spending,  (2) and (3)  The main interest of this paper is the identification of structural shocks exp g t e and tr t e , and the response of private consumption to these shocks. To identify these two structural shocks, there is need to impose further restrictions on the system above. To this ends, the paper used the observation made by Blanchard and Perotti (2002) that noted that it took policymakers and legislatures more than a quarter to reach to a macroeconomic shock (private consumption). This virtually eliminates the possibility of discretionary adjustment of fiscal policy in response to unexpected movement in private consumption.
Consequently, since this paper used quarterly data, the coefficient of However, with regards to the impact of government spending on private consumption, the paper assumed that the response of the spending residual to the private consumption structural residual is zero, 6 that is 0 exp  g pc  following the work of Blanchard and Perotti (2002) that observed that it is difficult to identify any automatic feedback mechanism from quarterly macroeconomic variable i.e. private consumption to quarterly government spending all within the same quarter due to decision making and implementation lags. The use of exogenous contemporaneous elasticities allows this paper to compute cyclically adjusted reduced-form residuals for government spending CA g t u exp, , and for taxes CA tr t u , . This represents the second step of this paper identification procedure: Equation (4) and (5)  Meanwhile, Perotti (2005) argued that neither of the alternatives has any theoretical or empirical basis. It is more plausible to assume that decisions on government expenditure dictate decisions on tax revenues, so the paper assumed that the 0 exp  tr g  . Although the above assumption is made in Perotti (2002), Heppke-Falk et al, (2006), Caldara and Kamps (2008), de Castro and de Cos (2008) as well as in Lozano and Rodriquez (2008), hence, this assumption would be tested. Thus, under the initial assumption that the 0 exp  g tr  , the cyclically adjusted reduced shocks are: Finally, to estimate the remaining coefficients in the equation for the reduced form residual of private consumption, the two structural shocks estimated are orthogonal to the other structural shocks of the economy and are used as instrumental variables. For instance, if private consumption is ordered first among the other variables, we can estimate the private consumption equation below: Hence, other equations of reduced form innovations are estimated using instrumental variables, where t e is used as instrument, since it is orthogonal (Heppke-Falk et al, 2006). Also, this procedure is also found in Perotti (2005), and other equations of reduced form innovations are: With the above mentioned four steps, the paper obtained all the coefficients needed to construct the estimates of A and B matrices, which can be used to compute the impulse responses to fiscal shocks: To make the system just-identified, this paper used 2 ) 1 ( 2 2   k k k , i.e. 35 constraints (k is the number of endogenous variables) are imposed in total in both matrices. Matrix B has 18 coefficients that are equal to zero, and the main diagonal of matrix A provides another 5 restrictions. All the coefficients associated with the equation of reduced innovation in government spending are set to zero, except for the impact of government debt on government spending, which is assumed to be -0.5, because the decision on government expenditure and government debt is taken by the central government and Central Bank Authority who negotiates the borrowing on behalf of the government. This is done to achieve check and balance of federation account. This relationship of government debt and government spending is taken from Caldara and Kamps (2008) and Lozano and Rodriquez (2008). All other coefficients associated with reduced innovation in government spending are zero, because government spending is entirely under the control of economic policy, which cannot react in the same period and the effect is not automatic because it is a variable whose dynamics is solely influenced by government decisions. These arguments give 4 additional restrictions. Furthermore, the assumption that the overnight interest rates-reduced innovation does not affect any one of the remaining four reduced innovations provides 3 more restrictions. As the impact of government expenditure on tax revenues can be modeled in a matrix B with structural innovations, the relationship in matrix A is assumed to be zero. Also, it is assumed that the reduced form innovation of government debt is not affected during the same period by the reduced form of private consumption, which gives 2 more restrictions. The remaining two restrictions necessary for the system to be just-identified are obtained by calculating the impact of reduced innovation of private consumption on reduced innovations of tax (exogenous elasticity 0.95) and the impact of reduced innovation of debt on reduced innovation of taxes (exogenous elasticity 0.89), exogenous elasticity methodology as presented below in line with the work of Ravnik and Zilic (2011). Also such exact SVAR identification was used by Lozano and Rodriquez (2008), Štikova (2006), and Caldara and Kamps (2008).
Structural shocks exp g t e and tr t e represent a one-time increase in government spending or tax revenue by one standard deviation compared to the average of the period. Although, Perotti (2005) has argued that due to the nature of the budget process there is only one fiscal shock per year but in practice the fiscal authorities with several revisions and changes in tax policy frequently change the course of fiscal policy. Since the fiscal shock cannot be viewed in the context of the variables in the model (except for the fiscal variables, private consumption, government debt and interest rate), the shock of fiscal variables cannot be interpreted through the initial movement of these variables and however caused by (1) imperfect information about the current state of the economy by the fiscal authorities, (2) changes in the relative weights placed on various budget spending by the fiscal authorities. The first cause is explicitly assumed in the model identification, while the second one is explained by the fact that the process of making decisions about government spending is largely influenced by the struggle of various interest or social groups for greater government spending, so the weights the state puts on various forms of spending are constantly changing.

Formulation of Exogenous Elasticities
In order to calculate the required exogenous elasticities, the paper used the method developed in Blanchard and Perotti (2002) and followed by Perotti (2002), Caldara and Kamps (2008) as well as in de Castro and de Cos (2006), Lozano and Rodriquez (2008), and Ravnik and Zilic (2011). The elasticity of government revenue to private consumption is consist of the elasticity of each tax category to their defined tax base, and the elasticity of each defined tax base to private consumption. The elasticities are calculated using quarterly data from CBN Statistical bulletin, 2014 and World Bank data base (WDI, 2014); GDP which is available only on a quarterly basis was used as a base for some categories of tax revenues (Oil Revenue, Non-Oil Revenue and Excise duty).
The private consumption elasticity of government revenues was calculated using the following formula:  is the elasticity of each tax category to its tax base, i tb pc  is the elasticity of each tax base to GDP, while W W i is the weight of type i tax in the sum of taxes, In Nigeria, there are six largest central government budgetary revenue but five were used in this paper because the value added tax begun in 1992 as against 1981 of the starting period. Thus the five categories of government revenues used were: oil revenue, non-oil revenue, income taxes, social contributions (such as grants, stabilization receipt and others) and excises. The base of income taxes and social contributions was aggregate wages; the base of oil revenue and non-oil revenue was gross operating surplus, while the base of indirect taxes was GDP. As noted above, is the sum of five revenues, so W W i is a simple weight of each revenue, which was multiplied with the private consumption elasticity to each base and the base elasticity to the corresponding revenue in order to obtain a single elasticity; thus, this result is presented in table 1 below. Similarly, the equation (15) was used to calculate the required elasticity of government revenue to government debt using the same categories of tax revenue and tax base as shown below:  is the elasticity of each tax category to its tax base, i tb b  is the elasticity of each tax base to GDP.
The result of the elasticity generated by the equation (15) is presented in table 2 below. 9 However, the impact of reduced innovation of private consumption on reduced innovations of tax revenue is 1.35 (that is, exogenous elasticity 1.35) and impact of reduced innovation of government debt on reduced innovations of taxes is 0.95 (that is, exogenous elasticity 0.95) using exogenous elasticity in line with the works of Caldara and Kamps (2008) and Ravnik and Zilic (2011).
Therefore, by incorporating the values of the calculated elasticities and all the assumptions stated earlier into the SVAR matrices A and B described above, the estimated matrices A and B becomes: Thus, the estimated elements in matrix A are nine (9) while the estimated elements in matrix B are seven (7). In total, there are sixteen elements to be estimated from matrices A and B.

Estimation of Lag Length of the SVAR Model
To avoid spurious results in the SVAR model is by adding a sufficient number of autoregressive lags. Therefore, the process of determining the number of lagged values needed to be included in the SVAR model is an integral part of specifying a stable SVAR model. However, incorrect specification of the lag length of a SVAR model can lead to inconsistent impulse responses and variance decompositions (Braun & Mittink, 1993). Also, over-fitting the model may result to inefficiency while under-fitting may cause some dynamics in the system to be unrealized. Several methods and tests were used to identify the true lag length of the unrestricted VAR model with constant and trend.
Therefore, the paper used various lag length identifiers such as Akaike information criterion (AIC), Schwarz's information criterion (SIC) and so on from VAR system to identify the appropriate lag length. The results of the lag length selection are shown in the table 3 with orders of m = 1,..., 8 with constant and linear deterministic terms. The information criterion presented in table 3 offers mixed results. Based on LR, FPE, AIS, HQ indicating the choice of 6-lag, whereas the SIC suggests an order of 2-lag length. Thus, this result showed that four out of five criteria indicated an optimal lag order 6, while only SIC supported an order of 2. To test the validity of the result, a lag exclusion Wald test was conducted to verify whether six lags was suitable for the unrestricted VAR model or not. The results of the 2 X statistic in table 4 above for the joint significance of all endogenous variables in the VAR at two lag lengths was jointly significant at the 5% level of significance, indicating that 2-lag lengths is the optimal. Thus this paper used 2-lag lengths for the analysis of the fiscal policy shocks.

Discussion of the Results of the Effect of Fiscal Policy Shocks on Private Consumption
Using the Structural Vector Autoregressive (SVAR) approach proposed by Blanchard and Perotti (2002) and Perotti (2004) to examine the impact of fiscal policy shocks on private consumption, the paper arranged the variables in the following order-government expenditure, tax revenue, private consumption government debt and interest rate. Thus, this ordering is based on the following assumptions and justifications: (a) government spending was placed first because it does not react contemporaneously to shocks to other variables in the system and was not affected by business cycle fluctuation; (b) tax revenue was ordered second, which implied that it does not react contemporaneously to private consumption, tax revenue and interest rate shocks but was affected by government spending shocks; (c) private consumption was ordered third meaning that it was contemporaneously affected by government spending, government debt and tax revenue shocks; (d) Government debt was ordered as fourth, meaning that it was not affected contemporaneously by private consumption and interest rate shocks but it reacted to government spending shocks; (e) interest rate was placed last, because it was affected by all shocks from the system since interest are not payable on fiscal variables and therefore not sensitive to interest rate changes. This could be taken as the justification for the placement of the interest rate among the variables. This method was also used by Caldara and Kamps (2008), Ravnik and Žilić (2011) who investigated fiscal policy shocks for Croatia. The basic point in this approach was that identification of fiscal policy shocks was achieved by exploiting decision lags in policy making and information about the elasticity of fiscal variables to private consumption.

Empirical Analysis of Fiscal Policy Shocks -Blanchard and Perotti (2002) Approach
The SVAR results of fiscal policy shocks on private consumption in line with Blanchard and Perotti (2002) approach was presented in table 5.

Source: Author, 2015
The estimated result showed that the estimated coefficient of government debt shock to private consumption was positively signed but statistically non-significant. It suggested that a positive one percent shock in government debt increased the private consumption by 38 percent. This finding supported the Keynesian proposition that says an expansionary fiscal policy through government debt would enhance income and aggregate demand and thus foster private consumption. This was in conformity with the study conducted by Ravnik and Žilić (2011) on Croatia's economy.
The coefficient of government spending shocks was negatively signed and statistically significance at 1 percent level. Thus, a shock in government spending hit private consumption negatively in Nigeria. This showed that a negative one percent shock in government expenditure decreased private consumption by 36 percent. Hence, the result indicated that government spending shock crowded-in private consumption in Nigeria. This finding was in conformity with non-Keynesian proposition which states that an expansionary fiscal policy in terms of increase in government expenditure would result to a contractionary effect on private consumption. This result supports the view of Giavazzi and Pagano (1996), Giavazzi, Jappelli and Pagano (2000) and Onodje, (2009).
The negative coefficient of pc,tr showed that a negative shock in tax revenue contributed to a change in private consumption and statistically significant. This showed that a unit negative shock in government revenue would reduce private consumption by 97 percent and hence, crowded-in private consumption. This might arise as a result of increase in tax revenue through increase in tax rate by the government in order to finance her excess spending or to pay back the debt incurred through increase in government expenditure on infrastructural facilities; such as provision of tidy road, hospitals and building of more schools as well as increase in transfer payment like bursary, pension and so on.
A positive and non-significant value of gexp,B indicated that increase in government debt would lead to increase in government expenditure. The negative and statistically significant value of pc,ir showed that decrease in interest rates would increase aggregate demand and thus, private consumption. A negative one percent shock in interest rates thus increased private consumption by 0.73 percent. This was in conformity with the result of Perotti (2008).
A negative value of B,ir suggested an inverse relationship between lending rate and government debt. This relationship was statistically significant and showed that a positive shock in interest rate contributed to decrease in lending rate and this would foster aggregate demand and private consumption in the country. Hence, as interest rates charges on debt reduce, government intends to borrow more at highly reduced interest rates in order to enhance aggregate demand and private consumption. A positive value of tr,B implies that an increase in government debt would result in an increase in government revenue and this estimate was theoretically consistent. The estimated coefficient of gexp,tr suggested a direct relationship between tax revenue shocks and government expenditure. Thus, since tax revenue was an income to the government, it would enhance both private and public consumption. This was in conformity with the result of Alfred et al (2013).

Variance Decomposition of Fiscal Policy Shocks
The decomposition of forecast variance was used to examine how much the fitted SVAR deviates from the actual values of the vector of endogenous variables. What percentage of a variable's deviation from its forecasted value was attributable to another variable provided additional insight into historical relationships. Evidence for contemporaneous correlation exists when one variable begins to explain the forecast variance in the other with a time lag. This occurs because the correlation takes time to work through the lags in the system. Table 6 to 9 reported the results of the forecast error variance decomposition for the SVAR model of the relationship between fiscal policy variables shocks (government expenditure, tax revenue and government debt) and private consumption in Nigeria. When interpreting the results, the ordering of the variables was important because the decomposition assumed that all the variances in the initial period were entirely due to the first variable in the ordering. As the forecast horizon expanded, the other variables in the system began to exert their influence. In Table 6, for example, if interest rates variable was first in the ordering, this means that it explained all of its forecast variance in the initial period. It was possible that results might be sensitive to potentially arbitrary variable orderings. Without strong theoretical guidance, a common recommendation was to switch the variable orderings in order to check for robustness. The first column in Table 6 listed the steps in the forecast with each step corresponding to one quarter. Thus, the first step represented the first quarter of the forecast while the tenth step represented the tenth lag. The total forecast horizon covered one hundred and thirty-two quarters (thirty-three years). The next four columns in the table reported the percentage of forecast variance in the government debt explained by government debt, tax revenue, government expenditure and private consumption, respectively. The response of government debt to shocks in tax revenue was zero in the first period and negative in the second period. These shocks continued to increase till the fifth period and decrease in the sixth period. The shocks continued to be volatile from the sixth periods to the tenth period. The response of government debt to shocks in government expenditure also recorded zero in the first period and continued to be volatile till the ninth years and negative in the tenth year. The negative shock that emanated from government expenditure to government debt indicated that discretionary government policy in terms of change in government expenditure reacted inversely to change in government debt. Furthermore, the response of government debt to shocks in private consumption was negative for all the periods except the ninth and tenth period. Thus, this indicated that changed in government debt in Nigeria responded to negative shocks from private consumption. Hence, increase in government debt deterred private consumption in Nigeria.

13
The response of tax revenue to shocks in interest rates, government debt, government expenditure and private consumption was presented in the table 7. The response of tax revenue to shock in interest rate showed that 87 percent changed in interest rate responded to a change in tax revenue and 10.6 percent change in government debt also responded to change in tax revenue in the first period. In the third period, tax revenue responded to negative shocks from all the variables in the system. The negative shocks are 10.2%, 17.6% and 5.8% from government debt, government expenditure and private consumption respectively. In addition, the response of tax revenue to shocks in both government expenditure and private consumption decreased as the periods increased until the ninth period and tenth period for government expenditure and private consumption respectively. Thus, this response indicated that the effect of fiscal policy shock on private consumption has non-Keynesian effect; this was in conformity with the work of Giavazzi and Pagano, (1996) and Giavazzi, Jappelli and Pagano, (2000). The response of government expenditure to shocks in government debt, tax revenue and private consumption as presented in table 8 showed that government expenditure responded to positive shocks in all the variables in the system but in different categories. The shock in government debt accounted for a maximum of 12.6% response to government expenditure. This showed that a unit percent shock in government debt would response to 12.6% changed in government expenditure. Shock in tax revenue accounted for approximately 30% changed in government expenditure. This showed that the Nigerian government financed larger part of her expenditure with the tax revenue generated either from oil or non-oil export. The shock in private consumption accounted for approximately 6% changed in government expenditure. The response of private consumption to shocks in government debt, tax revenue and government expenditure was presented in table 9. It showed that in the second period, the private consumption responded to negative shocks from all the variables in the system. Subsequently, government debt shock maintained a positive shocks for the rest of the periods likewise the tax revenue shock. This indicated that shocks from both government debt and tax revenue enhanced the change in private consumption positively. This was in conformity with the Keynesian effect of fiscal policy as indicated in the works of Schclarek, (2003); Omitogun and Ayinla, (2007) and; Medee and Nembee, (2011). However, private consumption responded to negative shock from government expenditure in the fourth period. This shock was more than 50%, indicating that changed in government expenditure would result in decrease in private consumption. Hence, this was in conformity with the non-Keynesian effect of fiscal policy on private consumption (Giavazzi & Pagano, 1996). Finally, at an average, private consumption responded to positive shocks in government debt, tax revenue and government spending. This showed a mixed results between Keynesian and non-Keynesian effect on private consumption in Nigeria.

Impulse Response Function of Structural VAR Result
The impulse response function (IRF) enables one to analyze the response of one variable to a random shock in another variable while maintaining the original units of the data as well as providing an estimate of uncertainty. The results presented here were based on a Structural decomposition of the estimated residual covariance matrix of the estimated SVAR. Substantively, the IRF is useful because it provides a more statistically principle means of measuring a variable's response to changes in another variable. In the present context, the IRF helped in determining how quickly private consumption (fiscal policy variables) adjusted after being shocked by an unanticipated change in fiscal policy variables (private consumption). Such test provided support for substantive hypothesis test with respect to variable dynamics over time. If, for example, theory suggests that a change in fiscal policy variables should enhance private consumption over time, this expectation can be tested using the IRF. It is important to remember, however, that the results presented here were purely exploratory and were intended to assist with theoretical development by giving an account of the dynamic behaviour of fiscal policy variables and private consumption.
Theory and evidence regarding the way an increase in government debts, government spending or a tax shock affect private consumption were not conclusive. In particular, neoclassical models predicted a negative response of this variable (Baxter & King, 1993) while the opposite was found in Keynesian and neo-Keynesian models. On empirical grounds, Fatas and Mihov (2001), Blanchard and Perotti (2002) and Gali, Valles and Lopez-Salido (2007) found that the reaction of private consumption to an unexpected government spending shock was positive and persistent. On the contrary, Mountford and Uhlig (2009) found that the response of private consumption was statistically non-significant, while Ramey (2007) provided evidence of a negative reaction of private consumption. As for taxes, Romer and Romer (2007) found that tax increase had a large negative effect on private consumption.   Response of IR to IR Figure 1 column 3 showed the impulse responses to one standard error shock to the private consumption equation at time t on expected values of the endogenous variables in the SVAR at time t + n. In response to a shock to private consumption, government expenditure declined by -0.03% at second quarter and this continued thereafter to -0.08% at the end of the two years six months horizon. The response of government expenditure was larger on impact when compared to the response of taxes to innovations to private consumption. On the other hand, the response of taxes to private consumption declined by -0.06% on impact but rose to a -0.05% point increase in the fourth quarter. Taxes then rose in the eighth quarter by 0.001%. Thereafter, taxes rose to 0.01% over the next two years. Thus, the impulse response of government spending and taxes to a shock to private consumption were both negative. This was in conformity with the result of Onodje (2009).

The Effects of Shocks on Government Expenditure
Under the assumption of perfect foresight, an unanticipated generalized one standard deviation innovation to government expenditure caused private consumption to rise to 0.04% in the first quarter which declined sharply to -0.04% in the second quarter. In the third quarter, shocks in government expenditure resulted in 0.016% increase in private consumption. In the fourth quarter, private consumption declined to 0.004% and thereafter continued to remain unstable till the tenth quarter when private consumption declined to -0.006%. These findings indicated a mixed result. Hence, it lied between Keynesian and Neo-classical proposition. Thus, this was in conformity with non-Keynesian proposition and similar to the result of Alfonso and Sousa (2009).
The shocks in government expenditure caused 0.03% increase in tax revenue in the first quarter and decrease in tax revenue by 0.001% in the second quarter. Hence, this continued to be unstable till eighth quarter and remained steady from the ninth quarter at zero percent till tenth quarter. Thus, the shocks on government expenditure crowded-out private consumption as argued in the Keynesian theory that increase in government expenditure would cause increase in private consumption through increase in income. This result was in conformity with the earlier studies (Onodje, 2009;Sousa, 2009;Mancellari, 2011).

4.5.3
The Effects of Shocks on Tax Revenue An unanticipated one standard deviation innovation to taxes caused private consumption to rise by 0.076% in the first quarter. The shock in taxes to private consumption declined to 0.006% in the second quarter. However, there was an unstable changed in the private consumption between third quarter and eighth quarter as a result of shocks in taxes. Meanwhile, in the tenth quarter the shock in taxes resulted to 0.023% stable in private consumption. Hence, these findings were against the Keynesian proposition that increase in tax revenue due to increase in tax rate would result to fiscal contraction.
A positive shock to taxes had a significant negative effect on government spending since the one standard deviation bands included negative values for the entire horizon as shown in figure 5.3, column 2. This might be explained by a deficit-reducing tax increase behaviour that aimed at stabilizing or reducing government debt (Caldara & Kampus, 2008). More specifically, the Nigerian government tends to raise taxes to ameliorate budget deficits and pay down debt incurred in previous periods (deficit-driven tax changes), rather than to raise taxes to increase public investment expenditure in future periods to enhance private consumption (spending-driven tax changes). Thus, a positive shock to taxes as a result of increase in government expenditure in the short-run resulted to an increase in government savings in other to reduce public debt accumulated in previous periods.
In summary, figure 1 displayed the responses of private consumption to government debts, government spending and net-tax shocks. The responses of private consumption in the baseline SVAR notably decreased after a positive government spending shock, in line with non-Keynesian models, although such positive response phased out rather quickly. In the same vein, increase in net taxes brought private consumption upwards in quarters following the shocks and the response became significant after the fourth year. Also, the response of private consumption in the baseline SVAR showed an increasing response from a government debt shock in line with Keynesian models. This positive response did not phase out quickly.

Robustness Check of Structural VAR Results
The study employed various measures to test the stability and robustness of the SVAR results. First, in testing the stability condition of the model, the study employed h the graphical root characteristic polynomial. The results of this technique indicated that all the roots of the characteristic polynomial were inside the unit circle signifying that the defined SVAR model was stable as shown in figure 2. Furthermore, in addition to stability and robustness check the study estimated the matrices A and B to obtain the coefficients of structural shocks based on the earlier stated assumptions. The results of the short-run response pattern of the matrices A and B were presented in tables 10 and 11 respectively.   , the coefficient of government expenditure shock was positively signed and statistically significant at 5% level. This finding indicated that a positive government shock would result to increase in private consumption by 2.2%. This result was in conformity with the Keynesian proposition that stated that increase in government spending would result in increased private consumption through increase in disposable income. Also, the coefficient of tax revenue shock was positively signed and statistically significant at 5% level. The positive tax revenue shock showed that increase in tax revenue through increase in tax rate would result in 1.5% increase in private consumption. Thus, this result was in conformity with non-

Inverse Roots of AR Characteristic Polynomial
Keynesian proposition which stated that contractionary fiscal policy would have expansionary effect on macroeconomic variables.  (ii) when 0 exp  tr g  , a similar result was shown but their magnitudes vary. The coefficient of government spending shock was 2.7% while the coefficient of tax revenue shock was 1.2%. Both coefficients were statistically significant at 5% level. The positive government spending shock was in conformity with Keynesian proposition while the positive tax revenue shock was in conformity with non-Keynesian proposition.
In addition, the long-run response pattern of the SVAR result was equally presented as shown in table 12.  The long-run response pattern of the SVAR result showed that the coefficient of government spending shock was positive while the coefficient of tax revenue shock was negative. The positive government expenditure shock indicated that a unit unexpected increase in government spending would result to 1.2 unexpected increase in private consumption. This was however similar to short-run response pattern. Hence, it was in conformity with the Keynesian proposition. At a final note, a negative tax revenue shock showed that a unit unexpected increase in tax rate would lead to 1.9 unexpected decrease in private consumption. This result was in conformity with the Keynesian proposition that stated that a contractionary fiscal policy in term of increase in tax rate would lead to decrease in consumer's income and this would result to decrease in private consumption in the long-run. This result was in conformity with the earlier studies by Onodje (2009) and Sousa (2009).

Conclusion and Recommendations
The paper concluded that a shock in government spending hit private consumption negatively while a positive shock in tax revenue contributed to a change in private consumption. This showed that a unit positive shock in government revenue reduced private consumption by 97 percent and this crowded-in private consumption in Nigeria. Also, the negative shock emanated from government expenditure to government debt indicated that discretionary government policy in terms of change in government expenditure reacted inversely to change in government debt. The response of tax revenue to shocks in both government expenditure and private consumption decreased, indicating that the effect of fiscal policy shock on private consumption has non-Keynesian effect in Nigeria. Furthermore, the private consumption responded to positive shocks in government debt, tax revenue and government spending. This showed a mixed results between Keynesian and non-Keynesian effect on private consumption in Nigeria and the response of private consumption in the baseline SVAR showed an increasing response from a government debt shock in line with Keynesian models. Hence, the results of the paper confirmed the existence of both the Keynesian and non-Keynesian effect of fiscal policy. In line with this, this paper recommended that government should exercise fiscal discipline; this can be achieved through reduction of wasteful spending. With this step, it will be relatively easy to determine the expenditure growth pattern in Nigeria.