The role of fiscal policy in affecting economic activity has been on the theoretical and applied research agenda for both academicians and policy makers since the evolution of macroeconomics. Fiscal Policy can affect an economy dynamically; this impact could differ across economies depending on the structure of the economies. In the context of developing economies, such as Pakistan, where active fiscal policy or a non-Ricardian fiscal policy is practiced, large seinorage revenues exist and ratchet-up effects of expenditures are found [Khalid, et al. (2007)], it becomes crucial to ascertain the fiscal policy effectiveness. This paper attempts to identify the fiscal policy effectiveness with respect to different budgetary components towards aggregate economic activity in Pakistan. Here Structural VAR estimation using Blanchard and Roberto (2002) methodology is applied to identify the impact of fiscal policy instruments on the economy for Pakistan. In Pakistan's case it is evident that fiscal policy has been playing a major role in providing policy options for the government throughout her history of economic management. It was observed that government expenditures at the aggregate level affect the economy in line with the theory, i.e. it affects the economic activity positively, whereas the tax variable shock affects economic activity opposite to the theory. This may be due to the fact that the tax elasticities are very low and government expenditures also behave in a ratchet up manner, as also pointed out by Khalid, et al. (2007) while estimating fiscal reaction function for Pakistan Hence when revenues increase the government expenditures also increase instead of paying off the debt, which may lead to a positive impact on the economic activity. We also estimated separately, the fiscal policy effectiveness for fiscal policy instrument sub-components i.e. defense and interest payment expenditures. In our analysis it turns out that defense expenditures have positive impact on economic growth while interest payments negatively affects it which is comparable to a number of studies. Similarly disaggregated analysis for the revenue variable by splitting the tax revenues in the broad categories of direct taxes and indirect taxes were also made. From that analysis it appears that direct taxes affect economic activity negatively while indirect taxes ambiguously affect economic activity positively.

JEL Classifications: E62, H3, H5

Keywords: Fiscal Policy, Multipliers, Government Expenditures, Taxes

1. INTRODUCTION

The role of fiscal policy in affecting economic activity has been on the theoretical and applied research agenda for both academicians and policy makers since the evolution of macroeconomics. Fiscal Policy instruments have different direct multiplier (Keynesian) effects on aggregate demand and its components. For instance increased government expenditures can lead to increased aggregate demand and thus activating the idle production factors in the economy and creating more employment and output growth. Moreover, in recessionary phases, when economy is in a liquidity trap (e.g., Japan), where private investment demand becomes inelastic, fiscal policy provides the necessary stimulus to the economy for coming out of that trap. Further, in developing economies, government expenditures also play a complimentary role for private investments. Although depending on the set of assumptions the outcomes of a fiscal intervention could be different. E.g. rule of thumb consumers, ricardian equivalence, longrun and shortrun policy frame etc. Still the option of fiscal policy has always been preffered over other alternatives. Such as the recent evidence from global economic crises, where discretionary fiscal policy was adopted due to weakened monetary policy transmission channels [Furceri and Annabelle (2010)].

One of the more recent examples could be the fiscal stimulus in financial crises in the USA and the Euro-zone. Similarly, monetary policy becomes ineffective in real business cycle supply-side theories, but fiscal policy is effective through the investment demand and labour supply channels. Fiscal Policy can affect an economy dynamically; this impact could differ across economies depending on the structure of the economies. Fiscal policy is considered to be the most active tool for macroeconomic stabilisation and growth achievements, especially in a developing economy context. This is also evident by the activeness of fiscal policy vis-a-vis monetary policy in Pakistan [Nahyun (2010)].

In Pakistan active fiscal policy or a non-Ricardian policy is practiced, large seinorage revenues exist and ratchet-up effects of expenditures are found [Khalid, et al. (2007)], it becomes crucial to ascertain the fiscal policy effectiveness. This paper attempts to identify the fiscal policy effectiveness with respect to different fiscal policy tools towards aggregate economic activity. Most of the studies reviewed have used either the cumulative variable of fiscal deficit as an indicator for fiscal policy or have not adjusted the fiscal variables for their automatic responses towards the economic activity, hence the results may be dubious for the impact and effectiveness of Fiscal policy for Pakistan. Our study here employs the novel Structural Vector Auto Regressive (SVAR) method of estimation by taking two levels of disaggregation, first by looking at the impact of government expenditures and taxes as instruments of fiscal policy instead of fiscal deficit, and then by using the second level of disaggregation i.e. their subcomponents. Secondly the use of institutional information of fiscal policy settings i.e. the automatic response of a fiscal shock is also incorporated. Further we have also considered atypical fiscal disaggregated indicators such as interest versus non-interest, defence and interest combined versus non-defence non-interest expenditures for evaluating the fiscal policy effectiveness as each of these may have a different multiplier.

For the case of fiscal shocks almost none of the studies has empirically tested the relationship of disaggregated fiscal policy instruments at this level with the aggregate macroeconomic variables in the context of Pakistan. By considering different fiscal policy variables a number of policy lessons can be derived which can be helpful in designing an effective fiscal policy mechanism.

The paper is organised as follows:

2. FISCAL POLICY EFFECTS: THEORETICAL MODEL

The model presented here is akin to the model presented by Perotti (1999) and Barro (1989) wherein they develop a standard macro-economic model depicting the effects of both revenue and expenditure shocks on output and its components. The model is based on the following key assumptions which are fairly standard in all macroeconomic model:

* Taxes are distortionary (Although it can be argued that distortions may not occur in case of an equivalent tax i.e. replacing the distortionary tax with a non-distortionary tax such as income tax, keeping the revenue target intact. But here for simplicity we are assuming taxes to be distortionary).

* The policy makers effectively discount future more than the households/private sector (This will result in a non-tax smoothing position for initial periods).

Economy can be divided in individuals by their levels of access to financial markets (credit constraint); i.e. there is a segment of society which is without the possibility of smoothing their consumption by effectively savings and dis-savings through the credit market (rule of thumb consumers, this can further be extended by relaxing the extreme position and assuming that their access to the financial market is more costly as compared to others such as, larger borrowers, because they have to borrow at a higher then market rate of interest due to their low credit worthiness, small credit demand and lack of collaterals).

Government expenditures have positive effect on output(This assumption is highly debatable now-a-day, especially with the emergence of expansionary fiscal contraction literature such as empirical evidence provided by Gavin and Perotti (1997)). There are no supply side effects of government interventions; i.e., supplies of labour and other factors of production are inelastic. Government expenditures are exogenously given; i.e. they do not follow a particular reaction function [Khalid, et al. (2007) estimated a Fiscal Reaction function for Pakistan is estimated and it holds],

[DELTA][C.sub.1] = [[sigma].sub.1], [[epsilon].sup.G.sub.1] [[sigma].sub.2][[epsilon].sup.T.sub.1] + [gamma]([Y.sub.1/0] - [Y.sub.0]) + [[mu].sub.1] ... (1)

Theoretical model capturing the impact of both taxation and government expenditures is developed for consumption. The final equation of the model describes the possible effects of expenditure and taxation policy actions depending on the share of population having access to the financiais markets ([gamma]). Where;

[[sigma].sub.1] = [[sigma].sup.un.sub.1] + [[sigma].sup.r.sub.1]

[[sigma].sub.2] = [[sigma].sup.un.sub.2] + [[sigma].sup.r.sub.2]

[[mu].sub.1] = [[mu].sup.un.sub.1] + [[mu].sup.r.sub.1] ... (2)

Here [[sigma].sub.1] and [[sigma].sub.2] captures the total impact of government expenditures and taxation shocks respectively. Some more explanatory equations to these are below:

[[sigma].sup.r.sub.1] = [gamma][beta] > 0

[[sigma].sub.2] = -{1 + 2[omega] [T.sub.1/0]} < 0

[[mu].sup.r.sub.1] = [gamma]([alpha] [[phi].sup.x.sub.1] + [[phi].sup.Y.sub.1]) ... (3)

Here [beta] and [omega] are the expenditure and tax multiplier for consumption and [phi] is the stochastic disturbance. Y is the disposable income and X is the vector of all other variables which effects the disposable income. 1/0 means information about period one in base period. [U.sub.n] means unconstrained households.

It is anticipated that with no wealth affects of the expected policy changes, expenditure shocks, with taxation being assumed to be constant, will have a positive impact on the economy (through consumption). Similarly, it will be the opposite in the taxation case, although it was not the case with the unconstrained households.

3. COMPARATIVE FISCAL POSITION OF PAKISTAN

Contrary to the general belief about the size of the government, Pakistan is not among the "large government" countries. Its share of expenditures in the total GDP is not very high. However the overall revenue and taxes in particular are very low. In fact amongst the lowest in the world. This puts the fiscal policy conduct in a suboptimal position. Because on the expenditure side also, although the overall size of the government is small but due to consistent increase in the interest payments and debt servicing the available resources are shrinking. This in turn leads to excessive cuts in the development expenditures. On the other side, our tax system is also not very elastic and/or buoyant. On top of it the overall tax incidence shows that the tax system is not progressive.

4. METHODOLOGY AND DATA

Large body of literature shows a common trend of showing a positive effect of fiscal change to macro outputs such as consumption. This is due to the presence of endogenity and issues specific to fiscal policy operations, so it becomes difficult to get reliable estimates [see Perotti (2004) for details]. Specifically while administering the fiscal operations, the changes in the fiscal stance come as a result of long and politically manipulated process, hence private agents are not surprised. This may result in even not affecting the fiscal variables itself in the first instance [what Lippi and Reichlin (1994) referred to as shocks being non-fundamental].

Secondly with the presence of automatic response from budgetary components (although the response may vary for each country and budget component) these problems are further exacerbated [Afonso and Peter (2008)]. Lastly since the fiscal shock may have a different originating base (such as direct taxes or indirect taxes, through expenses or transfers etc.), hence each of these may have a different short term and long term impact. This problem of endogenity is addressed by Blanchard and Roberto (2002); in their seminal paper they have used a Structural VAR that employs out of model institutional information, such as the elasticity of budgetary components and other timing issues of the precedence in fiscal policy making decisions.

Identification of the fiscal shock is achieved by considering the decision lags in fiscal policy and the institutional information about the elasticity of fiscal variables such as taxes and spending to economic activity. The present paper mainly relies on this approach, while extending it to the components of taxes and spending for Pakistan. We use the Structural VAR (SVAR) approach here to identify structural balances and impacts of various fiscal policy instruments following methodology used by Blanchard and Roberto (2002), which filters the cyclical response of the Fiscal policy to economic conditions.

Now for the aggregate tax elasticity ([[alpha].sub.ft]) estimation we can define total taxes to be

T = [summation over (i)][T.sub.i] ... (4)

and the relevant tax base to be [B.sub.i], then we can define the aggregate tax elasticities with respect to economic activity as:

[mathematical expression not reproducible] (5)

Here [T.sub.i] refers to ith tax, [mathematical expression not reproducible] refers to elasticity of ith type of tax to its relevant base and [mathematical expression not reproducible] refers to elasticity of ith tax base to total economic activity (1)

Data for elasticities the data is used from 1971-2010.The reason for limiting the analysis to this period is availability of estimated changes in revenues due to discretionary changes in tax structure. Another limitation of study is that such data are not available for provincial taxes or non-tax revenues. The data on estimated revenue impact of discretionary changes in the tax variables were personally obtained (earlier it was used to be published as an exploratory memorandum with the federal budget publications set, but later discontinued) from the Federal Board of Revenue. For direct taxes these data on disaggregated level were not available; hence the elasticity for each of the subhead of direct taxes could not be estimated. But availability of these data regarding indirect taxes made estimation of elasticity possible at a disaggregated level. Further, for provincial revenues this sort of data (estimates of revenue impact for discretionary changes) was not available; hence their buoyancy estimates were used in the analysis.

While estimating the elasticity by applying the method of Proportional Adjustment Method (PAM) developed by Prest (1962) and Mansfield (1972), each tax variable is cleaned for the discretionary changes starting from base year (which in our case is 1971). In the next step each cleaned tax variable is regressed upon a proxy tax base using appropriate econometric methodology. Similarly the tax base is regressed on the national income to obtain the tax base elasticity with respect to national income. However in the case of tax base to total income elasticity econometric problem of simultaneity was anticipated and corrected accordingly. As the tax base variables are simultaneously determined with total income, hence running a regression with endogenous right hand side variables does not provide consistent results. This problem was solved by using the 2SLS method.

We used the Structural VAR methodology to recover the parameter estimates for analysis of fiscal policy effectiveness in Pakistan. Once the parameters are identified then the impulse responses and variance decomposition is also done to comment on the transmission mechanism. Estimating the VAR in first difference to take away the stochastic trend as it also takes away the essential information, namely the error correction [Enders (2004)]. So if there is a problem of Stationarity and there is cointegration in the variables then there should be a cointegrating relationship imposition, i.e. to estimate VAR in an un-restricted form and use the error correction mechanism to look at the innovation accounting or to use the VAR in levels with appropriate lags if the parameters are not important and just the innovation accounting is required [Enders (2004)].

So we have used both the Deterministic trend and Stochastic trends in reduced form VAR estimation and also a cointegrating relationship approach (VECM) to compare the results. We first calculated three sub categories of aggregate tax elasticities i.e. federal government indirect tax elasticity, federal government direct tax elasticity and the total provincial government's tax elasticity. Then these sub categories were aggregated on the basis of their respective weights in total tax revenues using equation shown on earlier.

Low value of Consolidated Total Tax Elasticity (TTE) 0.53 represents peculiar characteristics of tax structure in Pakistan. The low tax effort, large tax gap and huge tax credits have resulted in a tax to GDP ratio which is abysmally very low. On the basis of Unit root tests we decided to use the linear trend and take the first difference to account for both the possibilities of deterministic trend and stochastic trend in the VAR estimations. After recovering the structural parameters of the primitive VAR by structural factorisation we have used the innovation accounting method by estimating the impulse responses using the structural factorisation used in recovering these estimates. These impulse responses provide average dynamic responses of fiscal shocks. To determine the impact of various fiscal shocks on the total economic activity (GDP), we have used a number of definitions for the revenue side and expenditure side variables. The GDP deflator is used to deflate the revenue side variables and other expenditure side variables as it allows us to show the impulse response in relation to GDP [Blanchard and Roberto (2002)].

5. FISCAL POLICY EFFECTS ON AGGREGATE MACROECONOMIC INDICATORS IN PAKISTAN

Case 1: Net Taxes (Net of Subsidies and Interest Payments)

From the above table it appears that there is not much difference in the estimates of contemporaneous coefficients in case of both specifications except when government expenditures are ordered first. The coefficient is still same in sign i.e. negative but becomes significant and increased in value in case of Stochastic Trend specification. If we look at the signs of these parameters then for (i.e. effect of g on x within a year) it is positive, in line with the theory and is significant. But the magnitude of the parameter is quite small and that's quite surprising. The low value of government expenditures effecting the output may indicate the poor planning and high proportions of budgets going on non-productive expenses.

On the other hand tax shocks also seem to affect the GDP positively and significantly although the coefficient is quite small. This is not surprising as the tax to GDP elasticity is very low and the government expenditures tend to behave in a ratchet up fashion. Secondly since we are using the net taxes (net of total transfer payments i.e. including interest payments and subsidies), which as compared to total taxes behave quite differently especially in the later years and also differs a lot in magnitude. The difference in magnitude and signs of and shows the importance of considering the causing factors of changes in budget deficit; as each instrument of fiscal policy has a different impact on the state of the economy. Especially in case of Pakistan like other less developed countries both instruments have altogether different implications and cannot be considered as an alternative to each other.

All other parameters are insignificant except for the tax response to government expenditures when government expenditures are ordered first. It is negative and significant in this case, which is a non-trivial result. This would also mean that there will not be much difference in the impulse (dynamic) responses when we evaluate them. One of the reason could be the budgeting method used in Pakistan, where first the expenditures are set, then all possibilities of donor and external financing is considered and then the left over of unmet expenditures are set to be the revenue targets. This type of revenue targets are ad-hoc, it has been a practice that they are under achieved, while on the other hand the expenditures are always under estimated and they tend to be higher when financial year closes on June 30th of each year.

The expenditures (more specifically the current expenditures) tend to be under estimated while the revenues are over estimated. The revenue over estimations leads to a higher expenditures setting which does not reduce once revenue target loss is realised, as the revenues accrue mostly in the last quarter of the fiscal year whereas the expenditures tend to start from the first quarter. Hence it may appear that with increase in expenditures the revenues tend to fall. Secondly again since we have netted the taxes with the total subsidies and interest payments hence the results are like that, but as we moved further we have also done estimations without netting these and there the results are quite in line with the theory.

6. DYNAMIC EFFECTS OF A TAX SHOCKS

The figure shows impulse responses of one standard deviation shock to the tax variable under the structural factorisation both for the Deterministic Trend (DT) and Stochastic Trend (ST) specifications and a cointegrating relationship model under the sub-case one where taxes are ordered first followed by sub-case two where government expenditures are ordered first. With shock in tax variable, the GDP variable shows a positive increase of around .03 but is insignificant, from there it increases to a positive value of .06 by the second and third year and then starts to decline. However except for the first year of impact the next values are highly significant.

A case of positive contemporaneous effects of a tax shock due to the ratchet up effect of the increase in government expenditures responding to the increase in taxes and abysmally low value of the tax elasticity and thus overall low tax system efficiency. This is also observable from the tax shock effect on the government consumption expenditures; in the first year it posted a decline and then it gradually increases upto the third year and then starts to fall slowly. It does not reach back to its normal value even beyond ten years of impact. Whereas correspondingly the GDP variable fall is sharp and has a tendency to move back to its normal pre-shock position. Here the tax variable itself also has a tendency to fall sharply after the initial shock and after the 10th year of shock the impact is almost zero. If we look at the peaks then we see that for GDP it's the second year, for government expenditures it's the third year of impact and for tax variable itself after the first year of shock it keeps on declining.

7. DYNAMIC EFFECTS OF A GOVERNMENT EXPENDITURE SHOCK

For both the tax and expenditure shocks, own variable response is quite significant (hence fundamental). Impact is positive in the long-run on the other instrument i.e. tax on expenditure and expenditure on tax, however the channels for the two may be different. For the first case it is direct impact (Ratchet up effect) and second is indirect (through increase in economic activity and increased tax collection). This is further analysed by taking other definitions of tax (i.e. without netting them out).

However in terms of impact on the economic activity, contrary to standard theory, the tax variable is having a positive impact on economic activity, which is mainly due to the construction of variable where large proportion of taxes are netted out for transfers of interest payments and subsidies. Secondly due to ratchet up effect, the negative impact of tax on the economic activity may be subdued. But for the case of government expenditures the sign of the shock is following the standard theory. Further the shocks via expenditures are more persistent as compared to the tax shock, which are also similar to the large body of literature which exists in this field [e.g. Blanchard and Roberto (2002) etc].

Case 2: Tax Revenues (Net of Interest Payment only, Net of Subsidies only and Total Taxes)

With the estimation results of taking different definitions of taxes (Netting all transfer payments i.e. both interest payments and subsides, just the interest payments, just the subsidies and without netting any of these), and different cases of estimations and ordering of variables (using deterministic trend, stochastic trend and cointegration method of estimations and ordering taxes first i.e. no contemporaneous response of taxes to government expenditures and then taking government expenditures first) have not differed fundamentally in the direction of impact when compared for same method of estimation and same shock across various definitions of taxes. Although the impact's persistence, peak, trough and magnitude of impact have varied a lot across these definitions.

Case 3: Government Expenditures (Defense and Non-Defense Spending, Interest Payment Spending and Non-Interest Payment Spending, Defense and Interest Payment Spending and Non-Defense-Non-Interest Payment Spending)

The figure above shows the dynamic responses of disaggregated government expenditures in defense (DG) and non-defense (NDG) expenditures. It is widely debated as to whether the defense expenditures are pro-growth or other wise. From our analysis by using the structural information on the fiscal structure it appears that there is a pro-growth impact. There is a positive impact on the total tax revenues, although at impact the increase in tax revenue is maximum and it gradually decreases, which almost becomes zero in a span of a decade.

For defense expenditures itself the impact is non persistent and once impacted it gradually decreases. On the other hand the non-defense expenditures decrease upon impact and falls upto the third year as there is no matching tax revenue increase after the first two years, then starts to recover as the defense expenditures are on the decline by then but does not touch 0 in ten years interval.

Finally the economic activity increases at the impact, but falls immediately in the next period, due to the reduction in non-defense expenditures and then gradually again picks up, but it never touches zero in our analysis interval of ten years. Hence quite contrary to the common concerns the impact of defense expenditures on economic growth turns out to be positive. Without going into theoretical details, what appears to be is that most of the defense production is also done within country, so it might be coming from expenditure side.

Case 4: Revenues (Direct taxes and Indirect Taxes)

Both taxes have a different dynamic impact which not only differs in direction but also in persistence. It appears that in case of direct taxes the government expenditures initially increase then decrease for the next 2 years and remains below zero till the seventh year after which again it is positive. While for the economic activity the impact is negative, which stays at a high negative value for about seven years then gradually truncate in the next three to a zero impact. This shows the persistence of a direct tax shock in case of economic activity. One of the reasons for such dampening effect could be because of the decline in the government expenditures for the same number of years with respect to a fiscal shock coming from direct taxes. But overall there is strong negative impact of a direct tax based fiscal shock to the economy. Which is quite opposite to what was observed in case of a total tax shock as explained above.

On the other hand in the lower panel of the above figure when we see the impact of a shock to indirect taxes then the things are more or less the same as of the total taxes. One of the reasons for this similarity is of course the higher proportion of indirect taxes in the total taxes as compared to the direct taxes. Unlike the direct taxes case the indirect taxes impact on its own variable quickly dies out in the 2nd year and remain very low till the tenth year.

Corresponding to this shock direct taxes fall, one of the reason could be the decline in reliance on them as indirect taxes are on the rise. As in case of Pakistan the potential tax gap is widening due to weak tax implementation and large undocumented economy. Government expenditure presents the ratchet up effect; as the tax revenues increase the corresponding government expenditures also increase and keep on increasing. Finally the economic activity is presenting a positive impact upon receiving a shock in indirect taxes, mainly due to weak tax elasticity (as earlier mentioned around 0.43 only) and because of the corresponding ratchet up effect in government expenditures which further exacerbate the situation. But clearly the anticipated negative impact on the economic activity due to shock in indirect taxes does not take place.

8. CONCLUSION

This study employs the novel procedure developed by Blanchard and Roberto (2002) in using the institutional information such as the elasticity of different fiscal instruments e.g. taxes and the decision precedence while opting for a fiscal instrument. By using the SVAR methodology it was observed that government expenditures at the aggregate level affect the economy in line with the theory, i.e. it affects the economic activity positively, whereas the tax variable shock affects economic activity oppositely.

The spending multipliers are larger than the tax shocks impact purporting the traditional Keynesian theory of larger expenditure multipliers than tax multipliers for Pakistan. This may be due to the fact that the tax elasticities are very low and government expenditures also behave in a ratchet up manner. Hence when revenues increase the government expenditures also increase instead of paying of debts, which may lead to a positive impact on the economic activity. Estimation results for different definitions of taxes (Netting all transfer payments i.e. both interest payments and subsides, just the interest payments, just the subsidies and without netting any of these), and different cases of estimations and ordering of variables (using deterministic trend, stochastic trend and cointegration method of estimations and ordering taxes first i.e. no contemporaneous response of taxes to government expenditures and then taking government expenditures first) have not differed fundamentally in the direction of impact when compared for same method of estimation and same shock across various definitions of taxes.

Although the impact's persistence, peak, trough and magnitude of impact have varied a lot across these definitions. Government expenditures have come out to be more effective towards desired outcomes. These results are in conformity of the earlier studies done for other countries however for Pakistan we could find only one study supportive of our results. The reasons for these differences with studies in Pakistan are attributed to the difference in estimation methodology and the fiscal policy indicator used. Further by looking at the dynamics in terms of sub components it was observed that there is a separate transmission mechanism of the sub-aggregates i.e. defense and interest payment expenditures. In our analysis it turns out that defense expenditures have positive impact on economic growth while interest payments negatively affects it. These results are comparable with a number of studies.

Similarly disaggregated analysis for the revenue variable by splitting the tax revenues in the broad categories of direct taxes and indirect taxes it appears that direct taxes affect economic activity negatively while indirect taxes ambiguously affect economic activity positively, again for the reason of ratchet up effect seems to be the reason.

9. POLICY IMPLICATIONS

Government expenditures as a policy instrument appear to be more effective as compared to taxes. Three possible reasons for such an outcome appears; low tax base, less elastic taxes and ratchet up effect on government expenditures. Hence there is a need to reform our taxation system. Secondly private investment is supplemented with government expenditures, hence increase in development is inevitable for increasing the pace of economic growth.

Finally aggregate indicators of policy intervention variables; here the Fiscal policy, such as budget deficit and the outcome variable, here the economic activity (such as the GDP) may give a picture which is different from what is happening at the disaggregate level for both the intervention and outcome variables. Hence fiscal policy conduct may incorporate the disaggregated level of instrumentation and outcome variables should also be seen in component wise effects.

Mahmood Khalid <[email protected]> is Senior Research Economist at Pakistan Institute of Development Economics, Islamabad. Ahsan ul Haq Satti <[email protected]> is Assistant Professor at Pakistan Institute of Development Economics, Islamabad.

Authors' Note: Authors acknowledge the contribution in paper through valuable comments provided by Dr. Wasim Shahid Malik (SBP-National Professor). Findings of the paper does not reflect their Institution's perspective.

REFERENCES

Afonso, Antonio and Peter Caleys (2008) The Dynamic Behaviour of Budget Components and Outputs. Economic Modeling 25, 93-117.

Barro, Robert J. (1989) The Ricardian Approach to Budget Deficits. Journal of Economic Perspectives 3:2, 37-54.

Bernanke, Ben S. (1986) Alternative Explanations of the Money-income Correlation. Carnegie-Rochester Conference Series on Public Policy. Elsevier 25:1, 49-99.

Blanchard, Oliver and Roberto Perotti (2002) An Empirical Characterisation of the Dynamic Effects of Changes in Government Spending and Taxes on Output. The Quarterly Journal of Economics, November, 1329-1368.

Christopher, A. Sims (1986) Are Forecasting Models Usable for Policy Analysis? Quarterly Review, Federal Reserve Bank of Minneapolis, Issue Win, 2-16.

Enders, Walter (2004) Applied Econometric Time-series, 2nd edition. New York: Willey.

Furceri, Davide and Annabelle Mourougane (2010) The Effects of Fiscal Policy on Output and Debt Sustainability in the EURO Area: A DSGE Analysis. (OECD Economics Department Working Papers).

Gavin, Michael and Roberto Perotti (1997) Fiscal Policy in Latin America. NBER Chapters. In NBER Macroeconomics Annual 12, 11-72. National Bureau of Economic Research, Inc.

Khalid, Mahmood, Wasim Shahid M. and Abdul Sattar (2007) Fiscal Reaction Function and Transmission Mechanism for Pakistan. The Pakistan Development Review 46:4, 435-447.

Mansfield, Charles Y. (1972) Elasticity and Buoyancy of a Tax System: A Method Applied to Paraguay. International Monetary Fund Staff Papers 19:2, 425-46.

Nahyun, Abdul Sattar (2010) Fiscal Theory of Price Level for Pakistan. (Unpublished PhD Thesis).

Pakistan, Government of (2005) Fiscal Responsibility and Debt Limitation Law. Islamabad: Ministry of Finance.

Perotti, R. (1999) Fiscal Policy in Goods Times and Bad. Quarterly Journal of Economics 114:4, 1399-1436.

Prest, Alan R. (1962) The Sensitivity of the Yield of Personal Income Tax in the United Kingdom. The Economic Journal 72:287, 576-596.

(1) Details about the methodology and tax bases used are provided in the next sections.

Caption: Fig. 1. Response to Tax Shock

Caption: Fig. 2. Response to Government Expenditures Shock

Caption: Fig. 3. Response to Government Expenditures Shock-Defense Expenditures

Caption: Fig. 4. Response to Tax Shock

Table 1
Cross Country Comparison: Tax to GDP Ratios

Country Name      Country Classification   2005
                  by World Bank

Singapore         High Income: Non-OECD    12.20%
Australia         High Income: OECD        23.70%
Austria           High Income: OECD        20.20%
France            High Income: OECD        22.30%
Germany           High Income. OECD        11.10%
The Netherlands   High Income: OECD        22.60%
New Zealand       High Income: OECD        31.30%
United Kingdom    High Income: OECD        27.30%
United States     High Income: OECD        11.40%
Brazil            Upper Middle Income      3.30%
Maldives          Lower Middle Income      18.00%
Pakistan          Lower Middle Income      9.60%
Sri Lanka         Lower Middle Income      13.70%
India             Lower Middle Income      10.20%
Bangladesh        Lower Income             8.20%
Nepal             Lower Income             9.20%

Country Name      2006     2007     2008

Singapore         12.60%   13.90%   14.60%
Australia         23.50%   23.10%   --
Austria           19.80%   20.20%   20.10%
France            22.40%   21.80%   --
Germany           11.30%   11.80%   --
The Netherlands   23.20%   23.60%   --
New Zealand       33.20%   31.70%   --
United Kingdom    28.10%   27.70%   28.60%
United States     12.10%   12.20%   10.30%
Brazil            15.40%   16.30%   16.40%
Maldives          19.90%   21.50%   21.00%
Pakistan          9.40%    9.80%    9.80%
Sri Lanka         14.60%   14.20%   --
India             11.50%   12.40%   12.90%
Bangladesh        8.20%    8.00%    8.80%
Nepal             8.80%    9.80%    10.40%

Source: The World Bank;
http://databank.worldbank.org/ddp/home.do?Step=3andid=4

Table 2
Cross Country Comparison: Fiscal Expenditure Indicators (2007)

                                           Expense
                  Country Classification    (% of
Country Name      by World Bank             GDP)

Singapore         High Income: Non-OECD     13.17
Austria           High Income: OECD         38.78
Australia         High Income: OECD         23.62
France            High Income. OECD         44.40
Germany           High Income: OECD         29.02
The Netherlands   High Income: OECD         40.28
New Zealand       High Income: OECD         32.93
United Kingdom    High Income: OECD         40.00
United States     High Income: OECD         21.80
Brazil            Upper Middle Income       24.84
Maldives          Lower Middle Income       48.56
Pakistan          Lower Middle Income       16.32
Sri Lanka         Lower Middle Income       20.05
India             Lower Middle Income       16.01
Bangladesh        Lower Income              10.08
Nepal             Lower Income             15.06 *

                  Interest   Interest     Subsidies
                  payments   payments     and other
                   (% of      (% of       transfers
Country Name      expense)   revenue)   (% of expense)

Singapore           0.17       0.10          0.32
Austria             7.04       7.22         70.33
Australia           3.74       3.48         69.72
France              5.63       5.93         62.38
Germany             5.90       5.96         81.62
The Netherlands     4.49       4.43         79.19
New Zealand         3.86       3.42         37.63
United Kingdom      5.44       5.77         53.28
United States      10.41      11.55         60.92
Brazil             16.75      17.51         51.73
Maldives            3.57       3.11          2.57
Pakistan           26.39      29.17         30.67
Sri Lanka          25.46      30.70         23.41
India              22.09      24.07         53.86
Bangladesh         21.73      20.70         29.32
Nepal              7.00 *      6.03           --

Source: The World Bank; http://databank.worldbank.org/,
* Figures pertain to Year 2005.

Table 3
Tax Elasticities (1971-2010)

                            Tax to Base         Tax Base to
                                               Total Income

                         [[alpha].sub.i,b]   [[alpha].sub.b,y]

FDT                            1.15                0.77
CD (with Imports only)         0.84                0.77
CD(with total trade)           0.90                0.49
FED                            0.77                0.24
ST                             0.27                1.20
SUR                            0.87                0.42
PDT *                          0.92                1.00
PIT **                         0.84                0.75

                            Tax to Total
                               Income

                          [[alpha].sub.i] =
                         [[alpha].sub.i,b] *
                          [[alpha].sub.b,y]

FDT                             0.88
CD (with Imports only)          0.65
CD(with total trade)            0.44
FED                             0.18
ST                              0.32
SUR                             0.36
PDT *                           0.92
PIT **                          0.63

* Since the provincial government can levy agricultural
income tax, hence their base was taken to be the total income.

** Trade taxes are administered by Federal Government only.

Table 4
Weighted Tax Elasticities (Average of 1971-2010)

Federal Indirect Tax Elasticity (FITE)               0.42
Federal Direct Tax Elasticity (FDTE)                 0.88
Federal Total Tax Elasticity (FTTE)                  0.52
Provincial Total Tax Elasticity (PTTE)               0.66
Consolidated Total Indirect Tax Elasticity (TITE)    0.43
Consolidated Total Direct Tax Elasticity (TDTE)      0.88
Consolidated Total Tax Elasticity (TTE)              0.53

Table 5
Contemporaneous Coefficients for Case 1

                       [[beta].sub.tg]   [[beta].sub.gt]

Deterministic Trend
Coefficients              -0.188348         -0.052100
P-Values                   0.1874            0.7153
Stochastic Trend
Coefficients              -0.262499         -0.063787
P-Values                   0.0690            0.6585

                       [[alpha].sub.xt]   [[alpha].sub.xg]

Deterministic Trend
Coefficients               0.033372           0.020641
P-Values                    0.0000             0.0892
Stochastic Trend
Coefficients               0.033080           0.042352
P-Values                    0.0000             0.0576

Notes:

Sample Period 1960-2009

[[beta].sub.tg] = effect of g on t within a year assuming
[[beta].sub.gt] = 0; i.e. when government expenditures
are ordered first

[[beta].sub.gt] = effect of t on g within a year assuming
[[beta].sub.tg] = 0; i.e. when taxes are ordered first

[[alpha].sub.xt] = effect of t on x within a year

[[alpha].sub.xt] = effect of g on x within a year

In unrestricted VAR single lag was selected using criteria
explained above and no-auto correlation was found using
the LM test.

In Structural VAR estimation convergence was achieved
in 58 iterations for sub-case 1 and 90 in second sub-case.

Table 6
Fiscal Forecasts and Actuals (Billion Rupees)

                                Budget
                              Estimates       Actual

                              FY09   FY10   FY09   FY10

Total revenue                 1809   2155   1851   2078
Tax revenue                   1308   1564   1331   1473
Non-tax revenue               501    592    520    605
Total expenditure             2391   2877   2531   3007
Current expenditure           1876   2104   2042   2386
Development and net lending   516    774    487    653
Fiscal balance                -582   -722   -680   -929
Total financing               582    722    680    929
External financing            165    312    150    189
Domestic financing            417    390    531    740
Non-bank financing            243    246    225    436
Bank financing                149    144    306    304
Privatisation proceeds         25     19     0      0

Source: Economic Survey of Pakistan FY09 and FY 10.