The Dynamic Effects of Countercyclical Fiscal Stimulus on Output in Tunisia

With the global financial crisis hitting many countries, policymakers around the world have been weighing different countercyclical policies to support aggregate demand and restore growth. The analysis in this paper estimates a Structural Vector Error Correction model for Tunisia in order to identify the impact of fiscal policy shocks on real output. The authors find that public investment has a small impact on output in the short run but is an important medium-term growth-enhancing countercyclical instrument that has a robust impact on growth. Raising public investment by 1 dinar yields 0.12 dinar the first year, 0.30 dinar the second year, half a dinar the third year, and 1.08 dinars the sixth year. An increase in recurrent expenditure has a smaller but positive and persistent impact on real output. For Tunisia to obtain a larger short-term impact of public spending on output, procurement processes should be made faster and simpler. Finally, the analysis finds a countercyclical pattern of real public investment vis-à-vis real output and a relative rigidity/inelasticity of recurrent expenditures to output fluctuations.

With the global financial crisis hitting many countries, policymakers around the world have been weighing different countercyclical policies to support aggregate demand and restore growth. The analysis in this paper estimates a Structural Vector Error Correction model for Tunisia in order to identify the impact of fiscal policy shocks on real output. The authors find that public investment has a small impact on output in the short run but is an important medium-term growth-enhancing countercyclical instrument that has a robust impact on growth. Raising public investment by 1 dinar yields This paper-a product of the Poverty Reduction and Economic Management, Middle East and North Africa Region-is part of a larger effort in the unit to assist Tunisia in its response to the global financial crisis. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at ndiop@worldbank.org. 0.12 dinar the first year, 0.30 dinar the second year, half a dinar the third year, and 1.08 dinars the sixth year. An increase in recurrent expenditure has a smaller but positive and persistent impact on real output. For Tunisia to obtain a larger short-term impact of public spending on output, procurement processes should be made faster and simpler. Finally, the analysis finds a countercyclical pattern of real public investment vis-à-vis real output and a relative rigidity/inelasticity of recurrent expenditures to output fluctuations.

Introduction
Because of its limited foreign exposure, Tunisia's financial sector was not directly affected by the global financial crisis that emanated in the US in late 2008. 1 However, Tunisia has built strong trade and investment links with the European Union economies over the past quarter century. The Although textiles and clothing, leather and footwear (henceforth TCLF) and the mechanical and electrical sectors combined account for about 65 percent of exports, the decline in external demand of these products is unlikely to create a balance of payment problem in Tunisia. Indeed, the import content of export in these sectors is about 85 percent and imports typically decline in tandem with exports. Furthermore, continuing growth in FDI allows a convenient financing of the current account. In 2008, FDI increased by 40 percent to 5.6 percent of GDP representing 119 percent of the current account deficit.
Instead, the main concern of Tunisian policymakers is lower GDP growth and declining employment in the manufacturing sector in the short term. The manufacturing sector accounts for 20 percent of total jobs, 63.7 percent of which are in "totally" exporting firms. 2 The TCLF sector alone accounts for 48 percent of manufacturing employment and 9 percent of total employment. The mechanical and electrical sector, on the other hand, represents 16% of manufacturing jobs and 3.5 percent of employment. 3 Clearly, given their weigh in employment within the manufacturing sector, the sharp decline in export in these two sectors is a threat not only to growth but also to employment.
A key question is thus what type of countercyclical policies can help offset the negative shock on exports and moderate the economic slowdown and job losses. Many countries have taken emergency measures in late 2008 to support the exporting sector. The Tunisian government is no 1 For instance, the stock market grew by 9 percent in 2008. 2 Totally exporting firms are those that export 100 percent of their production. In Tunisia, a law enacted in 1971 (offshore regime) grants several fiscal and financial incentives to these firms, including duty-free imports of raw materials and equipments entering the production, 10 year tax holiday and free repatriation of profits. There also are "partially" exporting firms. Exports (in value) over GDP equals 47 percent. 3 Hydrocarbons, Tunisia's third export item is capital-intensive but not labor intensive. 3 exception. It has reacted swiftly in December 2008 by adopting several temporary measures to support exporting firms affected by the crisis, including partial exemption of social security contributions, fiscal incentives, and credit guarantees. These measures combined with a strong desire by firm managers to keep their trained employees, seem to have helped reduce significantly job losses.
In addition to emergency measures directed to the exporting sector, governments around the world have considered countercyclical measures to support aggregate demand and restore growth.
To this effect, the tool considered by most countries is a fiscal stimulus (discretionary increase in public spending and/or tax reduction). The theoretical case for a fiscal stimulus rests on the fact that (i) the origin of the current crisis is well identified: strong adverse shocks to aggregate demand following the global financial crisis; (ii) conventional monetary policy -e.g. reducing the interest rate-is likely to be ineffective because the pre-crisis investment boom in many countries led to excess capacities and; (iii) monetary transmission from the central bank to the real sector activity is not effective or predictable because of confidence crisis in the credit markets. Factors (ii) and (iii) dramatically reduce the comparative advantage of monetary policy over fiscal policy (IMF 2008).
Empirically, there is a rich literature on the effects of fiscal policy on output and other macroeconomic variables (see, for instance, Blanchard and Perotti 2002, de Castro and de Cos 2008, Giordano, Momigliano, Neri, and Perotti 2005and Giavazzi and Pagano 1990. However, while the empirical literature for developed countries is rich, not many studies exist on developing countries (see Table A1 for a list of recent studies). Furthermore, the issue of the sign and magnitude of the effects of fiscal policies across different countries is very much an open empirical question.
We contribute to the literature by estimating the impact and profile of fiscal stimuli on Tunisia.
We start by describing the fiscal policy context. Then, we describe the empirical strategy used to estimate the impacts of fiscal variables on output. This allows one to gauge the magnitude of fiscal stimuli (in the form of discretionary public spending) needed to offset growth slowdown in Tunisia. We use for that an SVEM model, which is the suitable tool if the variables are co- Public debt reduction combined with pro-active expenditure reallocation has been the main approach used to manage the budget deficit. However, a question is whether the level of public debt itself is so high as to be a constraint to undertaking countercyclical fiscal policies. Clearly, at 47.5 percent of GDP, Tunisia's public debt remains relatively high (slightly above the median level for emerging economies with similar Sovereign Rating). However, the share of public debt exposed to exchange rate and rollover risks has declined significantly in recent years, owing to efficient debt management. Government-guaranteed loans were estimated at about 11 percent of GDP at end-2007. Risks are contained by the following factors: (i) almost 82 percent of the external debt stock is in medium and long-term liabilities, 70 percent of which is owed to multilateral and bilateral creditors, implying limited rollover risks; (ii) around 75 percent of the debt is contracted in fixed interest rates suggesting little interest rate risk; and (iii) the stock of reserves more than fully covers all short-term liabilities (remaining maturity) in 2007 (IMF 2008).
Thus, to the extent that additional debt is invested in growth-enhancing areas, the level of public debt is not a major constraint in undertaking countercyclical fiscal policies. A fiscal deficit of up to 4 percent of GDP would still leave the public debt-to-GDP ratio below 50 percent (its 2007 level). Furthermore, domestic borrowing is unlikely to crowd out private investment: the deposit to credit ratio of the banking system stood at 1.15 percent in 2008 and 1.10 percent in the first quarter of 2009, denoting a situation of excess liquidity (Central Bank of Tunisia).

Dynamic effects of fiscal policy on output
Our objective is to determine the extent to which increasing public expenditures can help offset the deceleration in growth driven by the decline in external demand. To that effect, we use a model to simulate the effect of 1 dinar of increase in public spending (public investment and recurrent expenditure) on Tunisia's GDP in the short and medium-term. In other words, how much additional GDP can be obtained from 1 dinar additional public spending?

The impact of fiscal stimulus: empirical strategy
The system to be estimated consists of four variables: iv I Pr private investment and Y output. The extent to which an increase in Inv G and R G affect Y , as well as the temporal profile of the impact (lag and persistence of the shock) is the main focus of this paper.
Most recent studies of the impact of fiscal shocks on macroeconomic variables use the structural vector autoregressive regression (VAR) approach proposed by Blanchard and Perotti (2002) and Perotti (2002). This approach consists of isolating the structural shocks in a VAR system by imposing restrictions based on economic theory and institutional features that constrain the behavior of policy makers. For instance, government spending categories are generally contemporaneously unaffected by GDP. Therefore, reactions of fiscal policy to output changes only result from so called "automatic" responses, which are defined by existing laws and regulations. All fiscal policy developments in a given quarter or a year in some instances, which 6 do not reflect automatic responses, are basically seen as structural fiscal policy shocks, which are exogenous to output. This approach allows the identification of genuine shocks that hit the economy. We describe below the specific restrictions imposed in the coefficients of the model used here.
In matrix form, a Structural VAR (SVAR) could be written as: It follows that the reduced-form VAR is equal to: or, equivalently: and the structural coefficients, once VAR is estimated, one needs to impose restrictions on the relationship between the innovations t u and the structural shocks t  . We will come back to this below. SVAR models are appropriate when there is co-integration between the variables of the system. If co-integration tests cannot reject the hypothesis of one or more co-integration relationships among the variables, a structural vector error-correction model (SVEC) should be used. An SVEC is a transformed SVAR to include an error correction term necessary when the variables are co-integrated.
The reduced-form of a first-order VEC can be written as:

Variable measurement and data source
All variables are expressed in real terms and in logarithm. Public investment (called Titre 2 in Tunisia's fiscal jargon) is the gross fixed capital formation by the public sector, including state-7 owned enterprises. It includes direct expenditures on infrastructure and equipment in all sectors, donor-financed projects and the state's external borrowing on behalf of the private sector. Current expenditures are expenses under Tunisia's Titre 1, i.e, salaries and wages, goods and services, interest payments and transfers and subsidies. In this paper, we exclude interest payments from recurrent expenditures. Private investment is measured as the gross fixed capital formation by the domestic private sector.
Both recurrent expenditures and GDP are converted to real terms by using the GDP deflator.

Unit roots and co-integration tests
Tables 2 and 3 in annex show the results of the unit root tests. They show that all the variables in the system have a unit root in level but are stationary in first difference. We then apply the Johansen (1988 and1991) and Juselius (1990 and1992) co-integration tests. 4 As Table 4 shows, both tests indicate that the null that there is no co-integration relationship between the two variables (r=0) is rejected at 5 percent. On the other hand, one cannot however reject the hypothesis that there is one co-integration relationship (r<=1) against r=2. An SVEC model is thus the correct specification.

Restrictions imposed on short-term multipliers
For the multiplier matrix to be identified, some assumptions should be made regarding the contemporaneous relationships between the variables of the system. In our case, we have four variables and have imposed the following restrictions in the SVEC short-term multiplier matrix based on economic theory and specific institutional constraints (see Table 5):  Output does not affect public investment or recurrent expenditures instantaneously. This assumption may be seen as strong a priori because we use annual data and Tunisia does undertake revised budget laws from time to time (twice in the 1990s and 4 times since 8 2000). However, all revisions of the budget laws are so far driven by response to exogenous shocks: oil and commodity price shocks chiefly.

 Public investment cannot react instantaneously to recurrent expenditures and vice-versa.
In other words, once credits for public investment and recurrent expenditures are allocated, these are not changed within the fiscal year; Tunisia's virement rules provide a little flexibility within and between major recurrent budget heads (wages, goods and services, etc.) but no flexibility regarding reallocations between recurrent and investment budgets. 5  Recurrent expenditure does not affect private investment instantaneously even though its non-contemporaneous effect is expected to be positive. Indeed, maintenance of road, electricity, communication infrastructures for instance increases the marginal productivity of existing private inputs (both capital and labor), thereby lowering marginal production costs and increasing the level of private investment and production.
 Private investment does not affect public investment and recurrent expenditures instantaneously.  The multiplier is greater than 1 in the 6 th years following the 1 dinar shock to public investment.

Impact of public investment and recurrent expenditures on output
Although public investment is not only about infrastructure, this finding is consistent with the empirical literature that finds a positive impact of infrastructure spending flows on output, growth, or productivity (Agenor, Nabli and Youssef 2005). We also find a positive and consistent impact of public investment on private investment. Public investment in road, electricity, communication infrastructures, public capital in education, health, etc. can increase the marginal productivity of existing private inputs (both capital and labor), thereby lowering marginal production costs and increasing the level of private investment and production.
Interestingly, recurrent expenditure has a small but positive impact on output. As defined above, recurrent expenditures include salaries and treatment, public spending in goods and services and transfers and subsidies. The positive impact on output reflects the first two items. An increase in salaries and treatment leads to either immediate rise in private consumption or an increase in private savings, ultimately transformed into demand for investment goods. Either way, a rise in salaries and treatment affects output. Public spending in goods and services can also support in a straightforward fashion. Indeed, some components of current spending such as maintaining of roads and schools are important to maintain the quality of the services financed by public capital and contribute to ensuring the profitability of infrastructure investments. Spending on operation and maintenance that keeps existing infrastructure in good condition or spending that contributes to health outcomes and the accumulation of human capital-can promote growth more effectively than capital expenditure per se (Moreno-Dodson 2008). Finally, a reallocation of subsidies to public investment or recurrent spending is growth-enhancing, all things being equal.
There are not many studies of the impact of dynamic fiscal shocks to output on countries similar to Tunisia. The study by Restrepo and Rincon (2006) on Chile and Colombia is an exception.
These authors use an SVAR and SVEC model respectively to calculate spending multipliers for Chile and Colombia using quarterly data over 1989 through 2005. They find that a one Chilean peso increase in government spending has a transitory positive effect of 1.9 pesos on real GDP growth, which stabilizes at 1.37 cents. For Colombia, the authors find that a one peso increase in public spending translates, at impact, into a 0.12 increase in the level of GDP and stabilizes at 0.15 peso.
The size of fiscal multipliers depends, among other things, on leakages into saving and imports.
Multipliers tend to be smaller for economies more open to trade (i.e., higher leakages into imports) and more susceptible to financial market constraints (i.e., upward pressure on real interest rates), or when it is subject to monetary policy that offsets the fiscal stimulus (e.g., IMF, 2008). Many cross-country studies have found small fiscal multipliers and in some cases multipliers with a negative sign (Christiansen 2008). The most notable studies with "negative multipliers" are found in the literature on expansionary fiscal contraction initiated by Giavazzi and Pagano (1990) and surveyed in Hemming, Kell, and Mahfouz (2002).
Finally, a result that requires further investigation is the small impact of private investment on growth. Indeed, real private investment has a positive but smaller impact on GDP that public spending (Table 1). This finding is consistent the widely held view that private investment in the period of investigation went mostly to low value-added sectors.

Impulse responses
Figures 3 to 5 provides more details regarding the profile of the response of output to shocks to public investment, recurrent expenditure and private investment. They suggest that public investment is strongly growth-enhancing countercyclical instrument in the medium-term that can effectively counterbalance the shortfall in growth driven by the sharp decline in external demand. Figures 6-8 show the feedback effect of real GDP to the fiscal variables (starting t+1) and on private investment. Public investment follows a contra-cyclical pattern in the medium-term in contrast with recurrent expenditures that display a relative rigidity and insensitivity of output fluctuations. The contra-cyclicality of public investment seems at odd with what is found in most developing countries where a pro-cyclical bias is more prevalent (Gavin and Perotti 1997, Kaminsky, Reinhart and Vegh 2004and Calderon and Schmidt-Hebbel 2008. Indeed, public investment tends to increase in good times because higher growth leads to increased public revenues and greater ease in financing public investment. The relationship between real GDP and real investment, on the other hand, displays an "accelerator" model pattern in the medium term. In his "accelerator" hypothesis, Samuelson (1958) showed that higher expected demand (a proxy for higher GDP) stimulates investment and capital accumulation, further boosting future growth.  This result provides some confidence regarding the effectiveness of a fiscal spending stimulus in the case of Tunisia. Indeed, our results are based on historical data and on an average level of efficiency of public spending. If Tunisia speeds up procurement procedures and targets highgrowth-enhancing projects, our spending multipliers can be considered as a lower band, i.e., one dinar spent today could yield a higher additional output than in the past. In addition to a rapid execution, to maximize impact, the public spending should be well-targeted. We also find that reallocation of public spending is not growth neutral. Indeed, reducing fuel subsidies to increase public investment or even maintenance or other non-subsidies items can help boost growth. A finding that deserves further investigation however is the contra-cyclical nature of public investment. Does it reflect an explicit or implicit fiscal rule or does it indicate the existence of strong stabilizers?