WPS6562 Policy Research Working Paper 6562 Fiscal Rules and the Pro-Cyclicality of Public Investment in the West African Economic and Monetary Union Sébastien Dessus Jose Luis Diaz Sanchez Aristomene Varoudakis The World Bank Africa Region Poverty Reduction and Economic Management Department & Development Economics Vice Presidency Operations and Strategy Team August 2013 Policy Research Working Paper 6562 Abstract Evidence from a large panel of low-income and lower become pro-cyclical since the introduction of the fiscal middle-income countries over the period 1995–2012 convergence criteria in 1994. The pro-cyclicality of public suggests that, contrary to other countries, public expenditure and the high asymmetry of shocks that affect investment in the West African Economic and WAEMU countries justify exploring options for greater Monetary Union (WAEMU) has been pro-cyclical. counter-cyclicality of rules-based fiscal frameworks and Public investment contracts more in “bad times” for risk-sharing. than it increases in “good times” and appears to have This paper is a product of the Poverty Reduction and Economic Management Department, Africa Region; and the Operations and Strategy Team Development Economics Vice Presidency. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at sdessus@worldbank.org, jdiazsanchez@worldbank.org, and avaroudakis@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Fiscal Rules and the Pro-cyclicality of Public Investment in the West African Economic and Monetary Union Sébastien Dessus, Jose Luis Diaz Sanchez, and Aristomene Varoudakis* The World Bank Keywords: Asymmetric shocks; fiscal rules; pro-cyclicality; public investment; risk sharing. JEL Classification: E620, H620, H770 ACKNOWLEDGEMENTS The authors are thankful to Eleonora Mavroeidi for excellent research assistance, and to Sanjeev Gupta, Aart Kraay, Darryl McLeod, Herve Joly, Tito Cordella, and the participants in a World Bank seminar and in the University of Oxford’s Conference on Economic Development in Africa for their comments and suggestions. *Corresponding author: e-mail: avaroudakis@worldbank.org, Phone No: +1202 458 5447, Fax No: +1202 522 1158 1. INTRODUCTION The debt crisis in the Eurozone revealed new challenges for monetary unions. Countries entering a monetary union relinquish national monetary and exchange rate policies for the benefit of greater integration associated with the union. If countries in the union are subject to large asymmetric shocks, and there are no transfers through a federal budget, national fiscal policy would be the only instrument left to cushion these shocks. Yet, despite the need for fiscal flexibility, existing monetary unions observe strict fiscal rules that typically limit the leeway of national fiscal policies to respond to shocks. Concerns about debt externalities of national fiscal policies and possibly weak incentives for fiscal restraint provide a common rationale for fiscal rules in monetary unions (De Grauwe, 1992).1 However, fiscal rules may also reduce the quality of fiscal policy because they disregard the composition of fiscal adjustment necessary for compliance (Blanchard and Giavazzi, 2004). The need to comply with fiscal rules may result in easy cuts in capital spending. These can have two main effects: First, they may amplify volatility through pro-cyclical cuts in expenditure and, in particular, public investment. Second, they may have a potentially negative impact on long-term growth if the level of public investment, or its quality, are negatively affected. Thus, among the most important challenges for monetary unions is to improve their capacity to ensure fiscal convergence, while also developing efficient mechanisms to mitigate asymmetric shocks, which do not affect all members at the same time. In this paper we study the impact of fiscal rules on public expenditure, with emphasis on public investment, in the West African Economic and Monetary Union (WAEMU). Fiscal convergence rules in the WAEMU impose strict limits on the budget deficit. Current rules include a balanced basic budget deficit defined as domestic revenue minus domestically financed expenditure; a public debt ceiling at 70 percent of GDP; and the non-accumulation of public expenditure 2 arrears. The rules have not been systematically enforced in the past, especially concerning the basic fiscal balance. Even so, they still leave only limited room for counter-cyclical response in countries affected by asymmetric shocks. Counter-cyclical response through national budgets is bound by foreign financing, originating from outside of WAEMU, of any additional spending or of the budget deficit resulting from an adverse revenue shock. Moreover, a large part of current public expenditures are non-discretionary (wages, transfers, debt service) and thus difficult to cut in the short run. One can therefore anticipate that an unexpected change in revenue, or current expenditure, resulting from a shock, will be to a large extent cushioned by changes in discretionary public investment. Furthermore, unlike the relative ease in cutting investment expenditures in bad times, rapidly converting unexpected revenues into new projects is often difficult, reflecting, for example, bottlenecks in project selection or in procurement. It is therefore possible that volatility of public investment may well affect its average level too. 2 We extend previous empirical work on pro-cyclical fiscal policies in Sub-Saharan Africa by Thornton (2008); Lledo, Yackovlev and Gadene (2011); and Guillaumont-Jeanneney and Tapsoba (2011) who found that total public expenditure is more pro-cyclical in WAEMU than in other African countries. We examine separately the cyclical patterns of public investment and current public expenditure, comparing WAEMU to a large sample of low-income countries (LICs) and lower middle-income countries (LMICs) in Sub-Saharan Africa and in other developing regions. We compare patterns estimated over 1995-2012 with earlier patterns in 1981-1994 to gain insight on the possible impact of the fiscal convergence criteria adopted in 1994, after the devaluation of the CFA Franc, the WAEMU’s common currency. We also examine how these patterns differ in recessions and booms of economic activity. Our findings justify exploring options for greater counter-cyclicality of rules-based fiscal frameworks and 3 options for risk-sharing mechanisms with a view to reducing the pro-cyclicality of public investment. Risk-sharing mechanisms could include a move towards fiscal federalism through greater centralization of national budgets, the design of group insurance schemes, or both. The rest of the paper is organized as follows. Section 2 presents some stylized facts on GDP growth, public investment, and the nature of shocks in the WAEMU. Section 3 discusses the empirical findings on the cyclical patterns of public investment and current public expenditure in the WAEMU and other LICs and LMICs. Section 4 discusses policy implications for greater counter cyclicality of fiscal rules and risk sharing. Section 5 concludes. 2. STYLIZED FACTS: GDP GROWTH, PUBLIC INVESTMENT, AND ASYMMETRIC SHOCKS We consider 67 low-income (LIC) and lower middle-income (LMIC) countries, according to the World Bank’s classification, including the 8 WAEMU members, over the period 1995-2012. We compare WAEMU to two other country groups: 28 other Sub Saharan Africa LICs and LMICs (excluding WAEMU), and 31 other LICs and LMICs in the rest of the world (see Appendix Table A1 for group definitions). As Table 1 shows, at 3.6 percent on (un-weighted) average, annual real GDP growth in WAEMU ranks well below comparator groups. Growth volatility in WAEMU, as measured by the coefficient of variation of annual GDP growth, has been higher than in other LICs and LMICs, but lower than in other SSA comparator countries. Public investment in proportion to GDP has also been low in WAEMU. At 6 percent of GDP on average, it ranks well below the other two groups. Reflecting weak public investment, unsurprisingly perhaps, WAEMU lags behind the rest of Sub Saharan Africa as a whole on almost all infrastructure indicators, with most notable gaps in in paved road density, mainline 4 density, and generation capacity (IMF, 2010). Such infrastructure gaps could be partly responsible for WAEMU’s “missing growth”.3 Table 1: GDP growth and Public Investment, 1995-2012 (in %) WAEMU Other SSA Other LIC and LMIC LIC and LMIC (8 countries) (28 countries) (31 countries) GDP growth per year Average 3.56 4.12 4.34 Standard deviation 4.32 6.58 4.15 Standard deviation/Average 1.21 1.60 0.96 Public Investment/GDP Average 6.02 7.85 7.15 Standard deviation 2.64 6.20 9.48 Standard deviation/Average 0.44 0.79 1.32 Note: Based on the country groups in Appendix, Table A1. Standard deviations reflect variations both within and between countries. Source: Authors’ calculations based on data from World Economic Outlook, IMF. Turning to the degree of synchronization of shocks, as Table 2 shows, over 1995-2012, the average correlation of each WAEMU member country’s annual GDP growth with the un- weighted average annual GDP growth of other WAEMU members was 0.117. Burkina Faso and Togo had the highest GDP growth correlation (0.56 and 0.47, respectively) with average GDP growth of other WAEMU members, while all others had no significant correlations with the rest of WAEMU members. The low correlations of real GDP growth among WAEMU members contrast with the much higher correlations of GDP growth observed in monetary unions among advanced economies where economic integration is much higher. 4 The use of terms of trade changes as a more direct measure of exogenous shocks points to a similar conclusion (Table 2, last column). The average correlation of individual WAEMU members’ terms of trade changes with the rest of the WAEMU is 0.21, with only Benin, Burkina Faso and Mali exhibiting relatively high positive correlations with the rest of WAEMU members. We next consider how fiscal variables respond to shocks. 5 Table 2: Asymmetric shocks in WAEMU, 1995-2012 Comparison with other WAEMU members GDP growth (in %) Correlation of GDP Correlation of Terms Average St. Dev. growth of Trade changes Benin 4.1 1.0 0.031 0.440 Burkina Faso 5.9 2.0 0.556 0.551 Côte d'Ivoire 1.7 3.6 0.157 -0.120 Guinea-Bissau 1.1 8.9 -0.257 -0.204 Mali 4.2 3.7 0.193 0.556 Niger 4.9 4.6 -0.169 -0.023 Senegal 3.9 1.7 -0.044 0.307 Togo 2.4 2.7 0.469 0.172 Average 3.5 3.5 0.117 0.210 Source: Authors calculations based on data from World Economic Outlook, IMF. 3. THE PRO-CYCLICALITY OF FISCAL POLICY Evidence suggests that, contrary to high-income countries where fiscal policy is mostly uncorrelated with the business cycle, in developing countries fiscal policy is pro-cyclical: it turns expansionary in good times and contractionary in bad times (Talvi and Vegh, 2005). The pro- cyclicality of fiscal policy is often explained by the loss of international capital market access during bad times, which, in the absence of fiscal space through accumulated savings, makes it expensive, if not impossible, to finance expansionary policies during downturns (Aizenman et al, 2000; Gavin and Perotti, 1997).5 The pro-cyclicality of fiscal policy has been also documented in African countries: Government consumption has been found to be pro-cyclical, the more so when dependence on foreign aid is high (Thornton, 2008). Pro-cyclicality of total public expenditure has been also found by Lledo, Yackovlev and Gadene (2011), but with a mitigating impact of foreign aid and debt relief. Whether fiscal rules exacerbate pro-cyclicality of fiscal policy is largely an empirical matter. The outcome will depend on the design of rules and the incentives they create for policymakers. 6 Strict fiscal rules that target the overall fiscal balance on an annual basis may, arguably, amplify pro-cyclicality as shocks would trigger immediate expenditure and tax adjustments to meet the fiscal targets. By contrast, fiscal rules targeting the structural deficit or the deficit over the cycle could mitigate pro-cyclicality. Importantly, fiscal rules could mitigate pro-cyclicality if they change the incentives of policymakers toward creating fiscal space in good times for counter- cyclical response in bad times. There is factual evidence that policy incentives do change over time: Experience with credit rationing during bad times, especially after the East-Asian crisis in the late 1990s, prompted many developing countries to self-insure by building buffers of savings during good times. This made it possible for several emerging economies to respond counter- cyclically to the 2008-09 global financial crisis, including, to some extent, in Africa (Krumm and Kularatne, 2012). In what follows, we extend previous research by comparing WAEMU to a large sample of other low-income and lower middle-income countries and by analyzing the pro-cyclicality of the two components of public expenditure, public investment and current expenditure, but also by analyzing the pro-cyclicality of the fiscal balance.6 Data is taken from the World Economic Outlook and the IMF’s IFS database. Our analysis spans over 1995-2012, covering the period since the introduction of the fiscal convergence rules. We compare the estimated patterns over this period with the period 1981-1994, preceding the fiscal convergence rules. In addition, we examine more closely the pro-cyclicality patterns in recessions and economic booms. As we are interested in how shocks affect public investment and current expenditure in the short run, we regress the annual growth rate of public investment and current public expenditure in real terms on the annual real GDP growth rate, without considering other potential determinants of these fiscal variables over the medium term (such as, for example, the level of public debt or 7 of foreign aid).7 We include country specific effects, to account for time-invariant country specificities, and time-specific effects, to capture the impact of possible symmetric shocks. The specification also includes a dummy for the years of armed conflict since these events could cause abrupt changes in both economic structure and in fiscal policy. As Blattman and Miguel (2010), among others, we use the UCDP/PRIO data on armed conflict first presented in Gleditsch et al. (2002) and updated in Themnér and Wallensteen (2014).8 Admittedly, a positive relationship between government expenditures and GDP growth could reflect not only a pro-cyclical behavior of fiscal variables but also an increase in infrastructure and/or social services demand resulting from a higher level of income.. To account for this possible simultaneity between GDP growth and the fiscal variables –-we use a Generalized Method of Moments (GMM) estimator with standard errors robust for both heteroskedasticity and autocorrelation. Lags of the independent variable (the real GDP growth) are used as instruments, together with an additional instrument constructed as the product of world GDP growth and each country’s ratio of exports to GDP.9 The over-identifying restrictions are tested through the Hansen-Sargan J test, which provides a test of the general validity of the instruments used. Regression coefficients for the WAEMU countries are estimated separately from the other countries of the sample. Results for the period 1995-2012 are reported in Table 3. The results confirm the validity of the over-identifying restrictions and thus of the instruments used.10 Public investment has been pro-cyclical in WAEMU over the estimation period, as the estimated elasticity to real GDP growth is positive at a 95% confidence level. By contrast, there is no significant pro-cyclicality of public investment in other LICs and LMICs (Table 3, column 1). The results for current public expenditure show an a-cyclical behavior of this fiscal policy variable for the two groups of countries (Table 3, column 2). Pro-cyclical public expenditure is 8 thus associated with public investment, rather than current expenditure, in the WAEMU. This supports the perception that public investment, more than current expenditure, is a major shock absorber, or residual fiscal variable. As to the fiscal balance, there is evidence of counter- cyclicality in other LICs and LMICs as the fiscal deficit (surplus) decreases (increases) when growth is stronger (Table 3, column 3). By contrast, in WAEMU the fiscal balance is a-cyclical. The absence of counter-cyclicality of fiscal deficits in WAEMU may reflect the large compensating changes in public investment when fiscal revenues are affected by shocks: In bad (good) times, when fiscal revenues shrink (expand), a contraction (increase) of public investment offsets the impact of the shock on the budget, resulting in only small changes in the fiscal deficit in proportion to GDP. Finally, it is worth noting that our dummy for armed conflicts does not reach statistical significance.11 Table 3: The Pro-cyclicality of Fiscal Policy (1995-2012) (1) (2) (3) Dependent variable DLNKFIG DLNKCURX DEF/NGDP W*DLNKGDP 7.213** -0.0386 -0.0152 (2.072) (-0.0382) (-0.0185) (1-W)*DLNKGDP 1.409 1.225 0.590*** (0.502) (1.522) (2.643) AC dummy 4.609 0.459 0.0034 (1.126) (1.535) (0.0301) Observations 1,086 1,108 1,147 Number of countries 67 78 78 Hansen J test: Statistic 2.958 4.609 6.030 Chi- sq (6), P-value 0.8140 0.5949 0.4199 Note. D and LN denote the first difference operator and the natural logarithm operator respectively. KFIG: real public investment; KCURX: current real public expenditure (total expenditure excluding public investment); DEF: fiscal balance; NGDP: nominal GDP; KGDP: real GDP; W: dummy variable for WAEMU countries. We convert nominal variables into real variables using the GDP deflator. AC is a dummy with value 1 for armed conflict episodes in a given year and 0 otherwise. Robust z-statistics are in parentheses. ***, **,* indicate 9 significance at 1%, 5%, and 10%, confidence levels respectively. The method of estimation is GMM with standard errors and statistics robust to both arbitrary heteroskedasticity and arbitrary autocorrelation. Instruments include: 1, 2, 3, and 4 lags of the two independent variables and the ratio of exports to GDP for each country multiplied by the world GDP growth. Accordingly, there are 6 over-identifying restrictions, equal to the total number of instruments minus the number of regressors. Each regression includes country fixed effects and time fixed effects. The null hypothesis of the Hansen J test (overidentification test) is that all moment conditions are valid, i.e. the instruments used are not correlated with the residuals. Estimating the same set of regressions over the period 1981-1994, preceding the introduction of the fiscal convergence criteria in WAEMU, provides evidence on the possible impact of this fiscal framework on the cyclical patterns of public investment and current expenditures. As Table 4 shows, public investment exhibits a similar pattern of pro-cyclicality in other LICs and LMICs in 1981-1994 (Table 4, column 1). However, in WAEMU, contrary to the more recent period 1995-2012, there is no evidence of pro-cyclicality of public investment over the period 1981-1994, while current public expenditure was also a-cyclical in the earlier period as in the more recent period (Table 4, column 2). This confirms the perception that, since the introduction of the fiscal convergence framework, public investment, more than current expenditure, has responded pro-cyclically in the face of shocks that affect the budget in WAEMU countries. Here again, the dummy for armed conflicts does not seem to play a role in the pro-cyclicality of the fiscal variables. Table 4: The Pro-cyclicality of Fiscal Policy (1981-1994) (1) (2) (3) Dependent variable DLNKFIG DLNKCURX DEF/NGDP W*DLNKGDP 0.230 2.587 -0.450* (0.111) (1.162) (-1.810) (1-W)*DLNKGDP 0.946 1.616** 0.209 (0.385) (2.140) (0.751) AC dummy 0.261 0.272 -0.0664 (0.490) (1.082) (-0.664) Observations 432 207 229 Number of countries 48 32 34 Hansen J test: 10 Statistic 6.117 3.029 11.521 Chi- sq (6), P-value 0.4102 0.8052 0.0735 Note. D and LN denote the first difference operator and the natural logarithm operator respectively. KFIG: real public investment; KCURX: current real public expenditure (total expenditure excluding public investment); DEF: fiscal balance; NGDP: nominal GDP; KGDP: real GDP; W: dummy variable for WAEMU countries. We convert nominal variables into real variables using the GDP deflator. AC is a dummy with value 1 for armed conflict episodes in a given year and 0 otherwise. Robust z-statistics are in parentheses. **,* indicate significance at 5%, and 10% confidence levels respectively. The method of estimation is GMM with standard errors and statistics robust to both arbitrary heteroskedasticity and arbitrary autocorrelation. Instruments include: 1, 2, 3, and 4 lags of the two independent variables and the ratio of exports to GDP for each country multiplied by the world GDP growth. Accordingly, there are 6 over-identifying restrictions, equal to the total number of instruments minus the number of regressors. Each regression includes country fixed effects and time fixed effects. The null hypothesis of the Hansen J test (overidentification test) is that all moment conditions are valid, i.e. the instruments used are not correlated with the residuals. The pro-cyclical changes in fiscal policy in developing countries have often been found to be asymmetric in good and bad times. For example, Gavin and Perotti (1997) found that fiscal balances in Latin America were more pro-cyclical in bad times, when negative deviations of GDP growth from average were large. In WAEMU, Guillaumont-Jeanneney and Tapsoba (2011) found total public expenditure to be more pro-cyclical in recessions than in good times. In our larger sample we further investigate this issue by examining whether the elasticity of public investment to GDP is different when countries face negative and positive shocks. For this exercise we concentrate on public investment, which, according to our results, was found to be pro-cyclical in the WAEMU over the recent period 1995-2012. For each country we identify periods of negative shocks as years with below-average real GDP growth (over 1995-2012) and periods of positive shocks as years with above-average real GDP growth. Regressions of real public investment growth on real GDP growth are estimated separately on periods of negative and positive shocks, while distinguishing the coefficients for WAEMU and non-WAEMU countries. We use a GMM estimator with robust standard errors for heteroskedasticity and autocorrelation, using the same instruments as for the previous regressions. Results are reported in Table 5. 11 Table 5: The pro-cyclicality of public expenditure in “good” and “bad” times (1995-2012) GDP growth>Avg GDP growth