Policy Research Working Paper 8963 Why Some Countries Can Escape the Fiscal Pro-Cyclicality Trap and Others Cannot? Santiago Herrera Wilfred A. Kouame Pierre Mandon Macroeconomics, Trade and Investment Global Practice August 2019 Policy Research Working Paper 8963 Abstract This paper analyzes the procyclicality of fiscal policy on with procyclicality. The findings also show that the quality the tax and spending sides in a sample of 116 developing of fiscal institutions is associated with procyclicality; coun- countries between 2000 and 2016. About 20 percent of the tries with fiscal rules have smaller procyclical bias, but the countries in the sample switched from procyclical to coun- effect is not homogeneous; and higher degrees of expendi- tercyclical policy stance. In Sub-Saharan Africa, 30 of 39 ture rigidity are associated with lower procyclical bias. The countries remained caught in the procyclicality trap and the study finds asymmetric policy stances along the business region has the highest degree of procyclicality. The Middle cycle, with procyclicality being more pronounced during East and North Africa region switched from a countercycli- recessions. Similarly, the political cycle affects procyclicality, cal policy stance to a procyclical one over time. The Europe as procyclical bias increases in electoral years. From the tax and Central Asia and Latin America and the Caribbean management perspective, procyclical bias is still present, but regions significantly reduced the degree of procyclicality. there are significant changes: most of the political economy The main economic variables that affect procyclicality are variables lose significance; the resource-dependence variable financial depth, tax base variability, and natural resource is not significant; external credit availability reduces procy- dependence. In line with the political economy literature, clicality; tax base variability increases procyclical bias; and the perception of corruption, social fragmentation, and expenditure rigidity is no longer significant, but fiscal space inequality in resource distribution are positively associated becomes determinant of procyclical bias. This paper is a product of the Macroeconomics, Trade and Investment Global Practice. 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://www.worldbank.org/prwp. The authors may be contacted at sherrera@worldbank.org, wkouame@worldbank.org, and pierr.mandon@gmail.com. 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 Why Some Countries Can Escape the Fiscal Pro-Cyclicality Trap and Others Cannot? Santiago Herrera1, Wilfred A. Kouame, and Pierre Mandon Keywords: Procyclicality, public spending, tax rate, rigidity JEL codes: E62, E32, E02, F41, Q32 1 This paper is part of the LAC Chief Economist Regional Study on Budget Rigidities in Latin America. A preliminary version was presented in the ERF workshop in Cairo, in February 2018, and authors thank comments by Raimundo Sojo. Authors are grateful to Carlos Vegh and Guillermo Vuletin their insightful comments.         I. Introduction This paper examines the major determinants of fiscal procyclicality identified in previous literature (Frankel et. al., 2013; Vegh and Vuletin, 2015; and Aizenman et al., 2019), expanding the database, including new variables, namely the fiscal space and degree of rigidity of spending, and measuring procyclicality of tax rates (Aizenman et al., 2019), in addition to spending. Discretionary fiscal policy can help stabilize the economy, but there is evidence that in some developing countries fiscal policy may play a destabilizing role (Suescun, 2007) and in developed economies, the stabilizing properties have been questioned (Debrun et al., 2008). The procyclical nature of fiscal policy, by which countries use contractionary policy during recessions and expand while in booms, has been associated with more volatility and has been studied for decades. Still, we observe procyclical fiscal behavior, more common in developing economies than in developed ones, more prevalent in some regions, such as Africa, than in others, and countries move from procyclical to countercyclical behavior over time. This paper describes procyclical behavior in a large set of countries, along the lines of previous literature (Frankel et. al., 2013) with several variations: 1) Focusing exclusively on developing countries; 2) expanding the time period to 2016 to include the fiscal expansion during the Great Recession and lack of adjustment in the growth recovery phase; 3) including more specific variables to examine the quality of institutions to provide more detailed discussion; 4) Adding a set of explanatory variables to capture social fragmentation or polarization which may complement the traditional macroeconomic explanations such as borrowing constraints or volatility of revenues (Ilzetzki, 2011). 4) focusing on regional differences, between Africa, Latin America and the Middle East where the more salient features and changes in procyclicality take place; and 5) accounting for effects of expenditure rigidity and fiscal space. The literature from the last two decades provides significant insights regarding procyclical fiscal policy, with most of the papers contrasting results between the developed economies and the developing world. The first generation of papers highlighted the lack of access to credit and international capital markets to explain such a behavior, with the seminal contribution of Gavin and Perotti (1997) focusing on Latin American countries. The second generation of papers, illustrated by Talvi and Vegh (2005) examined the role of political distortions, and particularly political polarization, to explain the “voracity effects” on the budget during booms. This explanation was found to be the most relevant empirically (Ilzetzki, 2011). Alesina et al. (2008) and showed that higher (perceived) levels of corruption (especially with a lack of fiscal transparency) led to a rational decision of the voters to “starve the Leviathan”, i.e., to reduce political rents by optimally demanding more public goods (and/or lower taxes) during booms. More recent developments in the literature have examined the resource-led boom of many developed economies and concluded that procyclical behavior was stronger in resource dependent nations (Arezki 2              and Bruckner, 2012), with further examination and focus on Sub-Saharan Africa (SSA) (Konuki and Villafuerte, 2016). The paper has seven sections. The next section describes the data and stylized facts of procyclicality over the period 2000-2016. Section III presents the empirical strategy. Section IV discusses the findings of procyclicality both on the taxation and spending sides. Section V presents a discussion on policy options to mitigate procyclicality, with a focus on the effect of fiscal institutions and fiscal rules on procyclicality. Section VI explains the cross-regional differences in procyclicality. Finally, Section VII presents concluding remarks. II. Data sets and stylized facts To examine the correlation between a country’s fiscal policy stance and the business cycle, the standard procedure is to regress the cyclical real GDP on the cyclical real primary general government expenditures.2 We run equation (1) below on a sample of 116 developing countries for the 2000-2016 period.3 . . , (1) where α refers to the intercept, and subscripts, i and t stand for country and year. and are country and year fixed effects (regional fixed effects are absorbed by country fixed effects). β, the coefficient of interest, captures the variation of cyclical spending (in local currency) due to changes in the cyclical GDP by x units of local currency. Later sections explore cyclicality in the tax rates as in Aizenman et al. (2019), using the Vegh and Vuletin (2015) database on tax rates. To examine how fiscal policy management evolved over time, we split the period into two subperiods: 2000-2008 (pre-global recession) and 2009-2016 (post-global recession) and compared how each country’s procyclicality coefficient changed over time. In the first subperiod, 64% of the countries had a procyclical fiscal stance, while in the second one the percentage fell to 60% (Figure 1).4 Hence, only 40% of developing countries ran countercyclical fiscal policy in the second subperiod, but in SSA the proportion was only 20% (8 of 38) (Fig. 2). In LAC, 50% of the countries ran countercyclical fiscal                                                              2 Alesina et al. (2008), Frankel et al. (2013). The cyclical components of GDP and expenditures are estimated using the Hodrick-Prescott (HP) filter with a smoothing parameter of 6.25, as done by Ravn and Uhlig (2002). Though correlation cannot be interpreted as causation, Ilzetsky and Vegh (2008) show that output causes government spending when properly instrumented. Konuki and Villafuerte (2016) also conclude that output shocks drive fiscal policy. 3 See Appendix A for a detailed description of the data set and the sample selection process. To make results comparable with previous literature, Frenkel, et al. (2013), Vegh and Vuletin (2015), and Aizenman et al., (2019). 4 Figures 1 and 2 replicate the format employed by Frankel et al. (2013). 3              policy. Over time, SSA countries did not change their procyclical stance, while ECA and LAC reduced their procyclical stance, and MENA switched from a countercyclical to a procyclical stance, while South Asia and East Asia maintained their countercyclical stance on average (Table 1). In the remainder of the paper, we seek to explain why procyclicality is more prevalent in some countries and regions than in others and why some countries can switch from procyclical to countercyclical fiscal policy management. Table 1: Evolution of (pro)cyclicality at the regional level, over time # Code Region Corr(G, GDP) 2000-2008 Corr(G, GDP) 2009-2016 1 ECA Europe and Central Asia 0.48 0.05 2 LAC Latin America and Caribbean 0.20 0.09 3 MENA Middle East and North Africa -0.23 0.12 4 SEAP South, East Asia, and Pacific -0.07 -0.02 5 SSA Sub-Saharan Africa 0.23 0.22 Figure 1: Progress in Fiscal Policy Management 2009-2016 (In red, resource-rich countries) A. Entire sample Back to procyclicality Still procyclical 1 CAF ECU LBY CIV NIC UKR GHA IRN SDN BOL IDN IND BRA HUN SWZ VEN YEM MLI MOZ UGA OMN UZB ERI MDG 0.50 URY TCD BFA GAB AGOROM BDI CZE MNG BEN THA POL HTI NER PAN HRVMDA BIH ARG BGD Corr(G, GDP) 09-16 NAM COL ETH ZMB NGA KAZ ISR TZA JAM TUR PAK JOR MUS SLE AZE MMR MKD SEN NPL TGO CMR DZA COGPHL GNB KEN MAR 0 EST BGR MEX GMB ALB SAU TJK VNM ZAR HND LBR LKA LAO LBN PER RUS SLV GTM LVA CHN TUN DOM KHM SVK TTO LTU PNG KGZ EGY -0.50 SVN RWA BLR LSO KWT MWI CRI MYSGIN ARE SRB GEO BWA PRY SGP ZAF CHL KOR -1 Established countercyclical Recent countercyclical -1 -0.50 0 0.50 1 Corr(G, GDP) 00-08 4              B. Latin America and the Caribbean (LAC) Back to procyclicality Still procyclical 1 ECU NIC BOL BRA VEN 0.50 URY PAN ARG HTI Corr(G, GDP) 09-16 COL JAM 0 MEX HND PER SLV GTM DOM TTO -0.50 CRI PRY CHL -1 Established countercyclical Recent countercyclical -1 -0.50 0 0.50 1 Corr(G, GDP) 00-08 C. Middle East and North Africa (MENA) Back to procyclicality Still procyclical 1 LBY IRN YEM OMN 0.50 Corr(G, GDP) 09-16 ISR JOR DZA MAR 0 SAU LBN TUN EGY -0.50 KWT ARE -1 Established countercyclical Recent countercyclical -1 -0.50 0 0.50 1 Corr(G, GDP) 00-08 5              D. Sub-Saharan Africa (SSA) Back to procyclicality Still procyclical 1 CAF GHA SDN CIV SWZ MLI MOZ UGA ERI MDG 0.50 TCD BFA GAB AGO BDI BEN NER Corr(G, GDP) 09-16 NAM ETH ZMB NGA TZA MUS SLE TGO SEN CMR COG GNB KEN 0 GMB ZAR LBR -0.50 RWA LSO MWI GIN BWA ZAF -1 Established countercyclical Recent countercyclical -1 -0.50 0 0.50 1 Corr(G, GDP) 00-08 E. Eastern and Central Europe (ECA) Back to procyclicality Still procyclical 1 UKR HUN UZB 0.50 ROM CZE HRVMDA BIH POL Corr(G, GDP) 09-16 KAZ TUR AZE MKD 0 EST BGR TJK ALB RUS LVA LTU KGZ SVK -0.50 SVN BLR SRB GEO -1 Established countercyclical Recent countercyclical -1 -0.50 0 0.50 1 Corr(G, GDP) 00-08 6              F. South, East Asia and Pacific (SEAP) Back to procyclicality Still procyclical 1 IDN IND 0.50 MNG THA BGD Corr(G, GDP) 09-16 PAK MMR NPL PHL 0 LKA VNM LAO CHN KHM PNG -0.50 MYS SGP KOR -1 Established countercyclical Recent countercyclical -1 -0.50 0 0.50 1 Corr(G, GDP) 00-08 III. Empirical strategy To analyze the determinants of fiscal procyclicality both on the tax and spending sides, regression (1) is expanded to include, as covariates, interacting terms of the output gap with the conditioning explanatory variables, following Alesina et al. (2008) and Frankel et al. (2013). Eq. (2) captures the effects of conditional factors on procyclicality as follows5: . . . (2) The variable refers to conditional effects of procyclicality and capture the non-linearity of the cyclical GDP on spending. We check for the following possible conditioning factors considered in the literature. Financial depth effects. The first generation of papers on the determinants of procyclicality focused on the role of liquidity constraints as the main explanatory factor. Gavin and Perotti (1997) and                                                              5 The cyclical component of expenditure is measured as in Frankel, et al. (2013), while the cyclicality of tax rates is based on the Vegh et. al. database and the method presented by Aizenman et al. (2019). 7              Mendoza and Oviedo (2006) are classic examples of this literature. Along these lines, several measures of financial depth or access to international credit markets are used. Financial depth is proxied by the ratio of credit to the private sector over GDP and capital openness with the Kaopen index by Chinn and Ito (2006), Tax base variability effects. Some of the pioneer studies on procyclicality stressed how the volatility of the revenue base, in the presence of political constraints of generating higher surpluses during boom periods, lead to borrowing less during recessions to maintain intertemporal budget constraints, and hence, configuring the procyclical nature of policy (Talvi and Vegh, 2005). Rigidity of spending. This paper defines rigidity of spending as the relative importance of factors explaining the current levels of spending. Rigid components of public expenditure are difficult to change due mainly to the associated costs. The rigidity of public spending is measured as the sum of the components of public expenditure in wages, pensions, and interest payment as a percentage of total expenditure. The expected sign is ambiguous as the less flexible spending is, the less it can be used as a countercyclical tool. But a higher degree of structural spending will stabilize spending and hence will not increase due to upturns in the cycle, which are temporary. Quality of institutions. Papers in the political economy literature examining factors that determine fiscal outcomes emphasized the role of perceived corruption in a system of low transparency and limited fiscal monitoring leading rational voters to “starve the Leviathan” in order to mitigate the principal- agent problem (Alesina et al., 2008). Hence, during booms, the rational voter would demand more public goods and services, leading to the procyclical bias. Besides perceived corruption, other institutional factors may affect procyclical bias. Some authors consider averages of several indicators (Frankel et al., 2013), which has the advantage of parsimony, though at the cost of reduced granularity for policy implications. Here we include fiscal institutions and examine the role of rigidities in the budget as well as that of fiscal rules and fiscal councils on procyclicality. Level playing field. A branch of the literature examines the role of political and social fragmentation or the poor rule of law and, more generally, unequal distribution of resources within societies, which may lead to “voracity effects” on the budget and accentuate the common pool problem, and hence the procyclical bias (Ilzetzki, 2011). Natural resources. As highlighted by the natural resource curse literature, natural resources imply large public rents originated without taxation and are subject to potential elite capture problem. Hence, it is 8              expected that the procyclicality would be even higher in resource-dependent and resources-rich countries due to exacerbated political distortions. Other potential factors. We examine whether the procyclical behavior is symmetric along the business cycle, i.e., if governments increase spending during recessions but do not cut it during booms. Hence, we explore whether there is an asymmetry along the business cycle. We also examine the impact of political cycles on the fiscal policy reaction function. Another factor which may be associated with procyclicality of fiscal management is the existence of buffers to accommodate shocks, such as the level of public debt or the stock of international reserves. Higher public debt reduces the fiscal space to absorb shock and hence would lead to a more procyclical policy (Frankel et al., 2013). IV. Estimation results A. Cyclicality of public spending Equation 2 is estimated in multiple stages to examine the role of each factor separately (Tables 1 and 2). The main results to be highlighted are as follows. Without conditioning for any factor, fiscal policy is procyclical (column (3)).6 The degree of procyclicality decreases with the level of development (column 4).7 The effect remains large even after introducing conditioning variables. The sign of the coefficient changes when results are conditioned on the perception of corruption variables, but when the average values of those variables are used, then the total effect of the business cycle is about the same order of magnitude, between 0.4 and 0.6. Below we discuss the impact of the different conditioning factors in the sequential stages summarized in Tables 3, 4, and 5. The tax base variability has the expected positive sign (Column (4)), as predicted by the literature.8 Talvi and Vegh (2005) show that, in the presence of political distortions that make it costly to generate budget surpluses, tax base fluctuations will lead to procyclical policy. We analyze the role of political economy variables later in this section. The hypothesis of credit constraints being associated with procyclicality cannot be rejected when using the domestic financial depth: larger ratios of credit to the private sector to GDP are associated with lower procyclicality (column 5). At the mean level of the ratio, 35 percent, the total effect of the business cycle is 0.62, which is procyclical. But in countries where the credit ratio is higher, for instance close                                                              6 All regressions control for a set of missing observations as described in Appendix A. 7 The level of development is a discrete categorical variable with values from 1 to 4, with 1 assigned to the least developed group and 4 to the most developed. 8 Variability was measured as the absolute deviation from the mean for each year. Given limited degrees of freedom, it was impossible to calculate other statistics. 9              to the maximum level (160 percent), fiscal policy is countercyclical, and the reverse happens in countries close to the minimum. The threshold level for switching from procyclical to countercyclical is 68%, which is significantly higher than the median of 21% for developing countries. The external credit availability, measured by the Chinn and Ito (2006) capital openness variable (column 6) was not significant, similar to previous findings for African countries (Konuki and Villafuerte, 2016), but different from studies with broader samples which include developed economies (Frankel et al., 2013). The role of the quality of institutions is proxied by two variables which capture the impact of the (perceived) corruption in both the political sphere and in the public services (columns 7 and 8). Higher indices of perceived corruption are associated with higher degrees of procyclicality, in line with previous literature (Alesina et al., 2008). Fiscal policy is procyclical when the perceived corruption indexes exceed .53 (in the political sphere) and .66 (in the public services) threshold levels, as summarized in Table 2. Other papers that explore the role of quality of institutions aggregate different variables to measure the quality of institutions, which has the benefit of parsimony in the analysis but does not allow granularity (Frankel et al., 2013). We also examine the role of institutions by including variables related to the rule of law, and the sign is the expected one (Table 3, column 9), with countries with higher scores in law and order having lower procyclicality bias. Table 2: Thresholds to reach a countercyclical policy, whole sample (2000-2016) Conditional effects Threshold level Median in sample Development level 3.47 2.00 Revenue.ratio (%) 22.84% 23.64% Credit.ratio (%) 82.31% 25.99% Pol.corruption (0-1) 0.53 0.37 Pub.service.corruption (0-1) 0.66 0.71 Law and order (1-6) 4.94 3.00 Resource.distribution (0-1) 0.80 0.59 Ethnicandrel.stability (0-6) 2.55 4.25 Notes: Median values are computed on the sample of the respective regression, not the whole sample of 116 countries. 10              Table 3: Fiscal (pro)cyclicality of expenditure, whole sample (2000-2016): Impacts of development level, tax base, credit constraints, corruption, rule of law and order (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Dependent variable: Cyclicality of real general government primary spending Cyclicality of real GDP 0.425** 2.124*** 0.107 0.279*** 1.070*** -0.089 -1.320** -0.829** 1.730*** -0.114 (0.199) (0.706) (0.378) (0.092) (0.313) (0.310) (0.653) (0.400) (0.118) (0.414) Development level Crgdp*dvpt.level (polytomic) -0.612*** (0.229) Tax base effect Crgdp*tax 0.041 (0.057) Crgdp*tax.base.variability 0.091*** (0.002) Credit constraints Crgdp*credit.ratio -0.013*** (0.004) Crgdp*kaopen 1.178 (0.920) Perceived corruption Crgdp*pol.corruption 2.808** (1.106) Crgdp*pub.service.corruption 2.452*** (0.927) Rule of law and order Crgdp*law.order -0.350*** (0.042) Crgdp*rule.of.law 1.213 (1.153) Constant -377.001 152.592 -1,680.682 -1,180.377 -66.903 -8.790 -1,044.820 -1,185.272 575.096 -158.364 (340.819) (518.802) (1,848.808) (956.057) (461.523) (582.628) (1,388.965) (1,020.610) (859.336) (467.021) Adjusted R-squared 0.067 0.135 0.083 0.114 0.148 0.105 0.150 0.157 0.122 0.108 Rmse 5,531 5,324 9,347 9,188 5,339 5,607 5,497 5,437 5,823 5,629 Joint significance (p-value) - 0.000 0.074 0.000 0.003 0.000 0.001 0.000 0.000 0.000 CD test (p-value) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Order of integration I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0)/I(1) I(0) Cluster country country country country country country country country country country Covariates Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 1,954 1,954 663 663 1,802 1,707 1,639 1,639 1,569 1,639 # countries 116 116 42 42 115 115 112 112 99 112 Notes. Standard errors (in parentheses) clustered at the country level. All interactive variables are also introduced in level but not reported to save space. All regressions include the volatility of the terms of trade, BD for missing data in real GDP and real expenditure series, BD for five years HP cycle correction. See Table A.1 in appendix A for further discussion on variables definition. *** p<0.01, ** p<0.05, * p<0.1. 11              We examine the role of variables derived from the political economy literature such as the degree of social fragmentation (Ilzetzki, 2011) or the absence of a level playing field, such as the power distribution among social groups, the distribution of resources within societies, and the level of ethnic- religious tensions (Table 4, columns 1 to 3). All these variables are considered as a proxy to the social polarization, to capture the impact of the “voracity effect” on the budget. Indeed, social polarization is mostly captured by ethnic/linguistic/religious polarization indicators. We find that more equal distribution of resources and more ethnic-religious stability are associated with less procyclicality. One of the more critical determinants of procyclicality is whether the country is resource-dependent or resource-rich (Table 4, columns 4, 5 and 6).9 In all cases, we find a strong procyclical bias in resources- dependent and resources-rich countries, suggesting a procyclical resource curse, in line with the recent literature (Arezki and Bruckner, 2012; Konuki and Villafuerte, 2016). One of the more interesting results is that procyclicality is arising from resource-dependence increases over time. We also found a more significant procyclicality bias during election years (Column 7 of Table 4), though the effect appears decreasing over time. This effect is more pronounced in AFR than in LAC or MENA. We find no impact of the debt level on the procyclicality l bias, contrary to findings by Frankel et al. (2013) but like Konuki and Villafuerte (2016) for Africa. To examine the correlation of procyclicality and fiscal institutions, we include a measure of the rigidity of the budget and variables that describe fiscal rules. The rigidity of spending variable dampens the procyclical bias. The rigid component of public expenditure reduces the procyclical bias (Table 4, column 9). But the effect is asymmetric over the business cycle: the mitigation of procyclicality through spending rigidity is larger during the upturns of the cycle (column 2 of Table 5). Finally, we found no evidence that fiscal space affects the procyclicality of public spending (Table 4, column 10). However, exploring potential heterogeneity over the period, we find that the fiscal space variable affects procyclicality, with lower fiscal space inducing higher procyclicality, but only when the sample is broken to include the most recent years 2009-2016.10 Finally, we find evidence of heterogeneous procyclical behavior along the cycle: at first, we find no direct evidence of the asymmetry (Table 5, Column 1), but when the rigidity and fiscal space variables                                                              9 Resources-dependent countries are countries for which commodity revenues (GGRC in WEO data set) account for at least 10 percent of total revenues minus grants (GGR and GGRG series in WEO data set) at least 50 percent of the time over the considered period. This definition is derived from Konuki and Villafuerte (2016) and enables us to capture countries with a resource-dependent revenue structure. For more details, see Table E.1. Resources-rich countries are defined by the IMF (2012), here is a direct access to the report (https://www.imf.org/external/np/pp/eng/2012/082412.pdf). According to this definition, we capture countries where production on natural resources reached more than 20% of exports. For more details, see Table A.2. 10 The findings are not presented in the paper but available upon request.   12              are included, the asymmetric impact becomes evident (Table 5, columns, 2 and 3).11 However, as discussed above, the level rigidity allows mitigating the procyclicality during the upturns of the business cycle.                                                              11The “good times” dummy was constructed as 1 when the cyclical component of GDP was above the potential level, and zero otherwise. 13              Table 4: Fiscal (pro)cyclicality of expenditure, whole sample (2000-2016): Impacts of the level playing field, natural resources, and other factors (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Dependent variable: Cyclicality of real general government primary spending Cyclicality of real GDP 0.496*** 1.557*** -0.159** -0.154** -0.159** 0.385** 0.393* 1.143** 0.473* 0.473* (0.174) (0.453) (0.073) (0.075) (0.073) (0.191) (0.232) (0.459) (0.268) (0.268) Level playing field Crgdp*power.distribution 0.364 (0.326) Crgdp*resource.distribution -2.291*** (0.624) Crgdp*ethnicandrel.stability -0.416*** (0.152) Natural resources Crgdp*resource.dependant (bd) 0.610** (0.243) Crgdp*resource.rich (bd) 0.605** (0.240) Crgdp*resource dependant and rich 0.610** (0.244) Other factors Crgdp*elections 0.161** (0.066) Crgdp*debt.ratio 0.001 (0.003) Crgdp*rigidity -0.037** (0.015) Crgdp*fiscal space -0.000 (0.000) Constant -269.523 -270.481 482.975 -330.074 -292.797 -298.574 -413.932 -434.781 537.497 -540.189 (502.504) (436.305) (807.001) (367.332) (328.276) (327.165) (355.390) (283.039) (376.005) (420.066) Adjusted R-squared 0.113 0.094 0.185 0.078 0.078 0.078 0.069 0.065 0.004 0.066 Rmse 5,615 5,672 5,610 5,497 5,498 5,497 5,525 5,593 824 5,632 Joint significance (p-value) 0.00 0.001 0.000 0.029 0.029 0.029 0.004 0.009 0.015 0.000 CD test (p-value) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Order of integration I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) Cluster country country country country country country country country country country Covariates Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 1,639 1,639 1,569 1,954 1,954 1,954 1,954 1,913 871 1,886 # countries 112 112 99 116 116 116 116 115 54 113 Notes. Standard errors (in parentheses) clustered at the country level. All interactive variables are also introduced in level but not reported to save space. All regressions include the volatility of the terms of trade, BD for missing data in real GDP and real expenditure series, BD for five years HP cycle correction. See Table A.1 in appendix for further discussion on variables definition. *** p<0.01, ** p<0.05, * p<0.1. 14              Table 5: Fiscal (pro)cyclicality of expenditure, whole sample (2000-2016): Impacts of rigidity, fiscal space and business cycle (1) (2) (3) Dependent variable: Cyclicality of real general government primary spending Cyclicality of real GDP 0.611*** -0.407 0.402 (0.028) (1.059) (0.376) Business cycle Crgdp*economic upturn -0.230 2.763*** 2.959* (0.299) (0.817) (1.549) Crgdp*economic upturn*rigidity -0.047*** (0.017) Crgdp*rigidity 0.005 (0.019) Rigidity of public spending -9.681 (6.120) Crgdp*economic upturn*fiscal space -0.001 (0.002) Fiscal space 0.150 (0.164) Crgdp*fiscal space 0.001 (0.002) Economic upturn 29.722 66.490* 21.955 (226.974) (34.059) (220.581) Constant -468.287 464.121 -453.536 (569.519) (371.214) (486.993) Adjusted R-squared 0.071 0.013 0.078 Rmse 5,518 820,2 5,597 Joint significance (p-value) 0.00 0.002 0.000 CD test (p-value) 0.000 0.000 0.000 Order of integration I(0) I(0) I(0) Cluster country country country Covariates Yes Yes Yes Observations 1,954 871 1,886 # countries 116 54 113 Notes. Standard errors (in parentheses) clustered at the country level. All interactive variables are also introduced in level but not reported to save space. All regressions include the volatility of the terms of trade, BD for missing data in real GDP and real expenditure series, BD for five years HP cycle correction. See Table A.1 in appendix A for further discussion on variables definition. *** p<0.01, ** p<0.05, * p<0.1. 15              Instrumental Variable Estimation Our identification strategy in the estimations above has two problems: the omitted variable bias and endogeneity of the cyclical real GDP. Regarding the omitted variable bias, a significant number of potential covariates might affect both cyclical spending and cyclical GDP, and here we have included a large number of covariates analyzed in the literature. Still, there is uncertainty regarding the true model underlying the determination of procyclicality. Future research could consider a Bayesian approach to estimation. Regarding the endogeneity of cyclical real GDP, one might infer the reverse causality in the empirical specification, due to multiplier effects (IMF, 2014). To address these problems, we adopt an instrumentation strategy in the spirit of Fatas and Mihov (2013), collapsing the panel data over the period 2000-2016 and using instruments from the year 1999, the year before the start of our sample period. The main instrument is the initial cyclical real GDP, and in successive columns we consider the initial values for each conditioning factor highlighted in Tables 1 and 2: the initial values for the tax ratio; the credit ratio; the perceived corruption of politicians; the perceived corruption in public services; the rule of law; the power distribution; the resource distribution; and the ethnic/religious stability. The multivariate regressions are summarized in Table 6. Irrespective to the set of instruments, procyclicality coefficients are statistically significant and robust.12                                                              12Note that with the collapse of the data, all binary dummy (BD) variables refer not to the presence of a condition (i.e. a missing variable), but to the percent of time with the considered condition (i.e. the percent of time with a missing variable over 2000-2016). 16              Table 6: Fiscal (pro)cyclicality, Instrumental Variables Regressions (2000-2016) Cyclicality of real general government primary spending (1) (2) (3) (4) (5) (6) (7) Variables IV 2SLS Cyclicality of 0.329*** 0.327*** 0.322*** 0.320*** 0.328*** 0.324*** 0.327*** real GDP (0.067) (0.063) (0.063) (0.063) (0.068) (0.063) (0.068) Constant -3.128 -3.862 -4.644 -4.715 -4.009 -4.556 -4.040 (24.320) (22.203) (22.544) (22.532) (25.390) (22.560) (25.250) Centered R- 0.141 0.145 0.147 0.148 0.144 0.146 0.145 squared Rmse 156 147 147 147 146 147 158 Weak ID test 103.953 119.327 118.917 120.440 100.334 119.050 100.753 (stat>10) OID test (p- 0.813 0.803 0.189 0.177 0.329 0.426 0.948 value) Intruments Initial Crgdp Initial Crgdp Initial Crgdp Initial Crgdp Initial Crgdp Initial Crgdp Initial Crgdp Initial Initial revenue.ratio Initial credit.ratio Initial pol.corruption Initial pub.service.corruption Initial law.order Initial resource.distribution ethnicandrel.stability Covariates Set2 Set2 Set2 Set2 Set2 Set2 Set2 Observations (# 95 108 108 108 108 108 94 countries) Notes: The year 1999 is considered as reference for the initial Crgdp and successive instruments. The full list of removed countries per columns is available upon request. With the collapse of the data, all BD variables refer not to the presence of a condition (i.e., a missing variable), but to the percent of time with the considered condition (i.e., the percent of time with a missing variable over 2000- 2016). 17              B. Cyclicality of policy: Tax rates The cyclical behavior of policy can also be examined from the tax side, as proposed by Vegh and Vuletin (2015) and done recently by Aizenman et. al. (2019). Accordingly, we study the cyclicality of policy based on three different taxes, namely the individual income tax rate, the corporate tax rate and the VAT rate, and the same conditioning factors described in the previous part of this paper. Results are summarized in Tables B.1 to B.9 in appendix B. We find unconditional procyclical fiscal policies, as coefficients in column 1 of Tables B.1, B.4 and B.7 are negative and statistically significant, implying lower tax rates during good times and higher tax rates during bad times. In addition, we find notable economic nonlinearities and heterogeneity across the different types of taxes. VAT tax rates show significantly lower procyclical bias, which would imply a benefit of using this type of tax from a purely stabilizing perspective. The development level of countries (column 2 of Tables B.1, B.4 and B.7) seems to reduce only the procyclicality of corporate tax rates. The level of resource mobilization reduces the procyclicality in the case of corporate tax and VAT rates but amplifies it in the case of individual income rates (Tables B.1, B.4 and B.7, column 3). Tax base variability is associated with more procyclical individual income and corporate tax rate but less procyclical VAT rates (column 4 of Tables B.1, B.4 and B.7). The financial openness, proxied with the capital account openness index mitigates the procyclicality of individual income and corporate taxes, but does not have any impact on VAT rates (column 6 of Tables B.1, B.4 and B.7). The levels of perceived political corruption and corruption in the public service are positively correlated with the procyclicality of VAT rates (columns 7, 8 of Table B.4). The ethnic and religious stability contributes to reduce the procyclicality of VAT rates (column 3 of Table B.5). Elections are also associated to less procyclicality of corporate income tax rates and VAT tax rates (column 7 of Tables B.2 and B.11). Note that the expenditure rigidity is not correlated with the cyclicality of tax rates, but larger fiscal space reduces the procyclicality of VAT rates, without any statistically significant effect on the cyclicality of individual income and corporate tax rates (columns 9, 10 of Table B.3, B.5 and B.8). Resource dependence is not statistically significant as a conditioning factor of procyclicality in the tax rates. 18              V. How to mitigate the procyclicality bias: Policy options budget composition, decentralization, and fiscal rules Top-down solution: Fiscal rules This section explores the correlation between the procyclical bias and the existence of fiscal rules and some of their features.13 National fiscal rules are generally recognized to be more effective than supranational fiscal rules (Tapsoba, 2012). Indeed, supranational rules generally suffer from a problem of insufficient enforcement and compliance. Fiscal rules (FRs) are often criticized to amplify the procyclical bias,14 but Guerguil et al. (2016) suggest that the design of the rules matters. Based on the fiscal rules data set from the Fiscal Affairs Department, we consider several characteristics of the national FRs, namely (i) if a fiscal council also exists; (ii) the presence of effective monitoring mechanisms; (iii) the presence of effective enforcement mechanisms; (iv) the coverage of general government for the rules; (v) the existence of a written legal basis for the rules (statutory or constitutional); (vi) the presence of escape clauses; and (vii) the presence of other kind of flexibility - investment-friendly rules and cyclically-adjusted rules (see Guerguil et al., 2016). The comparison of the procyclicality bias between countries with and without these features shows (Table C.1 and Figure C.3) that countries with national FRs have a lower procyclical bias, though still positive. The other characteristics that show statistically significant differences are (i) fiscal councils; (ii) monitoring mechanisms; (iii) enforcement mechanisms; (iv) a written legal basis; (v) the presence of escape clauses, and (vi) flexibility features are associated with less procyclicality than non-FR countries, highlighting the relevance of adequate design of national FRs to mitigate procyclicality. [insert Table C.1 and Figure C.1] Top-down solution: Fiscal councils The other potential top-down solution we explore is another fiscal institution, namely the fiscal councils. Fiscal councils (FCs) are defined as independent, non-partisan agencies with an official mandate to assess fiscal policies, plans, rules, and performance (Debrun et al., 2013). They do not have a direct role in setting policy instruments, but they can influence fiscal behavior through three main channels. First, by fostering transparency of fiscal policy, an FC might reduce the rationale for voters to “starve the Leviathan.” Second, they can increase the reputational costs for politicians on unsound policies and broken commitments. Third, they can provide direct inputs to the budget process, with forecasts or assessments of structural positions; as suggested in the previous paragraph they can close technical loopholes that allow governments to circumvent numerical fiscal rules (Debrun et al., 2017). Taking advantage of the FC data set from the Fiscal Affairs Department, we check (i) whether the FC                                                              13 We account for numerical fiscal rules. For a literature review on procedural rules, see Alesina and Perotti (1999). 14 On procyclical bias, a petition signed by 1,100 economists and 11 Nobel economists in the New York Times claimed that attempts to strictly keep the budget balance (in US states) would aggravate recession (see Levinson, 1998).  19              encompasses a large coverage (i.e. general government); (ii) whether the FC make forecasts; (iii) whether the FC makes forecasts on the preparation of the fiscal policy; (iv) whether the FC makes forecasts assessments on the fiscal policy; (v) whether the FC makes recommendations to the government; (vi) whether the FC makes long-term sustainability analyses; (vii) whether the FC establish consistency with the objectives; (viii) whether the FC evaluates the costs of fiscal measures; (ix) whether the FC makes ex-post analysis; (x) whether the FC publish public reports; and (xi) whether the FC has a high media impact. We focus on the period 2009-2016 to evaluate the efficiency of post- crisis policies. The comparison of the procyclical bias across groups of countries with and without these features (Table C.2 and Figure C.2) shows that countries with fiscal councils have lower procyclicality bias; but the largest impact comes from attributes of the council, especially those that examine sustainability and evaluate the costs of different measures. The features are associated with countercyclical spending. Transparency through public reports and the media exposure also matters, in line with predictions by Alesina et al. (2008). [insert Table C.2; Figure C.2] Bottom-up solution: Fiscal decentralization Though fiscal decentralization may be an effective mechanism to enhance the provision of public service delivery in developing countries, such as the access to primary education, the access of drinking water, the access of refuse and sewage disposal facilities. The discussion still exists around the optimal level of fiscal decentralization for both spending and expenditures. So, the decentralization might be also effective in the struggle against fiscal procyclicality of the general government because of the application of subsidiarity principle. Indicators of fiscal decentralization generally refer to the fiscal composition ratio established by the IMF’s Government Finance Statistics Yearbooks (see Blume and Voigt (2008) and Blume et al. (2009) for more details). Instead of establishing arbitrary thresholds for the fiscal ratio, we focus on indicators of political decentralization (PD) established by the Database of Political Institutions provided by the World Bank Group. Namely, we focus on (i) whether there are contiguous autonomous regions; (ii) whether local governments and legislature are elected; (iii) whether state/provinces governments and legislature are elected; (iv) whether states/provinces have authority over taxing, spending, or legislating; and (v) whether the constituencies of the upper house members are the states/provinces. We focus on the period 2009-2015 (2016 not available yet in data) to evaluate the efficiency of post-crisis policies. The correlations (Table C.3 and Figure C.3) show that the closest proxy of fiscal decentralization (i.e. the authority of states/provinces over spending, taxing or legislating) is negatively correlated with fiscal procyclicality, suggesting that decentralization may be a tool to mitigate the procyclicality bias. 20              [insert Table C.3; Figure C.3] Bottom-up solution: Direct democracy Since the classic Alesina et al. (2008) paper shows that procyclicality (associated with corruption) is a feature of democracies, we explore the role of democratic institutions on procyclicality. Taking advantage of the Varieties of Democracy (V-dem) database, we identify (i) whether the initiative is permitted at the subnational level, and or at the national level; and (ii) whether the popular referendum is permitted at the subnational level, and or at the national level. We focus on the period 2009-2015 (2016 not available yet in data). The correlations (Table C.4 and Figure C.4) show that both the initiative and the referendum are negatively correlated with fiscal procyclicality, but the effect seems stronger for the initiative rights. This finding echoes results of Matsusaka (2014) regarding the effects of initiative on fiscal congruence and fiscal conservatism in U.S. states. Hence, delegating more fiscal legislation to the citizens may be a tool to reduce the procyclical bias. [insert Table C.4; Figure C.4] Bottom-up solution: Spending and tax composition The literature shows that different types of spending behave differently along the cycle. Ardanaz and Izquierdo (2017) show that current spending is more procyclical during upswings of the cycle, whereas capital spending is more procyclical during downturns. Hence, for policy making purposes and institutional design, it would be beneficial to limit the growth of current spending in the good times, while in bad times protecting the capital spending would be beneficial for stabilization. In this paper, we show that when spending is more rigid, it will be less procyclical during the good times, but less countercyclical during the downturns. The composition of the tax structure also indicates that VAT rates are less procyclical than the other types of taxes., hence adding a potential benefit to using this type of taxes for economic stabilization, along with the proposal of sales tax holidays during the Great Recession (Furman, 2018). 21              VI. Differences across regions and countries Using the Mean Group (MG) estimator (Pesaran and Smith, 1995) we estimate the individual country coefficients of the cyclicality of the real GDP on the cyclicality of primary spending for the entire sample. Equation (2) is adapted for the MG estimator: . . (3) To facilitate comparison across regions, we first compute the MG coefficient for the entire sample and then estimate the same model for each region. That is, we compute the procyclicality coefficient for the 112 countries, similar to Table 1, but relaxing the fixed effects constraint.15 The disadvantage is the limited degrees of freedom when applying this estimator because the methodology consists of estimating OLS regressions at the country level.16 Table 7 summarizes the results for the entire sample and the different regions. For the overall sample, the effect is positive, indicating the procyclicality effect. Across different regions, the effect is larger in SSA, followed by LAC, and then ECA. The effect is not significant for the overall MENA region. Table 7: MG Estimator (2000-2016) (1) (2) (1) (2) (1) (2) Whole sample SSA LAC MENA ECA SEAP Cyclicality of real GDP 0.164*** 0.289*** 0.229** -0.042 0.188*** 0.167*** (0.049) (0.099) (0.109) (0.120) (0.059) (0.150) Observations 1,954 653 357 221 440 283 # countries 116 39 21 13 26 17 Notes. Clustered standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Table D.1 allows comparing the different determinants of procyclicality across regions. SSA and LAC, the more procyclical regions, show the lowest credit to private sector ratios. SSA has the highest perception of corruption. Also, it is the region with the highest degree of ethnic and religious fragmentation. Also, SSA has the most unequal distribution of resources, after MENA. All the conditioning factors point at SSA as the one with the highest procyclicality bias.                                                              15 This is equivalent to relaxing the common know-how (common intercept for each country), and of common technology (common coefficient for each country) which are critical when using macro data. 16 When balanced, we have 17 observations per country (2000-2016), and six coefficients to be estimated, excluding the intercept. The xtmg command by Markus Eberhardt (https://sites.google.com/site/medevecon/code) is considered to regress the MG estimator on STATA software. 22              MENA is a puzzling case. It has the highest inequality and second highest level of corruption, and it is, on average, not procyclical despite its resource wealth. The level of credit to the private sector is the second highest, suggesting that financial depth may be compensating the negative impact of the other two factors. Maybe the extreme cases of the United Arab Emirates, Saudi Arabia, and Kuwait, that are strongly countercyclical drive the results for the region. This result needs further analysis, probably by introducing the level of international reserves as a control variable, which provides a cushion that allows the policy to be managed countercyclically. Across countries We compute individual coefficients, and individual standard errors for the 38 SSA countries, which are available upon request. Using such a methodology, we find five to six countries which are countercyclical over the whole period: Guinea, Lesotho and Democratic Republic of Congo, Mauritius, South Africa, and Botswana.17 Based on the individual country-coefficients, we examine whether procyclicality is different in the sample of SSA countries for resource-dependent economies.18 We find that procyclical bias is higher in resource-rich or resource-dependent economies. The average coefficient for SSA economies is .31; for resource-dependent economies, it is .33; and for the other SSA economies, it is .27 (Figure D.1). The comparison between resource-dependent (and resource-rich) countries in SSA and resource- dependent countries in other regions shows that the procyclicality bias is bigger in SSA. While the SSA resource-dependent economies have a procyclicality coefficient of .33 the analogous category in other regions has .23. However, the resource-rich countries show an even larger contrast: the SSA resource- rich show a procyclicality coefficient of .33 while those of other regions register .08. This is due to the one-to-one mapping between the two groups of countries (resource-rich and resource-dependent), while in other regions the correspondence is not as clear (Figure D.2). In other regions there are resource- rich countries, such as Chile and Mexico, which are not resource-dependent; both show countercyclical behavior.19 [insert Tables D.1 and D.2; Figures D.1 and D.2]                                                              17 Cyclicality coefficients for the entire sample of countries in both subperiods are available upon request. 18 Table 1 shows that the variable is significant for the entire sample, assuming homogeneous coefficients and FE. 19 The evolution of the coefficient of procyclicality is available upon request.  23              VII. Concluding remarks Fiscal procyclicality has a substantial impact on macroeconomic stability and development outcomes in developing countries. This paper adds to the literature by (i) accounting for the effects of fiscal rigidity and fiscal space; (ii) expanding the time period to 2016 and hence including the fiscal expansion in many countries to tackle the Great Recession and the adjustment of lack of it during the growth recovery; (iii) including more granular examination of the quality of institutions, comparing policy cyclicality in countries with fiscal rules or fiscal councils with that of countries without those institutions; (iv) adding a set of explanatory variables to capture social fragmentation or polarization which may complement the traditional macroeconomic explanations such as borrowing constraints or volatility of revenues; and (v) focusing on regional differences between Africa, Latin America and the Middle East where the more salient features and changes in procyclicality take place. We find that fiscal policy in developing countries is procyclical both on the tax and spending sides. However, the degree of procyclicality of public spending is conditional on the level of development, the extent of domestic credit constraints, the perceived level of corruption, ethnic-religious stability, and equal distribution of resources. One of the more critical aspects that determine procyclicality is whether the country is natural resource-dependent or not, with the impact of this variable becoming larger over time. On the fiscal institutions, we find that the rigidity of public spending reduces procyclicality bias while the effect is asymmetric along the business cycle: The mitigation of procyclicality is larger during the upturn phase of the cycle. On the revenue side, we examined three types of tax rates (personal income tax rate, corporate tax rate, and value-added tax rate), and all of them show the procyclical behavior of policymakers. The VAT tax rate shows a smaller procyclical bias. As on the spending side, the development level and financial openness reduce the procyclicality of corporate tax rates. Tax base variability is associated with more procyclicality of the personal income and corporate tax rate while reducing the procyclicality of the VAT rate. The ethnic and religious stability and elections contribute to reducing the procyclicality of VAT rates. Finally, we find no evidence that the rigidity of fiscal spending affects the procyclicality of fiscal space, and the fiscal space does affect the procyclicality of VAT rates. The lower fiscal space increases the procyclicality of the impact on personal and corporate tax rates but operates in the opposite direction in the VAT rate management: lower fiscal space is associated with more countercyclical behavior. 24              References Alesina, A., and Perotti, R. (1999). Budget Deficits and Budget Institutions. In Fiscal Institutions and Fiscal Performance, 13-36. Cambridge, MA: National Bureau of Economic Research. Alesina, A., and Devleeschauwer, A., and Easterly, W., and Kurlat, S., and Wacziarg, R. (2003). "Fractionalization ." Journal of Economic Growth, 8(2), 155-194. Alesina, A., F.R. Campante, and G. Tabellini. (2008). Why is Fiscal Policy Often Procyclical?. Journal of the European Economic Association, 6, 1006-1036. Aizenman, J., Jinjarak, Y., Nguyen, H. T. K., and Park, D. (2019). Fiscal space and government- spending and tax-rate cyclicality patterns: A cross-country comparison, 1960–2016. Journal of Macroeconomics, 60, 229-252. Ardanaz, M. and A. Izquierdo. (2017). Current Expenditure Upswings in Good Times and Capital Expenditure Downswings in Bad Times? New Evidence from Developing Countries. Inter- American Development Bank, IDB-WP-838. Arezki, R., and M. Bruckner. 2012. Commodity Windfalls, Polarization, and Net Foreign Assets: Panel Data Evidence on the Voracity Efect. Journal of International Economics, 86(2), 318-326. Blume, L., and Voigt, S. (2008). Federalism and Decentralization- A Critical Survey of Frequently Used Indicators. MAGKS Joint Discussion Paper series in Economics, No. 21-2008, Marburg: Philipps-University Marburg. Blume, Lorenz, J. Muller, and S. Voigt. (2009). The Economic Effects of Direct Democracy—A First Global Assessment. Public Choice, 140(3-4), 431-461. Chinn, M., and H. Ito. (2006). What Matters for Financial Development? Capital Controls, Institutions, and Interactions. Journal of Development Economics, 81(1) 163-192. Debrun, X., Moulin, L., Turrini, A., Ayuso-i-Casals, J., & Kumar, M. S. (2008). Tied to the mast? National fiscal rules in the European Union. Economic Policy, 23(54), 298-362. Debrun, X., and Kinda, T., and Curisitine, T., and Eyraud, L., and Seiwad, J. (2013). The Functions and Impact of Fiscal Councils. IMF Policy Paper, Washington DC: International Monetary Fund. Debrun, X., and Zhang, X., and Lledo, V. (2017). The Fiscal Council Dataset: A Primer to the 2016 Vintage. IMF Note, Washington DC: International Monetary Fund. Fatas, A., and I. Mihov. (2013). Policy Volatility, Institutions and Economic Growth. The Review of Economics and Statistics, 95(2),362-375. Frankel, J., C. Végh, and G. Vuletin. (2013). On Graduation From Fiscal Procyclicality. Journal of Development Economics, 100(1), 32-47. Furman, J. (2018). The fiscal Response to the Great Recession: Steps Taken, Paths Rejected, and Lessons for Next Time. Mimeo. Brookings Institution. Gavin, M., and P. Roberto. (1997). Fiscal Policy in Latin America. NBER Macroeconomics Annual, vol. 12, pp 11-72, Cambridge, MA: National Bureau of Economic Research, Inc. 25              Guerguil, M., and Mandon, P., and Tapsoba, R., (2016). Flexible Fiscal Rules and Countercyclical Fiscal Policy. IMF Working Paper WP/16/8, Washington DC: International Monetary Fund. Ilzetzki, E., & Végh, C. A. (2008). Procyclical fiscal policy in developing countries: Truth or fiction? National Bureau of Economic Research, No. 14191. Ilzetzki, E. (2011). Rent-seeking distortions and fiscal procyclicality. Journal of Development Economics, 96(1), 30-46. IMF. (2012). Macroeconomic Policy Frameworks for Resource-Rich Developing Countries. IMF report, Washington DC: International Monetary Fund. IMF. (2014). Fiscal Multipliers : Size, Determinants, and Use in Macroeconomic Projections. IMF Report, Washington DC: International Monetary Fund. Konuki, T., and M. Villafuerte. (2016). Cyclical Behavior of Fiscal Policy among Sub-Saharan Countries. IMF report, Washington DC: African Department, International Monetary Fund. Levinson, A. (1998). Balanced Budgets and Business Cycles: Evidence from the States. National Tax Journal, 51 715-732. Matsusaka, J. G. (2014). Disentangling the Direct and Indirect Effects of the Initiative Process. Public Choice, 160(3-4), 345-366. Mendoza, E. G., and Oviedo, P. M. (2006). Fiscal policy and macroeconomic uncertainty in developing countries: The tale of the tormented insurer. National Bureau of Economic Research, No 12586. Pesaran, P. H., and R. Smith. (1995). Estimating Long-Run Relationships from Dynamic Heterogeneous Panels. Journal of Econometrics, 68(1), 79-113. Ravn, M. O., and H. Uhlig. (2002). On Adjusting the Hodrick-Prescott Filter for the Frequency of Observations. The Review of Economics and Statistics, 84(2), 371-380. Suescún, R. (2007). The size and effectiveness of automatic fiscal stabilizers in Latin America. Policy Research Working Paper, No 424, Washington DC: The World Bank. Talvi, E., Vegh, C. A. (2005). Tax Base Variability and Procyclical Fiscal Policy. Journal of Development Economics, 78(1), 156-190. Tapsoba, R. (2012). Do National Numerical Fiscal Rules Really Shape Fiscal Behaviours in Developing Countries? A Treatment Effect Evaluation. Economic Modelling, 29(4), 1356-1369. Vegh, C. and Vuletin, G. (2015). How is Tax Policy Conducted over the Business Cycle? American Economic Journal: Economic Policy. 7(3), 327-370. WEO. (2009). World Economic Outlook, April 2009. IMF Report, Washington DC: International Monetary Fund. 26              Appendix A: Data construction and variable description Data construction Our data set of reference is the WEO (October 2016 version) to capture the general government primary expenditures (GGX and GGEI series), gross domestic product, current prices (NGDP series), the GDP deflator (NGDP_D series), and the terms of trade (TT series).20 We consider the period of reference by Frankel et al. (2013) from 1960-2009, extended to 2016. To select our countries, we apply the following criteria: 1. We first drop every observation with strictly fewer than one million inhabitants, following Alesina et al. (2008). The underlying idea is those tiny countries are exposed to very large shocks, making the comparison with larger countries more difficult.21 2. Then, we drop every country with strictly fewer than 16 years of data (over 1960-2016) for general government primary spending and gross domestic product series, still following Alesina et al. (2008). The underlying idea is that we need to observe at least two or three cycles in each country as we study fiscal procyclicality. 3. We drop all the countries which are simultaneously (i) classified as “high income” for the year 2017 by the World Bank (https://datahelpdesk.worldbank.org/knowledgebase/articles/906519- world-bank-country-and-lending-groups); and (ii) were members to the OECD strictly before the fall of the Soviet Union (year 1992). Under this double selection process, we safely remove all established developed countries, and we keep Turkey, high-income non-OECD countries (such as Saudi Arabia or Singapore) and recent OECD members (such as Israel, Mexico, and Slovenia). By keeping these countries in a first step, we are able to capture more variability in our results and eventually to remove them in a second step for robustness checks. Applying these three criteria leaves us with a sample of 116 developing countries (the Syrian Arab Republic is removed due to the war) over the period 1960-2016. To maximize the number of countries displayed in our descriptive statistics, we focus mainly on the sub-period 2000-2016.22                                                              20 More details on the variables in Table A.5. We consider the general government primary expenditures, instead of central government, total expenditure and net lending (GCENL series) used by Frankel et al. (2013) as the last one is not available since WEO archives 2009. Accounting for the general government coverage enables us to have comparable spending figures between unitary states and federal states (such as the Democratic Republic of Congo and Nigeria). Accounting for primary spending enables us to account for current performance of administrations, irrespective to the debt service burden. 21 To capture the population size, we use the “Population, total” series from the WDI data set (updated February 1, 2017), and removed all observations with a blank in the series. We keep Taiwan, China, which in no longer covered in WDI data set, as the population size is over the threshold of one million inhabitants for the whole period. 22 For example, we do not have data on general government primary expenditures available before 2000 for Nigeria. To avoid confusion, we do not attempt to apply the criteria #2, on the sub-period 2000-2016; as recalled by Alesina et al. (2008), in general, the larger the cutoff for inclusion is –relative to the time horizon we have-, the stronger the results are.  27              We also highlight resource-dependent countries and resource-rich countries in our sample (note that the two lists are not mutually exclusive), as they are highlighted to be particularly procyclical in the recent literature (Arezki and Bruckner, 2012; Konuki and Villafuerte, 2016): 1. Resource-dependent countries are countries for which commodity revenues (GGRC in WEO data set) account for at least 10 percent of total revenues minus grants (GGR and GGRG series in WEO data set) at least 50 percent of the time over the considered period. This definition is derived from Konuki and Villafuerte (2016) and enables us to capture countries with a resource- dependent revenue structure. For more details, see Table E.1 in Appendix E. 2. Resource-rich countries are defined by the IMF (2012), here is direct access to the report (https://www.imf.org/external/np/pp/eng/2012/082412.pdf). According to this definition, we capture countries where production on natural resources reached (or expected to reach) significant levels. For more details, see Table E.2 in Appendix E. Treatment of missing data We expand equation 1 as follows: . . (A-1) The vector of covariates includes the standard deviation of the terms of trade (TT series in WEO database) measured with five years rolling window, two binary dummy (BD) variables equaling one if the original real primary expenditures and the cyclical components of real GDP were missing, zero otherwise. Last but not least, we include two additional BD variables equaling one if the cyclical components of expenditures and GDP were missing the previous year. This is the Set 1 of covariates. The Set 2 of covariates differs with the inclusion of two BD variables equaling one if the cyclical components of expenditures and GDP were missing from one to five years ago, zero otherwise. With this, we aim to capture inaccurate cycles due to the linear interpolation method. 28              Table A.1: Variable Definitions (2000-2016) # Name Description Souces N Mean S.D. Min Max Cyclicality real primary Cylcicality of real primary general government WEO (October, 2016) and Own 1 1,933 -20.66 5,755.94 -77,767.53 82,694.79 spending* expenditures (HP filter λ=6.25). Construction. Cylcicality of real gross domestic product (HP WEO (October, 2016) and Own 2 Cyclicality real GDP** 1,933 -26.67 4,832.80 -82,929.94 106,627.20 filter λ=6.25). Construction. 5y rolling window standard deviation of the 3 Volatility ToT WEO (October, 2016). 1,886 10.13 12.58 0.13 135.91 terms of trade (TT series). Real general government WEO (October, 2016) and Own 4 Total real revenue (% of GDP). 1,933 11.06 11.03 0.64 72.47 total revenues*** Construction. Real general government WEO (October, 2016) and Own 5 Total real tax (% of GDP). 651 13.07 5.83 0.00 31.48 total tax**** Construction. Annual absolute variation of real general WEO (October, 2016) and Own 6 Revenue base variability 632 1.06 1.15 0.00 12.44 government total internal revenue. Construction. 7 Credit to private sector Domestic credit to private sector (% of GDP). WDI (February, 2017). 1,779 35.15 30.33 0.00 160.12 Index measuring the degree of capital account 8 Kaopen Chinn and Ito (2006). 1,688 0.48 0.35 0.00 1.00 openness 9 Political corruption Political corruption (v2x_corr series) V-dem vers. 6.2. 1,620 0.41 0.23 0.03 0.96 Corruption of the public services (v2x_pubcorr 10 Public service corruption V-dem vers. 6.2. 1,620 0.41 0.25 0.04 0.97 series) 11 Rule of law v2xcl_rol series. V-dem vers. 6.2. 1,620 0.66 0.24 0.04 0.99 12 Political Power distribution v2pepwrsocl series. V-dem vers. 6.2. 1,620 0.48 1.08 -2.46 3.10 13 Distribution of resources v2xeg_eqdr series. V-dem vers. 6.2. 1,620 0.58 0.21 0.10 0.97 Ethnic and religious Average between the indicator of ethnic 14 ICRG indicators, PRS group. 1,530 4.11 1.09 0.00 6.00 Stability tensions and religious tensions. Binary variable if the country is ressource Adapted from Konuki and Villafuerte 15 Resource dependent 1,933 0.28 0.45 0.00 1.00 dependent. (2016). 16 Resource rich Binary variable if the country is ressource rich. IMF (2012). 1,933 0.40 0.49 0.00 1.00 Binary variable if the yearly growth rate is 17 Bad times Own Construction 1,932 0.09 0.29 0.00 1.00 negative. Binary variable for the year of the highest level 18 Elections DPI (2015) and Own Construction. 1,933 0.18 0.39 0.00 1.00 national election. 19 Debt ratio Gross debt ratio over GDP. WEO (October 2016). 1,892 40.83 51.99 1.01 789.83 Notes: *: To compute real general government primary spending we substract GGEI series to GGX series and deflate it with the NGDP_D series. **: To compute real GDP spending we deflate the NGDP series with the NGDP_D series. ***: To compute real revenue over GDP we divide the GGR series to NGDP series. The series are deflated with the NGDP_D series. ****: To compute real internal revenue over GDP we subtract GGRG series to GGR series and divide it with the NGDP series. The series are deflated with the NGDP_D series. Information regarding variables used in the section of ``Potential solutions'' are available upon request. 29              Appendix B: Regression tables – Tax rate procyclicality Table B.1: Fiscal (pro)cyclicality of individual income tax rates, whole sample (2000-2016): Impacts of development level, tax base, credit constraints, corruption, rule of law and order (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Dependent variable: Individual income tax Cyclicality of real GDP -0.338*** -0.709 0.166 0.403*** -0.695 -0.801** -1.641 -0.385 2.220 0.273 (0.041) (0.535) (0.313) (0.044) (0.598) (0.353) (2.008) (0.292) (2.557) (0.577) Development level Crgdp*dvpt.level (polytomic) 0.122 (0.170) Tax base effect Crgdp*tax -0.068* (0.037) Crgdp*tax.base.variability -1.268*** (0.066) Credit constraints Crgdp*credit.ratio 0.009 (0.012) Crgdp*kaopen 1.611* (0.898) Perceived corruption Crgdp*pol.corruption 2.496 (3.279) Crgdp*pub.service.corruption 0.329 (0.647) Rule of law and order Crgdp*law.order -0.598 (0.642) Crgdp*rule.of.law -1.152 (1.643) Constant 2.011** 66.846*** -1.163 2.659 6.789** 7.041** 41.887*** 42.290*** 2.591 32.860*** (0.890) (0.970) (3.887) (1.955) (2.998) (3.399) (3.616) (3.446) (4.488) (4.651) Adjusted R-squared 0.874 0.874 0.947 0.948 0.881 0.881 0.835 0.837 0.880 0.832 Rmse 4.325 4.328 3.175 3.154 4.237 4.246 4.282 4.246 4.292 4.320 Joint significance (p-value) - 0.000 0.000 0.000 0.000 0.000 0.046 0.001 0.002 0.027 CD test (p-value) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Order of integration I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) Cluster country country country country country country country country country country Covariates Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 826 826 289 289 761 725 689 689 746 689 # countries 50 50 19 19 50 50 47 47 48 47 Notes. Standard errors (in parentheses) clustered at the country level. All interactive variables are also introduced in level but not reported to save space. Coefficients attached to cyclicality of GDP or its interactions are multiplied by 10000. All regressions include the volatility of the terms of trade, BD for missing data in real GDP and real expenditure series, BD for five years HP cycle correction. See Table A.3 for further discussion on variables definition. *** p<0.01, ** p<0.05, * p<0.1. 30              Table B.2: Fiscal (pro)cyclicality of individual income tax rates, whole sample (2000-2016): Impacts of level playing field, natural resources, and other factors (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Dependent variable: Individual income tax Cyclicality of real GDP -0.200** 0.121 0.408 -0.096 -0.065 -0.096 -0.384*** -0.060 8.195 0.377*** (0.083) (0.434) (0.816) (0.163) (0.169) (0.163) (0.084) (0.442) (12.065) (0.034) Equality of conditions Crgdp*power.distribution -0.051 (0.138) Crgdp*resource.distribution -0.396 (0.684) Crgdp*ethnicandrel.stability -0.201 (0.293) Natural resources Crgdp*resource.dependant (bd) -0.256 (0.182) Crgdp*resource.rich (bd) -0.288 (0.188) Crgdp*resource dependant and rich -0.255 (0.182) Other factors Crgdp*elections 0.176 (0.188) Crgdp*debt.ratio -0.021 (0.036) Crgdp*rigidity -1,598.647 (2,139.051) Crgdp*fiscal space -30.251*** (1.268) Constant 36.452*** 22.366** -0.766 2.008** 2.007** 2.008** 2.006** 3.433*** 44.874*** 2.934*** (1.904) (8.656) (4.602) (0.891) (0.891) (0.891) (0.892) (1.067) (4.441) (0.877) Adjusted R-squared 0.831 0.835 0.880 0.874 0.874 0.874 0.874 0.879 0.778 0.879 Rmse 4.326 4.279 4.287 4.327 4.327 4.327 4.330 4.270 5.559 4.240 Joint significance (p-value) 0.003 0.097 0.000 0.002 0.000 0.000 0.000 0.000 0.433 0.000 CD test (p-value) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Order of integration I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) Cluster country country country country country country country country country country Covariates Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 689 689 746 826 826 826 826 814 399 823 # countries 47 47 48 48 50 50 50 50 26 50 Notes. Standard errors (in parentheses) clustered at the country level. All interactive variables are also introduced in level but not reported to save space. Coefficients attached to cyclicality of GDP or its interactions are multiplied by 10000. All regressions include the volatility of the terms of trade, BD for missing data in real GDP and real expenditure series, BD for five years HP cycle correction. See Table A.3 for further discussion on variables definition. *** p<0.01, ** p<0.05, * p<0.1. 31              Table B.3: Fiscal (pro)cyclicality of individual income tax rates, whole sample (2000-2016): Impacts of rigidity, fiscal space and business cycle (1) (2) (3) Dependent variable: Individual income tax Cyclicality of real GDP 0.052 55.509 -2.194 (0.188) (35.893) (3.394) Business cycle Crgdp*good times -5.960e+07** -5.781e+09 2.369e+08 (22995752.015) (5.978e+09) (3.385e+08) Crgdp*good times*rigidity 1.431 (1.227) Crgdp*rigidity -13,856.972 (8,135.989) Rigidity of public spending -0.089 (0.061) Crgdp*good times*fiscal space -0.015 (0.017) Fiscal space 116.726 (170.358) Crgdp*fiscal space -0.001* (0.001) good times 1.740 3.339 1.306 (1.310) (3.066) (1.100) Constant 0.375 42.418*** 1.648 (1.731) (5.691) (1.373) Adjusted R-squared 0.876 0.781 0.880 Rmse 4.300 5.520 4.230 Joint significance (p-value) 0.00 0.002 0.000 CD test (p-value) 0.000 0.000 0.000 Order of integration I(0) I(0) I(0) Cluster country country country Covariates Yes Yes Yes Observations 826 399 823 # countries 50 26 26 Notes. Standard errors (in parentheses) clustered at the country level. All interactive variables are also introduced in level but not reported to save space. Coefficients attached to cyclicality of GDP or its interactions are multiplied by 10000. All regressions include the volatility of the terms of trade, BD for missing data in real GDP and real expenditure series, BD for five years HP cycle correction. See Table A.3 for further discussion on variables definition. *** p<0.01, ** p<0.05, * p<0.1. 32              Table B.4: Fiscal (pro)cyclicality of corporate tax rates, whole sample (2000-2016): Impacts of development level, tax base, credit constraints, corruption, rule of law and order (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Dependent variable: Individual income tax Cyclicality of real GDP -0.338*** -1.221** -4.267*** 0.085 -1.349 -1.181*** -2.312 -0.634 -0.534 0.389 (0.043) (0.521) (0.936) (0.099) (0.896) (0.391) (3.248) (0.773) (4.002) (0.962) Development level Crgdp*dvpt.level (polytomic) 0.290* (0.169) Tax base effect Crgdp*tax 0.476*** (0.113) Crgdp*tax.base.variability -0.679*** (0.153) Credit constraints Crgdp*credit.ratio 0.018 (0.018) Crgdp*kaopen 2.144** (0.979) Perceived corruption Crgdp*pol.corruption 3.437 (5.331) Crgdp*pub.service.corruption 0.758 (1.612) Rule of law and order Crgdp*law.order 0.054 (0.991) Crgdp*rule.of.law -1.741 (2.745) Constant 59.535*** 18.554*** 59.612*** 63.086*** 64.350*** 60.164*** 40.269*** 37.649*** 60.042*** 29.966*** (1.136) (1.251) (2.950) (3.590) (2.151) (3.093) (2.894) (1.805) (3.233) (4.539) Adjusted R-squared 0.732 0.732 0.639 0.636 0.741 0.730 0.724 0.723 0.715 0.728 Rmse 4.369 4.371 6.055 6.080 4.323 4.388 3.695 3.703 4.468 3.673 Joint significance (p-value) - 0.000 0.000 0.000 0.000 0.000 0.001 0.000 0.000 0.000 CD test (p-value) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Order of integration I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) Cluster country country country country country country country country country country Covariates Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 831 831 291 291 766 730 700 700 751 700 # countries 50 50 19 19 50 50 47 47 48 47 Notes. Standard errors (in parentheses) clustered at the country level. All interactive variables are also introduced in level but not reported to save space. Coefficients attached to cyclicality of GDP or its interactions are multiplied by 10000. All regressions include the volatility of the terms of trade, BD for missing data in real GDP and real expenditure series, BD for five years HP cycle correction. See Table A.3 for further discussion on variables definition. *** p<0.01, ** p<0.05, * p<0.1. 33              Table B.5: Fiscal (pro)cyclicality of corporate tax rates, whole sample (2000-2016): Impacts of level playing field, natural resources, and other factors (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Dependent variable: Individual income tax Cyclicality of real GDP -0.258** -1.359* 0.788 -0.416 -0.441 -0.416 -0.401*** -0.043 -31.549 -0.146** (0.104) (0.770) (1.492) (0.425) (0.440) (0.424) (0.047) (0.281) (23.212) (0.058) Equality of conditions Crgdp*power.distribution 0.098 (0.172) Crgdp*resource.distribution 1.740 (1.243) Crgdp*ethnicandrel.stability -0.372 (0.539) Natural resources Crgdp*resource.dependant (bd) 0.083 (0.420) Crgdp*resource.rich (bd) 0.110 (0.435) Crgdp*resource dependant and rich 0.083 (0.420) Other factors Crgdp*elections 0.208** (0.102) Crgdp*debt.ratio -0.021 (0.024) Crgdp*rigidity 5,727.836 (3,927.079) Crgdp*fiscal space -8.396*** (1.575) Constant 39.587*** 40.695*** 59.060*** 59.537*** 59.537*** 59.537*** 59.533*** 58.880*** 38.483*** 59.068*** (2.361) (9.174) (5.738) (1.138) (1.138) (1.138) (1.133) (1.180) (2.696) (1.195) Adjusted R-squared 0.723 0.723 0.716 0.732 0.732 0.732 0.732 0.737 0.819 0.743 Rmse 3.705 3.705 4.465 4.371 4.371 4.371 4.374 4.359 3.021 4.288 Joint significance (p-value) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.139 0.000 CD test (p-value) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Order of integration I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) Cluster country country country country country country country country country country Covariates Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 700 700 751 831 831 831 831 819 407 828 # countries 47 47 48 48 50 50 50 50 26 50 Notes. Standard errors (in parentheses) clustered at the country level. All interactive variables are also introduced in level but not reported to save space. Coefficients attached to cyclicality of GDP or its interactions are multiplied by 10000. All regressions include the volatility of the terms of trade, BD for missing data in real GDP and real expenditure series, BD for five years HP cycle correction. See Table A.3 for further discussion on variables definition. *** p<0.01, ** p<0.05, * p<0.1. 34              Table B.6: Fiscal (pro)cyclicality of corporate tax rates, whole sample (2000-2016): Impacts of rigidity, fiscal space and business cycle (1) (2) (3) Dependent variable: Individual income tax Cyclicality of real GDP 0.560** -9.308 -2.741 (0.227) (12.916) (5.671) Business cycle Crgdp*good times -1.230e+08*** -3.576e+09 2.221e+08 (26210474.704) (3.773e+09) (5.662e+08) Crgdp*good times*rigidity 0.535 (0.652) Crgdp*rigidity 2,743.200 (2,791.339) Rigidity of public spending 0.034 (0.046) Crgdp*good times*fiscal space -0.017 (0.028) Fiscal space 162.312 (279.729) Crgdp*fiscal space 0.003** (0.001) good times 0.453 -0.164 0.932 (1.037) (0.775) (1.115) Constant 59.088*** 38.562*** 58.122*** (1.508) (2.595) (1.662) Adjusted R-squared 0.733 0.818 0.744 Rmse 4.359 3.029 4.275 Joint significance (p-value) 0.000 0.168 0.009 CD test (p-value) 0.000 0.000 0.000 Order of integration I(0) I(0) I(0) Cluster country country country Covariates Yes Yes Yes Observations 831 407 828 # countries 50 26 50 Notes. Standard errors (in parentheses) clustered at the country level. All interactive variables are also introduced in level but not reported to save space. Coefficients attached to cyclicality of GDP or its interactions are multiplied by 10000. All regressions include the volatility of the terms of trade, BD for missing data in real GDP and real expenditure series, BD for five years HP cycle correction. See Table A.3 for further discussion on variables definition. *** p<0.01, ** p<0.05, * p<0.1. 35              Table B.7: Fiscal (pro)cyclicality of VAT rates, whole sample (2000-2016): Impacts of development level, tax base, credit constraints, corruption, rule of law and order (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Dependent variable: Individual income tax Cyclicality of real GDP -0.057*** -0.168 -1.107* -0.263*** -0.125*** -0.473* 0.023 0.118 -0.329* -0.121** (0.019) (0.176) (0.569) (0.036) (0.031) (0.273) (0.070) (0.123) (0.172) (0.057) Development level Crgdp*dvpt.level (polytomic) 0.036 (0.057) Tax base effect Crgdp*tax 0.125* (0.068) Crgdp*tax.base.variability 0.328*** (0.017) Credit constraints Crgdp*credit.ratio 0.001 (0.000) Crgdp*kaopen 0.948 (0.684) Perceived corruption Crgdp*pol.corruption -0.209* (0.123) Crgdp*pub.service.corruption -0.452* (0.266) Rule of law and order Crgdp*law.order 0.055 (0.044) Crgdp*rule.of.law 0.056 (0.132) Constant -0.262 33.047*** -0.139 -0.259 0.575 -0.647 19.695*** 19.651*** -1.505* 19.376*** (0.249) (0.566) (0.637) (0.388) (0.446) (0.975) (0.564) (0.429) (0.818) (1.391) Adjusted R-squared 1.000 1.000 0.971 0.971 1.000 1.000 1.000 1.000 1.000 1.000 Rmse 1.008 1.009 1.148 1.144 0.978 0.963 1.002 0.995 0.992 1.006 Joint significance (p-value) - 0.012 0.041 0.000 0.000 0.000 0.000 0.001 0.000 0.000 CD test (p-value) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Order of integration I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) Cluster country country country country country country country country country country Covariates Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 823 823 291 291 758 722 687 687 743 687 # countries 50 50 19 19 50 50 47 47 48 47 Notes. Standard errors (in parentheses) clustered at the country level. All interactive variables are also introduced in level but not reported to save space. Coefficients attached to cyclicality of GDP or its interactions are multiplied by 10000. All regressions include the volatility of the terms of trade, BD for missing data in real GDP and real expenditure series, BD for five years HP cycle correction. See Table A.3 for further discussion on variables definition. *** p<0.01, ** p<0.05, * p<0.1. 36              Table B.8: Fiscal (pro)cyclicality of VAT rates, whole sample (2000-2016): Impacts of level playing field, natural resources, and other factors (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Dependent variable: Individual income tax Cyclicality of real GDP -0.082*** -0.103 -0.217*** 0.001 -0.013 0.001 -0.070*** -0.111* -3.662 -0.205*** (0.022) (0.173) (0.046) (0.035) (0.036) (0.035) (0.019) (0.057) (3.669) (0.026) Equality of conditions Crgdp*power.distribution 0.047 (0.028) Crgdp*resource.distribution 0.013 (0.295) Crgdp*ethnicandrel.stability 0.035*** (0.012) Natural resources Crgdp*resource.dependant (bd) -0.063 (0.040) Crgdp*resource.rich (bd) -0.047 (0.041) Crgdp*resource dependant and rich -0.063 (0.040) Other factors Crgdp*elections 0.042** (0.019) Crgdp*debt.ratio 0.004 (0.004) Crgdp*rigidity 670.223 (655.163) Crgdp*fiscal space 6.248*** (0.392) Constant 20.068*** 19.372*** -0.731 -0.263 -0.263 -0.263 -0.262 -0.416 20.722*** -0.350 (0.705) (1.708) (1.314) (0.249) (0.249) (0.249) (0.248) (0.334) (0.670) (0.287) Adjusted R-squared 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 Rmse 1.003 1.007 0.997 1.009 1.009 1.009 1.009 1.014 0.865 1.007 Joint significance (p-value) 0.000 0.002 0.011 0.011 0.012 0.011 0.000 0.007 0.548 0.000 CD test (p-value) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Order of integration I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) I(0) Cluster country country country country country country country country country country Covariates Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 687 687 743 823 823 823 823 811 410 820 # countries 47 47 48 48 50 50 50 50 26 50 Notes. Standard errors (in parentheses) clustered at the country level. All interactive variables are also introduced in level but not reported to save space. Coefficients attached to cyclicality of GDP or its interactions are multiplied by 10000. All regressions include the volatility of the terms of trade, BD for missing data in real GDP and real expenditure series, BD for five years HP cycle correction. See Table A.3 for further discussion on variables definition. *** p<0.01, ** p<0.05, * p<0.1. 37              Table B.9: Fiscal (pro)cyclicality of VAT rates, whole sample (2000-2016): Impacts of rigidity, fiscal space and business cycle (1) (2) (3) Dependent variable: Individual income tax Cyclicality of real GDP -0.105*** -10.364 -0.776 (0.036) (10.232) (0.848) Business cycle Crgdp*good times 6725549.158* 8.833e+08 58844199.784 (3520752.871) (1.607e+09) (83654229.792) Crgdp*good times*rigidity -0.190 (0.302) Crgdp*rigidity 2,166.815 (2,028.945) Rigidity of public spending 0.004 (0.011) Crgdp*good times*fiscal space -0.003 (0.004) Fiscal space 33.016 (42.190) Crgdp*fiscal space 0.000 (0.000) good times 0.033 0.001 0.062 (0.123) (0.151) (0.125) Constant -0.293 20.710*** -0.411 (0.270) (0.695) (0.320) Adjusted R-squared 1.000 1.000 1.000 Rmse 1.009 0.868 1.008 Joint significance (p-value) 0.012 0.413 0.000 CD test (p-value) 0.000 0.000 0.000 Order of integration I(0) I(0) I(0) Cluster country country country Covariates Yes Yes Yes Observations 823 410 820 # countries 50 26 50 Notes. Standard errors (in parentheses) clustered at the country level. All interactive variables are also introduced in level but not reported to save space. Coefficients attached to cyclicality of GDP or its interactions are multiplied by 10000. All regressions include the volatility of the terms of trade, BD for missing data in real GDP and real expenditure series, BD for five years HP cycle correction. See Table A.3 for further discussion on variables definition. *** p<0.01, ** p<0.05, * p<0.1. 38              Appendix C: How to mitigate procyclicality bias? Table C.1: Potential top-down solutions: fiscal rules (correlation 2009-2015) Fiscal rule (FR) without N (without) with N (with) p-value ≠ 0 0.14 0.06 National FR 580 214 0.02 (0.02) (0.03) 0.15 0.08 FR with fiscal council 572 128 0.12 (0.02) (0.04) 0.14 -0.05 FR with monitoring 580 81 0.00 (0.02) (0.06) 0.14 -0.03 FR with enforcement 580 96 0.00 (0.02) (0.04) 0.14 0.20 FR with large coverage 573 105 0.25 (0.02) (0.04) 0.14 -0.01 FR with written legal basis 573 185 0.03 (0.02) (0.03) 0.14 0.02 FR with escape clause 580 74 0.04 (0.02) (0.04) 0.14 -0.04 FR with flexibility 573 99 0.00 (0.02) (0.05) Notes: For consistency, the control group systematically exclude FRers countries, so not any country in the control group has a national FR, irrespective to the specification. 39              Figure C.1: Correlation conditional to FRs characteristics (2009-2015) 0.15 0.14 0.14 0.13 0.12 0.11 0.1 0.09 0.08 0.07 0.06 0.06 0.05 No FR FR 0.15 0.14 0.15 0.14 0.14 0.13 0.10 0.12 0.11 0.1 0.05 0.09 0.08 0.08 0.00 No FR FR with monitoring 0.07 0.06 ‐0.05 0.05 ‐0.05 No FR FR with fiscal council ‐0.10 0.15 0.14 0.20 0.19 0.1 0.17 0.15 0.14 0.05 0.13 0 0.11 No FR FR with enforcement 0.09 ‐0.05 ‐0.03 0.07 0.05 ‐0.1 No FR FR with large coverage 0.15 0.14 0.15 0.14 0.13 0.1 0.11 0.09 0.05 0.07 0.05 0 No FR FR with written legal basis 0.03 0.02 ‐0.01 0.01 ‐0.05 No FR FR with escape clause 0.15 0.14 0.1 0.05 0 No FR FR with flexibility ‐0.05 ‐0.04 40              Table C.2 : Potential top-down solutions: fiscal councils (correlation 2009-2016) Fiscal council (FC) without N (without) with N (with) p-value ≠ 0 0.13 -0.02 FC 811 96 0.00 (0.02) (0.06) 0.13 0.07 FC with large coverage 811 64 0.25 (0.02) (0.08) 0.13 0.03 FC with forecast prep. 811 66 0.07 (0.02) (0.07) 0.13 0.04 FC with forecast asses. 811 79 0.10 (0.02) (0.06) 0.13 -0.06 FC with recommendation 811 74 0.00 (0.02) (0.07) 0.13 -0.38 FC with sustainability 811 40 0.00 (0.02) (0.07) 0.13 -0.10 FC with consistency 811 73 0.00 (0.02) (0.07) 0.13 -0.29 FC evaluating costs of measures 811 54 0.00 (0.02) (0.06) 0.13 -0.08 FC with ex-post analysis 811 67 0.00 (0.02) (0.07) 0.13 0.01 FC with reports 811 93 0.01 (0.02) (0.06) FC with media impact 0.13 0.02 811 67 0.06 (0.02) (0.07) Notes: For consistency, the control group systematically exclude FCers countries, so not any country in the control group has a FC, irrespective to the specification. 41              Figure C.2: Correlation conditional to FCs characteristics (2009-2016) 0.15 0.13 0.10 0.05 0.00 No FC FC ‐0.02 ‐0.05 0.15 0.13 0.15 0.13 0.13 0.13 0.11 0.11 0.09 0.09 0.07 0.07 0.05 0.04 0.05 0.04 0.03 0.03 0.01 No FC FC with large coverage 0.01 No FC FC with forecast prep. 0.15 0.15 0.13 0.13 0.13 0.10 0.11 0.09 0.05 0.07 0.00 0.05 0.04 No FC FC with recommendation 0.03 ‐0.05 0.01 ‐0.06 No FC FC with forecast asses. ‐0.10 0.20 0.15 0.13 0.13 0.10 0.10 0.00 No FC FC with sustainability 0.05 ‐0.10 0.00 ‐0.20 No FC FC with  consistency ‐0.05 ‐0.30 ‐0.40 ‐0.10 ‐0.38 ‐0.10 0.15 0.13 0.15 0.13 0.10 0.05 0.10 0.00 No FC FC evaluating costs of 0.05 ‐0.05 measures ‐0.10 0.00 ‐0.15 No FC FC with  ex‐post analysis ‐0.20 ‐0.05 ‐0.25 ‐0.08 ‐0.30 ‐0.10 ‐0.29 0.15 0.14 0.13 0.13 0.13 0.12 0.11 0.10 0.09 0.08 0.06 0.07 0.04 0.05 0.02 0.03 0.02 0.01 0.00 0.01 No FC FC with reports No FC FC with  media impact 42              Table C.3: Potential bottom-up solutions: Political Decentralization (correlation 2009-2015) Political Decentralization (PD) without N (without) with N (with) p-value ≠ 0 0.10 0.13 PD 332 462 0.54 (0.02) (0.02) 0.14 0.17 with autonomous regions 318 126 0.43 (0.02) (0.04) 0.16 0.18 with local governments elected 192 266 0.69 (0.03) (0.03) 0.16 0.17 with state governments elected 213 112 1.00 (0.03) (0.04) with state fiscal and legislative 0.28 0.15 94 119 0.04 authority (0.05) (0.04) with state constituencies in the -0.02 0.12 60 147 0.05 upper house (0.06) (0.04) Notes: For consistency, the control group systematically exclude PDers countries, so not any country in the control group has decentralization, irrespective to the specification. 43              Figure C.3: Correlation conditional to PD characteristics (2009-2015) 0.15 0.13 0.13 0.11 0.10 0.09 0.07 0.05 0.03 0.01 No PD PD 0.18 0.17 0.18 0.18 0.16 0.16 0.16 0.14 0.13 0.14 0.12 0.12 0.10 0.10 0.08 0.08 0.06 0.06 0.04 0.04 0.02 0.02 0.00 No PD with autonomous regions 0.00 No PD with local governments elected 0.30 0.28 0.18 0.17 0.16 0.16 0.25 0.14 0.20 0.12 0.15 0.10 0.15 0.08 0.10 0.06 0.05 0.04 0.02 0.00 No PD with state fiscal & legislative 0.00 authority No PD with state governments elected 0.15 0.12 0.10 0.05 0.00 No PD with state constituencies in the upper house ‐0.02 ‐0.05 44              Table C.4 : Potential bottom-up solutions: direct democracy (correlation 2009-2016) Direct democracy (DD) without N (without) with N (with) p-value ≠ 0 0.16 0.04 DD 448 184 0.00 (0.02) (0.04) 0.16 0.04 Initiatives permitted* 448 184 0.00 (0.02) (0.04) 0.16 0.11 Referendums permitted 448 103 0.10 (0.02) (0.05) Notes: For consistency, the control group systematically exclude DDers countries, so not any country in the control group has DD irrespective to the specification. *: Same as DD. 45              Figure C.4: Correlation conditional to DD characteristics (2009-2015) 0.18 0.16 0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.04 0.02 0.00 No DD DD 0.18 0.16 0.18 0.16 0.16 0.16 0.14 0.14 0.12 0.12 0.11 0.10 0.10 0.08 0.08 0.06 0.04 0.06 0.04 0.04 0.02 0.02 0.00 No DD Initiatives permitted 0.00 No DD Referendums permitted 46              Appendix D: Difference across regions Figure D.1: (Pro)cyclicality in SSA, Resource Rich vs. Other Countries Figure D.1.a: (Pro)cyclicality in SSA countries, separating for resource-dependent countries 0.34 0.33 0.33 0.32 0.31 0.31 0.30 0.29 0.29 0.28 0.27 0.26 0.25 SSA Resource‐dependent Others SSA country Figure D.1.b: (Pro)cyclicality in SSA countries, separating for resource-rich countries 0.34 0.33 0.33 0.32 0.31 0.31 0.30 0.29 0.28 0.28 0.27 0.26 0.25 SSA Resource‐rich country Others SSA 47              Figure D.2: (Pro)cyclicality in Resource- Rich SSA countries vs Resource Rich in other regions Figure D.2.a: (Pro)cyclicality in SSA countries, separating for resource-dependent countries 0.35 0.33 0.30 0.25 0.23 0.20 0.15 0.10 Resource‐dependent SSA Resource‐dependent others regions Figure D.2.b: (Pro)cyclicality in SSA countries, separating for resource-rich countries 0.35 0.33 0.30 0.25 0.20 0.15 0.10 0.08 0.05 Resource‐rich SSA Resource‐rich others regions 48              Table D.1: Key indicators by regions Revenue ratio (%) # Code Region 2000-2008 2009-2016 1 ECA Europe and Central Asia 34.48 36.52 2 LAC Latin America and Caribbean 22.01 24.01 3 MENA Middle East and North Africa 36.97 33.72 4 SEAP South, East Asia and Pacific 17.72 19.76 5 SSA Sub-Saharan Africa 22.27 22.49 Tax (%) 2000-2008 2009-2016 1 ECA Europe and Central Asia 28.07 24.39 2 LAC Latin America and Caribbean 24.39 20.16 3 MENA Middle East and North Africa 8.11 8.50 4 SEAP South, East Asia and Pacific 13.98 17.27 5 SSA Sub-Saharan Africa 11.12 12.56 Credit ratio (%) 2000-2008 2009-2016 1 ECA Europe and Central Asia 28.96 47.09 2 LAC Latin America and Caribbean 32.50 38.75 3 MENA Middle East and North Africa 43.98 56.29 4 SEAP South, East Asia and Pacific 48.30 66.43 5 SSA Sub-Saharan Africa 16.70 23.15 Political corruption (0-1) 2000-2008 2009-2016 1 ECA Europe and Central Asia 0.46 0.48 2 LAC Latin America and Caribbean 0.48 0.51 3 MENA Middle East and North Africa 0.40 0.46 4 SEAP South, East Asia and Pacific 0.35 0.36 5 SSA Sub-Saharan Africa 0.33 0.35 Public service corruption (0-1) 2000-2008 2009-2016 1 ECA Europe and Central Asia 0.48 0.49 2 LAC Latin America and Caribbean 0.52 0.54 3 MENA Middle East and North Africa 0.38 0.42 4 SEAP South, East Asia and Pacific 0.38 0.38 5 SSA Sub-Saharan Africa 0.30 0.33 Law and order (1-6) 2000-2008 2009-2016 1 ECA Europe and Central Asia 4.13 3.92 2 LAC Latin America and Caribbean 2.69 2.40 3 MENA Middle East and North Africa 4.23 4.19 4 SEAP South, East Asia and Pacific 3.43 3.43 5 SSA Sub-Saharan Africa 2.96 2.90 Resource distribution (0-1) 2000-2008 2009-2016 1 ECA Europe and Central Asia 0.77 0.75 2 LAC Latin America and Caribbean 0.57 0.59 3 MENA Middle East and North Africa 0.49 0.52 4 SEAP South, East Asia and Pacific 0.55 0.54 5 SSA Sub-Saharan Africa 0.51 0.53 Ethnic and religious stability (0-6) 2000-2008 2009-2016 1 ECA Europe and Central Asia 1.02 0.99 2 LAC Latin America and Caribbean 4.96 5.04 3 MENA Middle East and North Africa 3.79 3.97 4 SEAP South, East Asia and Pacific 3.75 3.46 5 SSA Sub-Saharan Africa 3.61 3.67 Notes: Average levels are computed for the subperiods 2000-2008; 2009-2016. 49              Table D.2: Key indicators by group Revenue ratio (%) # Code Region 2000-2008 2009-2016 1 EG Established Graduates 24.63 27.16 2 RG Recent Graduates 27.14 28.93 3 BS Back to School 28.30 26.08 4 SS Still in School 25.07 25.73 Tax (%) 2000-2008 2009-2016 1 EG Established Graduates 11.53 13.04 2 RG Recent Graduates 18.18 20.15 3 BS Back to School 8.25 8.85 4 SS Still in School 12.79 14.50 Credit ratio (%) 2000-2008 2009-2016 1 EG Established Graduates 44.15 60.93 2 RG Recent Graduates 33.28 45.02 3 BS Back to School 39.59 56.29 4 SS Still in School 22.38 31.58 Political corruption (0-1) 2000-2008 2009-2016 1 EG Established Graduates 0.44 0.47 2 RG Recent Graduates 0.47 0.54 3 BS Back to School 0.38 0.38 4 SS Still in School 0.35 0.37 Public service corruption (0-1) 2000-2008 2009-2016 1 EG Established Graduates 0.42 0.45 2 RG Recent Graduates 0.49 0.55 3 BS Back to School 0.36 0.36 4 SS Still in School 0.37 0.38 Law and order (0-6) 2000-2008 2009-2016 1 EG Established Graduates 3.72 3.54 2 RG Recent Graduates 3.35 3.20 3 BS Back to School 3.48 3.48 4 SS Still in School 3.19 3.06 Resource distribution (0-1) 2000-2008 2009-2016 1 EG Established Graduates 0.54 0.55 2 RG Recent Graduates 0.68 0.71 3 BS Back to School 0.53 0.52 4 SS Still in School 0.57 0.58 Ethnic and religious stability (0-6) 2000-2008 2009-2016 1 EG Established Graduates 4.44 4.44 2 RG Recent Graduates 4.31 4.32 3 BS Back to School 3.55 3.66 4 SS Still in School 4.05 4.01 Notes: Average levels are computed for the subperiods 2000-2008; 2009-2016. 50              Appendix E: List of resource-dependent and resource-rich countries Table E.1: List of Resource-dependent countries in our sample (2000-2016) Resource-dependent Percent time with # Region Code countries RD revenues Oil exporters Mineral exporters Other exporters 1 Sub-Saharan Africa AGO Angola 100% yes 2 Middle East and North Africa ARE United Arab Emirates 100% yes 3 Sub-Saharan Africa BFA Burkina Faso 86% yes 4 Latin America and Caribbean BOL Bolivia 100% yes 5 Sub-Saharan Africa BWA Botswana 100% yes 6 Sub-Saharan Africa CIV Côte d'Ivoire 100% yes (cocoa) 7 Sub-Saharan Africa CMR Cameroon 100% yes 8 Sub-Saharan Africa COG Congo, Rep. 100% yes 9 Middle East and North Africa DZA Algeria 100% yes 10 Latin America and Caribbean ECU Ecuador 100% yes 11 Sub-Saharan Africa GAB Gabon 100% yes 12 Sub-Saharan Africa GIN Guinea 100% yes 13 Sub-Saharan Africa GMB Gambia, The 100% yes (artificial filament) 14 Sub-Saharan Africa GNB Guinea-Bissau 65% yes (coconut, Brazil nuts, cashew) 15 South, East Asia and Pacific IDN Indonesia 88% yes 16 Middle East and North Africa IRN Iran, Islamic Rep. 100% yes 17 Middle East and North Africa KWT Kuwait 100% yes 18 Sub-Saharan Africa LBR Liberia 53% yes 19 Sub-Saharan Africa MLI Mali 100% yes 20 Sub-Saharan Africa NGA Nigeria 100% yes 21 Middle East and North Africa OMN Oman 100% yes 22 Latin America and Caribbean PER Peru 71% yes 23 Europe and Central Asia RUS Russian Federation 94% yes 24 Middle East and North Africa SAU Saudi Arabia 100% yes 25 Sub-Saharan Africa SDN Sudan 88% yes 26 Sub-Saharan Africa SEN Senegal 94% yes 27 Sub-Saharan Africa SWZ Eswatini 100% yes 28 Sub-Saharan Africa TCD Chad 100% yes 29 Latin America and Caribbean TTO Trinidad and Tobago 100% yes 30 Latin America and Caribbean VEN Venezuela, RB 88% yes 31 South, East Asia and Pacific VNM Vietnam 88% yes (telephone sets) 32 Middle East and North Africa YEM Yemen, Rep. 100% yes Notes: The repartition is based on the principal exportation according to OEC (http://atlas.media.mit.edu/en/visualize/tree_map/hs07/export/fra/all/show/2014/ ; dataset HS07) for the last year available (2014). 51              Table E.2: List of Resource-rich countries in our sample (2000-2016) # Region Code Resource-rich countries Oil exporters Mineral exporters Other exporters 1 Sub-Saharan Africa AGO Angola yes 2 Europe and Central Asia ALB Albania yes 3 Europe and Central Asia AZE Azerbaijan yes 4 Latin America and Caribbean BOL Bolivia yes 5 Sub-Saharan Africa BWA Botswana yes 6 Sub-Saharan Africa CAF Central African Republicⁱ yes (wood) 7 Latin America and Caribbean CHL Chile yes 8 Sub-Saharan Africa CIV Côte d'Ivoire yes (cocoa) 9 Sub-Saharan Africa CMR Cameroon yes 10 Sub-Saharan Africa COG Congo, Rep. yes 11 Middle East and North Africa DZA Algeria yes 12 Latin America and Caribbean ECU Ecuador yes 13 Sub-Saharan Africa GAB Gabon yes 14 Sub-Saharan Africa GIN Guinea yes 15 Latin America and Caribbean GTM Guatemalaⁱ yes (can and beet sugar) 16 South, East Asia and Pacific IDN Indonesia yes 17 Middle East and North Africa IRN Iran, Islamic Rep. yes 18 Europe and Central Asia KGZ Kyrgyz Republicⁱ yes 19 South, East Asia and Pacific LAO Lao PDR yes (wood) 20 Sub-Saharan Africa LBR Liberia yes 21 Middle East and North Africa LBY Libya yes 22 Latin America and Caribbean MEX Mexico yes 23 Sub-Saharan Africa MLI Mali yes 24 South, East Asia and Pacific MNG Mongolia yes 25 Sub-Saharan Africa MOZ Mozambiqueⁱ yes 26 Sub-Saharan Africa NER Niger yes 27 Sub-Saharan Africa NGA Nigeria yes 28 Middle East and North Africa OMN Oman yes 29 Latin America and Caribbean PER Peru yes 30 South, East Asia and Pacific PNG Papua New Guinea yes 31 Europe and Central Asia RUS Russian Federation yes 32 Middle East and North Africa SAU Saudi Arabia yes 33 Sub-Saharan Africa SDN Sudan yes 34 Sub-Saharan Africa SLE Sierra Leoneⁱ yes 35 Middle East and North Africa SYR Syrian Arab Republic yes 36 Sub-Saharan Africa TCD Chad yes 37 Sub-Saharan Africa TGO Togoⁱ yes 38 Latin America and Caribbean TTO Trinidad and Tobago yes 39 Sub-Saharan Africa TZA Tanzaniaⁱ yes 40 Sub-Saharan Africa UGA Ugandaⁱ yes (coffee) 41 Europe and Central Asia UZB Uzbekistan yes 42 Latin America and Caribbean VEN Venezuela, RB yes 43 South, East Asia and Pacific VNM Vietnam yes (telephone sets) 44 Middle East and North Africa YEM Yemen, Rep. yes 45 Sub-Saharan Africa ZAR Congo, Dem. Rep. yes 52