Chapter

Chapter 8. The Cyclicality of Fiscal Policies in the CEMAC Region

Author(s):
Bernardin Akitoby, and Sharmini Coorey
Published Date:
August 2012
Share
  • ShareShare
Show Summary Details
Author(s)
Gaston K. Mpatswe, Sampawende J.-A. Tapsoba and Robert C. York 

To promote and sustain the Coopération Financière en Afrique Centrale (CFA) franc–pegged exchange rate regime and enhance regional economic integration, Central African Economic and Monetary Community (CEMAC) member countries (Cameroon, the Central African Republic, Chad, Equatorial Guinea, Gabon, and the Republic of Congo) are called upon to support a number of policy objectives. These objectives are translated into convergence criteria, which include achieving a nonnegative basic fiscal balance, maintaining total debt of less than 70 percent of GDP, nonaccumulation of domestic and external arrears, and annual inflation of no more than 3 percent. The adequacy and effectiveness of fiscal convergence for supporting the fixed exchange rate and economic integration are frequently questioned, because member countries have rarely observed these criterion mainly because they are flawed in that they do not account for oil production in five of the six member countries and other factors (York and others, Chapter 2 of this volume).

The convergence criteria also fail to recognize an important dimension of fiscal policy—the possibility of procyclicality. The CEMAC Commission’s narrow focus on a nonnegative fiscal balance could send misleading signals about fiscal policy performance in member countries because it does not prevent countries from spending windfall oil or commodity-export receipts when prices in these sectors rise. Such spending goes against good fiscal management, which suggests that expenditure not be influenced by temporary changes in oil (or other commodity) prices. Of course, if a price shock is permanent, a structural shift in the expenditure envelope could be prudent and consistent with long-term sustainability. Like others (Iossifov and others, 2009), this chapter argues that fiscal surveillance in the region and in individual CEMAC countries should not rely solely on the current convergence criteria but should also include fiscal indicators that consider cyclical factors, the fiscal impulse, and the behavior of non-oil aggregates and the non-oil primary balance. Yet the analysis remains cognizant of the broader concerns about procyclical fiscal policies, which can exacerbate economic fluctuations, hamper growth, and have long-term welfare implications (IMF, 2005). Also, if expansionary fiscal policies in good times are not fully offset in bad times, they may lead to large fiscal deficits and a buildup of debt and its associated problems, including possible default.

The aim of this chapter is to shed further light on the nature of fiscal policies across the CEMAC, and in particular, to assess whether there is procyclical bias—an important question for the region, but one that has attracted little empirical interest. The analysis uses a panel data approach to address this question and assesses the cyclicality of fiscal policies of CEMAC countries during 1980–2008 as compared with other countries in the subregion. It examines the potential driving factors behind fiscal policies across the CEMAC and extends the empirical literature by employing a time-varying coefficient model to look at procyclicality coefficients over time.

The next section reviews the empirical literature devoted to the CEMAC and is followed by a section that lays out the empirical model and its estimation. The final section draws conclusions and discusses policy implications. Overall, the findings support most existing literature demonstrating that fiscal policies are strongly procyclical across the region with significant evidence of persistence.

Review of the Empirical Literature

Only a few empirical studies have investigated the cyclicality of fiscal policies in CEMAC member countries. The studies that do exist have followed the mainstream literature in focusing on comovements of various indicators of fiscal policy and output cycles. Although the various studies use different econometric techniques, the modeling strategy is generally similar. A fiscal reaction function of the following generic form is estimated by

in which i and t represent country and period, respectively; gi,t and yi,t are, respectively, indicators of fiscal policy (defined, for example, as a function of the fiscal balance as share of GDP or total government expenditure) and output cycles (which could be defined either as output gaps or growth rates);χi,t represents a vector of control variables; and vi,t captures fiscal shocks. The disturbance term νi,t can be decomposed into three orthogonal components: a country-specific fixed effect νi, period random terms ut, and an idiosyncratic fiscal shock ξit. The fiscal shocks are thus expressed as (νi,t = μi + ut + ξit). The cyclicality of fiscal policy is determined by looking at the sign and the size of the coefficient β. When the indicator of fiscal policy is expressed in terms of government expenditure, procyclicality is inferred from the data if β > 0, that is, a cyclical upturn (downturn) is associated with an increase (decrease) in government spending; countercyclical if β < 0 and acyclical if β = 0. A few authors examine the fiscal balance in equation (8.1), but the majority of studies focus on the growth or level of public spending when testing for cyclicality, especially in low-income countries with limited data and in oil-producing countries (Box 8.1).

Adedeji and Williams (2007) and Weigand (2004) are the only two studies that focus on CEMAC countries. Although their interests were not on investigating fiscal cyclicality directly, the issue is raised in their analyses but the conclusions are not clear cut. Rather than government spending as the dependent variable, Adedeji and Williams (2007) regress the basic fiscal balance on output, debt-to-GDP ratio, lagged dependent variable, and other control variables. Their model is estimated using both fixed effects and the difference and system generalized method of moments (GMM) estimators on annual panel data for the six countries covering 1990–2006. They find that the coefficient on the lagged basic primary fiscal balance is positive and significant, suggesting a high degree of persistence in fiscal policy. Because current fiscal performance is strongly determined by performance in the previous year, they conclude that exogenous shocks—in the absence of automatic stabilizers—would result in procyclical fiscal policies across the region (in other words, implying comovement between the policy stance and output, which is the essence of procyclicality).

Box 8.1Fiscal Indicators and Economic Cycles

A common methodology for determining whether a fiscal policy is expansionary or contractionary is to test comovements between changes in various summary indicators of the fiscal balance (often expressed in proportion to output) and cyclical changes in output. In this context, countercyclical fiscal policy is defined as running fiscal deficits in bad times and surplus in good times, whereas procyclical fiscal policy is defined as running fiscal deficits, or a surplus lower than what would otherwise have been achieved given the cycle of the economy, during good times.

Reinhart, Kaminsky, and Végh (2004), among others, have questioned the accuracy of fiscal balance indicators in assessing the cyclicality of the fiscal stance, mainly on two grounds. First, the fiscal balance and other indicators like the revenue- and expenditure-to-GDP ratios reflect the outcomes of policy, and are affected only endogenously by the actions of policymakers. For this reason, the direction of comovements between these fiscal indicators and economic cycles might be ambiguous. Second, expressing fiscal variables as proportions of output could yield misleading results because the cyclical fiscal stance may be dominated by the cyclical behavior of output.

The assessment of the fiscal stance of oil-producing countries is a particular case in point: for example, an expansion through increased spending is masked by an improvement in the overall fiscal balance resulting from rising oil revenue. Moreover, although devised to correct such volatility from biasing policy analysis, the applicability of indicators of discretionary policy—including cyclically adjusted fiscal balances and the fiscal impulse—has limitations in the case of oil-exporting countries. These indicators rely heavily on estimates of output gaps and the elasticity of the budget to changes in output, which could raise specific issues for oil producers because (i) they are subject to substantial and frequent terms-of-trade shocks, making it difficult to identify business cycles; and (ii) unlike most industrial non–oil-producing countries, in which output fluctuates around a relatively stable trend, shocks to trend growth are the primary source of fluctuations in oil-producing countries, blurring the simple distinction between trend and cycle (Aguir and Gopinath, 2004).

Reinhart, Kaminsky, and Végh (2004) therefore advocate an approach that involves judging fiscal cyclicality by assessing the direction of comovements between fiscal policy instruments—tax rates and government spending—and economic cycles. They propose that governments pursue

  • countercyclical fiscal policy when they lower (raise) spending and raise (lower) tax rates in good (bad) times, because, for example, a reduction of spending and raising of tax rates tends to stabilize the business cycle;

  • procyclical fiscal policy when they raise (lower) spending and lower (raise) tax rates in good (bad) times; and

  • acyclical fiscal policy when spending and tax rates remain constant over the economic cycle—a pattern that will neither reinforce nor stabilize the business cycle.

For many developing countries—including those in the Central African Economic and Monetary Community—the lack of comprehensive and systematic data on tax rates suggests that government spending is probably the most reliable indicator for judging fiscal cyclicality.

In his assessment of the fiscal convergence criteria in the CEMAC, Wiegand (2004) finds some contrary evidence. He computes the elasticity of the non-oil basic fiscal balance with respect to the CFA franc oil price over time to check whether procyclicality is a matter of concern. On this basis, no significant procyclicality is found over the period 1994–2001. Weigand (2004) suggests that large short-term changes in the non-oil balances are more closely aligned with political events than with oil price movements.

Akitoby and others (2004) assess the short- and long-term behavior of government expenditure in 51 developing countries, including Cameroon and the Republic of Congo, during 1970–2002. Using an error-correction model, they find a positive short-term elasticity of government spending with regard to output and conclude that fiscal policy is procyclical in more than half the countries, including the two CEMAC members. Also, they show that output and government spending for most countries, and for at least one aggregate measure, are cointegrated, indicating a long-term relationship between them in line with “Wagner’s law” (see Akitoby and others, 2004).

With government consumption responding more than proportionately to fluctuations in output in many cases (the mean across the region was 0.91, with a standard deviation of 0.39). For CEMAC countries, procyclicality is found to be significant and low for Chad and Gabon, and high for the Republic of Congo and Equatorial Guinea.

Diallo (2008) investigates the impact of institutions on fiscal policy; he finds procyclical fiscal policies across the SSA region, but concludes these procyclical policies can be reversed by strong institutions. His empirical approach is slightly different from the generic equation (8.1), using instead a type of fiscal Taylor rule. He postulates that the fiscal stance can be captured by deviations of government spending from its trend; these deviations are themselves driven by deviations in output captured in exogenous shocks, which Diallo proxies by movements in the terms of trade from their (Hodrick-Prescott filtered) trend.1 He uses system GMM on a panel of 47 countries, including all six CEMAC members, during 1989–2002.

Lledó, Yackovlev, and Gadenne (2009) use several methods (ordinary least squares [OLS], instrumental variables, and GMM) to estimate equation (8.1) for 174 countries, including 44 in sub-Saharan Africa (SSA) (and all six CEMAC members) during 1970–2008. Multiple methods were used to account for the possible reverse causality between growth in output and government spending. They find that β is positive and statistically significant for all developing countries and more pronounced for those in SSA, although the procyclicality bias declined over time, especially after the late 1990s.

Carmignani (2010) looks at correlation coefficients between filtered real government spending and filtered real GDP growth for 37 African countries, including three from the CEMAC (Cameroon, Chad, and Gabon). He finds a statistically significant and positive relationship, which he concludes is evidence of procyclicality. CEMAC countries were classified among the procyclical countries.

Finally, using measures of the fiscal stance and fiscal impulse (both including and excluding oil revenue), Iossifov and others (2009) suggest that procyclicality was a prominent feature of fiscal policies in the CEMAC region, particularly after 2000, coinciding with the run-up in world oil prices over nearly a decade. The study noted that the ripple effects of terms-of-trade and real effective exchange rate shocks on the business cycles in the region are potentially magnified through fiscal policies. In their assessment, procyclical policies may exaggerate these effects whereas countercyclical policies could potentially mitigate them.

Empirical Model

To assess fiscal cyclicality, the ideal would be to estimate a fiscal reaction function in which real government expenditure responds to changes in real output cycles and other factors noted in the literature. A form of this reaction function is the Taylor-type rule without the inflation terms, as used by Diallo (2008). This reaction function assesses the direction and level of comovements between government spending and output deviations from their respective steady state trends:

in which (Gi,tGi,t*) denotes the deviation of actual government spending from its long-run trend, and (Yi,tYi,t*) is the deviation of output (or real GDP) from its long-term trend.2 The terms Xi,t and νi,t are defined as in equation (8.1). As above, the cyclicality of fiscal policy is determined by the sign and size of the coefficient on output β, which represents the short-run fiscal response to the economic cycle. Note that equation (8.2) is formulated with respect to comovements between spending and economic cycles, and not fiscal balance indicators. For the reasons outlined in Box 8.1, fiscal balance indicators may not be appropriate for judging cyclicality in low-income and oil-producing countries.

To estimate this reaction function, adequate representations for the unobserved long-run values of government spending and output would be needed. One approach would be to use a filter to estimate these values and then run the model based on spending and output gaps (e.g., Diallo, 2008).3 Another approach would be to estimate a dynamic equation with the lagged values of these variables used as proxies for potential output and run the regression in first difference (e.g., Thornton, 2008)4:

in which Δ is the first difference operator, Log is the logarithm, coefficient β would capture the short-term overall cyclical behavior of government spending without differentiating between discretionary fiscal actions and automatic responses to the economic cycle, and coefficient γ would capture possible inertia effects or long-term mean reversion in government spending. This coefficient is expected to have a positive sign and be less than 1.5 Because automatic stabilizers are likely to be small in CEMAC countries, equation (8.3) is a reasonable approximation of discretionary fiscal policy. In addition, data to estimate equation (8.3) are easily available, thus avoiding the shortcomings of using the filtering approach to estimate potential output (Box 8.1).

A consistent identification of β requires addressing the potential endogeneity bias stemming from two factors: (i) the reverse causality between public expenditure and output given that output is likely to be responsive to a fiscal stimulus; and (ii) the simultaneity bias caused by the fact that omitted variables, including country-specific factors, likely are correlated with government spending. Following the work of Lledó, Yackovlev, and Gadenne (2009), and others, the present analysis addresses the endogeneity problem by estimating equation (8.3) with GMM procedures (difference and systems) adapted for dynamic-panel estimation and fixed-effects estimators, which can control for country-specific factors. It should be noted that the lagged dependent variable will be correlated with the regression error when using fixed effects. To determine the consistency of the estimated coefficients, the results are reported in this chapter with and without the lagged dependent variable. Endogeneity is less a problem if the results do not change significantly. This analysis uses the two methods to improve on previous empirical work and to determine which leads to the most robust results.6 Year-specific dummy variables are included in the estimates to control for the CEMAC’s covariant shocks, thus ensuring that the identified fiscal behavior is specific to fiscal authorities only.

The estimation of equation (8.3) is carried out in two steps. The first step focuses solely on addressing the null hypothesis to determine whether fiscal policy is acyclical across the region (β = 0). This coefficient is then allowed to be country specific and time varying, as specified in equation in (8.4)

which is a modified form of equation (8.3), excluding the vector of control variables. Following Aghion and Marinescu (2008), the procedure estimates equation (8.4) using local Gaussian-weighted OLS. This technique determines the time-varying cyclicality coefficient β^i,t for country i at year t by using all observations and assigning greater weights to those observations closest to the reference year. This is achieved by giving a Gaussian-centered weight to the reference period. A 10-year rolling window approach is also applied to ensure that the cyclicality captured is a result of transitory discretionary fiscal policies. If τ denotes the length of the rolling window, then the error term ϑit follows a normal distribution function: 0,σ2Wt(τ) with Wt(τ)=1σ2Πexp((τt)22σ2),τϵ(t5,t+4). As suggested in Aghion and Marinescu (2008), the smoothing parameter σ is arbitrarily set to 5 because the results are qualitatively robust to slight changes of this parameter.7

In the second step, the analysis pools the estimated time-varying cyclicality coefficients (β^it) and regresses them against the vector of control variables that are thought to be the main causes of cyclicality in developing countries (as noted in the literature review above). This is represented in equation (8.5)

in which Zi,tk denotes the explanatory variables and ηit is the classical error term encompassing a country-specific fixed effect and an idiosyncratic fiscal shock. The explanatory variables include political and institutional factors to capture rent-seeking behavior that could drive profligacy in public spending, weaknesses in governance (and corruption), and public financial management, election cycles, and financing constraints.8 This chapter proxies political institutional factors and governance through an election variable (measuring a political or electoral cycle), Freedom House’s Civil Liberties and Political Rights indices (assessing the extent of democratization and the role of checks and balances on the fiscal authorities), and the World Bank’s Country Policy and Institutional Assessment index (measuring the quality of policies and institutions). Financing constraints are captured by aid-to-GDP ratios. Aid dependency is important and remains the main source of financing for developing countries when they are shut out of capital markets or during economic downturns.

In addition to these usual factors, the analysis tests the potential influence of variables measuring fiscal space, which may affect a government’s ability to conduct active spending policy. Fiscal space is defined as the room to maneuver without jeopardizing fiscal sustainability.9 The role of IMF programs in creating such space is also explored. IMF-supported programs are viewed as policy anchors because they are generally accompanied by economic and structural reforms that could influence the conduct of fiscal policy. Accordingly, fiscal space is proxied by the current inflation rate, the lagged public-debt-to-GDP ratio, and IMF program status.

Given the heterogeneity across member countries and the prominence of oil in the CEMAC region (five of the six countries are oil exporters), the analysis controls for terms-of-trade shocks and the level of development using real GDP per capita. These factors could strongly influence the ability of the countries in the region to implement countercyclical fiscal policies.

Finally, a dummy variable is introduced among the regressors to take account of the effects of the 1994 devaluation of the CFA franc, with the value 1 for the subsequent period. The assumption is that structural adjustments that accompanied the devaluation—as well as the subsequent adoption of regional convergence criteria—might have imposed additional constraints on the fiscal authorities’ policy behavior.

Data Set

The macroeconomic data for the six CEMAC countries and other SSA countries come from the IMF’s World Economic Outlook database and cover 1980–2008. Missing data caused an unbalanced panel of 164 observations. From this database are extracted general-government total expenditure (including the breakdown into consumption and investment),10 the GDP deflator, GDP per capita, the consumer price index, trade shares, and terms-of-trade of goods and services. Fiscal data and explanatory variables are converted into real terms using the GDP deflator, when necessary, with 2000 as the base year. All indices are scaled to 2000 as well. Terms-of-trade shocks are specifically computed by weighting the unexpected change of the terms-of-trade variable—using the Hodrick-Prescott filter with the smoothing parameter set to 100—with trade shares. The intention is to model the exposure and vulnerability of each country to these shocks, especially ones arising from the volatility of oil prices.

The other variables are taken from various sources. The election variable (taking the value 1 in an election year) is retrieved from the website African Elections Database (http://africanelections.tripod.com). The Freedom House Index (ranging between 1 and 7) is a measure of the average of the Civil Liberties and Political Rights subindices and comes from its website (http://www.freedom-house.org). A higher value of the index is associated with less civil and political liberty. The Country Policy and Institutional Assessment score is from the World Bank and was rescaled between 0 and 1 to make different years’ assessments comparable. A higher score corresponds to better-quality institutions and policies. The IMF program variable (taking the value 1 for countries under a program for a given year) and the public-debt-to-GDP ratio are compiled from other IMF databases.11 Finally, aid-to-GDP ratios are drawn from Organization for Economic Cooperation and Development data.

Estimation Results

Table 8.1 reports the system and difference GMM estimates of the cyclicality of real public expenditure, consumption, and investment across SSA and the CEMAC region as specified in equation (8.4).12 The optimal lag structure of internal instruments in the regional data set is four. Based on the classical appraisal criteria, the GMM results are internally consistent for the SSA estimates but not for the CEMAC subregion alone.

  • For the SSA region as a whole, the results align with the empirical literature: there is strong evidence that fiscal policies are highly procyclical. The coefficient on output growth is close to 1 and is statistically significant at the 1 percent level. The analysis finds that procyclicality is also evident (and statistically significant at the 1 percent level) when considering public consumption and public investment as the dependent variables.13

  • The behavior of public consumption is less procyclical than public investment, which has an estimated coefficient on output growth of close to 2. This implies that investment is extremely responsive to economic cycles (i.e., a 1 percent increase in output growth leads to a 2 percent increase in the growth of public investment).

  • When CEMAC countries are identified separately in the regression (through a dummy variable ΔLogY x CEMAC; the CEMAC dummy variable takes the value 1 for member countries and 0 otherwise), the results are similar. The corresponding coefficient—which approximates the difference in fiscal cyclicality—is not statistically significant. This implies that CEMAC countries’ fiscal policies also exhibit procyclicality over the sample period and that public investment is highly reactive to the economic cycle. In other words, governments in the CEMAC region tend to increase (cut) public investment rather than other types of spending during good (bad) times.

  • Consistent with most of the empirical literature cited above, the inertia effect captured by the coefficient on the lagged fiscal policy variable is negative and statistically significant for public investment. From the spending side of the budget, this provides some evidence that the behavior of fiscal policies is consistent with long-term sustainability, although not for public investment.

Table 8.1GMM Estimates of Cyclical Fiscal Policy in Sub-Saharan Africa, 1980-2008
Difference GMMSystem GMM
Dependent variable: ΔLogGTotal Public ExpenditurePublic ConsumptionPublic InvestmentTotal Public ExpenditurePublic ConsumptionPublic InvestmentTotal Public ExpenditurePublic ConsumptionPublic InvestmentTotal Public ExpenditurePublic ConsumptionPublic Investment
ΔLogY0.893***0.444*1.956***0.887**0.3722.102***0.950***0.516**1.925***0.954**0.4642.088***
(0.244)(0.249)(0.419)(0.372)(0.369)(0.484)(0.238)(0.239)(0.437)(0.358)(0.344)(0.490)
ΔLog/interacted0.0240.212–0.402–0.0050.152–0.470
with CEMAC(0.427)(0.414)(0.691)(0.412)(0.422)(0.782)
ΔLogG-,–0.058–0.038–0.137***–0.059–0.038–0.138***–0.054–0.038–0.141***–0.055–0.039–0.141***
(0.044)(0.057)(0.046)(0.043)(0.057)(0.047)(0.041)(0.058)(0.046)(0.041)(0.057)(0.046)
Constant0.007

(0.035)
0.011

(0.035)
–0.011

(0.049)
0.007

(0.037)
0.013

(0.037)
–0.015

(0.049)
Year dummiesYesYesYesYesYesYesYesYesYesYesYesYes
Observations905921921905921921943959959943959959
Countries383838383838383838383838
Instruments323232363636353535404040
Hansen0.3310.2930.8350.4720.08430.7360.6010.3950.6260.7720.1230.783
AR(1)0.0000.0060.0000.0000.0060.0000.0000.0060.0000.0000.0060.000
AR(2)0.1420.3660.270.1390.3840.2530.1940.3870.2540.1920.3970.235
Source: Authors’ estimates.Note: Standard errors in parentheses; AR(1) and AR(2) denote the p values of Arellano-Bond tests for the first and second order autocorrelation. CEMAC = Central African Economic and Monetary Community;GMM = generalized method of moments.***significant at the 1 percent level.**significant at the 5 percent level.*significant at the 10 percent level.
Source: Authors’ estimates.Note: Standard errors in parentheses; AR(1) and AR(2) denote the p values of Arellano-Bond tests for the first and second order autocorrelation. CEMAC = Central African Economic and Monetary Community;GMM = generalized method of moments.***significant at the 1 percent level.**significant at the 5 percent level.*significant at the 10 percent level.

The procedure also used the GMM for the CEMAC only; however, these results are not internally consistent. The inconsistency occurs because the number of internal instruments far exceeds the number of countries; consequently, the equation is overidentified and the validity of using GMM (to address the endogeneity bias) disappears.14 Instead, the analysis relies on the fixed-effects estimates for the CEMAC countries, which are reported in Table 8.2 for real public expenditure, consumption, and investment. These results mimic those from the GMM estimates over the entire SSA sample. The table also displays the results without the lagged dependent variable to assess the extent of the possible correlation with the error component. The results remained qualitatively robust in these different specifications, and the table reports those including the lagged dependent variable to address the notion of fiscal sustainability (or more correctly, inertia).

  • For the CEMAC, there is strong evidence of procyclicality because the coefficient on output growth is positive and statistically significant, with a magnitude similar to that of SSA (presented in Table 8.1). Once again, public investment is shown to be the most reactive to economic cycles, with elasticity above 1. Public consumption is also less responsive, in line with wider sample results.

  • The coefficient on the lagged fiscal variable is negative, albeit weakly significant (at the 5 and 10 percent levels) for total expenditure; it is also negative but not statistically significant for public investment. This reveals the possibility of inertia and variability in public spending, but the effect could disappear over time (as is implied by the absolute value of the estimated coefficient of less than 1). York and Zhan (2009) reinforce the notion of persistence in government spending in considering the long-term sustainability of CEMAC oil-producing countries under a permanent-income framework. According to that study, public spending has driven the non-oil primary fiscal deficits too high, beyond levels consistent with the region’s proven oil reserves and the relatively short time horizon of future production (or long-term wealth, which is the discounted present value of the future stream of oil production).

Table 8.2Fixed Effects Estimates of Cyclical Fiscal Policy in the CEMAC, 1980–2008
Dependent Variable: ΔLogGTotal Public ExpenditurePublic ConsumptionPublic InvestmentTotal Public ExpenditurePublic ConsumptionPublic Investment
With the Lagged Dependent VariableWithout the Lagged Dependent Variable
ΔLogY0.932***0.652***1.105**0.918***0.674***1.092**
(0.133)(0.098)(0.356)(0.121)(0.111)(0.308)
ΔLogG–1−20.200*

(0.090)
0.040

(0.063)
−20.151

(0.145)
Constant−0.049−0.011−0.1370.255***0.1230.041
(0.030)(0.065)(0.131)(0.016)(0.079)(0.195)
Year dummiesYesYesYesYesYesYes
Observations155164157161170163
Countries666666
R20.3330.2980.1540.3220.2980.138
Source: Authors’ estimates.Note: Standard errors in parentheses. CEMAC = Central African Economic and Monetary Community.***significant at the 1 percent level.**significant at the 5 percent level.*significant at the 10 percent level.
Source: Authors’ estimates.Note: Standard errors in parentheses. CEMAC = Central African Economic and Monetary Community.***significant at the 1 percent level.**significant at the 5 percent level.*significant at the 10 percent level.

equation (8.5) is used to shed light on the time-varying cyclicality coefficient by country and the OLS estimates for equation (8.4) to derive individual country coefficients. These coefficients are summarized in Table 8.3 and plotted in Figure 8.1. Both the coefficients15 and the figure show that CEMAC countries, except Chad, exhibit a high degree of procyclicality over a long period. There are, however, some significant cross-country differences in the cyclical behavior of fiscal policies. The cyclicality coefficient moves widely in the Central African Republic, the Republic of Congo, and Gabon, whereas Cameroon exhibits a sustained upward trend.

Table 8.3Fiscal Cyclicality in the CEMAC
Total Public ExpenditurePublic ConsumptionPublic Investment
Country estimates1
Cameroon1.207**1.306**1.012***
Central African Republic2.039*1.0365.741**
Chad20.48020.66520.129
Congo0.5690.7180.748
Equatorial Guinea0.861***0.736***0.880**
Gabon4.967***0.717**9.099***
Time-varying estimates
Mean1.8791.0392.863
Standard deviation1.7891.1013.621
Maximum6.1264.55510.842
Minimum−1.008−1.531−1.092
Number of times (β > 0148 (92%)155 (91%)138 (85%)
Number of times (β > 183 (52%)67 (39%)103 (63%)
Number of times (β > 013 (8%)15 (9%)25 (15%)
Total161170163
Source: Authors’ estimates.Note: Percentage of total coefficients estimated in parentheses.

Estimates include lagged dependent variable in right-hand side variables.

***significant at the 1 percent level.**significant at the 5 percent level.*significant at the 10 percent level.
Source: Authors’ estimates.Note: Percentage of total coefficients estimated in parentheses.

Estimates include lagged dependent variable in right-hand side variables.

***significant at the 1 percent level.**significant at the 5 percent level.*significant at the 10 percent level.

Figure 8.1Time-Varying Fiscal Cyclicality Coefficients in the CEMAC, 1980–2008

Source: Authors’ estimates.

Note: CEMAC = Central African Economic and Monetary Community.

The result for Chad was unexpected and requires future investigation. The estimates suggest that after 1994, all components of public spending tended to be countercyclical, while the empirical literature is ambiguous. For example, Carmignani (2010) estimates a negative but insignificant coefficient on government consumption during 1990–2007 (closer to the time frame of this chapter), while Thornton (2008) finds strong evidence of procyclicality for Chadian expenditure over a longer time frame (1960–2003).

Estimation of equation (8.5) also allows the factors driving the procyclicality of total government expenditure to be explored. Table 8.4 reports the results of the regression of six sets of variables that have been shown in the empirical literature to influence the cyclicality of public spending: political and institutional factors, financing constraints, fiscal space, the level of economic development, external shocks, and structural change.

  • Political and institutional factors. The investigation anticipated that elections would have a positive influence on the procyclicality of government spending and that higher-quality institutions would dampen profligacy, while political freedom would have an ambiguous effect. The prior expectations are only partially borne out by the data. The quality of institutions plays an important and quantitatively significant role in controlling total government expenditure and public investment in CEMAC countries, as does an improvement in political freedom. Election spending cycles are absent in the data, which is unexpected but not empirically significant.

  • Financing constraints. The impact of financing constraints (proxied by the foreign-aid-to-GDP ratio) on fiscal cyclicality is ambiguous. On the one hand, high dependency on aid for government spending heightens fiscal procyclicality because external assistance is empirically viewed as volatile, unpredictable, and overwhelmingly procyclical in many recipient countries (Pallage and Robe, 2001; and Bulir and Hamann, 2008). Disbursements increase during expansion episodes and may stop suddenly in downturns or during political instability and subsequent periods of slow growth or recession. Meanwhile, foreign aid is likely to become the only available financing for government spending during a sharp decline in commodity exports and could help the conduct of countercyclical policy by relaxing the financing constraint. As in Thornton (2008), this analysis finds that aid drives the procyclicality of total expenditure, especially public investment. The coefficients on these fiscal variables are strongly positive and statistically significant. The claim that foreign aid is itself procyclical and triggers the profligacy of public spending is, therefore, evident across the CEMAC region.

  • Fiscal space. The notion of fiscal space is accounted for through the debt-to-GDP ratio (lagged), inflation, and the existence of an IMF-supported program. The existence of an IMF-supported program was expected to dampen procyclicality, whereas lower inflation and debt might lead to the opposite effect. It was found that the debt-to-GDP ratio and inflation do not have an empirically significant impact on public spending when the analysis controls for the existence of an IMF-supported program. Indeed, the data show that an IMF-supported program encourages fiscal discipline and mitigates procyclical behavior, with a statistically significant impact on total expenditure and public consumption but not significant on public investment.

  • Level of economic development. Following the conclusions of most of the previous empirical studies that fiscal policies are less procyclical in wealthier countries (especially advanced countries), both less procyclicality in richer CEMAC countries and a negative coefficient on GDP per capita would be expected. The data, however, reveal a positive and statistically significant relationship in the regression for total expenditure, suggesting that the wealthier CEMAC countries (notably Congo and Gabon in Figure 8.1) behave more procyclically. Perhaps this reflects their historical tendency to overspend oil windfall revenue, as noted in York and Zhan (2009).

  • External (terms-of-trade) shocks. Frequent and large terms-of-trade shocks stemming from oil price volatility would be expected to spur fiscal profligacy in CEMAC countries. This analysis finds that an expansion of public expenditure is heightened by unanticipated changes in terms of trade. This is inconsistent with the work of Talvi and Végh (2005), who postulate that under an unstable environment (i.e., one with frequent shocks) and institutional weakness, it might be optimal for budgetary authorities in developing oil-producing countries to spend the windfall revenue during the good times, partly in response to demands from different political and social groups.

  • Structural change. The impact of structural changes on the cyclicality of regional fiscal policies is ambiguous. The analysis introduces a dummy variable for the period after the 1994 devaluation of the CFA franc, with a value of 1 in the post-1994 period. The dummy variable is not statistically significant, as reported in Table 8.4. This finding is not consistent with suggestions from others (e.g., see Lledó, Yackovlev, and Gadenne, 2009) that after 1996 and during the recent global financial crisis, public spending has tended to be countercyclical across the SSA countries more generally.

Table 8.4Determinants of Cyclical Fiscal Policy in the CEMAC, 1980–2008
Estimator: Fixed Effects

Dependent Variable:

β-Coefficient in equation (8.5)
Total PublicExpenditurePublicConsumptionPublicInvestment
Political and institutional factors
Election−0.263−0.222−0.197−0.221−0.243−0.047
(0.192)(0.207)(0.286)(0.267)(0.208)(0.502)
Freedom House index−0.933***−0.912***−0.294−0.284−2.401**−2.344**
(0.353)(0.312)(0.257)(0.278)(1.157)(1.010)
CPIA score−3.544** (1.675)−0.300 (2.154)−12.264** (5.004)
Financing constraints Aid-to-GDP ratio2.782**4.526**1.1772.3955.5629.436***
(1.168)(1.818)(2.246)(2.647)(3.658)(3.124)
Fiscal space Debt-to-GDP ratio (- 1)0.4200.0650.5560.456−0.391−1.474
(0.393)(0.307)(0.349)(0.288)(1.150)(1.132)
Inflation0.2682.8922.7853.957−3.0044.157
(2.472)(3.617)(1.824)(3.174)(6.236)(6.855)
IMF program−1.006**

(0.455)
−0.967*

(0.561)
−1.699

(1.112)
Level of development
Log Y per capita0.924*1.337***0.4890.6730.2621.387
(0.491)(0.501)(0.573)(0.603)(1.760)(1.509)
External shocks
Terms-of-trade shocks0.278**0.415*0.2880.4440.3010.482
(0.138)(0.212)(0.220)(0.325)(0.588)(0.685)
Structural break
Dummy 19940.0630.334−0.0610.3460.6810.473
(0.836)(0.979)(1.123)(1.746)(1.340)(1.591)
Constant−5.350−8.410−4.518−6.32413.2266.130
(8.287)(7.609)(7.285)(7.368)(–6.173)(–1.839)
Year dummiesYesYesYesYesYesYes
Observations121121121121121121
Countries666666
R20.5060.5890.2960.3660.3810.484
Source: Authors’ estimates.Note: CPIA = Country Policy and Institutional Assessment.Standard errors in parentheses.***significant at the 1 percent level.**significant at the 5 percent level.*significant at the 10 percent level.
Source: Authors’ estimates.Note: CPIA = Country Policy and Institutional Assessment.Standard errors in parentheses.***significant at the 1 percent level.**significant at the 5 percent level.*significant at the 10 percent level.

Conclusions and Policy Implications

The empirical analysis confirms the results of previous studies addressing the behavior of government spending over the economic cycle. The panel data for the six CEMAC countries—Cameroon, the Central African Republic, Chad, the Republic of Congo, Equatorial Guinea, and Gabon—during 1980–2008 provide strong evidence of procyclicality. This result is consistent with public spending behavior observed in other studies across the sub-Saharan region more generally, although this analysis provides further insights. The estimated cyclicality coefficients vary by country and over time—an interesting result that should be assessed more thoroughly to deepen the understanding of fiscal policies across these six countries.

For three of the six CEMAC countries, total public expenditure responds more than proportionately to fluctuations in output, with the reaction of the public investment component extremely high. This has probably amplified economic cycles across the subregion and prevented countercyclical fiscal policies from playing an important role in mitigating a decline in output caused by external shocks (Iossifov and others, 2009). In a region buffeted by frequent external shocks stemming from volatility in the prices of oil and other commodities, this is cause for concern. The highly responsive behavior of public investment to output fluctuations suggests weaknesses in public financial management across the CEMAC. The tendency to use windfall revenue to boost such spending when oil prices rise may not be appropriate, given the importance of maintaining tight expenditure control over public investment programs, especially in light of the capacity and institutional constraints in the subregion. This ramped-up spending can easily be of poor quality and wasted, with adverse implications for building the physical and human capital needed to generate long-term wealth in a region endowed with considerable natural resources but burdened by relatively high poverty.

This analysis shows that institutional weaknesses and poor governance partly explain the procyclicality of CEMAC fiscal policies, as does foreign aid, probably because of its own procyclical behavior. In contrast, the existence of an IMF-supported program could help provide a counterbalancing influence by attenuating—but not totally eliminating—this procyclical bias. The level of debt is not found to be a statistically significant driver of procyclicality across the CEMAC region. However, the coefficient on the lagged dependent variable is negative, suggesting some concern about the persistence of and possible swings in government spending, which could have adverse consequences for the accumulation of public debt.

Finally, the CEMAC Commission does not address the possibility of cyclicality in assessing its convergence criteria, which could send misleading signals about fiscal policy performance in member countries. As shown, the focus on a nonnegative fiscal balance alone can accommodate procyclical behavior because it does not prevent members from spending windfall oil or commodity-export receipts as prices rise. This violates principles of good fiscal management and fails to recognize the broader concerns about procyclical fiscal policies, including the potential for exacerbating economic volatility and hindering growth, and the policies’ long-term welfare implications.

References

    AbbasA.S.M.N.BelhocineA.ElGanainy and M.Horton2010A Historical Public Debt Database,IMF Working Paper 10/245 (Washington: International Monetary Fund).

    AdedejiO. and O.Williams2007Fiscal Reaction Functions in the CFA Zone: An Analytical Perspective,IMF Working Paper 07/232 (Washington: International Monetary Fund).

    AghionP. and I.Marinescu2008Cyclical Budgetary Policy and Economic Growth: What Do We Learn from OECD Panel Data?NBER Macroeconomic Annual 2007 pp. 25193.

    AguirM. and G.Gopinath2004Emerging Market Business Cycles: The Cycle is the Trend,NBER Working Paper No. 10734 (Cambridge, Massachusetts: National Bureau of Economic Research).

    AkitobyB.B.ClementsS.Gupta and G.Inchauste2004The Cyclical and Long-Term Behavior of Government Expenditures in Developing Countries,IMF Working Paper 04/202 (Washington: International Monetary Fund).

    AlesinaA.F.Campante and G.Tabellini2008Why Is Fiscal Policy Often Procyclical?Journal of the European Economic Association Vol. 6.No. 5 pp. 100636.

    BulirA. and A.Hamann2008Volatility of Development Aid: From the Frying Pan into the Fire?World Development Vol. 36No. 10 pp. 204866.

    CarmignaniF.2010Cyclical Fiscal Policy in Africa,Journal of Policy Modeling Vol. 32No. 2 pp. 25467.

    DialloO.2008Tortuous Road Toward Countercyclical Fiscal Policy: Lessons From Democratized Sub-Saharan Africa,Journal of Policy Modeling Vol. 31No. 1 pp. 3650.

    GavinM. and R.Perotti1997Fiscal Policy in Latin America,NBER Macroeconomics Annual Vol. 12 (Cambridge, Massachusetts: MIT Press) pp. 1171.

    HellerP.S.2005Understanding Fiscal Space,IMF Policy Discussion Paper 05/4 (Washington: International Monetary Fund).

    International Monetary Fund2005Cyclicality of Fiscal Policy and Cyclically Adjusted Fiscal Balances,SM/05/393 (Washington).

    International Monetary Fund2010Emerging from the Global Crisis: Macroeconomic Challenges Facing Low-Income Countries (Washington).

    IossifovP.N.KinoshitaM.TakebeR.York and Z.Zhan2009Improving Surveillance Across the CEMAC Region,IMF Working Paper 09/260 (Washington: International Monetary Fund).

    LledóV.I.Yackovlev and L.Gadenne2009Cyclical Patterns of Government Expenditures in Sub-Saharan Africa: Facts and Factors,IMF Working Paper 09/274 (Washington: International Monetary Fund).

    PallageS. and M.Robe2001Foreign Aid and the Business Cycle,Review of International Economics Vol. 9No. 4 pp. 64172.

    ReinhartC.G.Kaminsky and C.Végh2004When It Rains It Pours: Procyclical Capital Flows and Macroeconomic Policies,NBER Macroeconomics Annual 2004 (Cambridge, Massachusetts: National Bureau of Economic Research).

    TalviE. and C.Végh2005Tax Base Variability and Procyclical Fiscal Policy in Developing Countries,Journal of Development Economics Vol. 78No. 1 pp. 15690.

    ThorntonJ.2008Explaining Procyclical Fiscal Policy in African Countries,Journal of African Economies Vol. 17No. 3 pp. 45164.

    TornellA. and P.Lane1999The Voracity Effect,American Economic Review Vol. 89No. 1 pp. 2246.

    WeigandJ.2004Fiscal Surveillance in a Petro Zone: The Case of the CEMAC,IMF Working Paper 04/8 (Washington: International Monetary Fund).

    YorkR. and Z.Zhan2009Fiscal Sustainability and Vulnerability in Oil-Producing Sub-Saharan African Countries,IMF Working Paper 09/174 (Washington: International Monetary Fund).

In Diallo (2008), terms-of-trade shocks are taken as an explanatory variable in place of the deviations of real GDP from its long-run trend. His formulation is built on the argument that shocks, such as terms-of-trade volatility—which is among the most important sources of shocks with which African countries are confronted—negatively affect the design and implementation of sound economic policies.

The deviation terms could also be defined as spending gaps and output gaps, respectively, if related variables are expressed in logarithms or if the cyclical components are normalized onto the filtered trends.

Box 8.1 highlights the difficulties of estimating output gaps, particularly for oil-exporting countries.

This methodology differs slightly from Thornton (2008), however, because it uses the panel dimension of the data.

For a given time series process kt = γkt–1 or equivalently Δkt = γΔkt – 1, mean reversion implies that 0 < γ < 1. The series will either oscillate around the mean or drift away from the mean unless |γ| < 1.

The system and difference GMM procedure is particularly adapted for dynamic panels because it enables correction for any potential correlation between explanatory variables and country-specific factors. It does not require external instruments and uses lagged variables to address the endogeneity bias. The difference GMM yields similar results to the system GMM while imposing fewer restrictions on the correlations between the instruments and the error term. The within fixed-effects estimator only controls for country-specific factors and mitigates the simultaneity bias. It does not address problems such as reverse causality.

The β coefficients without control variables would be biased upward. However, estimating the coefficients without control variables allows us to explain fiscal procyclicality using the control variables.

For instance, Tornell and Lane (1999) have shown that if there are no institutional controls to limit policy discretion, the risk of overspending the windfall revenue during good times is high. Such profligacy tends to be ubiquitous in volatile environments (Talvi and Végh, 2005), corrupted regimes (Alesina, Campante, and Tabellini, 2008), and where weak institutions prevail with fewer checks on the executive (Diallo, 2008). Also, developing countries’ lack of access to capital markets triggers fiscal profligacy during downturns as concerns increase about government creditworthiness and fiscal sustainability (Gavin and Perotti, 1997).

Heller (2005) defines fiscal space as the availability of budgetary room that allows a government to provide resources for a desired purpose without any prejudice to the sustainability of a government’s financial position or the stability of the economy.

Across the CEMAC, data on general government mainly reflect the central government.

In particular, debt data are taken from Abbas and others (2010).

SSA includes 44 countries that are grouped in the IMF’s African Department. It does not include the countries in North Africa.

An IMF staff paper (IMF, 2010) concluded that low-income countries, including those in SSA, were able to use built-up fiscal space to conduct countercyclical fiscal policies during the recent global financial crisis. This difference in conclusions from this chapter could reflect the data sample—the data for the present analysis runs only through 2008, just before the crisis.

Simply by being numerous, instruments overfit the instrumented variables, failing to expunge their endogenous components and biasing the coefficient estimates toward those from nonincremental estimators. As a consequence, the overidentification test is biased toward the null hypothesis of the validity of instruments. We found that the probability of the Hansen J-statistic test equals 1 in all cases. Moreover, the first- and second-order autocorrelation tests are also biased toward the null hypothesis of the presence of autocorrelation.

Individual country coefficients should be viewed as indicative only, given the limited time series and the use of OLS.

    Other Resources Citing This Publication