Fiscal Politics
Chapter

Chapter 3. Fiscal Policy over the Election Cycle in Low-Income Countries

Author(s):
Vitor Gaspar, Sanjeev Gupta, and Carlos Mulas-Granados
Published Date:
April 2017
Share
  • ShareShare
Show Summary Details
Author(s)
Christian Ebeke and Dilan Ölçer 

Introduction

A growing literature assesses the detrimental effects of policy volatility on long-term growth and aggregate welfare (Fatás and Mihov 2003, 2012). One source of policy volatility might be related to national elections and the incumbent’s incentive to use economic policy instruments for reelection purposes. Reelection-minded incumbents might have an incentive to use fiscal and monetary policy in such a way that during election years, public spending or money aggregates increase to satisfy the median voter despite potential adverse effects on fiscal sustainability and aggregate macroeconomic stability.1 These cycles appear because of asymmetric information—voters lack full information about the incumbent’s competencies.

Empirical studies on the political business cycle from the 1970s until the 1990s focused almost entirely on advanced economies and generally did not find a regular statistically significant evidence of cycles.2 However, more recent studies have shown the effects of politically driven economic cycles in developing countries on government current expenditures, indirect tax revenues, budget deficits (Brender and Drazen 2005; Shi and Svensson 2006; Block 2002; Schuknecht 2000; Vergne 2009; Drazen and Eslava 2010; Ehrhart 2012), and monetary aggregates (Fouda 1997; Block 2002).

Several limitations and pending issues remain. First, most of these studies did not explicitly focus on low-income countries (LICs) but rather pooled together developing countries. The analysis in this chapter focuses on LICs because they are particularly vulnerable during election cycles. With weaker institutional capacity and poor transparency in budgets, these countries face greater risks that election-related fiscal policies will be conducted. By depleting their fiscal buffers during election years, LICs further increase their macroeconomic vulnerability and limit their ability to guard themselves from external shocks. Therefore, it is important to better understand the composition of political budget cycles in these countries and consider ways to mitigate related fiscal policy volatility.

Second, although various studies provide insights into what happens to specific variables during elections, they do not typically focus on the composition of the postelection adjustment. Block (2002) analyzes, in a sample restricted to sub-Saharan Africa, a number of fiscal and monetary variables during and after elections and concludes that government spending shifts toward more visible, current expenditures and away from public investment. However, his analysis has several limitations. The period under study is restricted to 1980 to 1995, although for many countries in sub-Saharan Africa, elections were not competitive before 1990, and even during the first half of the 1990s. Moreover, the analysis is limited to presidential systems to address issues related to endogenously timed elections, yet this restriction induces an important selection bias.3

This chapter investigates the behavior of a comprehensive set of fiscal variables during and two years after national elections. It seeks to shed light on the main form of fiscal expansion during the election year and the composition of the fiscal retrenchment (if any) in the subsequent years. This analysis uses a recent data set, National Elections across Democracy and Autocracy (NELDA) (Hyde and Marinov 2012), for the election variable and follows the convention in this literature (for example, Shi and Svensson 2006; Brender and Drazen 2008) by focusing on the highest level of national elections. Therefore, it only includes legislative elections for countries with parliamentary political systems and executive elections for countries with presidential elections.

Third, to the best knowledge of the chapter authors, no empirical work examines the effects of fiscal rules and IMF programs on the likelihood, the size, and the composition of political budget cycles in LICs. Therefore, the chapter explores the efficiency of these two main constraints on the ability of the incumbent to pursue a politically motivated fiscal impulse in election years. It formally tests whether active national fiscal rules and IMF programs in LICs help dampen the magnitude of the political budget cycle by limiting the incumbent’s incentives to significantly modify fiscal policy for reelection purposes. Inspired by the pioneering work of Rose (2006) in the case of U.S. states, the analysis tests the extent to which national fiscal rules matter in LICs using the IMF data set on fiscal rules (Schaechter and others 2012). This is an interesting question to investigate given that experts always express doubts about the effectiveness of these rules in the LIC context.4 The issue of the effect of IMF programs on the political budget cycle in LICs is also policy relevant. With several LICs having experienced various waves of IMF programs over the past decades, one important question could be the extent to which IMF program conditions have constrained incumbents’ election-year extravagances. The chapter follows the work by Hyde and O’Mahony (2010) but focuses on the dampening role of IMF programs on the political budget cycle in LICs.

Fourth, taking advantage of the NELDA data set on elections compiled by Hyde and Marinov (2012), this analysis is able to address the endogeneity of election timing within countries. Moreover, it also factors in the self-selection bias in the decision to adopt fiscal rules or participate in IMF programs. The econometric models are used to control for several variables that ensure that the election effects are well identified so that any shift in fiscal variables associated with a national election must be interpreted as a discretionary fiscal policy by the incumbent. These control variables include external sources of financing (grants and loans), the real GDP growth rate, the inflation rate, and other covariates.

The results indicate that during election years, government consumption increases, leading to higher fiscal deficits by about 1 percentage point. During the two years following elections, fiscal retrenchment takes the form of increased revenue effort in trade taxes and cuts to government investment. However, this postelection partial fiscal adjustment is not translated into reduced current spending envelopes, or sufficient revenue mobilization efforts to fully offset the deviation allowed during the election year. The chapter also finds that national fiscal rules help mitigate the cycles in government consumption. Results also show that LICs with active IMF programs during a national election experience a much lower political budget cycle compared with years in which there is no IMF program.

The remainder of the chapter is organized as follows: The second section sets up the main empirical framework and presents the baseline results. The third section tests the robustness of the baseline results by factoring in the endogeneity of election timing. The fourth section investigates the role of fiscal rules and IMF programs as mitigating factors. The fifth section concludes.

How is Fiscal Policy Conducted Over the Election Cycle? Preliminary Evidence

This section presents the general framework and the data used to assess the dynamic of fiscal variables during and after the occurrence of a national election in LICs. Several fiscal outcomes (government consumption, public investment, breakdown of tax revenues, and budget balance as percentage of GDP) are used to assess the magnitude of the shocks on the budget during and after elections. The section also discusses the baseline econometric results.

Baseline Specification and Data

This chapter estimates several dynamic panel equations linking a given fiscal outcome with the election dummy while controlling for standard determinants of the given fiscal variable. As has now become standard in the literature on fiscal policy, dynamic equations are specified to control for the inertia characterizing fiscal variables over time. The analysis uses a panel data set that covers 68 LICs over 21 years, 1990–2010; 51 of these LICs have had at least one election.5 A country is classified as an LIC if it benefits from the IMF Poverty Reduction and Growth Trust as of 2010. The choice of time period is based on available data after the democratic reforms that many countries, particularly in sub-Saharan Africa, implemented in 1990 that made elections more competitive.

The baseline specification is as follows:

where Yi,t is the fiscal outcome in each country i at year t. Xi,t is the vector of country-level covariates that follow and slightly augment the empirical literature on the determinants of government consumption, tax revenues, and budget deficits in developing countries (Rodrik 1998; Keen and Lockwood 2010; Combes and Saadi-Sedik 2006). More specifically, models control for variables such as real GDP growth rate, inflation rate, trade openness, foreign aid, external debt, natural resource rents, agriculture value added, and fiscal rules. All the fiscal variables, inflation rate, external debt, and foreign aid are drawn from the IMF World Economic Outlook database; real per capita GDP growth series were downloaded from Penn World Table 7.1 data set. Natural resource rents and total population data are drawn from the World Bank World Development Indicators database.6

The main variables of interest are the three election dummies ELEi, t, ELEi,t–1, and ELEi,t–2, which take the value 1 in case of a national election and 0 otherwise. From equation (3.1), the three coefficients θ1 θ2, and θ3, measure the percentage point change in the fiscal variable during, one year after, and two years after a national election, respectively.7 The analysis uses the NELDA data set (Hyde and Marinov 2012) for the election variables and follows the convention in this literature (for example, Shi and Svensson 2006; Brender and Drazen 2008) by focusing on the highest level of national elections. Therefore, only legislative elections for countries with parliamentary political systems and executive elections for countries with presidential elections are included. The binary election indicator, ELE, takes the value 1 depending on the year in the election cycle, and 0 otherwise, as described above. There were 191 national elections during the sample period. Annex Table 3.1.1 shows the distribution of national elections across LICs over the period of analysis.

The IMF data set is a comprehensive source of information on budget composition based on data collected by economist desks in the field and officially approved by countries. With regard to government spending, this chapter differentiates between current expenditures (proxied by government final consumption) and public investment to assess the effects of elections on the composition of spending.

A certain granularity for tax expenditures is also accommodated.8 Instead of using only overall tax revenues as many other papers in this literature do, this analysis decomposes tax revenue into three categories—direct, indirect, and trade taxes.9 Distinguishing between these types of taxes makes the analysis richer by illustrating which taxes the government will make a particular effort to collect during the different years of the election cycle. Ehrhart (2012) bases her analysis on direct and indirect taxes, but there might be a political economy story behind trade taxes as well. Because trade involves crossing borders, it is potentially easier for the government to vary tax effort on these specific locations on the borders (Stotsky and WoldeMariam 1997).10 By looking at tax revenue ratios at a more disaggregated level, this chapter provides additional insights into the shift in the composition of tax revenue efforts over the whole election cycle. The focus on LICs is another important difference from existing papers.

Equation (3.1) is a dynamic specification and is used because of the strong inertia characterizing the fiscal variables of interest. Government administrations are constrained by budgets, and the current budget largely determines the next period’s appropriations. Although such inertia has been argued to provide some stability and predetermines fiscal spending (Schuknecht 2000), the presence of lagged dependent variables and country-specific effects renders the ordinary least squares estimator biased because the lagged dependent variable is correlated with the error term (Nickell 1981). To deal with this issue, there are two commonly used estimators: the difference–generalized method of moments (GMM) estimator (Arellano and Bond 1991) and the system-GMM estimator (Arellano and Bover 1995; Blundell and Bond 1998). In the difference-GMM estimator, equation (3.1) is taken in first differences (to remove country fixed effects), and the first differentiated variables are instrumented by their lagged values in levels. However, Arellano and Bover (1995) and Blundell and Bond (1998) have shown that when the explanatory variables are persistent over time, the lagged values of variables in levels risk being poor instruments for variables in first differences. To improve the efficiency, they propose the system-GMM estimator, which increases the moment conditions. The equation in levels and the equation in differences are combined in a system and then are estimated with an extended GMM system that allows for the use of lagged differences and lagged levels of the explanatory variables as instruments. Hence, the system-GMM estimator controls for unobserved country-specific effects as well as potential endogeneity of the explanatory variables. The chapter uses Windmeijer’s (2005) correction of standard errors for finite sample bias. Two specification tests check the validity of the instruments. The first is the GMM standard Sargan/Hansen test of overidentifying restrictions. The second test examines the hypothesis that there is no second-order serial correlation in the first-differenced residuals. The number of lags of the explanatory variables used as instruments is usually limited to reduce the “overfitting” bias (Roodman 2009).

Baseline Estimates

The chapter first reports the baseline findings for the expenditure side, then for the revenue side, and finally for the fiscal balance.

Composition of Expenditures

Table 3.1 presents the results for the system-GMM estimator for the various fiscal outcomes. Column 1 reports that government consumption increases during an election year, with no significant decrease the two years after an election. The coefficient on ELEt is significant and shows that, on average, consumption as a share of GDP increases by 0.8 percentage point during the election year. The result for government investment is reported in column 2, which shows that government investment as a share of GDP decreases by almost 0.4 percentage point the year following an election. This result is statistically significant. Although the sign of the coefficients for ELEt-2 is also negative, it is not significant.

Table 3.1.Estimates of the Political Budget Cycle across Selected Fiscal Variables in Low-Income Countries, 1990–2010
(1)(2)(3)(4)(5)(6)(7)
GITTGSTDTTBal
Electiont0.841***−0.1940.209−0.0750.1020.112*−1.047*
[0.258][0.210][0.152][0.061][0.083][0.066][0.542]
Electiont-1−0.059−0.371**0.340*0.1340.0230.173***0.275
[0.223][0.159][0.187][0.084][0.084][0.064][0.330]
Electiont-2−0.049−0.0810.290*0.026−0.0460.212*−0.262
[0.197][0.208][0.159][0.101][0.061][0.114][0.322]
Lagged Dependent Variable0.730***0.813***0.893***1.029***0.831***0.976***0.260***
[0.103][0.089][0.168][0.057][0.105][0.064][0.094]
Real per Capita GDP Growth−0.0320.0270.041**0.033***−0.0030.016***0.084***
[0.034][0.023][0.020][0.006][0.010][0.006][0.031]
Official Development Assistance to GDP0.073**0.038**−0.001−0.000−0.0020.005−0.019
[0.037][0.015][0.014][0.003][0.003][0.003][0.018]
External Debt to GDP−0.002−0.001−0.004−0.000−0.001−0.001−0.005
[0.003][0.001][0.004][0.001][0.001][0.001][0.003]
Trade Openness0.028**0.011**0.0130.0000.008*0.0020.014
[0.012][0.005][0.012][0.002][0.004][0.001][0.010]
ln (100 + inflation rate)−0.1490.419−0.0490.0050.046−0.164*0.582
[0.443][0.335][0.214][0.083][0.088][0.094][0.458]
Fiscal Rule Dummyt-11.625***0.3450.179−0.0360.281−0.232**−0.400
[0.470][0.327][0.541][0.170][0.206][0.093][1.603]
Natural Resource Rents to GDP−0.003−0.0050.010−0.003
[0.006][0.005][0.013][0.003]
ln (total population)0.020−0.0100.0350.004
[0.104][0.029][0.023][0.049]
Intercept1.999−2.0860.7600.040−0.5560.711−5.941**
[2.348][1.785][2.320][0.590][0.653][1.225][2.385]
Observations1,2341,140815679705679970
Number of Countries60575652535261
m1: p-value0.0040.0010.0030.0000.0220.0000.017
m2: p-value0.5160.4160.5300.3500.2940.1040.260
Hansen Overidentification Test: p-value0.8290.0600.0310.8390.2490.5400.114
Number of Instruments17172121212523
Source: Authors’ estimates.Note: All equations are estimated using the two-step system-GMM with Windmeijer (2005) correction of standard errors. Standard errors are in brackets. Bal = overall fiscal balance ratio; G = government consumption ratio; I = public investment ratio; T = total tax revenue ratio; TD = tax revenues on income ratio; TGS = tax revenues on goods and services ratio; TT = trade tax revenues ratio.*p < .1; **p < .05; ***p < .01.
Source: Authors’ estimates.Note: All equations are estimated using the two-step system-GMM with Windmeijer (2005) correction of standard errors. Standard errors are in brackets. Bal = overall fiscal balance ratio; G = government consumption ratio; I = public investment ratio; T = total tax revenue ratio; TD = tax revenues on income ratio; TGS = tax revenues on goods and services ratio; TT = trade tax revenues ratio.*p < .1; **p < .05; ***p < .01.

These results indicate that political budget cycles have an impact on government expenditures in LICs. More specifically, governments in LICs tend to increase consumption expenditures during election years, while investments are unchanged. The postelection adjustment takes the form of decreased government investment. These results confirm previous claims (Vergne 2009) that government spending shifts toward more visible consumption during election years. In addition, the analysis shows that the negative effect on government investment appears with a lag and implies that publicly financed projects, for example, in infrastructure, stagnate the year after an election. From the politicians’ point of view, this is strategic, because stagnating investments during election years would probably have a negative impact on reelection prospects. Although not studied explicitly in this chapter, this postelection investment stagnation may have serious consequences for economic growth.

Composition of Tax Revenues

Column 3 of Table 3.1 shows that government’s overall tax effort improves significantly in the years following an election. A look at the composition of tax revenues reveals a more detailed picture of government resource mobilization efforts over the election cycle. In column 4, where indirect taxes (taxes on goods and services) are reported, the results do not suggest the existence of an election-related cycle. Although the coefficient on the ELEt variable is negative, it is not statistically significant. The coefficients on ELE-1 and ELE-2 are not significantly different from zero either. These results contrast with findings by Ehrhart (2012), who finds a significant and negative impact of elections on indirect taxes using a broader sample of all developing countries. The results in this chapter suggest the opposite, implying that LICs’ tax policies differ from those of other developing countries over the election cycle. The different results compared with existing studies can be explained by at least two factors. First, this analysis decomposes total tax revenues into various components to get better granularity and finds that at least for LICs, the impact of national elections is observed in trade tax revenues during the run to rebuild eroded fiscal buffers. Second, the results provide a more detailed assessment of the impact of elections on government indirect tax revenues because it does not pool taxes on goods and services with trade taxes, an approach that is different from previous papers.11 The results show that within the broad definition of indirect taxes, it is the trade tax revenue ratio that matters, not taxes on goods and services.

In addition, neither does direct tax effort (on income, profits, and capital gains) vary along the election cycle, as shown in column 5. Column 6 shows the results for tax revenues on international trade. Econometric estimates indicate that governments change efforts in collecting trade taxes during election years. There is a slight (barely statistically significant) increase in the ratio of trade taxes to GDP of about 0.11 percentage point during the election year, an effort that is maintained and strengthened during the two postelection years. This explains why total tax revenues increase one and two years after elections. The results clearly suggest that LICs tend to partially rebuild the eroded policy buffers on the revenue side through increased discretionary tax revenue mobilization on international trade. This finding may occur because trade taxes tend to be relatively easier to collect in LICs, given that these countries tend to find it harder than advanced and emerging economies to mobilize revenue, particularly domestic tax revenue.

Overall Fiscal Balance

The dynamic of the overall fiscal balance throughout the election cycle mirrors the behavior of the expenditure and revenue variables (column 7 of Table 3.1). The overall fiscal deficit ratio increases by about 1 percentage point of GDP during the election year, mainly driven by the observed increase in government current spending.12 In the postelection years, attempts are certainly made to rebuild the eroded fiscal buffers, but it does not appear large and balanced enough to generate any significant statistical impact. The decline in public investment and the observed tax revenue increases in the postelection years constitute the main adjustment package in LICs, but fall short of fully rebuilding the eroded fiscal buffers. The irreversibility of government current expenditures and the ratcheting effect are the main factors behind the protracted pressure exerted by elections on overall fiscal performance throughout the years.

Dealing with the Endogeneity of Election Timing

One potential critique of the baseline results discussed above is that the analysis treats the election variables as exogenous relative to fiscal policy, which may not be the case. Timing of both elections and fiscal policies could, for example, be influenced by a number of unobserved variables that are not included in the regressions. There may be a bias if, for example, the timing of the election is strategically chosen by the incumbent politician to coincide with favorable economic conditions. One way to address this potential bias is to distinguish between elections whose timing is predetermined relative to current fiscal policies (Shi and Svensson 2006) and election timing that is not predetermined. Using information provided in the NELDA data set, an election is classified as predetermined if it took place on the date fixed by an established constitution or procedure. Conversely, election timing is considered endogenous if the election was early or late relative to the date it was supposed to be held per established procedure.13

New election indicators, ELEPREi,t and ELEENDOi,t, are created to replace ELEi,t. The variable ELEPREi,t equals 1 in country i and year t when an election was held at a predetermined time, and 0 otherwise. The variable ELEENDOi,t equals 1 in country i and year t if an election whose timing was not predetermined took place, and 0 otherwise. The postelection indicators were coded accordingly. Among the 191 elections in the sample, 56.5 percent are classified as predetermined.14 The baseline regressions are reestimated with the new election indicators. If the baseline results are robust, they should also hold for predetermined elections. The revised model takes the following form:

The coefficients of interest are φp, which capture the impact of elections after ruling out the effects of elections that occurred on an unpredicted schedule compared with the constitutional calendar.

Table 3.2 presents the econometric results. They are very similar to the previous ones in magnitude and impact on fiscal outcomes. Government current expenditures deviate significantly from their normal levels during election years, leading to an increase in the overall fiscal deficit of about 1.3 percentage points of GDP. The postelection years are characterized by an effort to partially rebuild fiscal buffers, but this comes with a price. Public investment is reduced by about 0.4 percentage point of GDP. The result that governments increase their efforts to mobilize trade tax revenues still holds. Two years after the election, the estimates indicate a reduction of the fiscal deficit by about 0.5 percentage point of GDP.

Table 3.2.Addressing the Endogeneity of Election Timing in Low-Income Countries, 1990–2010
(1)(2)(3)(4)(5)(6)(7)
GITTGSTDTTBal
Predetermined Electiont0.758**−0.2310.095−0.0740.1470.098*−1.278**
[0.327][0.183][0.167][0.089][0.116][0.055][0.618]
Predetermined Electiont-10.224−0.398**0.1310.0530.0930.112*0.253
[0.305][0.188][0.255][0.108][0.086][0.058][0.438]
Predetermined Electiont - 2−0.031−0.1590.076−0.020−0.0570.171*−0.466*
[0.218][0.308][0.148][0.108][0.096][0.094][0.278]
Observations1,1311,043758633657631900
Number of Countries60575652535261
m1: p-value0.0070.0020.0040.0000.0090.0010.019
m2: p-value0.6530.3900.2360.2690.2500.0890.210
Hansen Overidentification Test: p-value0.5830.0220.0310.8640.2380.7050.217
Number of Instruments20202424242820
Source: Authors’ estimates.Note: Windmeijer (2005) corrected standard errors are in brackets. All specifications include the exact control variables per Table 3.1. The models also control for the endogenous election dummies (dated at year t, t- 1, and t- 2, respectively) that identify whether the election was early or late relative to the date it was supposed to be held per an established constitution or procedure. The predetermined election dummies identify elections that took place on the date fixed by an established constitution or procedure. All equations are estimated using the two-step system–GMM with Windmeijer (2005) correction of standard errors. Bal = overall fiscal balance ratio; G = government consumption ratio; I = public investment ratio; T = total tax revenue ratio; TD = tax revenues on income ratio; TGS = tax revenues on goods and services ratio; TT = trade tax revenues ratio.*p < .10; **p < .05; ***p < .01.
Source: Authors’ estimates.Note: Windmeijer (2005) corrected standard errors are in brackets. All specifications include the exact control variables per Table 3.1. The models also control for the endogenous election dummies (dated at year t, t- 1, and t- 2, respectively) that identify whether the election was early or late relative to the date it was supposed to be held per an established constitution or procedure. The predetermined election dummies identify elections that took place on the date fixed by an established constitution or procedure. All equations are estimated using the two-step system–GMM with Windmeijer (2005) correction of standard errors. Bal = overall fiscal balance ratio; G = government consumption ratio; I = public investment ratio; T = total tax revenue ratio; TD = tax revenues on income ratio; TGS = tax revenues on goods and services ratio; TT = trade tax revenues ratio.*p < .10; **p < .05; ***p < .01.

Domestic and International Scrutiny

A developing literature has tried to identify the role played by various macroeconomic factors on the magnitude of political budget cycles in developing countries. O’Mahony (2010) examines the role played by openness (globalization). Vergne (2009) and Faye and Niehaus (2012) consider the role of media and financing variables such as natural resource rents and official development assistance. Combes, Ebeke, and Maurel (2013) examined the role of migrant remittances inflows on the magnitude of political budget cycles in developing countries. This section examines two main factors not fully analyzed in the LIC context that could dampen electoral fiscal manipulation. The discussion distinguishes between a domestic institutional constraint on fiscal policy and participation in a program with the IMF.

Do Fiscal Rules Matter?

The political budget cycle may be reduced in the presence of national fiscal rules if the rules prevent the incumbent from fiscal extravagances during national elections. This analysis focuses on national fiscal rules because they are more effective and better enforced than supranational rules in the LIC context.15

One main challenge in isolating the impact of fiscal rules is to address the obvious endogeneity (self-selection) of the adoption and stability of these rules. This issue is addressed in the empirical specifications. The literature on the role of fiscal rules on the reduction of political budget cycles is not large. Based on a study on the U.S. states, Rose (2006) shows that balanced budget rules help dampen politically driven cycles in overall spending, taxes, and deficits. Not surprisingly, Rose (2006) finds that the stricter the rules are, the weaker are the cycles. Inspired by this study, this analysis tests whether national fiscal rules act as a domestic scrutiny factor that helps dampen political budget cycles in LICs. However, it is important to be prudent when interpreting the results because only a few LICs use national fiscal rules, and enforcement and compliance are limited.

The econometric model exploits the interaction term between the national election dummy and a dummy for the presence of a fiscal rule to quantify the dampening impact (if any) of the presence of a national fiscal rule during election times. Because the adoption and presence of a fiscal rule are likely to be nonrandom, this issue is addressed by using a dummy variable capturing whether a national fiscal rule has been in place for at least five years. Basically, the strategy consists in interacting the election dummy with the five-year lag of the fiscal rule dummy (FR).16 More formally, the specification is the following:17

The magnitude of the effect of the political budget cycle on public consumption in the absence of a national fiscal rule is measured by θ1. In the presence of a rule, the size of the electoral fiscal manipulation is captured by θ1 + σ1. The main hypothesis is that θ1 > 0; σ1 < 0, suggesting that the amplitude of the political budget cycle is higher when the country lacks a fiscal rule compared with when the country has one.

Estimation results are presented in Table 3.3. Results indicate that for LICs without national fiscal rules (most countries), the size of the effect of the political budget cycle on government consumption is about 1 percent of GDP. This result is not much different from the estimations performed earlier in the chapter. However, once the election dummy is interacted with the national fiscal rule dummy, the coefficient turns negative and statistically significant. The significance of the coefficient of the interaction term is low, however, suggesting that the strength of the dampening role of national fiscal rules is still low in LICs, possibly because of lack of enforcement, lack of compliance, and limited number of LICs using national numerical fiscal rules. When focusing on the marginal effect of elections in LICs that have adopted national fiscal rules, the coefficient estimates in Table 3.3 suggest that the size of the fiscal deviation during an election year is close to 0.13 percentage point of GDP (1 – 0.87 = 0.13).

Table 3.3.Do Fiscal Rules and IMF Programs Dampen the Political Budget Cycle in Low-Income Countries?
(1)(2)
GG
Election Dummy1.009***0.963***
[0.331][0.293]
Election × Lagged Fiscal Rule Dummy−0.870*
[0.511]
Election × IMF Program Dummy−0.623*
[0.342]
Lagged Fiscal Rule Dummy1.986***1.621***
[0.662][0.376]
IMF Program Dummy−0.712
[0.444]
λ (predicted selection correction factor)0.621**
[0.253]
Lagged Dependent Variable0.695***0.837***
[0.089][0.041]
Real GDP Growth−0.0170.004
[0.041][0.014]
Official Development Assistance to GDP0.073*0.045***
[0.037][0.017]
External Debt to GDP−0.003−0.001
[0.003][0.003]
Trade Openness0.032***0.015***
[0.011][0.005]
ln (100 + inflation rate)−1.189−0.442
[0.976][0.344]
Intercept7.8093.488*
[5.321][1.896]
Observations1,234864
Number of Countries6059
Joint Significance of Election Coefficients: p-value0.0070.002
m1: p-value0.0020.019
m2: p-value0.3410.114
Hansen Overidentification Test: p-value0.8080.132
Number of Instruments1718
Source: Authors’ estimates.Note: Windmeijer (2005) corrected standard errors in brackets. All equations are estimated using the two-step system-GMM with Windmeijer (2005) correction of standard errors. In column 2, the model controls for the self-selection bias associated with participation in IMF programs through a two-step approach following Maddala 1983, Vella and Verbeek 1999, and Keen and Lockwood 2010. G = government consumption ratio.*p < .10; **p < .05; ***p < .01.
Source: Authors’ estimates.Note: Windmeijer (2005) corrected standard errors in brackets. All equations are estimated using the two-step system-GMM with Windmeijer (2005) correction of standard errors. In column 2, the model controls for the self-selection bias associated with participation in IMF programs through a two-step approach following Maddala 1983, Vella and Verbeek 1999, and Keen and Lockwood 2010. G = government consumption ratio.*p < .10; **p < .05; ***p < .01.

IMF Program Engagement

The chapter further tests whether countries engaged in programs with the IMF are less likely to experience a political budget cycle. In other words, do IMF programs act as an international scrutiny mechanism that constrains incumbents from using fiscal policy for electoral motives? There are several reasons why IMF programs may contribute to reducing the magnitude of the political budget cycle in LICs. LICs that enter into IMF agreements are subject to conditionality. One key component of programs’ conditionality is the adoption of sustainable macroeconomic policies. As a result, if implemented, conditionality constrains government finances, making it more difficult for governments to engage in expansionary fiscal policies during elections. This issue has been discussed and tested by Hyde and O’Mahony (2010) in the context of a large panel of developing countries (94 countries) mixing LICs with other developing nations. The authors find that IMF scrutiny of the economy and pressure on governments to maintain a sustainable fiscal policy make pre-electoral manipulation of the government balance less likely. This result appears robust to the treatment of the selection bias characterizing the decision to request a program with the IMF. This chapter follows the pioneering work by Hyde and O’Mahony (2010) in the case of a large sample of countries but departs from it in several ways that are outlined below.

One important issue in this literature is the potential endogeneity of IMF programs with respect to both elections and macroeconomic outcomes. Scholars have argued that governments prefer not to be under IMF agreements during elections (Dreher 2004), and research has shown that governments are more likely to enter into IMF agreements after elections (Przeworski and Vreeland 2000). This issue is explored in detail in the first-stage selection equation estimated to purge the endogeneity of IMF programs with respect to both election timing and macroeconomic developments.18 The chapter therefore tries to robustly investigate the effect of IMF programs on the size of the political budget cycle by focusing only on LICs.19 It also departs from the previous literature in that the main dependent variable is government consumption, the budget item that was found to be strongly correlated with elections throughout the chapter. As further explained below, the selection bias associated with IMF programs has been carefully accounted for using an improved version of the standard first-stage probit model identifying the correlates of IMF programs that take into account LICs’ specific characteristics.

To assess the effect of IMF programs (IMF), the following model is specified:

where λi,t is the selection-correction factor associated with the IMF program dummy. More specifically, the model includes the selection factor in addition to the IMF dummy so that γ1 can be interpreted with limited risk of selection bias. The magnitude of the effect of the political budget cycle on public consumption in the absence of an IMF program is measured by θ. In the presence of an IMF arrangement, the size of the electoral fiscal manipulation is captured by θ + γ1. The main hypothesis is that θ > 0; γ1 < 0, suggesting that the political budget cycle is higher when the country is not currently under an arrangement with the IMF compared with when the country is engaged with the IMF.

The correction of the self-selection associated with the decision to participate in an IMF program proceeds as follow: A pooled probit model on the determinants of IMF programs in LICs over the period 1990–2010 is estimated. Standard determinants of IMF programs include previous levels of external reserves, fiscal balance, trade openness, and inflation rate. Also added to this list are the size of natural resource rents, as well as a dummy variable indicating whether a national election is scheduled in the next year, and the election variable interacted with an indicator of electoral competitiveness. These variables are added to the selection model to capture the extent to which LICs are less likely to request IMF programs in the year before national elections, conditional on the degree of competitiveness in the considered election.20 Controlling for resource rents, for the electoral calendar, and for the degree of electoral competitiveness before the year of an IMF program helps factor in some specifics of the LIC context.

Once the probit is estimated on the group of control variables Zi,t, the selection correction factor is computed as follows (see Maddala 1983; Vella and Verbeek 1999; Keen and Lockwood 2010):21

where φ(·) and Φ(·) represent the probability density and cumulative density functions of the standard normal distribution, respectively.

Estimation results are presented in Table 3.3. Results suggest that IMF programs dampen the magnitude of the impact of the political budget cycle on consumption in LICs. In the absence of an IMF program, government consumption deviates by about 1 percentage point of GDP during national elections, whereas the size of the deviation drops to 0.34 percentage point of GDP in the presence of an active IMF program. Results also indicate a positive and significant effect of the selection factor (λi,t), which suggests that accounting for the selection bias in the estimates was crucial and legitimate. The results indicate that both fiscal rules and IMF programs play important roles in LICs by limiting the propensity of incumbents to allow large deviations in government consumption during election years as a means of maximizing their chance of reelection. Although the results seem appealing, they should be interpreted with caution. Indeed, the coefficients associated with the interaction terms of elections interacted with fiscal rules and IMF programs exhibit low significance, which suggests that the dampening role is at play but is not strong enough to generate more precise estimates.22

Concluding Remarks

This chapter investigates political budget cycles in LICs by analyzing the behavior of the following variables throughout the election cycle: government consumption, government investment, tax revenue composition, and fiscal balance. The analysis finds that during election years, government consumption increases and leads to higher fiscal deficits. During the two years following elections, fiscal adjustment takes the form of increased revenue effort in trade taxes and cuts to government investment. The chapter shows that the size of the political budget cycle is much lower in countries that have adopted national fiscal rules and in those participating in IMF programs during the election period.

The behavior of these variables throughout the election cycle is analyzed because the way governments decide to manipulate fiscal policy may have implications for future economic growth. The results in this chapter show that elections not only incur a macroeconomic cost when they take place, but also trigger a painful fiscal adjustment in which public investment is largely sacrificed and trade put at risk. Although economic growth is not explicitly studied in this chapter, the different policy tools that LICs seem to be using during the political budget cycle suggest a negative effect on economic growth. One reason for this negative effect is the overall volatility in fiscal policy that elections trigger. Another is that trade may be hampered by the increased postelection effort to mobilize trade taxes. Similarly, the decrease in investment may directly hamper growth.

This chapter uses a novel data set on fiscal rules to highlight that such rules may help dampen election-driven cycles in the budget. Although the mere existence of fiscal rules does not mean that they will be enforced, they may be a first step toward tying the hands of politicians or governments who have incentives to influence political budget cycles. The chapter also shows that IMF programs in LICs have contributed to lowering the magnitude of the political budget cycle.

Annex 3.1. Supplementary Tables
Annex Table 3.1.1.Countries in the Sample and Number of National Elections, by Country, 1990–2010
CountryNo. of National Elections
Afghanistan2
Armenia5
Bangladesh3
Benin4
Bolivia6
Burkina Faso4
Burundi2
Cambodia3
Cameroon3
Central African Republic4
Chad3
Comoros5
Democratic Republic of the Congo1
Republic of Congo3
Côte d’Ivoire4
Djibouti3
Ethiopia3
The Gambia4
Georgia5
Ghana5
Guinea4
Guinea-Bissau5
Haiti6
Honduras5
Kenya4
Kyrgyz Republic5
Lesotho3
Liberia2
Madagascar5
Malawi4
Mali4
Mauritania5
Moldova8
Mozambique4
Nepal3
Nicaragua4
Niger4
Nigeria4
Papua New Guinea4
Rwanda2
Senegal3
Sierra Leone3
Solomon Islands1
Tajikistan4
Tanzania5
Timor-Leste1
Togo4
Uganda3
Uzbekistan3
Yemen2
Zambia5
Source: National Elections across Democracy and Autocracy (NELDA) data set (Hyde and Marinov 2012).Note: Only legislative elections for countries with parliamentary political systems, and executive elections for countries with presidential elections, are included.
Source: National Elections across Democracy and Autocracy (NELDA) data set (Hyde and Marinov 2012).Note: Only legislative elections for countries with parliamentary political systems, and executive elections for countries with presidential elections, are included.
Annex Table 3.1.2.Descriptive Statistics, Low-Income Country Sample, 1990–2010
VariableNo. of ObservationsMeanStandard DeviationMinimumMaximum
Election Dummy1,3300.130.3401
Government Consumption Ratio1,14515.207.521.5362.17
Public Investment Ratio1,0447.375.690.0859.85
Total Tax Revenue Ratio83214.686.511.2758.11
Taxes on Goods and Services Ratio6665.473.190.0416.96
Direct Taxes Ratio6974.313.070.0123.89
Trade Taxes Ratio6663.882.87014.12
Overall Fiscal Balance Ratio1,004−2.486.74−72.3561.83
Real per Capita GDP Growth1,2731.237.68−71.2464.20
Official Development Assistance Ratio1,27613.2211.74−2.56146.89
External Public and Publicly Guaranteed Debt Ratio1,22685.47110.300.582,394.86
Trade Openness Ratio1,17976.2837.180.19213.22
ln (100 + inflation)1,2795.400.654.6110.35
National Fiscal Rule Dummy1,3300.020.1401
ln (total population)1,33015.161.9211.1318.86
Total Natural Resource Rents Ratio1,3304.8512.290105.73
Reserve Coverage (in months of imports)1,2043.502.58019.75
Political Globalization (KOF Institute Index)1,31346.4018.466.5990.90
Change in Real per Capita GDP Growth1,2660.0010.00−60.55132.33
Source: Authors’ calculations from various sources.Note: All variables expressed as ratios denote nominal values normalized by nominal GDP of each country.
Source: Authors’ calculations from various sources.Note: All variables expressed as ratios denote nominal values normalized by nominal GDP of each country.
Annex Table 3.1.3.Correlates of Participation in IMF Programs in Low-Income Countries
Dependent Variable: IMF Program DummyLPMProbit
Period: 1990–2010(1)(2)
International Reserves Coverage (in months of imports),t-10.00496−0.0425**
[0.0152][0.0178]
Fiscal Balance to GDP,t-10.00286−0.00296
[0.00349][0.00966]
ln (100 + inflation rate),t-1−0.04940.00539
[0.0867][0.0650]
Official Development Assistance to GDP,t-10.00882***0.0417***
[0.00261][0.00547]
ln (population),t-1−0.08560.146***
[0.313][0.0307]
(Election × Competition),t-10.0408**0.163**
[0.0198][0.0665]
Election Dummy,t-1−0.244*−0.790*
[0.133][0.410]
Political Globalization (KOF Institute Index),t−0.001150.0204***
[0.00382][0.00325]
Change in Real per Capita GDP Growth,t−0.00204**−0.00654
[0.000880][0.00499]
Intercept2.085−3.430***
[4.483][0.532]
Country Fixed EffectsYesNo
Observations916916
R20.0430.177
Countries6363
Source: Authors’ estimates.Note: Robust standard errors in brackets. The variable “Competition” is not included additively because of its mechanical perfect colinearity with the election dummy variable. LPM = linear probability model with country fixed effects.*p < .1; **p < .05; ***p < .01.
Source: Authors’ estimates.Note: Robust standard errors in brackets. The variable “Competition” is not included additively because of its mechanical perfect colinearity with the election dummy variable. LPM = linear probability model with country fixed effects.*p < .1; **p < .05; ***p < .01.
References

    AlesinaA.N.Roubini and G. D.Cohen.1997. Political Cycles and the Macroeconomy.Cambridge, MA: MIT Press.

    ArellanoM. and S.Bond.1991. “Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations.Review of Economic Studies58 (2): 27797.

    ArellanoM. and O.Bover.1995. “Another Look at the Instrumental Variable Estimation of Error-Components Models.Journal of Econometrics68 (1): 2951.

    BlockS. A.2002. “Political Business Cycles, Democratization, and Economic Reform: The Case of Africa.Journal of Development Economics67 (1): 20528.

    BlundellR. and S.Bond.1998. “Initial Conditions and Moment Restrictions in Dynamic Panel Data Models.Journal of Econometrics87 (1): 11543.

    BrenderA. and A.Drazen.2005. “Political Budget Cycles in New versus Established Democracies.Journal of Monetary Economics52 (7): 127195.

    BrenderA. and A.Drazen.2008. “How Do Budget Deficits and Economic Growth Affect Reelection Prospects? Evidence from a Large Panel of Countries.American Economic Review98 (5): 220320.

    CombesJ.-L.C.Ebeke and M.Maurel.2013. “The Effect of Remittances Prior to an Election.Working Paper CERDI Etudes et Documents 2013-07. https://halshs.archives-ouvertes.fr/halshs-00826999/document/.

    CombesJ.-L. and T.Saadi-Sedik.2006. “How Does Trade Openness Influence Budget Deficits in Developing Countries?Journal of Development Studies42 (8): 140116.

    DrazenA.2001. “The Political Business Cycle after 25 Years.NBER Macroeconomics Annual 200015: 75138.

    DrazenA. and M.Eslava.2010. “Electoral Manipulation via Voter-Friendly Spending: Theory and Evidence.Journal of Development Economics92 (1): 3952.

    DreherA.2004. “The Influence of IMF Programs on the Re-election of Debtor Governments.Economics and Politics16 (1): 5376.

    DreherA. and R.Vaubel.2004. “Do IMF and IBRD Cause Moral Hazard and Political Business Cycles? Evidence from Panel Data.Open Economies Review15 (1): 522.

    EhrhartH.2012. “Elections and the Structure of Taxation in Developing Countries.Public Choice156 (1): 195211.

    FatásA. and I.Mihov.2003. “The Case for Restricting Fiscal Policy Discretion.Quarterly Journal of Economics118 (4): 141947.

    FatásA. and I.Mihov.2012. “Policy Volatility, Institutions and Economic Growth.Review of Economics and Statistics95 (2): 32576.

    FayeM. and P.Niehaus.2012. “Political Aid Cycles.American Economic Review102 (7): 351630.

    FoudaS. M.1997. “Political Monetary Cycles and Independence of the Central Bank in a Monetary Union: An Empirical Test for a BEAC Franc Zone Member Country.Journal of African Economies6 (1): 11231.

    HydeS. D. and N.Marinov.2012. “Which Elections Can Be Lost?Political Analysis20 (2): 191210.

    HydeS. D. and A.O’Mahony.2010. “International Scrutiny and Pre-Electoral Fiscal Manipulation in Developing Countries.Journal of Politics72 (3): 690704.

    KeenM. and B.Lockwood.2010. “The Value-Added Tax: Its Causes and Consequences.Journal of Development Economics92 (2): 13851.

    KhemaniS.2004. “Political Cycles in a Developing Economy: Effect of Elections in the Indian States.Journal of Development Economics73 (1): 12554.

    MaddalaG. S.1983. Limited-Dependent and Qualitative Variables in Econometrics.Econometric Society Monographs. Cambridge, U.K.: Cambridge University Press.

    NickellS.1981. “Biases in Dynamic Models with Fixed Effects.Econometrica49 (6): 141726.

    O’MahonyA.2010. “Engineering Good Times: Fiscal Manipulation in a Global Economy.British Journal of Political Science41 (2): 31540.

    PerssonT. and G.Tabellini.2003. The Economic Effects of Constitutions: What Do the Data Say?Cambridge, MA: MIT Press.

    PrzeworskiA. and J. R.Vreeland.2000. “The Effect of IMF Programs on Economic Growth.Journal of Development Economics62 (2): 385421.

    RodrikD.1998. “Why Do More Open Economies Have Bigger Governments?Journal of Political Economy106 (5): 9971032.

    RoodmanD.2009. “A Note on the Theme of Too Many Instruments.Oxford Bulletin of Economics and Statistics71 (1): 13558.

    RoseS.2006. “Do Fiscal Rules Dampen the Political Business Cycle?Public Choice128 (3/4): 40731.

    SchaechterA.T.KindaN.Budina and A.Weber.2012. “Fiscal Rules in Response to the Crisis—Toward the Next-Generation Rules. A New Dataset.Working Paper 12/187International Monetary FundWashington, DC.

    SchuknechtL.2000. “Fiscal Policy Cycles and Public Expenditure in Developing Countries.Public Choice102 (1/2): 11328.

    ShiM. and J.Svensson.2006. “Political Budget Cycles: Do They Differ across Countries and Why?Journal of Public Economics90 (8–9): 136789.

    StotskyJ. G. and A.WoldeMariam.1997. “Tax Effort in Sub-Saharan Africa.Working Paper 97/107International Monetary FundWashington, DC.

    VellaF. and M.Verbeek.1999. “Estimating and Interpreting Models with Endogenous Treatment Effects.Journal of Business and Economic Statistics17 (4): 47378.

    VergneC.2009. “Democracy, Elections and Allocation of Public Expenditures in Developing Countries.European Journal of Political Economy25 (1): 6377.

    WindmeijerF.2005. “A Finite Sample Correction for the Variance of Linear Efficient Two-Step GMM Estimators.Journal of Econometrics126 (1): 2551.

The authors thank Chris Lane, Christian Mumssen, Catherine Patillo, Noah Ndela Jean-Frederic, Susan Yang, Alejandro Guerson, colleagues in the Low-Income Countries Division of the Strategy, Policy, and Review Department at the IMF, as well as participants at the 6th CESifo Workshop on Political Economy and the lunch seminar at Sciences Po (France).

See Brender and Drazen (2008) for an analysis of whether fiscal outcomes affect incumbents’ reelection prospects. They find that fiscal deficits do not improve incumbents’ reelection prospects in general. In developed countries, a deficit even punishes the incumbent.

Alesina, Roubini, and Cohen (1997) and Drazen (2001) provide excellent reviews on the empirical results.

Block (2002) analyzes elections that take place at regular times in presidential systems. However, presidential regimes have endogenously timed elections too, particularly in developing countries (Shi and Svensson 2006). Also, countries with presidential regimes have characteristics that systematically distinguish them from parliamentary regimes (Persson and Tabellini 2003).

The predominant national fiscal rules for LICs are debt rules, possibly reflecting institutional weaknesses that would complicate, for example, the implementation of expenditure rules or cyclically adjusted budget balance rules. However, only four LICs use national fiscal rules. Other LICs operate under supranational rules.

See Annex Table 3.1.1 for the list of countries. The panel data set is unbalanced because some countries have missing values.

Descriptive statistics for the dependent variables are provided in Annex Table 3.1.2.

When investigating how elections affect government spending, one critique may be that the political budget cycles observed are due to the extra cost of running elections, and not necessarily a strategic allocation that is driven by reelection incentives. This is a fair point, though not a real concern for LICs. These countries are very poor and highly dependent on aid. Most of the expenses related to elections are also borne through aid. Therefore, we investigate the effect of elections on government spending while controlling for aid as a share of GDP. The fact that political budget cycles are observed even when keeping aid constant shows that there is a political incentive even if elections were costly to run.

Nonresource tax revenue mobilization is a major challenge in many LICs. While overall tax revenues equate to more than 50 percent of GDP in some countries in the sample, others barely manage to collect 1 percent of GDP. The mean overall tax revenue in the sample is 14.8 percent of GDP. The largest contribution to revenues comes from indirect taxes, followed by direct taxes, and trade taxes.

Indirect taxes, which are broad-based taxes on goods and services, are paid by most citizens and correspond to 5.6 percent of GDP in the sample. Direct taxes represent taxes on income, profits and capital gains, correspond to 4.4 percent of GDP, and are mostly paid by corporations since personal income taxes are almost nonexistent. Finally, trade taxes, which correspond to 3.8 percent of GDP, are taxes on trade and international transactions paid by corporations.

For the Indian states, Khemani (2004) provides an analysis with subcategories of commodity taxes: sales, excise and trade. Data are not available on this level for the 68 low-income countries that we study in this chapter.

However, it remains true that trade taxes also include value-added tax revenues collected at the border.

The magnitude of the deviation in the fiscal balance attributed to elections is similar to previous results by Shi and Svensson (2006).

This coding is done using the variable NELDA6 in the NELDA data set. An established procedure is one contained in the constitution.

Out of the 191 elections, 108 are classified predetermined and 38 endogenous. We were unable to classify 45 elections.

However, enforcement of and compliance with supranational fiscal rules has been, at best, mixed in most European Union member states, West African Economic and Monetary Union countries, and in the Central African and Economic Monetary Community region (Schaechter and others 2012).

The reader may wonder whether the proposed identification strategy for assessing the impact of a fiscal rule is the best available. For example, it could be interesting to proceed with an instrumental variable strategy to tackle the potential endogeneity of fiscal rules. However, finding such instrumental variables, which need to be fully exogenous to fiscal outcomes, is very challenging. Another strategy might be to pursue a two-step approach in which a selection equation explaining the decision to have a fiscal rule is estimated and used to control for the self-selection bias in the fiscal equation. However, with such a small number of LICs having national fiscal rules, performing the two-step approach does not seem suitable.

Also, we will disregard the postelection dummies used previously and concentrate the analysis on the election year since the cycles in government consumption are observed during election year and cycles on revenues are less robust. Moreover, we do not need to break down the fiscal rule dummy into subcomponents because national fiscal rules in LICs are primarily dominated by debt rules.

However, the strength of the bias due to the potential link between an IMF arrangement and an election is attenuated by one stylized fact. As discussed by Hyde and O’Mahony (2010), the majority of elections in the developing world are held while countries are already under an IMF agreement. LICs are more likely to have intensive program engagement because of their prolonged balance of payments needs.

This literature has typically used country samples that mix LICs and middle-income economies, tending to overlook the distinct characteristics of LICs as well as the distinct nature and objectives of IMF engagement in these countries. LICs face a number of challenges that differentiate them from other economies, such as the nature of shocks, access to financing, and long-term challenges (poverty reduction, infrastructure needs, institutional and capacity building, and others) that typically imply that the type of IMF facilities and their goals are quite different from other IMF financial instruments available to emerging market or advanced economies.

The selection equation also helps deal with the potential bias that could arise if IMF lending were significantly higher during months before elections. Dreher and Vaubel (2004) found that that is indeed the case. This analysis rules out this effect by always controlling for official development assistance in the regressions and by explicitly controlling for the electoral calendar and timing in the selection equation. Therefore, the risks that these results are fully driven by IMF lending dynamics before and after elections are limited. Moreover, the direction of this bias would work to lower the estimated effect of the IMF programs, leading to underestimated effects instead of inflated effects.

Results of the probit selection model identifying the determinants of LICs’ participation in IMF programs are available in Annex Table 3.1.3.

There are several reasons for these results. First, only four LICs have adopted active national fiscal rules, an issue that certainly contributes to reducing the explanatory power of the fiscal dummy in the model. It could also be that these rules are not sufficiently enforced, exacerbating the credibility problems faced by these institutional arrangements in many LICs. Second, even though the selection equation associated with IMF programs explicitly controls for several covariates, the self-selection bias is always only partially controlled for. In addition, because the majority of elections in LICs are held while countries are already in an IMF agreement, the statistical power of the IMF program dummy in dampening the political cycle is limited.

    Other Resources Citing This Publication