Fiscal Politics
Chapter

Chapter 11. Fiscal Discipline and Exchange Rates: Does Politics Matter?

Author(s):
Vitor Gaspar, Sanjeev Gupta, and Carlos Mulas-Granados
Published Date:
April 2017
Share
  • ShareShare
Show Summary Details
Author(s)
João Tovar Jalles, Carlos Mulas-Granados and José Tavares 

Introduction

The issue of what type of exchange rate regime is better for fiscal discipline has a long history in macroeconomics. Traditionally, the view that fixed exchange rate regimes could be associated with increased fiscal discipline was widely accepted. However, more recent theoretical models, together with mixed empirical evidence, have paved the way to an alternative policy view according to which flexible exchange rate regimes are also compatible with healthy public finances.1 In parallel, many countries have moved from fixed to flexible exchange rates in the past two decades.2 On the economic front, two opposing views are at stake: the first considers a fixed peg to be a provider of credible discipline and associates fixed exchange rates with enhanced fiscal restraint; the second defends the notion that flexible rates induce better fiscal performance instead because they expose the costly economic consequences of fiscal profligacy and do not allow policymakers to hide the deterioration of fiscal balances behind a loss of domestic reserves.

The empirical evidence that examines the relationship between fiscal discipline and exchange rate regimes is mixed. Gavin and Perotti (1997) uncover an association between fixed exchange rate regimes and public deficits in Latin American economies, but they do not find similar evidence for advanced, industrialized economies. Fatás and Rose (2001) find that, while adherence to a common currency area is not associated with increased fiscal discipline, adoption of a currency board is. Tornell and Velasco (1995, 1998) and Sun (2003) present evidence of the effects of fixed exchange rate regimes on fiscal discipline. Their analysis is limited to a specific set of countries, notably the Communautè financiére d’Afrique (CFA) zone in Africa and the pegged currency union in the Caribbean. In sum, there are divergent empirical results as to the relationship between exchange rate regimes and fiscal discipline.

This chapter posits that these mixed findings stem from the literature neglecting the importance of the political context in which fiscal policy decisions are made. On the political front, this analysis proposes that two dimensions affecting fiscal policy decisions must be taken into account: first, the electoral calendar, a timing dimension that directly affects the policymaker’s horizon; and second, the degree of cohesion or political fragmentation that policymakers enjoy—or not—within the government coalition structures or the legislature. In other words, considering the underlying political conditions, both electoral timing and political cohesion, under which fiscal policy choices are made is crucial to properly understanding how alternative exchange rate regimes interact with fiscal outcomes. This chapter examines the question of how politics affects the interaction between currency regime and fiscal discipline using the broadest possible sample yet available.

Determining whether fixed or flexible exchange rates are better for enhanced fiscal discipline in the presence of different political conditions is important for several reasons. First, the exchange rate regime may affect the incentives to use deficits with electoral motives (Tanzi and Schuknecht 1997) or limit the sustainability of fiscal adjustments (Lane and Perotti 2003; Lambertini and Tavares 2005). Second, if flexible exchange rates are associated with closer public scrutiny of policymakers and this in turn encourages fiscal discipline, shifting to a flexible regime may be a “low-cost” institutional fix to the absence of fiscal discipline. This idea is particularly important in countries where credible economic and political institutions do not exist or are incapable of monitoring fiscal authorities. A flexible exchange rate regime can in these cases be seen as a substitute, albeit an imperfect one, for the harder task of building better institutions. Third, as the experience of the euro area during the sovereign debt crises illustrates how hard and credible pegs, even in advanced economies, are not necessarily associated with fiscal restraint. Here too, one of the determinants of the fiscal stance may be the quality of institutions that are charged with ensuring accountability and transparency at the national level. Finally, many of the developing countries adopting fixed pegs are small open economies, vulnerable to considerable exogenous shocks, so that determining how internal politics affects the relationship between fixing the exchange rate and fiscal performance becomes a key issue.3

The results in this chapter show that, in general, fixed exchange rate regimes are associated with less fiscal discipline. When political conditions are taken into account, strong political environments help improve fiscal discipline, and more so under flexible exchange rate regimes. In other words, policymakers not facing immediate elections and who can operate without political fragmentation are associated with better fiscal performance, especially when under a flexible exchange rate regime.

The remainder of the chapter is organized as follows: The second section reviews the literature. The third section presents a generic model illustrating the mechanisms by which political conditions (as measured by political Horizon and Cohesion) affect the relationship between exchange rate regimes and fiscal discipline. The fourth section describes the data and presents the main results. The fifth section provides several robustness exercises. The sixth section discusses fiscal discipline in the euro area using synthetic control analysis. The final section summarizes the main findings and concludes.

Literature Review: Exchange Rate Regimes and Fiscal Discipline

The Traditional Argument

The traditional argument holds that fixed exchange rates encourage fiscal discipline. This argument is steeped in a long tradition, going back at least a century,4 according to which adhering to a specie standard, or a stable currency, would be associated with sound money and predictable policies that would keep inflation under control and lead to fiscal restraint. The idea is that lax fiscal policies could eventually lead to a collapse of the peg, which is a very costly scenario (Giavazzi and Pagano 1988; Frankel, Goldstein, and Masson 1991). This traditional argument emphasizes how anticipation of the harsh economic impact of defaulting from a fixed exchange rate regime disciplines policymakers today and leads to sound fiscal policy.

In most theoretical setups, the fiscal and monetary policymakers care for both government expenditure and the usual trade-off is in place, whereby larger government transfers today or tomorrow translate into higher inflation either today or tomorrow. In a sense, what makes default inevitable in some instances of fixed rates and fiscal laxity is that the central bank eventually abandons the peg to avoid hyperinflation, a deeper output decline, or both. In other words, the monetary authority has access only to a limited commitment technology, where the limitation concerns time.5

The Dynamic Approach

Several authors have challenged the traditional argument and suggest, instead, that a fixed exchange rate regime may actually induce fiscal indiscipline (Tornell and Velasco 1995, 1998; Sun 2003; Duttagupta and Tolosa 2006). In the presence of impatient policymakers who heavily discount the future, the fact that the economic cost of fiscal indiscipline and default takes time to occur leads policymakers to spend more today.6

Whereas the traditional argument did not consider the relevance of time and how policymakers discount the future, Tornell and Velasco (1995, 1998) show that if the economic costs of fiscal indiscipline were sufficiently delayed, a fixed exchange rate regime would not limit the policymaker’s tendency to overspend. In their setup, policymakers completely discount the future and thus, under fixed exchange rates, run reserves down through higher current spending. In contrast, flexible exchange rates would bring immediate economic punishment to fiscal laxity through currency depreciation and higher inflation.7

Sun (2003) develops a dynamic model integrating the traditional argument and Tornell and Velasco’s (1995, 1998) dynamic counterargument, showing how each one overemphasizes part of a larger story. According to Sun, the economic costs of fiscal indiscipline exist under both exchange rate regimes. While in the short term fixed exchange rates shelter policymakers from the consequences of lax fiscal policies, a higher future punishment, or a more balanced consideration of the future can induce fiscal discipline under fixed rates. However, given the temporally uneven structure of incentives under fixed exchange rate regimes, there is greater incentive for lower expenditures under flexible exchange rate regimes.8

Political Economy Dimensions: Horizon and Cohesion

In this chapter’s view, existing models focus on the economic costs of fiscal indiscipline, while not carefully spelling out the political dimension. In particular, when these models discuss the policymakers’ time horizon, they refer to how long it takes them to face economic disaster as a result of a fiscal crisis that would force their countries to abandon the currency peg. This chapter postulates that, in this time horizon, policymakers also take into account the distance to elections and the amount of time that they have been in government. This is because we consider politicians to be office seekers rather than policy seekers, and their prime objective is to maximize their probabilities of staying in power.9 In addition, policymakers are forced to take into consideration the degree of political fragmentation they face when designing budget measures to be approved by parliament. In this context, this chapter proposes the introduction of an explicit electoral component in the policymaker’s time horizon defined by Tornell and Velasco (1995, 1998) and Sun (2003). In addition, the chapter also proposes the inclusion of a new political dimension, that is, this degree of political cohesion. The two components of the revised theoretical framework can be summarized as follows:

  • Political Horizon: The time policymakers have before forthcoming elections—or regime change in autocratic regimes. Politicians facing longer horizons have fewer incentives to overspend and thus associate with more fiscal discipline.

  • Political Cohesion: The number of political actors participating in budgetary decisions who exhibit conflicting budgetary demands. These actors could be parties in government, or in opposition, in interest groups, or, more generally, veto players. Politicians who operate in more cohesive political environments are likely to be subject to less strident spending demands and be associated with tighter fiscal discipline.

In this framework, strong politics—a long horizon for the policymaker and high cohesion of the political body—add credibility to a flexible exchange rate system, leading to fiscal discipline. The next section spells out these hypotheses formally in an illustrative model.

Theoretical Framework

Consider an economy in which there are two periods and three economic agents: a monetary authority, which decides the exchange rate, and two political players with authority over public spending.

The central bank cares about the nominal exchange rate, E1, which it desires to be close to a target exchange rate, E*, in both periods. It also cares for output, Y which it wishes to be as close as possible to a target level, Y*. This target level can be interpreted as potential output or some other desired reference level of output.

The expression for the central bank’s loss function would be

where βM is the discount factor of the central banker, and aM weighs the losses stemming from deviations in the exchange rate relative to losses from deviations in output. The term aM > 1, so that, given the central bank’s nature, deviations in the exchange rate are weighted more heavily than deviations in output. Output levels Y1 and Y2 are the weighted sum of the output of the two political constituencies, where the weights γ1 and γ2 sum to 1, and accommodate a possible differentiated response of the central bank to the two political players.10 The central bank’s loss function becomes

where 1 stands for constituency and 2 stands for period, in Y1,1 and Y1,2.

The fiscal players care only about their constituency’s output, and the higher it is the better for them. Consequently, the loss function of constituency 1 is represented by:

where Yi,1 and Yi,2 have been defined before, and βF is the policymakers’ discount factor, which may be different from βM. The parameter βF is a proxy for political horizon, so that its value is smaller the shorter the policymakers’ time horizon due to the proximity of elections or higher likelihood of turnover due to a longer stay in power.

Output in periods 1 and 2 positively depends on the exchange rate, respectively, E1 and E2. The policymakers vie for transfers to their constituencies, TRF1, and TRF2, which increase output. However, these transfers have to be paid for in period 2, so that D = TRF1 + TRF2. Debt is fully repaid in period 2, and the debt burden is shared by the two constituencies according to weights α1 and α2 = 1 − α1, each greater than 0 and smaller than 1. Thus, output in periods 1 and 2 can be summarized as

In period 0, before the start of the game, the monetary authority sets a fixed exchange rate E1 We do not endogenize this choice of the exchange rate in period 0. In period 1, the two political players engage in a fiscal game whereby each decides the degree of transfers to their own constituencies. The deterrent to spending “too much” in period 1 is the fact that transfers, which increase utility in period 1, also increase debt that has to be fully repaid in period 2 and thus decreases utility in period 2. The extent to which each fiscal policymaker engages in transfers today depends on the amount of future debt that it will have to be repaid in the future. In period 2, because the central bank cares for output as well as for the exchange rate, it will default on the exchange rate. But the extent to which the central bank will accommodate fiscal profligacy depends on aM, and the policymakers take that factor into consideration.

Two key political economy parameters affect the degree of indebtedness. The first is the political Horizon, that is, the extent to which the fiscal policymakers value the future. The horizon is related to electoral and other incentives that may make the policymaker less responsive to the future consequences of indebtedness. When the political horizon is longer, policymakers value the decrease in utility tomorrow due to debt repayment relatively more than the current increase in utility, driven by transfers. The second factor affecting indebtedness is the degree of political Cohesion, which is a function of the share of debt to be repaid by each fiscal actor in period 2. When each political actor is responsible for 50 percent of debt repayment, the political system is more cohesive, and both policymakers behave responsibly in fiscal terms.

The model is now solved by backward induction. First, we determine the rate of devaluation chosen in period 2 by the central banker for each level of indebtedness. In other words, in period 2, the central bank decides how much debt it will accommodate through devaluation. The central bank’s reaction function to debt accumulation is the following:

As expected, the higher the debt incurred in period 1 by the policymakers, the higher the exchange rate depreciation by the central bank in period 2. Also as expected, the combination of parameters associated with D shows that the accommodation is less than one for one, given that aM > 1, and (α1 × γ1 + α2 × γ2) is generally smaller than 1. The higher aM, that is, the more averse the central bank is to exchange rate deviations, the less it accommodates debt. Furthermore, as aM increases to infinity, E2 tends to E* and there is no debt accommodation whatsoever. This is the case of a fully credible and unchangeable fixed exchange rate.

Working backward, in period 1, the policymakers take into consideration the expected future behavior of the central bank and each constituency’s share of the debt burden to decide the extent of the transfers to mobilize. Total transfers—and thus debt, one period later—are given by the intersection of two fiscal reaction functions where each constituency’s level of transfers depends on the others’, as below:

A one-monetary-unit increase in transfers to a specific constituency leads to a decrease in transfers by the other player by less than one unit, so that an equilibrium exists. Figures 11.1 and 11.2 illustrate the two main political economy propositions that can be derived from this model. Other things being equal,

  • Proposition 1: A decrease in the discount factor, βF, of the fiscal policymakers so that they act under a longer political Horizon, leads to lower indebtedness (Figure 11.1).

  • Proposition 2: An increase in political Cohesion, expressed by a more even distribution of the debt burden between political constituencies, leads to a decrease in public debt (Figure 11.2).

Figure 11.1.Political Horizon and Indebtedness

Source: Authors’ calculations.

Note: A lower discount of the future (that is, higher political Horizon) leads to lower indebtedness (that is, more fiscal discipline).

Figure 11.2.Political Cohesion and Indebtedness

Source: Authors’ calculations.

Note: A higher level of burden sharing (that is, higher political Cohesion) leads to lower indebtedness (that is, more fiscal discipline).

Empirical Analysis

Data Description

To test the propositions above, an unbalanced panel of 79 countries, including 31 advanced countries and 48 emerging and low-income economies, between the years 1975 and 2012, is used to estimate the following specification:

where FDit is the proxy for fiscal discipline, given by the primary balance, measured as a percentage of GDP;11ERit is a dummy variable taking the value one when a country is under a fixed exchange rate regime and zero if under a flexible exchange rate regime;12Xit is a vector of control variables expected to affect the fiscal policy stance and comprises the real GDP growth rate, trade openness (exports plus imports as share of GDP), terms of trade, log of total reserves in

U.S. dollars, private credit as a percentage of GDP, and the consumer price inflation rate.13 The parameters αi, δt are country and time effects that capture unobserved heterogeneity across countries and time-unvarying factors. The term εit is a white noise identically and independently distributed disturbance term satisfying standard assumptions of zero mean and constant variance. Finally, POLit accounts for political factors and is constructed using principal component analysis, to obtain the common factor(s) of each block of variables comprising two components termed Horizon and Cohesion. These are defined as follows:

  • Political Horizon: A longer political horizon is associated with fewer years in office, more years left in current term, the chief executive’s party with a long time in office, and more months to next election.14 Only the first principal component is retained.15

  • Political Cohesion: Stronger political cohesion is associated with a high margin of the parliamentary majority supporting the cabinet, low cabinet fragmentation, executive control of all houses, and a weak opposition. Only the first principal component is retained.

Horizon and Cohesion variables are each represented by one factor composed of four underlying variables.16 The resulting principal components indices are described in Table 11.1, while Table 11.2 lists the corresponding factor loadings.17 The principal components can be interpreted by focusing on the factor loadings and the uniqueness of each variable.18 With regard to political Horizon, uniqueness is relatively low for all variables, which implies that the retained factor spans the original variables adequately. As to political Cohesion, the factor appears to describe mostly the margin of majority and cabinet strength. In principle, both factors should enter with positive coefficients in the regressions.

Table 11.1.Summary of Political Composite Variables and Descriptive Statistics
ConceptVariables
HorizonYears in office
Years left in current term
Party of chief executive more time in office
Months to next election
CohesionMargin of majority
Cabinet strength
Executive control of all houses
Weak opposition
Source: Authors’ calculations.Note: The average and standard deviation of the respective principal component are normalized to 0 and 1, respectively.
Source: Authors’ calculations.Note: The average and standard deviation of the respective principal component are normalized to 0 and 1, respectively.
Table 11.2.Factor Loadings and Uniqueness
Factors
VariablesHorizonCohesionUniqueness
Years in Office0.390.29
Years Left in Current Term0.410.28
Party of Chief Executive More Time in Office0.370.29
Months to Next Election0.450.28
Margin of Majority0.930.12
Cabinet Strength0.900.17
Executive Control of All Houses0.760.42
Weak Opposition0.720.47
Percent Explained0.390.69
Source: Authors’ calculations.
Source: Authors’ calculations.

A first look at the data shows that it is in line with the findings of Tornell and Velasco (1995) and Sun (2003) so that fiscal discipline is lower under fixed exchange rate regimes than under flexible ones (Figure 11.3). Also, stylized facts seem to confirm the propositions of the theoretical model presented above: in contexts of long political Horizon, the average primary surplus is only 0.01 percent of GDP under fixed exchange rate regimes, and it reaches 0.3 percent of GDP under flexible exchange rates. These differences are smaller in contexts of high political Cohesion. In such cases, the average primary deficit is -0.2 percent of GDP under fixed exchange rates and at -0.1 percent of GDP.

Figure 11.3.Primary Balance, Exchange Rates, and Political Conditions

Source: Authors’ calculations.

Baseline Panel Results

First, equation (11.9) is estimated using ordinary least squares with heteroskedastic and serial correlation robust standard errors clustered at the country level. Alternative estimators are also used that correct for the standard econometric pitfalls, including possible reverse causality between exchange rate regimes and the fiscal stance, and omitted variable bias simultaneously affecting both the choice of the exchange rate regime and the fiscal stance. To correct for serial correlation and possible cross-sectional heteroskedasticity, feasible generalized least squares regressions are run using estimated cross-section residual variances as weights, thus attaining a more efficient estimate than under ordinary least squares.

Second, having in mind potential cross-sectional dependencies, the main regression in equation (11.9) is run with Driscoll-Kraay (1998) robust standard errors. This nonparametric technique assumes the error structure to be heteroskedastic, autocorrelated up to some lag, and possibly correlated between the groups.

Third, as a closer inspection of the data suggests, influential outliers potentially play a role in cross-section analysis. It is important to consider the extent to which outliers drive the results, particularly in such a heterogeneous sample as this one, which includes emerging market and low-income economies characterized by spells of exchange rate volatility. The analysis uses the method of moments that fits the efficient high breakdown estimator proposed by Yohai (1987).19 The model described by equation (11.9) is a reduced form; therefore, it does not legitimize causal statements or even immediate quantification of the effect of exchange rate regimes and politics on fiscal discipline. Because causality can run in both directions, some of the right-hand-side regressors may be correlated with the error term. The fixed-effects approach is complemented with a panel instrumental variable–generalized least squares approach. As instruments, the main variables used are those proposed by Acemoglu and others (2003) and Fatás and Mihov (2013). The first instrument—labeled constraints—captures potential veto points on the decisions of the executive. A variation of this measure of constraints is a variable constructed by Henisz (2000) called political constraints (labeled polcon). This variable differs from our measure in two ways: (1) Henisz (2000) adjusts for the ideological alignment across political institutions; and (2) he argues that each additional constraint has a diminishing marginal effect on policy outcomes, therefore, the link between the overall measure and the veto points should be nonlinear. In addition, dummies are used to control for the presence of rules on expenditure, taxes, and debt.

The first set of results concerns the estimation of equation (11.9) to empirically test Propositions 1 and 2 from the theoretical section. The procedure starts by independently analyzing the variables that measure the exchange rate regime, political Horizon and political Cohesion. Table 11.3 shows the estimates of the coefficient associated with the fixed exchange rate regime to be always negative and almost always statistically significant at conventional levels. These results are in line with the latest findings in the empirical literature: being in or moving to a fixed exchange rate regime is associated with less fiscal discipline. This confirms our theoretical propositions 1 and 2. A longer political Horizon and more political Cohesion are both associated with a higher primary balance and more fiscal discipline. The remaining controls are mostly statistically significant and with the expected sign: higher GDP growth, higher inflation, higher reserves, better terms of trade, and trade and financial openness are all associated with higher fiscal discipline.20

Table 11.3.Baseline, Fixed Effects Regressions
Specification
(1)(2)(3)(4)(5)(6)(7)
Fixed Exchange Rates−0.432**−0.412**−0.366*−0.388*−0.406*−0.472**−0.370
(0.180)(0.192)(0.219)(0.230)(0.215)(0.226)(0.234)
Growth0.087***0.098***0.116***0.100***0.116***0.135***0.145***
(0.019)(0.020)(0.021)(0.024)(0.020)(0.022)(0.023)
Trade Openness0.017***0.015***0.021***0.019***0.021***0.012*0.012*
(0.005)(0.005)(0.006)(0.006)(0.006)(0.006)(0.007)
Terms of Trade0.017***0.018***0.020***0.028***0.016***0.016***0.018***
(0.003)(0.003)(0.004)(0.004)(0.004)(0.005)(0.005)
Reserves0.438***0.410***0.451***0.398***0.271***0.217**0.226**
(0.093)(0.093)(0.108)(0.101)(0.099)(0.102)(0.107)
Inflation0.034***0.024**0.030***0.027***
(0.011)(0.011)(0.010)(0.010)
Financial Openness0.017*
(0.009)
Credit0.045***
(0.013)
Horizon0.346***0.310***
(0.096)(0.116)
Cohesion0.241*0.241
(0.142)(0.164)
Observations2,4042,2161,8881,7791,8081,5601,456
R20.4420.4560.4760.4710.4860.4780.478
Source: Authors’ calculations.Note: The estimation with country and time effects is omitted for reasons of parsimony. Robust standard errors are clustered at the country level in parentheses. The constant term is estimated but omitted.*p < .1; **p < .05; ***p < .01.
Source: Authors’ calculations.Note: The estimation with country and time effects is omitted for reasons of parsimony. Robust standard errors are clustered at the country level in parentheses. The constant term is estimated but omitted.*p < .1; **p < .05; ***p < .01.
Table 11.4.Interaction Terms, Fixed Effects Regressions
Specification
(1)(2)(3)(4)(5)(6)
Fixed Exchange Rates−0.637***−0.826***−0.720***−0.547**−0.696***−0.589**
(0.208)(0.236)(0.263)(0.214)(0.262)(0.267)
Growth0.111***0.126***0.135***0.117***0.135***0.145***
(0.020)(0.022)(0.023)(0.020)(0.022)(0.023)
Trade Openness0.025***0.014**0.016**0.022***0.011*0.012*
(0.006)(0.006)(0.007)(0.006)(0.006)(0.007)
Terms of Trade0.016***0.017***0.019***0.016***0.016***0.018***
(0.003)(0.005)(0.005)(0.004)(0.005)(0.005)
Reserves0.260***0.208**0.214**0.273***0.207**0.224**
(0.095)(0.098)(0.103)(0.099)(0.102)(0.107)
Horizon0.534***0.451***0.519***0.433***
(0.114)(0.144)(0.117)(0.148)
Horizon × Fixed Exchange Rates−0.435***−0.308−0.417**−0.278
(0.167)(0.225)(0.171)(0.233)
Cohesion0.370**0.2880.364**0.296
(0.169)(0.181)(0.175)(0.181)
Cohesion × Fixed Exchange Rates−0.397*−0.178−0.343−0.146
(0.231)(0.287)(0.259)(0.294)
Inflation0.024**0.029***0.027***
(0.010)(0.010)(0.010)
Observations1,9041,6551,5341,8081,5601,456
R20.4850.4700.4720.4880.4780.479
Source: Authors’ calculations.Note: The estimation with country and time effects is omitted for reasons of parsimony. Robust standard errors are clustered at the country level in parentheses. The constant term is estimated but omitted.*p < .1; **p < .05; ***p < .01.
Source: Authors’ calculations.Note: The estimation with country and time effects is omitted for reasons of parsimony. Robust standard errors are clustered at the country level in parentheses. The constant term is estimated but omitted.*p < .1; **p < .05; ***p < .01.

Table 11.4 includes two interaction terms between the dummy variable for fixed exchange rate regimes and the two variables measuring political Horizon and Cohesion. Again, the analysis delivers a statistically significant and negative estimate of the coefficient on fixed exchange rate regimes and positive coefficients for both political indicators. Interestingly, the interaction terms between the key variables are negative, suggesting that the positive effect of politics (longer Horizon or higher Cohesion) on the degree of fiscal discipline are particularly relevant for flexible exchange rate regimes. In other words, it is in flexible exchange rate settings that politics seems to matter most.

Note that these results shed light on the questions that motivate this chapter. In line with the model’s prediction, both exchange rates and politics matter for fiscal discipline. In line with the model’s prediction, flexible (not fixed) exchange rates, and strong politics (long Horizon and high Cohesion) are associated with better fiscal positions. The exchange rate regime seems to be quantitatively more important than the underlying political conditions. When considered together, strong politics attenuates the damaging effects of fixed rates on fiscal discipline but is insufficient to reverse fiscal profligacy. Moreover, the positive effects of longer political horizon and more political cohesion on fiscal performance are amplified under flexible exchange rate regimes.

Robustness Checks

Sensitivity to Alternative Estimators, Outliers, and Endogeneity

To test the robustness of the baseline results, a series of robustness checks are performed. First, equation (11.9) is reestimated using feasible generalized least squares and Driscoll-Kraay cross-sectional dependence robust standard errors, as well as the method of moments estimator to check for outliers. Results in Table 11.5 show again that fixed exchange rates are associated with less fiscal discipline, while a longer Horizon and higher Cohesion21 are associated with better fiscal performance. The interaction terms between the key variables are also negative and significant. Second, the regressions are estimated using instrumental variables to address endogeneity concerns. Results are reported in Table 11.6 and confirm the baseline set of findings.

Table 11.5.Sensitivity to Alternative Estimators
Specification
Feasible Generalized Least SquaresDriscoll-KraayOutlier-Robust
(1)(2)(3)(4)(5)(6)(7)(8)(9)
Fixed Exchange Rates−0.579***−0.635***−0.686***−0.662***−0.812***−0.956***−0.540***−0.476*−0.601**
(0.164)(0.181)(0.181)(0.240)(0.267)(0.277)(0.207)(0.269)(0.259)
Growth0.119***0.124***0.125***0.187***0.202***0.203***0.160***0.182***0.192***
(0.011)(0.012)(0.012)(0.041)(0.048)(0.048)(0.031)(0.032)(0.030)
Trade Openness0.026***0.026***0.026***0.0070.0130.0140.007***0.009***0.010***
(0.004)(0.005)(0.005)(0.006)(0.008)(0.008)(0.002)(0.002)(0.002)
Terms of Trade0.008***0.008**0.008**0.013***0.012**0.012**0.0040.010*0.010*
(0.003)(0.004)(0.004)(0.004)(0.005)(0.005)(0.003)(0.006)(0.005)
Reserves−0.026−0.004−0.0050.1890.1240.1250.379***0.420***0.407***
(0.067)(0.073)(0.073)(0.191)(0.204)(0.205)(0.044)(0.055)(0.056)
Inflation0.041***0.037***0.036***0.041***0.029***0.029***0.038***0.042***0.476***
(0.015)(0.012)(0.012)(0.009)(0.009)(0.008)(0.004)(0.004)(0.056)
Horizon0.215***0.315***0.344***0.521***0.0130.285
(0.059)(0.072)(0.106)(0.129)(0.158)(0.185)
Horizon × Fixed Exchange Rates−0.247**−0.423***−0.631**
(0.111)(0.157)(0.266)
Observations2,2141,8071,8072,2161,8081,8082,2161,8081,808
Source: Authors’calculations.Note: Estimation is made by either FGLS, Driscoll-Kraay cross-sectional dependence-robust standard errors, or method of moments–estimator for outliers. The estimation with country and time effects is omitted for reasons of parsimony. Robust standard errors are clustered at the country level in parentheses. The constant term is estimated but omitted.*p < .1; **p < .05; ***p < .01.
Source: Authors’calculations.Note: Estimation is made by either FGLS, Driscoll-Kraay cross-sectional dependence-robust standard errors, or method of moments–estimator for outliers. The estimation with country and time effects is omitted for reasons of parsimony. Robust standard errors are clustered at the country level in parentheses. The constant term is estimated but omitted.*p < .1; **p < .05; ***p < .01.
Table 11.6.Endogeneity: Instrumental Variables
Specification
(1)(2)(3)(4)(5)
Fixed Exchange Rates−0.730***−0.743***−0.888***−0.754***−0.873***
(0.203)(0.217)(0.227)(0.225)(0.303)
Growth0.208***0.214***0.214***0.247***0.247***
(0.019)(0.020)(0.020)(0.022)(0.022)
Trade Openness0.008*0.010**0.011**0.0020.002
(0.005)(0.005)(0.005)(0.005)(0.005)
Terms of Trade0.014***0.011***0.011***0.012**0.012**
(0.003)(0.004)(0.004)(0.005)(0.005)
Reserves0.0880.0640.0660.231***0.229***
(0.073)(0.078)(0.077)(0.087)(0.087)
Inflation0.035*0.0290.0290.036*0.036*
(0.019)(0.019)(0.019)(0.018)(0.018)
Horizon0.365***0.536***
(0.111)(0.136)
Horizon × Fixed Exchange Rates−0.431**
(0.200)
Cohesion0.1330.190
(0.196)(0.218)
Cohesion × Fixed Exchange Rates−0.175
(0.297)
Observations1,8721,6341,6341,3961,396
Source: Authors’ calculations.Note: The estimation is by instrumental variable. Country and time effects are omitted for reasons of parsimony. Robust standard errors are clustered at the country level in parentheses. Instruments include constraints on the executive and political constraints (see text for more details). The constant term is estimated but omitted.*p < .1; **p < .05; ***p < .01.
Source: Authors’ calculations.Note: The estimation is by instrumental variable. Country and time effects are omitted for reasons of parsimony. Robust standard errors are clustered at the country level in parentheses. Instruments include constraints on the executive and political constraints (see text for more details). The constant term is estimated but omitted.*p < .1; **p < .05; ***p < .01.

Sensitivity to Alternative Definitions of Fiscal Stance

The robustness of the results to a different definition of the dependent variable is also tested by identifying fiscal discipline with strong improvements in the fiscal stance, associated with fiscal consolidations. The dependent variable is now a dummy taking the value one during years of fiscal consolidation. The literature addressing the identification of fiscal adjustment episodes is vast and has relied on changes in the cyclically adjusted primary balance (CAPB) as a share of GDP. Some caveats surrounding this approach have been highlighted recently.22 To maximize country coverage, this analysis relies on Alesina and Ardagna’s (1998) method of identifying fiscal adjustments as years where the CAPB is at least 2 percent of GDP in one year, or at least 1.5 percentage points on average in the past two years.

Because we now have a dependent variable expressed in binary terms, we use a logistic regression to estimate equation (11.1). More precisely, we estimate the following model:

where ϕ,θ,β are vectors of the parameters to be estimated and Φ(.) is the logistic function.23

Given that the analysis relies on panel data, the structural model can be written as follows:

with i = 1, . . ., 78; t = 1980, . . ., 2013; λi captures the unobserved individual effects; and εit is the error term.

We take the analysis one step further and also assume that a fiscal adjustment is successful if the improvement in the CAPB for two consecutive years is at least η-times the standard deviation of the CAPB in the full panel (Afonso and Jalles 2012):

This analysis uses a threshold value of η = 1.

The results suggest that fixed exchange rate regimes decrease the likelihood of a given government engaging in a fiscal consolidation (see Table 11.7).24 More important, the likelihood is even smaller if the sample is restricted to successful fiscal episodes, as shown by the higher magnitude—in absolute value—of the estimated coefficients in columns (6)–(10) of Table 11.7. The Horizon of the policymakers is never relevant, in line with results in Alesina, Perotti, and Tavares (1998), who find that engaging in fiscal adjustments does not increase the likelihood of cabinet turnover. In this setting, political Cohesion becomes statistically irrelevant, and the interaction term between Cohesion and exchange rate regime also loses statistical significance.25

Sensitivity to Alternative Indicators of Exchange Rate Regime

The robustness of the empirical exercise to a different measure of the exchange rate regime is also tested. Instead of building the dummy variable using a normative classification of exchange rate systems from Reinhart and Rogoff (2004), a positive approach is used. Now, the new dummy variable ERitnew takes value one, denoting a fixed exchange rate regime, if the five-year rolling volatility of the real effective exchange rate is smaller than one-third of the five-year rolling average volatility.26 Results shown in Table 11.8 suggest that the new definition of fixed exchange regimes (Fixed ER-new) still yields negative and, when Cohesion is controlled for, statistically significant, coefficients when explaining the degree of fiscal discipline.

Table 11.7.Discretionary Fiscal Consolidations as Proxy for Fiscal Discipline: Logit Estimations
Specification
All Fiscal ConsolidationsSuccessful Fiscal Consolidations
(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)
Fixed Exchange Rates−0.405***−0.587***−0.526***−0.482***−0.607*−1.173**−1.720***−1.631**−1.797***−3.353*
(0.149)(0.165)(0.181)(0.176)(0.314)(0.556)(0.646)(0.665)(0.649)(1.877)
Growth0.002−0.002−0.003−0.033−0.0330.292***0.467***0.463***0.285**0.293**
(0.022)(0.024)(0.024)(0.024)(0.024)(0.102)(0.119)(0.120)(0.135)(0.137)
Trade Openness0.005*0.0040.0040.009***0.009***−0.037**−0.066***−0.065***−0.068*−0.067*
(0.003)(0.003)(0.003)(0.003)(0.003)(0.018)(0.023)(0.024)(0.039)(0.039)
Terms of Trade−0.005−0.006−0.006−0.002−0.002−0.086**−0.152***−0.151***−0.081*−0.080*
(0.004)(0.005)(0.005)(0.006)(0.006)(0.034)(0.031)(0.031)(0.048)(0.048)
Reserves0.148**0.0860.0890.0970.093−0.629*−1.103***−1.078***−0.204−0.226
(0.072)(0.078)(0.078)(0.087)(0.088)(0.346)(0.403)(0.395)(0.459)(0.471)
Horizon0.0980.0280.0460.219
(0.100)(0.134)(0.277)(0.527)
Horizon × Fixed Exchange Rates0.1490.239
(0.181)(0.584)
Cohesion0.782***0.737***0.7230.017
(0.189)(0.207)(0.888)(1.607)
Cohesion × Fixed Exchange Rates−0.155−1.553
(0.342)(1.777)
Observations985857857764764136134134100100
Source: Authors’calculations.Note: This is a logit estimation. The binary type dependent variable is identified in the second row. The constant term is estimated but omitted. Robust standard errors are clustered at the country level in parentheses.*p < .1; **p < .05; ***p<.01.
Source: Authors’calculations.Note: This is a logit estimation. The binary type dependent variable is identified in the second row. The constant term is estimated but omitted. Robust standard errors are clustered at the country level in parentheses.*p < .1; **p < .05; ***p<.01.
Table 11.8.Robustness to Alternative Definition of Exchange Rate Regime
Specification
(1)(2)(3)(4)
Fixed Exchange Rates—New−0.515−0.515−0.952***−0.907**
(0.356)(0.356)(0.356)(0.358)
Growth0.233***0.233***0.217***0.214***
(0.033)(0.033)(0.034)(0.034)
Trade Openness−0.005−0.005−0.008−0.009
(0.007)(0.007)(0.008)(0.008)
Terms of Trade0.023***0.023***0.033***0.035***
(0.007)(0.007)(0.009)(0.009)
Reserves0.2260.2230.335**0.349***
(0.137)(0.138)(0.133)(0.133)
Horizon0.301**0.312**
(0.135)(0.139)
Horizon × Fixed Exchange Rates—New−0.230
(0.369)
Cohesion0.992***1.044***
(0.259)(0.261)
Cohesion × Fixed Exchange Rates—New−0.852*
(0.440)
Observations1,0331,033976976
R20.5600.5610.5610.562
Source: Authors’ calculations.Note: The estimation with country and time effects is omitted for reasons of parsimony. Robust standard errors are clustered at the country level in parentheses. The constant term is estimated but omitted.*p < .1; **p < .05; ***p < .01.
Source: Authors’ calculations.Note: The estimation with country and time effects is omitted for reasons of parsimony. Robust standard errors are clustered at the country level in parentheses. The constant term is estimated but omitted.*p < .1; **p < .05; ***p < .01.

Case Study: Fiscal Discipline and Fixed Exchange Rates in the Euro Area

The introduction of a single currency in the European Union between 1999 and 2002 was a unique policy experience.27 An initial set of 11 countries made a voluntary and simultaneous transition from flexible to fully fixed exchange rates after joining the euro area. This experience is relevant to the analysis in this chapter for two reasons: First, because all euro area countries adhered to the euro at the same time and as part of a common move, the endogeneity of exchange rate regime choice is partially controlled for. In addition, it is normally in well-established democracies that politics seems to matter most for fiscal policy, which points to a case such as the euro area’s as worthy of attention.

To understand the effect of this decision on fiscal discipline, a case study analysis is performed using the synthetic control methodology (SCM) on the Euro-11 countries. The SCM is a formal data-driven procedure used to quantify the effect of an event—in this case, the introduction of a fully fixed exchange rate in the euro area—on an outcome variable, here, the primary budget balance. The method relies on the creation of an artificial counterfactual that mimics the primary budget balance before the introduction of the single currency using data on a set of variables from the years before the event, and then comparing the actual outcomes with the counterfactual for the years after the event.28 The counterfactual—the synthetic unit—is constructed as a weighted average of the primary budget balance in countries with characteristics similar to those of the country under consideration but subsequently unaffected by the event.29 Country weights are chosen to minimize the distance between the country under consideration and its counterfactual in terms of the primary balance variable and its predictors (Abadie, Diamond, and Hainmueller 2010). The effect of an event is obtained as the difference between the outcome variable for the country in question and the weighted average of the outcome variable for the synthetic control group.

Figure 11.4 plots a number of important results from the SCM exercise that confirm the theoretical predictions and the main empirical results of this chapter. Panel 1 in Figure 11.4 shows that, in the aftermath of moving to fixed exchange rates, the Euro-11 maintained a lower primary balance than in the synthetic counterfactual. In addition, Figure 11.4 shows results for four countries with extreme scores on the political Horizon and Cohesion variables vis-à-vis the principal component’s average value. Countries with high relative scores in at least one of the two political dimensions, such as Austria (higher-than-average horizon) or Portugal (higher-than-average cohesion), display a narrower distance from their synthetic counterfactuals, even if they experienced less fiscal discipline after joining the euro area. In contrast, countries with lower-than-average scores in both political dimensions, such as Belgium, saw their distance to the counterfactual increase. Finally, countries with higher-than-average scores on both political horizon and cohesion (Spain) managed to improve fiscal discipline significantly above their synthetic counterfactual, proving that strong politics partially compensates for the negative fiscal effect of moving to fixed exchange rates.30

Figure 11.4.Synthetic Control Analysis, Euro-11

Source: Authors’ calculations.

Note: Austria (panel 3) is a country with above-average political Horizon and below-average Cohesion. Portugal (panel 4) is a country with below-average political Horizon and above-average political Cohesion. Belgium (panel 5) is a country with below-average political Horizon and below-average political Cohesion. Spain (panel 6) is a country with above-average political Horizon and above-average political Cohesion. Euro- 11 = European Union 11 (Austria, Belgium, Finland, France, Germany, Ireland, Italy, Luxembourg, Netherlands, Portugal, and Spain). Data labels in panel 2 use International Organization for Standardization country codes.

Conclusion

This chapter presents both theoretical and empirical evidence showing that, contrary to the traditional argument, flexible exchange rate regimes are associated with more fiscal discipline. The fiscal implications of exchange rate regime choice do not occur in a vacuum, but within a specific political context. By bringing politics into the picture, this chapter contributes to the literature by carefully uncovering how the exchange rate regime interacts with the political context to affect fiscal policy outcomes. The results draw on the longest and widest cross-section of country experiences available. The analysis finds that strong political environments characterized by long Horizon and high Cohesion among policymakers (that is, where elections are not imminent and where there is little political fragmentation) are associated with better fiscal performance.31

These results offer important policy lessons from the point of view of fiscal policymaking. First, if policymakers operate in weak political contexts, flexible exchange rates are the preferred option because they are best suited to secure enhanced fiscal discipline. Second, the virtuous effect of flexible exchange rates on fiscal discipline is strengthened by strong political environments, characterized by long political Horizon and high political Cohesion. Third, in mixed political contexts, policymakers face a difficult choice because moving to a fixed exchange rate regime may negatively affect fiscal performance in a way that political institutions are unable to attenuate.

Annex 11.1. Robustness
Table 11.1.1.Robustness to Alternative Measures of Fiscal Discipline
Dependent VariableSpecification
Nominal Primary BalanceReal Primary Balance per Capita
(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)
Fixed Exchange Rates−0.926***−0.385−0.424−0.618*−0.233−0.881***−0.369−0.411−0.603*−0.206
(0.316)(0.301)(0.338)(0.393)(0.369)(0.313)(0.298)(0.335)(0.389)(0.366)
Growth−0.0060.0300.0300.0420.041−0.0050.0300.0300.0430.042
(0.034)(0.029)(0.029)(0.036)(0.036)(0.034)(0.029)(0.029)(0.035)(0.036)
Trade Openness−0.034***−0.022*−0.022***−0.039***−0.038***−0.034***−0.022***−0.022***−0.039***−0.038***
(0.007)(0.006)(0.006)(0.010)(0.010)(0.007)(0.006)(0.006)(0.010)(0.010)
Terms of Trade−0.006−0.002−0.002−0.008−0.008−0.006−0.002−0.002−0.008−0.008
(0.006)(0.004)(0.004)(0.006)(0.006)(0.006)(0.004)(0.004)(0.006)(0.006)
Reserves0.625***0.460*0.462***0.832***0.840***0.592***0.427***0.429***0.813***0.821***
(0.138)(0.114)(0.114)(0.152)(0.152)(0.137)(0.113)(0.113)(0.150)(0.151)
Horizon−0.154−0.121−0.149−0.114
(0.104)(0.140)(0.104)(0.139)
Horizon × Fixed Exchange Rates−0.088−0.093
(0.202)(0.201)
Cohesion0.909***0.746**0.943***0.775**
(0.291)(0.307)(0.288)(0.304)
Cohesion × Fixed Exchange Rates0.5660.584
(0.387)(0.383)
Observations1,1218968968008001,121896896800800
R20.7400.7620.7620.7370.7380.7410.7560.7560.7360.737
Source: Authors’calculations.Note: The estimation with country and time effects is omitted for reasons of parsimony. Robust standard errors are clustered at the country level in parentheses. The constant term is estimated but omitted.*p < .1; **p < .05; ***p < .01.
Source: Authors’calculations.Note: The estimation with country and time effects is omitted for reasons of parsimony. Robust standard errors are clustered at the country level in parentheses. The constant term is estimated but omitted.*p < .1; **p < .05; ***p < .01.
Annex 11.2. Synthetic Control Methodology

Step 1: We choose the country cases (Euro-11) and the potential comparator countries; we also choose the explanatory variables.

The group of country cases (Euro-11) is made up of European countries that belong to the OECD and that joined the euro area in 1999 (Annex Figure 11.2.1). The countries under analysis are Austria, Belgium, Finland, France, Germany, Ireland, Italy, Luxembourg, Netherlands, Portugal, and Spain.

Annex Figure 11.2.1.

Source: Authors’ calculations.

Note: The green boxes signify country cases; the red boxes signify the donor pool. OECD = Organisation for Economic Co-operation; Euro-11 = European Union 11 (Austria, Belgium, Finland, France, Germany, Ireland, Italy, Luxembourg, Netherlands, Portugal, and Spain).

The group of comparator countries is made up of European countries that belong to the OECD but did not join the euro area in 1999. Also from that group we exclude transition economies (because of lack of comparable data before the mid-1990s). The countries ultimately in the control group are Denmark, Greece, Iceland, Israel, Norway, Sweden, Switzerland, Turkey, and United Kingdom.

The outcome variable is the primary budget balance (as a percentage of GDP). As predictor variables to match the behavior of the primary balance, we choose the primary balance in 1981, 1993, and 2000; total reserves minus gold (in million U.S. dollars); real GDP growth (percent); terms of trade (U.S. dollars; index); trade openness (percent in U.S. dollars); change in nominal exchange rate; and change in debt-to-GDP ratio. All variables take three-year moving average values to smooth the fluctuations.

The year of the event is 2001 (last year before actual circulation of the single currency).

Step 2: Given the group of comparator countries and the outcome and predictor variables, we calculate the synthetic series.

Given country and variable selection, the procedure calculates weights of the predictor variables and comparator countries to reproduce as closely as possible the values of the outcome variable pre-event. Not all comparator countries have to receive a positive weight to create the synthetic comparator. The procedure is based on an iterative optimization algorithm as follows:

  • Start from some initial vector of weights of the predictor variables V, choose the vector W* of country weights to minimize a distance || X1− X0W || where X1 and X0 are matrices of predictor variables for the unit of interest and its comparator units, respectively, subject to weight constraints (weights have to be between 0 and 1). In particular, W* minimizes

  • Once the country weights are chosen, the variable weight V* is chosen among all positive definite and diagonal matrices such that the mean square prediction error of the outcome variable is minimized over pre-event periods. In particular,

where Z1 and Z0 are matrices of the outcome variable for the unit of interest and its comparator units, respectively. The resulting V* is used as input to equation (11.2.1) for the next round of optimization.

  • This iterative process continues until V* and W* converge. In summary, the synthetic series is constructed by solving a nested optimization problem that minimizes equation (11.2.2) for given W* (V) given by equation (11.2.1) (Abadie, Diamond, and Hainmueller 2010).

  • Using the weights thus obtained, we then use the synthetic comparator to create a counterfactual path of the outcome variable post-event.

Step 3: Compare actual post-event outcome variable series with the synthetic comparator.

The difference between the two series is the estimated impact of the event (assuming that all other factors potentially affecting the variable of interest have been controlled for successfully).

References

    AbadieA.A.Diamond and J.Hainmueller. 2010. “Synthetic Control Methods for Comparative Case Studies: Estimating the Effect of California’s Tobacco Control Program.Journal of the American Statistical Association105 (490): 493505.

    AcemogluD.S.JohnsonJ.Robinson and Y.Thaicharoen. 2003. “Institutional Causes, Macroeconomic Symptoms: Volatility, Crises and Growth.Journal of Monetary Economics50: 49123.

    AfonsoA. and J. T.Jalles. 2012. “Measuring the Success of Fiscal Consolidations.Applied Financial Economics22 (13): 105361.

    AfonsoA. and J. T.Jalles. 2014. “Assessing Fiscal Episodes.Economic Modelling37: 25570.

    AlesinaA. and S.Ardagna. 1998. “Tales of Fiscal Contractions.Economic Policy27: 487545.

    AlesinaA.R.Perotti and J.Tavares. 1998. “The Political Economy of Fiscal Adjustments.Brookings Papers in Political Economy1: 197266.

    BattagliniM.S.Nunnari and T. R.Palfrey. 2016. “The Political Economy of Public Debt: A Laboratory Study.Discussion Paper 11357Centre for Economic Policy ResearchLondon.

    BordoM.2003. “Exchange Rate Regime Choice in Historical Perspective.NBER Working Paper 9654 National Bureau of Economic ResearchCambridge, Massachusetts.

    CruzC.P.Keefer and C.Scartascini. 2016. “Database of Political Institutions Codebook, 2015 Update (DPI 2015).Inter-American Development Bank. Updated version ofT.BeckG.ClarkeA.GroffP.Keefer and P.Walsh. 2001. “New Tools in Comparative Political Economy: The Database of Political Institutions.World Bank Economic Review15 (1): 16576.

    DriscollJ. C. and A. C.Kraay. 1998. “Consistent Covariance Matrix Estimation with Spatially Dependent Panel Data.Review of Economics and Statistics80: 54960.

    DuttaguptaR. and G.Tolosa. 2006. “Fiscal Discipline and Exchange Rate Regimes: Evidence from the Caribbean.Working Paper 06/119International Monetary FundWashington, DC.

    FatásA. and I.Mihov. 2013. “Policy Volatility, Institutions and Economic Growth.Review of Economics and Statistics95 (2): 36276.

    FatásA. and A. K.Rose. 2001. “Do Monetary Handcuffs Restrain Leviathan? Fiscal Policy in Extreme Exchange Rate Regimes.IMF Staff Papers47: 4061.

    FrankelJ.M.Goldstein and P.Masson. 1991. “Characteristics of a Successful Exchange Rate System.Occasional Paper 82International Monetary FundWashington, DC.

    GavinM. and R.Perotti. 1997. “Fiscal Policy in Latin America.” In NBER Macroeconomics Annual 1997 edited by B. S.Bernanke and J.Rotemberg. Cambridge, MA: MIT Press.

    GiavazziF. and M.Pagano. 1988. “The Advantage of Tying One’s Hands: EMS Discipline and Central Bank Credibility.European Economic Review32 (5): 107577.

    GiavazziF. and M.Pagano. 1996. “Non-Keynesian Effects of Fiscal Policy Changes: International Evidence and the Swedish Experience.Swedish Economic Policy Review3 (1): 67103.

    GreeneW. H.2012. Econometric Analysis 7th edition. Upper Saddle RiverNJ: Prentice Hall.

    HeniszW. J.2000. “The Institutional Environment for Economic Growth.Economics and Politics12 (1): 131.

    LambertiniL. and J.Tavares. 2005. “Exchange Rates and Fiscal Adjustments: Evidence from the OECD and Implications for the EMU.BEJM Contributions in Macroeconomics5 (1): Article 11.

    LaneP. and R.Perotti. 2003. “The Importance of Composition of Fiscal Policy: Evidence from Different Exchange Rate Regimes.Journal of Public Economics87 (910): 225379.

    MorrisR. and L.Schuknecht. 2007. “Structural Balances and Asset Windfalls: The Role of Asset Prices Revisited.Working Paper 737 European Central Bank Frankfurt.

    MullerW. C. and K.Strom eds. 1999. Policy Office or Votes? How Political Parties in Western Europe Make Hard Decisions.Cambridge, U.K.: Cambridge University Press.

    ReinhartC. M. and K. S.Rogoff. 2004. “The Modern History of Exchange Rate Arrangements: A Reinterpretation.Quarterly Journal of Economics119 (1): 148.

    RousseeuwP. J. and K.van Driessen. 1999. “A Fast Algorithm for the Minimum Covariance Determinant Estimator.Technometrics41 (3): 21223.

    RousseeuwP. J. and V.Yohai. 1984. “Robust Regression by Means of S Estimators.” In Robust and Nonlinear Time Series Analysis edited by J.FrankeW.Härdle and R. D.Martin256—74 Lecture Notes in Statistics 26. New York: Springer Verlag.

    SunY.2003. “Do Fixed Exchange Rates Induce More Fiscal Discipline?Working Paper 03/78International Monetary FundWashington, DC.

    TanziV. and L.Schuknecht. 1997. “Reconsidering the Fiscal Role of Government: The International Perspective.American Economic Review87 (2): 16472.

    TornellA. and A.Velasco. 1995. “Fiscal Discipline and the Choice of Exchange Rate Regime.European Economic Review 39 (34): 75970.

    TornellA. and A.Velasco. 1998. “Fiscal Discipline and the Choice of a Nominal Anchor in Stabilization.Journal of International Economics46 (1): 30.

    WooldridgeJ. M.2002. Econometric Analysis of Cross Section and Panel Data.Cambridge, MA: MIT Press.

    YohaiV. J.1987. “High Breakdown-Point and High Efficiency Robust Estimates for Regression.Annals of Statistics15 (2): 64256.

J. Jalles and C. Mulas-Granados are with the International Monetary Fund. J. Tavares is with the Nova School of Business and Economics in Lisbon and the Center for Economic Policy Research in London. The authors are grateful to Vitor Gaspar, Sanjeev Gupta, and Ben Clements for useful comments and suggestions. Participants at the FAD Seminar Series also contributed with valuable discussions. The authors also thank Jaime Marques Pereira, Michela Schena, and Carolina Correa-Caro for excellent research assistance. The opinions expressed herein are those of the authors and do not necessarily reflect those of the authors’ employers.

For instance, the 1990s witnessed a series of crises in emerging market economies, as well as European countries under the Exchange Rate Mechanism, that suggested financial openness, monetary independence, and pegs were incompatible.

As reviewed in Bordo (2003), exchange rate regime choice evolved considerably over the past century. In the early twentieth century, adopting the gold standard—and, thus, a hard peg—seemed the obvious policy choice, followed by most advanced economies. Today the policy choice is the opposite, though equally “obvious.” All advanced economies, with the exception of euro area countries among themselves, have adopted flexible exchange rates. Developing countries, with some exceptions, seem more or less constrained to follow the prevailing policy view, and mimic, if not follow, advanced economies’ actions.

Duttagupta and Tolosa (2006) explore the sample of small economies in the Caribbean, while Tornell and Velasco (1995) present evidence for the CFA franc community in Africa.

See Bordo (2003) for a review of exchange rate regime adoption.

Limited commitment can be interpreted as a response to the need to keep the government solvent, an optimal response given the cost of commitment, or both. The likelihood of this happening relates to fiscal dominance.

For this intertemporal choice to be available to the policymaker, it is important that the policymaker either has sufficient reserves or access to credit. Otherwise, deficits would immediately lead to a currency depreciation.

Tornell and Velasco (1995, 1998) do not explicitly consider the trade-off between present and future punishment in fixed and flexible regimes; they completely discount the future in the former case and do not explicitly consider the future in the latter.

Duttagupta and Tolosa (2006) suggest a comparative model of fiscal policy for countries that either go it alone or are integrated in a currency union.

In political science, political parties are often seen as primarily office-seeking or policy-seeking parties. Office-seeking parties maximize their control over political office benefits, while policy-seeking parties maximize their impact on public policy (for further analysis on this issue, see Muller and Strom 1999).

Note that the central bank cares about both keeping the exchange rate stable and avoiding steep drops in output. Because higher borrowing levels by fiscal authorities would have a greater output cost, the central bank would react to debt increases by gradually accommodating them through currency depreciation.

From the IMF’s International Financial Statistics.

We use the classification of exchange rate regimes provided by Reinhart and Rogoff (2004) and generate a dummy variable that takes the value one (fixed exchange rate) if the country/year observation has any value between 1 and 8 in Reinhart and Rogoff’s classification, and takes the value zero otherwise—original Reinhart and Rogoff values of 9 to 15. We also test the robustness of the results by generating another dummy variable that uses Coarse’s classification from the IMF.

This is in line with Duttagupta and Tolosa (2006).

This latter indicator refers to actual months left to next election, after the fact, while the variable “more years left in current term” is observed ex ante. Both are informative.

A likelihood ratio test was used to examine the “sphericity” case, allowing for sampling variability in the correlations. This test comfortably rejects sphericity at the 1 percent level. The first factor explains almost 40 percent of the variance in the standardized data (see Table 11.2).

The source for each component variable is the Database on Political Institutions (Cruz, Keefer, and Scartascini 2016).

Principal component analysis is based on the classical covariance matrix, which is sensitive to outliers. Here we conduct a robust estimation of the covariance matrix. A well-suited method is the minimum covariance determinant (MCD) that considers all subsets containing h percent of the observations and estimates the variance of the mean on the data of the subset associated with the smallest covariance matrix determinant. Specifically, we implement Rousseeuw and van Driessen’s (1999) algorithm. When we compute the same indices with the MCD version, we obtain similar results, suggesting that outliers are not driving the factor analysis.

Uniqueness of a variable is the share of its variance that is not accounted for by all the factors.

In the first stage, it takes the S estimator, a high breakdown value method introduced in Rousseeuw and Yohai (1984) applied to the residual scale. It then derives starting values for the coefficient vectors, and on the second stage applies the Huber-type bi-square M-estimator using iteratively reweighted least squares to obtain the final coefficient estimates.

For reasons of parsimony, the remainder of the chapter refrains from commenting on the controls.

Results for Cohesion are available from the authors upon request.

In particular, the CAPB approach could bias empirical estimates toward finding evidence of non-Keynesian effects (see, for example, Afonso and Jalles 2014). Many nonpolicy factors influence the CAPB and can lead to erroneous conclusions regarding fiscal policy changes. For example, a stock price boom raises the CAPB by increasing capital gains tax revenue and tends to coincide with an expansion in private demand (Morris and Schuknecht 2007). Even when the CAPB accurately measures fiscal actions, these actions include discretionary responses to economic developments.

Because probit models do not lend themselves well to the fixed-effects treatment because of the incidental parameter problem (Wooldridge 2002, 484), we estimate a logit model with fixed effects.

As a robustness check to Alesina and Ardagna’s (1998) method, Giavazzi and Pagano’s (1996) method was also used. They propose using the cumulative changes in the CAPB that are at least 5, 4, and 3 percentage points of GDP in, respectively, 4, 3, or 2 years, or 3 percentage points in one year. Results, available upon request, did not qualitatively change.

An alternative dependent variable was also tried by using fiscal discipline defined in nominal levels (instead of percentage of GDP) and in real per capita terms. Results, available in Annex Table 11.1.1, confirm the findings.

Volatility is measured by the standard deviation. Note that changing the threshold slightly above or below one-third does not change the results.

In 1999, 11 European Union member states joined the final stage of monetary union and fixed their bilateral exchange rates. The euro was introduced then as a virtual currency, while national currencies remained in circulation for three more years. In 2002, the euro was substituted for national currencies in circulation. The Euro-11 countries are Austria, Belgium, Finland, France, Germany, Luxembourg, Ireland, Italy, Netherlands, Portugal, and Spain.

See Annex 11.2 for more details.

The control countries included in the synthetic unit are, in this case, OECD member countries other than those under analysis (that is, the Euro-11). We also exclude transition economies. See Annex 11.2.

For Spain, we replicated the analysis using the structural balance concept to control for the potential effect on the primary balance stemming from the real estate bubble that increased fiscal revenues considerably during the late 1990s. Results remain unchanged: although Spain could have adjusted further, it still performed better after the introduction of the euro.

In a recent paper making use of a laboratory experiment, Battaglini, Nunnari, and Palfrey (2016) find evidence that more inclusive requirements for fiscal decision making and more vulnerability to shocks (which can be equated to longer horizons) are associated with lower debt accumulation.

    Other Resources Citing This Publication