Fiscal Politics

Chapter 8. It’s Politics, Stupid! Political Constraints Determine Governments’ Reactions to the Great Recession

Vitor Gaspar, Sanjeev Gupta, and Carlos Mulas-Granados
Published Date:
April 2017
  • ShareShare
Show Summary Details


After the collapse of Lehman Brothers in September 2008, the world economy was hit by an economic crisis of a scale not seen since the Great Depression; during the winter half-year 2008/2009 world trade collapsed by almost 20 percent while world industrial production shrank by about 12 percent. The Great Recession, as the shock came to be known, did—at least in terms of its size—come as a surprise to virtually everyone. Governments around the world were surprised too, yet many reacted quickly by introducing fiscal stimulus packages.

However, the size of these packages varied considerably across countries. The UNCTAD Trade and Development Report (UNCTAD 2009) highlights that countries such as Kazakhstan, Saudi Arabia and Singapore had scheduled to implement discretionary packages amounting to 11.1, 9.2 and 8 percent of GDP, respectively. On the other end of the spectrum, the packages scheduled by the governments in Italy and Switzerland were 0.3 and 0.5 percent of their respective GDP levels.2

What explains these differences? Policy-oriented organizations and recent academic research have so far concentrated on three factors: “need,” the size of the shortfall in aggregate demand that discretionary spending aims to compensate for; “fiscal space,” the government’s fiscal ability to spend in time of need; and “effectiveness,” the fraction of the fiscal spending that translates into aggregate demand.3

In contrast, the role of domestic political factors is often neglected. Given that fiscal policy is enacted within a political environment, this environment should be expected to influence the outcome. In this paper we attempt to address this shortcoming by explicitly taking politics into account. We do this by estimating the effect of political constraints on the size of stimulus packages that were enacted in the wake of the crisis. We approximate the degree of political constraints by looking at whether a country’s executive party had control over the majority of legislative branches that are relevant for policy making. If it did, we consider it to have been free of political constraints as it had unilateral law-making power and was not required to cooperate with the opposition in order to enact fiscal stimulus measures.

We find that the effect of political constraints on the size of fiscal stimulus packages that governments have enacted in reaction to the shock of 2008–09 is large, statistically significant and robust to alternative dependent variables, alternative model specifications and changes in the sample. Our results suggest that on average, governments without political constraints have implemented stimulus packages that were—depending on the fiscal stimulus measure used—about 1 to 2.7 percentage points of GDP larger in size than packages enacted by governments that faced political constraints, and thus did have to cooperate with the opposition.

What our results do not and cannot show is whether these stimulus packages were appropriate responses to the crisis in the sense that they were effective in supporting economic recovery. For that we would have to analyse the consequences of different fiscal measures on the business cycle. There is a substantial, at times ideologically driven, literature on these questions,4 but our concern here is a different one: we are interested in the drivers behind the size of these stimulus packages, not in their effectiveness.

The remainder of this paper is organized as follows. The second section discusses our conceptual framework. In the third section a selective overview of the related literature is given. Whereas the data and the empirical model are introduced in the fourth section, the fifth section presents the empirical results. The sixth section offers some concluding remarks.

Conceptual Framework

Why should political constraints have an impact on a government’s response to an economic shock? A vast literature on economic voting finds that if voters are satisfied with the economic performance prior to an election, they re-elect the incumbent government while if they are not, they do not.5Bartels (2011) looks at the electoral consequences of economic stimulus packages during the Great Recession and finds that

voters consistently punish . . . incumbent governments for bad economic conditions, with little apparent regard for the ideology of the government or global economic conditions at the time of the election. [There is also] some evidence of electoral responses to specific fiscal policy choices, most notably, a boost in incumbent governments’ electoral support associated with spending on economic stimulus programs. (p. 1)

These findings have strong implications for political incentives. If incumbent governments expect to be punished for bad economic performance and to be rewarded for enacting stimulus packages in the wake of economic downturns, then we should expect them to enact stimulus packages of a size they deem optimal given need, fiscal space and the effectiveness of such packages. For the same reason, we would expect opposition parties to try to block, delay or reduce the size of such packages. In addition to political calculus, any type of fiscal stimulus will have distributional consequences that the opposition may oppose based on ideological differences. Hence, in countries where the opposition has the political means to influence legislation, we should, everything else being equal, expect stimulus packages to be smaller, at least initially.

What about autocratic regimes? To the extent that the legitimacy of the political regime depends on its delivery of economic progress, the same logic for enacting fiscal stimulus packages applies. Olson (2000) argues that a stable and durable autocratic regime has a strong interest to provide prosperity-enhancing public goods to protect the economic system from which it extracts taxes. When faced with a shortfall in aggregate demand, the goal of such a regime is the preservation of its rent, creating the incentive to introduce fiscal stimulus measures. What is different, of course, is the absence of an opposition that can delay or negotiate down the size of such packages or attempt to change its composition. All else equal, we therefore expect packages of non-democracies, like those in democracies that do not face political constraints, to be larger than those of democracies that do face such constraints.

The above reasoning rests on the premise that voters hold the government responsible for poor economic conditions, and reward it for enacting stimulus packages as a reaction to crises, regardless of whether the government faces political constraints or not. If this assumption is relaxed, then alternative interpretations for the negative relationship between the size of stimulus packages and political constraints emerge.

Suppose that voters realise that politically constrained governments should not be (fully) blamed for poor economic conditions. In such a case, constrained governments might spend less political capital and effort on enacting the stimulus package they deem optimal, because they know that they will not be blamed (as much) for a poor economic recovery. The observed outcome would remain the same: we would expect stimulus packages to be smaller in the presence of political constraints.

A further explanation for the negative relationship is that political constraints might prevent governments from enacting stimulus packages that are larger than what is socially optimal. If unconstrained governments are faced with an exogenous economic shock that reduces their re-election chances, this might shorten their time-horizon substantially. As a result, such governments might try to enact stimulus packages that are larger than what is socially optimal as a high-risk strategy to secure re-election. Political constraints could prevent this kind of behaviour. Not only would they make the implementation of oversized packages more difficult, but also the political burden of facing poor economic conditions would, as argued above, be shared with the opposition and so the time-horizon of such governments would not be shortened as much.

All of these explanations have in common that they lead to a negative relationship between stimulus size and political constraints and, given the data at hand, we cannot empirically discriminate between them. Nevertheless, they all underline the main message of this paper: political constraints matter.

Related Literature

There are three strands of the literature on the interaction between politics and economics that are related to our argument.

First, there is research that highlights the importance of politics for both fiscal and monetary policy outcomes: Porteba (1994) finds that one-party governments can and do react faster to unexpected fiscal deficit shocks than their divided-government counterparts. Weise (2012) concludes that the political environment in the United States in the 1970s was a main determinant of the Federal Reserve’s too moderate anti-inflationary policy, and that a change in the political environment was also behind the Federal Reserve’s switch to a more aggressive policy after 1979. Spolaore (2004) argues that cabinet systems in which there is a single decision maker adjust faster to shocks than systems with multiple decision makers.

A second strand highlights political economy considerations as a major drawback for discretionary fiscal policy. Blinder (1997) outlines the merits of moving a greater number of policy decisions away from the realm of politics into the realm of technocracy, so as to make them the result of a deliberative and objective process rather than the outgrowth of political considerations. Blanchard et al. (2010) mention the limits that political constraints impose on the de facto usefulness of discretionary fiscal policy. Cecchetti (2002) argues that when it comes to fiscal policy, political considerations tend to collide with economic prescriptions, while Romer (2012) mentions political-economy aspects to be important in understanding fiscal policy responses to the crisis.

Finally, Armingeon (2012) directly investigates the importance of politics in government’s reaction to the Great Recession. He finds that a unified government was a necessary condition for deviating from what he calls the default reaction to the crisis: a moderate fiscal expansion. In particular, in his qualitative and categorical analysis, he finds that it was only unified governments that enacted large fiscal stimulus packages. While these findings indicate that politics has played a role in determining the size of fiscal stimulus packages, they provide limited information on the size and strength of this relationship. It is this literature to which our paper contributes most directly.

Empirical Model and Data Description

Our estimation relies on a simple OLS framework, with stimulus package size as the dependent variable, political constraints as the main explanatory variable, and a set of control variables to capture need, fiscal space, and effectiveness. This section discusses the precise definition, measurement and data sources for each of these variables.

Size of Stimulus Packages

To measure the size of the fiscal stimulus we rely on two different sources and construct four different variables. All four of these variables have in common that they concentrate on fiscal policy measures initiated or carried out in the crisis year 2009.6 We consider the bankruptcy of Lehman Brothers in autumn 2008 and the subsequent collapse in world trade as a largely exogenous shock and do not want to mix this up with events, like the euro crisis, happening after an initial recovery in the second half of 2009 and early 2010.

Our first variable is directly taken from Table 1.8 in UNCTAD (2009). This table was compiled by the UNCTAD secretariat using a number of different sources.7 The variable corresponds to discretionary measures on public spending or revenues in response to the financial crisis, excluding so-called automatic stabilizers and scheduled to be implemented across a one- to three-year window. Hence, it covers discretionary “promises” of governments in selected countries as percentage of GDP over a somewhat varying implementation horizon. There are a few caveats when using this data: time horizons of these stimulus packages differ substantially and the exact definition of what is part of a stimulus package is likely to be country- and source-dependent to some extent.8 Furthermore, this particular data set only allows us to use a sample of 44 OECD and emerging market countries. Both data quality and coverage have led us to also look for other data sources.

The second variable is taken from Appendix Table 5 in Horton et al. (2009). It compares primary deficit forecasts for 2009 as published by the IMF in its July 2009 Update (IMF 2009b) and its October 2007 release of the World Economic Outlook (IMF 2007). We view this as a measure for the forecasted change in fiscal policy induced by the Great Recession and not related to changing interest payments of the government. The difference with the UNCTAD measure is twofold. First, it includes both discretionary measures as well as changes caused by automatic stabilizers.9 Second, it has a fixed time horizon: it reflects “promises” for the year 2009. These differences notwithstanding, in both cases, we are looking at forecasts, i.e., “promises,” and not at actual realizations.

But there might be a difference between the political promises for spending made during the crisis year and the spending that was actually implemented. To take this into account, our two remaining variables focus on actual realizations. Focusing on actual spending also has the advantage that it avoids issues surrounding the definition of stimulus packages, which, as discussed above, are likely to differ between countries. To construct our variables we use information released in the April 2013 IMF World Economic Outlook (IMF 2013) and take actual changes in primary fiscal deficits between 2008 and 2009. To increase the sample size, we also look at actual changes in (total) fiscal deficits during the crisis year.

Table 8.1 summarizes our four main dependent variables. Overall, the size of the fiscal stimulus is substantial with averages ranging from close to 2.5 to almost 5 percent of (pre-crisis) GDP. Although it covers up to three years, the UNCTAD variable contains the lowest values. A likely explanation for this is that by construction, it is the only variable that does not include the effect of automatic stabilizers. The table also reveals that, on average, democracies have enacted smaller fiscal stimulus measures than autocracies.10 Finally, with standard deviations between 3.3 and 4.5 percent of GDP, it is also safe to say that there is wide variation in the size of stimulus packages initiated during the Great Recession.

Table 8.1.Descriptive Statistics and Correlation Matrix for the Dependent Variables
Descriptive statistics
Obs.AverageStandard DeviationMinimumMaximumSource
Promised Stimulus 2008–12Discretionary measures442.443.39–8.38.0UNCTAD
Promised Stimulus 2008–09Change forecasted primary deficit404.973.29–0.714.9IMF 2009b/2007
Realized Stimulus in 2009Change primary deficit 2009 (% 2007-GDP)1084.164.31–3.421.6IMF 2013
[only democracies)773.583.36–3.415.7IMF 2013
Realized Stimulus in 2009Change deficit 2009 (% 2007-GDP)1514.004.48–4.125.3IMF 2013
[only democracies)1003.523.50–4.117.5IMF 2013
Variable description(1)(2)(3)(4)(5)(6)
(1) Promised Stimulus 2008–12Discretionary measures3640364439
(2) Promised Stimulus 2008–09Change forecasted primary deficit–0.0137344036
(3) Realized Stimulus in 2009Change primary deficit 2009 (% 2007-GDP)0.140.757710877
(4)[only democracies)–0.010.791.007777
(5) Realized Stimulus in 2009Change deficit 2009 (% 2007-GDP)0.030.740.980.96100
(6)[only democracies)–0.110.770.960.961.00
Correlation with the political constraints dummy:–0.31–0.29–0.16–0.18–0.10–0.11

Political Constraints

Political constraints are captured by a binary variable that equals one if during the Great Recession (i.e., during the winter of 2008/2009) a country’s executive party did not have a majority in the legislative branches that have law-making power. Conversely, the variable is equal to zero if throughout that same period, the party of the executive did have a majority in these branches and could therefore unilaterally enact law. All political variables—including this one—are taken from the 2012 version of the Database of Political Institutions (Beck et al. 2001). This particular variable is based on the variable ALLHOUSE.11,12

Given the exogenous character of the shock we are analyzing, we are convinced that we can treat our political constraint dummy as exogenous. Nonetheless, in Annex 8.1, we present robustness exercises where we investigate possible endogeneity issues. The results support our view.

While the constraints dummy likely captures the most direct dimensions of political constraints, there are more subtle constraints that, by virtue of being a dummy variable, it cannot capture. Consider, as an example, the events in the United States in early 2011. At the time the American Recovery and Reinvestment Act was enacted and signed into law by the Democratic president Barack Obama, the Democratic Party also controlled both the Senate and the House of Representatives. So, according to the definition of our constraints dummy, the Democrats were free of political constraints. And yet, there is evidence to suggest that both the Democratic Party’s internal disputes as well as public pressure prevented the stimulus package from being even larger than the actual $787 billion. Alter (2011) and Wallace-Wells (2001) report, for instance, that Christina Romer and Larry Summers, the President’s key economic advisers at the time, both believed that to close the entire output gap, the stimulus package would need to be above the politically incendiary 1 trillion dollar mark. So, as in the case of our stimulus measures, it is important to realize that while the variable captures an important part of what we aim to measure, it cannot account for all the country-specific subtleties.

Need for Fiscal Stimulus

The need for discretionary measures depends on both the expected size and type of the shock and the expected degree to which automatic stabilizers will alleviate it.

To proxy the size of the shock, we use the realised drop in exports during the winter half-year 2008/2009 relative to the winter half-year 2007/2008, measured as a percentage of 2007 GDP levels. For this we resort to the monthly export figures published in the IMF Direction of Trade Statistics. The timing of when stimulus measures were announced and implemented makes it very unlikely that they had a substantial impact on the size of this export shock, so that we can treat the variable as exogenous.

To proxy the role of automatic stabilizers we follow Gali (1994) and use the pre-crisis level of government expenditure as a percentage of GDP, as measured for 2007 and published by the IMF in April 2013 (IMF 2013). We thus assume that a larger public sector is more stabilizing than a smaller one. Depending on the dependent variable, we expect either a positive or a negative effect of this variable: for a given output gap, a higher level of government expenditure should reduce the size of discretionary measures, while it should increase the change in the deficit (i.e., in the total fiscal stimulus). The change in the deficit should increase because for a given size of the discretionary stimulus, higher government expenditures automatically alleviate the negative consequences of the shock, independent of the political decision-making process.

Fiscal Space

To capture a government’s fiscal space, we use two variables: the gross public debt-to-GDP ratio as measured for 2007, and the deficit-to-GDP ratio for 2007. Both are taken from the IMF World Economic Outlook published in April 2013 (IMF 2013). The differences across countries, particularly in pre-crisis deficit levels, are substantial. These reflect, among other things, differences in natural resources. In particular, those countries that export substantial amounts of oil or gas tend to have much smaller deficits or even substantial surpluses.13

Effectiveness of Fiscal Stimulus

To take the effectiveness of any fiscal stimulus into account, we include a broad measure of economic globalization as part of the KOF Globalization Index. We refer to figures for the year 2007. Small open economies have fewer opportunities to stimulate their own economy because a larger part of a given measure evaporates away to the rest of the world. At the same time, they also benefit more from measures undertaken by large trading partners. Both of these mechanisms reduce incentives to undertake large fiscal stimulus measures.

Table 8.2 shows the descriptive statistics for the above-mentioned right-hand-side variables as well as for all variables used in our robustness exercises and discussed in Annex 8.1.14 Regarding our main variable of interest, about half of the countries in our sample face political constraint, in the sense that the executive and legislative bodies are controlled by different parties. Quite a number of the countries in our sample are non-democratic. When focusing on democracies only, around 70 percent of the governments were not able to enact law unilaterally and were thereby politically constrained during the crisis period. The constraints dummy is not highly correlated with any of the control variables, so that including these variables into the model will most likely only have the effect of increasing the precision by which we can estimate the effect of the constraints dummy. There is also hardly any correlation among the control variables themselves, with the natural exception being the dummies for European Union (EU) and euro area membership, where the correlation coefficient is 0.71. Apart from that, the second highest correlation coefficient is between narrow money growth and official reserves and equals 0.56. Furthermore, our economic globalization measure and our measure of government size have a high absolute correlation of 0.45; more globalized economies, which often are European, also tend to have higher government expenditure shares.15

Table 8.2.Descriptive Statistics and Correlation Matrix for the Main Explanatory Variables
Descriptive Statistics
Obs.AverageStandard DeviationMinimumMaximumSource
Political Constraint Dummy1510.540.500.01.0DPI2012
Government Expenditures in 200715130.2510.387.652.6IMF, WEO April 2013
Change in 2009 Growth Forecast in April 2009 r.t. April 2008148–0.471.23–5.04.1IMF, WEO April 2009/October 2008
Change of Exports in Winter 2008/09 (%2007-GDP)143–4.035.31–44.22.6IMF, DOTS 2013
Percent Change Local Currency to USD between 2008:Q2 and 2008:Q414012.4311.68–8.063.5IMF
Growth Official Reserves (in USD) between 2008:Q2 and 2008:Q4142–4.6518.27–48.194.9IMF
Government Debt in 2007 (percent of GDP)14549.2250.171.3494.9IMF, WEO April 2013
Government Deficit in 2007 (percent of GDP)151–0.486.91–57.115.7IMF, WEO April 2013
Lending Rate in Winter 2008/0911613.467.951.052.6IMF
Central Bank Independence, Legal Measure880.620.200.20.9Crowe and Meade (2008)
Central Bank Governor Irregular Turnover Rate1240.
Change in the Lending Rate between August and December 20081160.462.33–8.18.2IMF
Growth Rate of M1 between August and December 2008785.1410.25–18.431.7Datastream, central banks
KOF Economic Globalization in 200713163.8216.9523.996.4KOF
KOF Political Globalization in 200715069.4519.6423.498.0KOF
G20 Dummy1510.120.330.01.0G20
Dummy for EU Membership1510.180.380.01.0EU
Dummy for EMU/Euro Area Membership1510.100.300.01.0ECB
Under an IMF Program1510.540.500.01.0ECB
Variable Description(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)
(1) Political Constraint Dummy1511481431401421451511168812411678131150151151151151
(2) Government Expenditures in 20070.101481431401421451511168812411678131150151151151151
(3) Change in 2009 Growth Forecast in April 2009 r.t. April 20080.04–0.131401381391421481148612111477129147148148148148
(4) Change of Exports in Winter 2008/09 (%2007-GDP)–0.10–0.01–0.271341341381431118611911177127143143143143143
(5) Percent Change Local Currency to USD between 2008:Q2 and 2008:Q40.240.24–0.170.021381351401098512210976125140140140140140
(6) Growth Official Reserves (in USD) between 2008:Q2 and 2008:Q4––0.11–0.231361421108712211078125141142142142142
(7) Government Debt in 2007 (percent of GDP)–0.45–0.180.121451108711911076127144145145145145
(8) Government Deficit in 2007 (percent of GDP)–0.120.17–0.070.34–0.15–0.140.211168812411678131150151151151151
(9) Lending Rate in Winter 2008/090.08–0.11–0.010.19–0.010.13–0.050.116610011659101116116116116116
(10) Central Bank Independence, Legal Measure0.280.22––
(11) Central Bank Governor Irregular Turnover Rate0.16–0.15–0.030.20–0.010.02–
(12) Change in the Lending Rate between August and December 20080.09––0.17–0.12–
(13) Growth Rate of M1 between August and December 2008–0.07––0.030.560.210.200.28–0.040.23–0.11747778787878
(14) KOF Economic Globalization in 20070.240.48–0.31–0.370.31–0.09–0.11–0.26–0.230.27–0.100.02–0.35131131131131131
(15) KOF Political Globalization in 20070.100.20––0.20–0.100.14––0.190.27150150150150
(16) G20 Dummy–0.030.12–––0.04–
(17) Dummy for EU Membership0.250.54–0.26–0.130.32–0.19–0.050.04–0.220.45–0.090.03–0.450.620.440.04151151
(18) Dummy for EMU/Euro Area Membership0.130.44–0.15–0.050.17––0.190.40–0.05–0.10–0.240.450.340.080.71151
(19) Under an IMF Program0.11–––0.31–0.10–0.21–0.18–0.19

Empirical Results

Table 8.3 presents our main results. Columns (1), (2), (3) and (5) report results for each of the four dependent variables using the full sample for which data is available.16 Columns (4) and (6) of the table restrict the sample of our two realised deficit measures to only democratic countries.

Table 8.3.Main Results
VariablesPromised Discretionary Measures 2008–12Promised Stimulus 2008–09Realized Change Primary Deficit 2009Realized Change Primary Deficit 2009: DemocraciesRealized Change Deficit 2009Realized Change Deficit 2009: Democracies
Political Constraint–0.971–2.629**–2.712***–2.413**–1.613**–1.730*
Government Expenditures in 2007 (percent of GDP)–0.0483–0.002250.05670.0848**0.05270.0864**
Change of Exports in Winter 2008/09 (%2007–GDP)–0.1060.188–0.212–0.215–0.154–0.135
Government Debt in 2007 (percent of GDP)–0.003560.000598–0.00845–0.0242*–0.00451–0.0182
Government Deficit in 2007 (percent of GDP)–0.165–0.284***–0.428***–0.206*–0.419***–0.229*
KOF Economic Globalization in 2007–0.08030.0684–0.0150–0.0285–0.00534–0.0247
Under an IMF Program–6.272**–1.147–1.216–1.816*–0.733–0.888
Adjusted R20.3890.2000.3430.2530.2870.168
Mean Dependent Variable2.4444.9734.1113.5943.8963.535
Standard Deviation Dependent Variable3.4323.2924.3183.2893.9983.090
Notes: t–statistics in parentheses. Huber–White robust standard errors are used.***p < 0.01, **p < 0.05, *p < 0.1.
Notes: t–statistics in parentheses. Huber–White robust standard errors are used.***p < 0.01, **p < 0.05, *p < 0.1.

The political constraints variable has a strong impact on the size of each of these fiscal stimulus measures, although it is only marginally significant when using our first measure of promises.17 Depending upon the dependent variable the results suggest that, on average, political constraints decrease the size of the fiscal stimulus by between 1 and 2.7 percentage points of GDP. The next-to-last row of the table reports the average size of the stimulus packages within each sample. The average stimulus packages range from 2.4 to 5.0 percent of GDP. Relative to that, the average impact of such political constraints amounts to between 25 and 80 percent of this average size. Figure 8.1 visualises these results. It compares the average sizes of our different stimulus measures for governments that do face political constraints and for those that do not. Whereas unconstrained governments did initiate stimulus packages of on average around 5 percent of GDP, this is roughly reduced to 3 percent for those that were politically constrained. Compared to the remaining variables in the model, the political constraint variable is by far the most robust and has a high explanatory power. When removing the political constraint variable the adjusted R-squared drops by 0.09 points.18

Figure 8.1.Differences in Average Size of Stimulus Packages across Politically Constrained and Unconstrained Countries

(Percent of GDP)

Note: This figure is based upon the results presented in Table 8.3. The numbers at the bottom of the graph refer to the columns in Table 8.3.

Of the other variables, only the initial government deficit turns out to be significant with the expected sign as often as our political constraints dummy; countries with high deficits enacted smaller stimulus packages, on average. The initial debt level has the expected negative sign, but is not statistically significant in most specifications. Nevertheless, fiscal space indeed appears to have been an important factor when explaining the size of the fiscal stimulus measures.

Perhaps surprisingly, “need” does not appear to have been that important. The effect of the change in exports during the winter half-year 2008/2009 mostly has the expected negative sign—a stronger drop has led to larger stimulus measures—but it is not significant.19

The initial size of the government sector, as measured by government expenditures as a share of GDP, turns significantly positive when focusing on realised changes in primary deficits. In line with the argument that government size largely reflects the importance of automatic stabilizers, and that larger automatic stabilizers reduce the need for discretionary stimulus in a crisis, the measure has a (insignificant) negative effect on the size of discretionary stimulus packages in column (1).20

Our measure of the effectiveness of the stimulus packages, the degree of globalization of a country, mostly has the expected negative sign, albeit never significant. The correlation with our measure of government size might be causing a multicollinearity problem. However, also without the government expenditures as a share of GDP included, the KOF Economic Globalization Index is never significant (and the political constraints coefficient is hardly affected by this).21

Being under an IMF program reduces at least the promises made by the government. Regarding actual realization it is less often significant. Nevertheless, these results indicate that this kind of international pressure does have an effect on the fiscal policy stance.

These results could be sensitive to both alternative specifications of factors we do include into our model and to the inclusion of different variables. Furthermore, the underlying sample of countries might have consequences. In Annex 8.1 we discuss a number of alternatives and present a large battery of robustness checks. In a nutshell: changing the set of explanatory variables, the sample of countries, or removing (potentially) extreme observations does not alter our conclusion. We always find very similar results to those presented in Table 8.3.

These robustness checks do provide some indication regarding the channel through which the effect is likely to emerge. When including a dummy for the occurrence of executive elections before June 2009 and interacting that with our political constraint variable, we find that in an environment without political constraints the realised primary deficit (and thus the fiscal stimulus) turns out to be about 2.4 percentage points larger than without upcoming elections. Conversely, in a country where the government faces political constraints, the occurrence of an executive election leads to a reduction of the fiscal stimulus by about 1 percentage point of GDP. This suggests that especially, but not only, during election times, political constraints tie the hands of the incumbent government. Hence, political budget cycles are more likely to occur in countries in which the executive party has control over the legislative branches. Note, though, that these interaction results should not be overemphasized (and are therefore not included in Table 8.3), as they rest upon only a handful of observations.22

Overall the conclusion of all our robustness tests is that our results are highly robust to changing the dependent variable, the use of alternative sets of explanatory variables and changing the sample of countries.

Concluding Remarks

In this paper, we use a simple framework to assess the impact of political constraints on the size of fiscal stimulus packages. We find that on average, political constraints reduce the size of fiscal stimulus packages by about 1 to 2.7 percentage points of GDP—an effect that is large, statistically significant and robust to alternative specifications. The results are thus in line with the widespread perception that political realities limit the de facto usefulness of discretionary fiscal policies as a tool to ameliorate negative economic shocks. To our knowledge it is, however, the first paper that quantifies that effect. Whether this implies that fiscal packages have been too small under a politically constrained government, or too large under a politically unconstrained one cannot be answered by the data at hand. For this a thorough analysis of the effectiveness of different fiscal stimulus programmes is needed. Whereas the United States, as an example of a country having an unconstrained government in our set-up, appears to have successfully implemented large fiscal stimulus measures during 2009, a politically constrained country like Switzerland has also fared well while implementing hardly any fiscal stimulus. Already this anecdotal evidence makes clear that analysing how to most successfully bring an economy back on its feet is not going to be an easy task.

The result that political constraints matter is important because in trying to make sense of policy decisions, we naturally focus on what we deem important. The accuracy of growth forecasts and, even more so, the role of fiscal space are omnipresent in policy discussions since the outset of the crisis. What our findings suggest is that discussing how legislative procedures can be designed to allow for optimal reactions to an economic crisis would be important as well.

Annex 8.1. Alternative Specifications and Robustness Checks

Our main results could be sensitive to alternative specifications of factors we do include into our model, the inclusion of different explanatory variables and to changes in the underlying sample. In this annex, we briefly discuss a number of alternatives and present robustness exercises along these lines. In doing so, we concentrate on the dependent variable measuring the realised change in primary deficits for the year 2009. The results using other dependent variables are very much in line with those shown below. Overall, the important message from the robustness exercises is that the results confirm our hypothesis and show that the effect of political constraints is large, statistically significant and robust along different dimensions.

Using Forecasts to Capture the Size of the Economic Shock

Auerbach and Gorodnichenko (2012) find that the size of fiscal multipliers varies considerably over the business cycle: 0 to 0.5 in expansions, and 1 to 1.5 during recessions. This suggests that also the size of the demand shortfall could matter for the effectiveness of fiscal stimulus. An alternative to the change in export variable, we can capture the size of the economic shock based on changes in growth forecasts for the year 2009. To do so, we compare IMF projections in April 2008 (IMF 2008a) with those in October 2008 (IMF 2008b), i.e., after the collapse of Lehman Brothers, for the year 2009. This measure should capture the economic shock as perceived in the early days after the collapse of Lehman Brothers, but only little, if anything, of the stimulus measures that were enacted in reaction to it.23

By considering both the export and forecast measure we also, in an admittedly crude way, correct for two different types of shocks; the change in exports clearly reflects a trade shock, while the change in the growth forecast captures other types of shocks as well. To also capture a balance-of-payments crisis we take into account both the percentage change of the exchange rate vis-à-vis the US dollar and the growth in official reserves between the second and fourth quarters of 2008.24

Monetary Policy

Besides fiscal policy, monetary policy is another way in which the public sector can try to stimulate its economy. Hence, in those countries where—given the severity of the crisis, fiscal space and effectiveness of fiscal policy—monetary policy has reacted more strongly, the pressure on fiscal policy to act might be lower. Using both the change in policy rates, approximated by the change in the lending rate, from the beginning of the third quarter of 2008 to the start of 2009 and the growth rate of M1 during the same period, we try to capture this dimension of the overall policy reaction to the crisis.

Government Lending Rate

Although monetary policy turned expansionary around the globe and thereby also reduced refinancing costs of governments, substantial differences in interest rates still existed during the winter of 2008/2009. To reflect such cross-country differences, we include the average lending rate during the winter half-year of 2008/2009 as published by the IMF in its International Financial Statistics.25

Central Bank Independence

From a political-institutional point of view, the probability that the money printing press might ultimately be used to deal with high public debt levels could alleviate worries of the current government regarding the unsustainability of future higher debt levels and reduce fiscal constraints. Thus, countries in which the central bank is politically less independent from the government might be willing to increase deficits substantially more than other countries. To take this into account, we use two different indicators for central bank independence, both of which are available for a relatively large number of countries. The first one measures legal independence and goes back to the work of Cukierman (1992) and Cukierman et al. (1992). It is based on how a central bank works internally (how is the central bank governor appointed and is an explicit policy target defined) and how its relationship with the government is arranged (how are disputes settled and are there rules limiting the amount of lending to the government). Crowe and Meade (2008) have updated this de jure indicator of central bank independence to reflect the year 2003. Especially for emerging and developing countries such a legal measure might, however, deviate substantially from actual practice. For that reason, we follow the literature and also construct a de facto measure of central bank independence based on the frequency of irregular central bank governor turnovers.26,27

International Policy Environment

Countries more sensitive to international political pressure or that are strongly integrated in international policy coordination activities might put greater effort into stimulating their own, and thereby also foreign, economies.

After the collapse of Lehman Brothers, the general fear of an overall meltdown generated a substantial amount of political pressure on governments to act in a timely and substantial manner. As indicated by the Leader’s Statement after the London Summit, the G20 very much pushed for strong coordinated actions on the side of its partners (G20 Information Centre 2009). To take this into account, we experiment with both a G20 dummy and a variable measuring the degree to which a country is politically integrated with the rest of the world, which we proxy with the political globalization measure from the KOF Globalization Index.

Whereas international pressure might have induced countries to spend more than they otherwise would have, one could also argue that an international political constraint like the Maastricht Treaty or the Stability and Growth Pact (SGP), which force member countries of the Economic and Monetary Union to focus on certain deficit and debt targets, had exactly the opposite effect. By using EU and EMU dummies, we check whether this international political constraint had an influence on the size of average fiscal stimulus measures in the euro area.

Hence, whereas international policy coordination (via the G20) might have reduced the free-rider problem during the Great Recession, the existence of other international arrangements like the SGP could have had the opposite effect. The involvement of the IMF in domestic (fiscal) policy also belongs to this latter category. In case a country was already under a program of the IMF at the start of the Great Recession, this is likely to have limited its fiscal space.

Annex Table 8.1.1 presents the results in the case in which each of these alternative variables is added to the specification listed in Table 8.3. None of these variables turns out to be significant and most importantly, the results regarding our main explanatory variable, political constraints, does not qualitatively change.

Annex Table 8.1.1.Robustness Tests for Extended Versions of the Baseline Model Using the Realized Change in Primary Deficits as Dependent Variable
VariablesChange Growth Forecast for 2009Percent Change Exchange RateGrowth Official ReservesLending Rate in Winter 2008/09CBI, Legal Measureirr.CB Government Turnover RateChange in Lending RateGrowth Rate of M1KOF Political GlobalizationG20 DummyEU DummyEuro area Dummy
Political Constraint–2.844***–2.555***–2.589***–2.119**–2.915***–2.178***–2.059**–2.367***–2.696***–2.711***–2.621***–2.767***
Government Expenditures in 2007 (percent of GDP)0.05340.06360.05560.06260.01950.06710.06480.02260.06620.05640.06640.0447
Change of Exports in Winter 2008/09 (%2007–GDP)–0.189–0.0704–0.0688–0.315*–0.0423–0.0868–0.358*0.100–0.199–0.213–0.219–0.217
Government Debt in 2007 (percent of GDP)–0.00872–0.0177*–0.0188*–0.0120–0.00728–0.0150–0.01030.00679–0.00841–0.00851–0.00898–0.00994
Government Deficit in 2007 (percent of GDP)–0.414***–0.368**–0.356**–0.287*–0.503***–0.391***–0.287*–0.451***–0.421***–0.428**–0.421**–0.427**
KOF Economic Globalization in 20070.002790.007780.00488–0.05660.00176–0.0247–0.05370.0384–0.00462–0.0149–0.00836–0.0216
Under an IMF Program–1.006–1.686**–1.987**–1.412–1.517*–1.362–1.620–0.806–1.430–1.207–1.221–1.148
Additional Variable (see column header)0.518–0.04510.0153–0.07092.012–3.469–0.0990–0.0266–0.03690.0388–0.6201.076
Adjusted R20.3540.3690.3660.2400.4430.3140.2280.4770.3560.3350.3370.341
Mean Dependent Variable4.1113.8353.7664.1133.9453.8204.1133.8234.1114.1114.1114.111
Standard Deviation Dependent Variable4.3183.8883.9034.2933.9353.4824.2933.4504.3184.3184.3184.318
Note: t–statistics in parentheses. Huber–White robust standard errors are used. Dependent variable: Change in primary deficit in 2009 (% 2007–GDP). CBI = ???***p < 0.01, **p < 0.05, *p < 0.1.
Note: t–statistics in parentheses. Huber–White robust standard errors are used. Dependent variable: Change in primary deficit in 2009 (% 2007–GDP). CBI = ???***p < 0.01, **p < 0.05, *p < 0.1.

In a next step we add political-institutional variables. Including these variables only makes sense when we look at democracies. Hence, we will now restrict our attention to that particular subset.

Political System

Our main explanatory variable, the degree of political constraints, will generally be determined by institutional choices and a complex political game, both of which seem unlikely to be systematically related to the size of fiscal stimulus packages. There is a considerable body of literature in political science that shows that the two most important factors influencing the probability of one party controlling both executive and legislative bodies are the decision between presidential and parliamentary system and the choice of the voting system.28

In a presidential system, such as the United States, where there are separate elections for both executive and legislative bodies, the probability of one party controlling both bodies is smaller than in a parliamentary system, such as the United Kingdom, where winning a majority in the House of Commons allows a party to appoint the prime minister.

At the same time, a plurality voting system, as is being used in the United Kingdom, makes it more likely for a single party to win a majority than in the case of a proportional system, such as in Germany. Within any given system, whether one party rules both bodies further depends on a host of factors such as election dates and the political climate, all of which are unlikely to be systematically correlated with any factor determining the size of fiscal stimulus packages.29 By including dummies for plurality and parliamentary systems we control for what might be more underlying causes of differences in fiscal policy.

Political Orientation of Government

There is a substantial literature on whether a government’s political orientation has an effect on its fiscal policy.30 Partisan theory suggests that left-wing governments implement more expansionary policies and intervene more heavily in the economy in general (Dreher and Sturm 2012). We therefore control for partisan composition of the government by including a dummy that equals one in the case in which the executive is considered to be from a left-wing party.

Minority Governments

Edin and Ohlsson (1991) argue that minority governments have more difficulties than majority (coalition) governments to reduce deficits and debt levels. In a similar vein, Falcó-Gimeno and Jurado (2011) argue that minority governments have to negotiate with the opposition over the budget. Furthermore, Brück and Stephan (2006) find that minority governments tend to make overly optimistic budget forecasts. We include a minority government dummy and a variable measuring the fraction of seats held by the government to capture such potential effects.

Coalition Governments and Fragmentation

Game theory suggests that cooperation is more difficult when the number of players is large. In this view, coalition governments will find it more difficult to close budget deficits after adverse shocks, since parties in the coalition will veto spending cuts or tax increases that impinge on the interests of their respective constituencies. Roubini and Sachs (1989a; 1989b) find that broad coalition governments experience higher deficits than one-party governments. Subsequent research by Edin and Ohlsson (1991) and De Haan and Sturm (1994; 1997) found less support for this hypothesis. We nonetheless include a coalition dummy control for this in our setting. Perotti and Kontopoulos (2002) subsequently broadened this approach somewhat by arguing that this overlooks what they call size fragmentation. One possible source of fragmentation of fiscal policy making is the number of decision makers. The larger the number of decision makers, the less each will internalize the costs that a certain policy will impose on others. It can be argued that the relevant group here is each political party in government. Indeed, Perotti and Kontopoulos (2002) find evidence that the higher the number of parties in government, the looser fiscal policy is. Although De Haan et al. (1999) do not find that coalition governments generally have more difficulty in keeping their budgets in line after an adverse economic shock, they also report that more fractionalized governments experience larger government debt growth. To capture possible effects of government fragmentation, we include a variable measuring the probability that two members of government do not belong to the same party. In a similar vein, we also take into account how fractionalized the opposition is by taking on board the probability that two members of the opposition are not of the same party.31

Political Budget Cycles

The final political-institutional variable that we include reflects the findings of the political budget cycles (PBC) literature and is closely linked to our motivation for why political constraints are relevant in democracies. PBC research examines the existence and determinants of election cycles in public spending, taxes and government budget deficits. Older theoretical PBC models emphasize the incumbent’s intention to secure re-election by maximizing his expected vote share at the next election (Nordhaus, 1975). It is assumed that the electorate is backward looking and the government is evaluated on the basis of its past track record. As a result, these models imply that governments, regardless of ideological orientation, adopt expansionary fiscal policies before elections in order to stimulate the economy.32 More recent PBC models emphasize the role of temporary information asymmetries regarding the politicians’ level of competence in explaining electoral cycles in fiscal policy. In these models, signalling is the driving force behind the PBC (see, e.g., Rogoff and Sibert, 1988; Tabellini and Persson, 2003; and Shi and Svensson, 2006). Pina and Venes (2011) and Jong-a-Pin et al. (2012) show that in OECD countries, there is evidence of electoral effects in revisions of official revenue and spending statistics. To capture possible effects from political business cycles, we include dummies for both executive and legislative elections that took place in the period between October 2008 and June 2009.

Annex Table 8.1.2.Robustness Tests with Additional Political Variables, One at a Time, while Using the Realised Change in Primary Deficits as Dependent Variable
VariablesPlurality SystemParliamentary SystemLeft-Wing ExecutiveMinority GovernmentFraction Seats Held by GovernmentCoalition GovernmentP(Government Member Not Same Party)P(Opposition Member Not Same Party)Executive ElectionExecutive ElectionLegislative ElectionLegislative Election
Political Constraint–2.410**–2.465**–2.414**–2.495**–2.661**–2.430*–2.531**–2.571**–2.413**–2.221*–2.414**–2.179*
Government Expenditures in 2007 (percent of GDP)0.0850**0.0991**0.0845*0.0886**0.05640.0842*0.06440.04690.0848*0.0791*0.0752**0.0778**
Change of Exports in Winter 2008/09 (%2007-GDP)–0.216–0.235–0.214–0.227–0.211–0.214–0.230–0.200–0.215–0.229–0.247–0.253
Government Debt in 2007 (percent of GDP)–0.0241*–0.0230*–0.0242*–0.0245*–0.0188–0.0242*–0.0198–0.0153–0.0242*–0.0227–0.0277**–0.0295**
Government Deficit in 2007 (percent of GDP)–0.206*–0.209*–0.206*–0.211*–0.215*–0.205*–0.216*–0.216*–0.206*–0.200*–0.186*–0.192*
KOF Economic Globalization in 2007–0.0284–0.0231–0.0284–0.0299–0.00412–0.0282–0.01190.00370–0.0285–0.0276–0.0283–0.0314
Under an IMF Program–1.813*–1.957**–1.814*–1.938**–1.470–1.810*–1.579–1.382–1.816*–1.836*–1.739*–1.734*
Additional Political Variable (see column header)0.0248–0.935–0.03120.6911.1230.0390–0.4280.1580.003912.430***1.330*2.127
Additional Political Variable × Political Constraint Dummy–3.352***–1.128
Adjusted R20.2410.2570.2410.2460.2630.2410.2620.2460.2410.2400.2690.261
Mean Dependent Variable3.5943.5943.5943.5943.5203.5943.5203.4373.5943.5943.5943.594
Standard Deviation Dependent Variable3.2893.2893.2893.2893.2543.2893.2543.2023.2893.2893.2893.289
See first part of Table 8.2; authors’ calculations.Note: t-statistics in parentheses. Huber-White robust standard errors are used. Dependent variable: Change in primary deficit in 2009 (% 2007-GDP).***p < 0.01, **p < 0.05, *p < 0.1.
See first part of Table 8.2; authors’ calculations.Note: t-statistics in parentheses. Huber-White robust standard errors are used. Dependent variable: Change in primary deficit in 2009 (% 2007-GDP).***p < 0.01, **p < 0.05, *p < 0.1.

Adding these additional political-institutional variables to our regressions is not affecting our conclusion. The political constraint variable remains significant and its coefficient of a similar order of magnitude. With only two exceptions, none of the additional political-institutional variables turns out to be significant in explaining our stimulus measure. These exceptions relate to elections and further strengthen our results. In case of executive elections in an environment without political constraints, the actual change in the primary deficit is significantly larger than when the same elections take place in a politically constrained environment. In the latter case, the overall reduction is more than 3 percentage points of GDP larger than without any elections (but still facing political constraints). The occurrence of legislative elections also seems to stimulate running a larger deficit. Albeit statistically not significant, this again appears to be largely due to elections in countries without political constraints.33

Changes in (Primary) Deficits over a Longer Time Horizon

Our empirical analysis is set up such that all control variables are measured before the start of the crisis and the crisis is treated as an exogenous shock. By at the same time focusing on the year 2009, which circumvents interference of other events (like the euro crisis and the Fukushima catastrophe), non-conventional policy reactions (like monetary Quantitative Easing programmes in certain countries), or the euro area debt crisis, we try to get as close as possible to a causal interpretation of our results. Nevertheless, one could argue that political constraints cause policymakers to react slower, but not necessarily less. By 2010, many countries, however, already had moved into recovery mode and first stimulus measures were put to a halt. Hence, it could well be argued that policy reactions by that time would already have been “behind the curve.” To have some first suggestive evidence on whether policy action was merely postponed, we have extended our dependent variables measuring realised changes in (primary) deficits to not only capture the change during 2009, but also that including the year 2010. Annex Table 8.1.3 summarizes the results. Whereas the sign of our political constraint variable remains negative, its impact declines and is no longer statistically significant.

Annex Table 8.1.3.Changing the Time Horizon of the Realized Changes in (Primary) Deficits, While Using the Realized Change in Primary Deficits as Dependent Variable
VariablesRealized Change Primary Deficit 2009Realized Change Primary Deficit 2009–10Realized Change Primary Deficit 2009: DemocraciesRealized Change Primary Deficit 2009–10: DemocraciesRealized Change Deficit 2009Realized Change Deficit 2009–10Realized Change Deficit 2009: DemocraciesRealized Change Deficit 2009–10: Democracies
Political Constraint–2.712***–1.325–2.413**–1.206–1.613**–0.747–1.730*–0.848
Government Expenditures in 2007 (percent of GDP)0.05670.04200.0848**0.06120.05270.03860.0864**0.0590
Change of Exports in Winter 2008/09 (%2007-GDP)–0.212–0.0782–0.215–0.187–0.1540.0749–0.135–0.150
Government Debt in 2007 (percent of GDP)–0.00845–0.0108–0.0242*–0.0161–0.00451–0.00616–0.0182–0.0152
Government Deficit in 2007 (percent of GDP)–0.428***–0.316***–0.206*–0.190***–0.419***–0.335***–0.229*–0.227***
KOF Economic Globalization in 2007–0.0150–0.0324–0.0285–0.0269–0.00534–0.00615–0.0247–0.0207
Under an IMF Program–1.216–2.013**–1.816*–2.703***–0.733–1.357–0.888–1.892**
Adjusted R20.3430.1330.2530.1030.2870.0830.1680.091
Mean Dependent Variable4.1113.5903.5943.3753.8963.2973.5353.326
Standard Deviation Dependent Variable4.3184.1853.2893.8053.9984.2673.0903.738
Note: t-statistics in parentheses. Huber-White robust standard errors are used.***p < 0.01, **p < 0.05, *p < 0.1.
Note: t-statistics in parentheses. Huber-White robust standard errors are used.***p < 0.01, **p < 0.05, *p < 0.1.

Changing the Sample of Countries

Given that we use what is arguably an unexpected exogenous shock hitting all countries around the world, we are working in a cross-section framework. This limits our degrees of freedom and we therefore concentrate most of our analysis on those samples that are as large as possible. Nevertheless, differences across different country groups might exist. To check the robustness of our results in this sense, Annex Table 8.1.4 varies the underlying sample in different ways.

Annex Table 8.1.4.Robustness Tests with Changing the Underlying Set of Countries, While Using the Realized Change in Primary Deficits as Dependent Variable
VariablesFull SampleDemocraciesMultiparty SystemsOECD CountriesNon-OECD CountriesG20 CountriesNon-G20 CountriesEU Member StatesNon-EU Member StatesEuro Area MembersNon-Euro Area Members
Political Constraint–2.712***–2.413**–2.254***–3.364***–2.159**–2.929**–2.708***–1.965*–2.308**–3.337**–2.665***
Government Expenditures in 2007 (percent of GDP)0.05670.0848**0.0614–0.002130.07290.07150.0555–0.02590.0704–0.1050.0480
Change of Exports in Winter 2008/09 (%2007-GDP)–0.212–0.215–0.2360.00201–0.2330.134–0.2200.0240–0.291–0.0381–0.250
Government Debt in 2007 (percent of GDP)–0.00845–0.0242*–0.005590.0100–0.01240.00562–0.01190.0265–0.01680.0257–0.0121
Government Deficit in 2007 (percent of GDP)–0.428***–0.206*–0.338**–0.296***–0.489*–0.554*–0.426**–0.183–0.408**–0.454*–0.427**
KOF Economic Globalization in 2007–0.0150–0.0285–0.006250.0687–0.0244–0.0104–0.01530.0523–0.007260.0381–0.0266
Under an IMF Program–1.216–1.816*–0.780–0.515–1.443–1.192–4.076***–0.812–1.192
Adjusted R20.3430.2530.2590.3360.3390.4690.3210.3750.358–0.1980.346
Mean Dependent Variable4.1113.5943.7874.3703.9964.1944.0963.9694.1664.9243.969
Standard Deviation Dependent Variable4.3183.2893.8901.8185.0602.3684.6072.1914.9071.4374.634
Note: t-statistics in parentheses. Huber-White robust standard errors are used. Dependent variable: Change in primary deficit in 2009 (% 2007-GDP). Only democratic countries are included in the sample. OECD = Organisation for Economic Co-operation and Development.***p < 0.01, **p < 0.05, *p < 0.1.
Note: t-statistics in parentheses. Huber-White robust standard errors are used. Dependent variable: Change in primary deficit in 2009 (% 2007-GDP). Only democratic countries are included in the sample. OECD = Organisation for Economic Co-operation and Development.***p < 0.01, **p < 0.05, *p < 0.1.

The first two columns repeat the main results from Annex Table 8.1.3. Whereas in the main text we rely on the definition of Cheibub et al. (2010) to distinguish between democracies and autocracies, column (3) uses a different split. Parts of our reasoning assume competitive elections. As an alternative, we therefore include in column (3) only those countries in which multiple parties did win seats.34 Although this increases the number of observations slightly, it does not affect our conclusions. The subsequent columns (4) to (11) distinguish between OECD, non-OECD, G20, non-G20, EU, non-EU and euro area and non-euro area members. In all of these subsamples our political constraint variable remains significantly negative.

Checking for Outlying Observations

Perhaps some extreme and thereby potentially outlying observations might drive our results. Particularly noteworthy are the three negative stimulus packages of Hungary, Ireland and Iceland in the UNCTAD (2009) data. As reported by UNCTAD (2009), these countries did all commit large financial resources to rescue their financial sectors while, at the same time, imposing restrictive fiscal policies such as tax increases and cuts in public expenditures of more than 7 percent of GDP. The extraordinary conditions in these countries might have an influence on our results. Column (2) of Annex Table 8.1.5, however, shows that dropping these observations has no real impact on our regression results.

Countries with (in an absolute sense) large values of either the dependent variable or any of the control variables might also have a substantive influence on our regression results. For that reason, we exclude in each of the remaining columns in Annex Table 8.1.5 the upper and lower 10 percent of the distribution regarding either the dependent variables, or in turn each of the control variables. Each time discarding around 20 percent of our (potentially influential) observations does not change our results in any qualitatively meaningful way. The political constraint variable remains highly significant and negative.

Annex Table 8.1.5.Robustness Tests by Removing Potentially Influential Observations, While Using the Realized Change in Primary Deficits as Dependent Variable
VariablesFull SampleExcluding Hungary, Iceland, and IrelandExcluding Tails of Change in Primary DeficitExcluding Tails of Government Expenditure SharesExcluding Tails of Change in GrowthExcluding Tails of Change in Export SharesExcluding Tails of Government Debt SharesExcluding Tails of Government Deficit Shares
Political Constraint–2712***–2.731***–1.262**–2.026**–3.396***–3.075***–1.971**–2.158**
Government Expenditures in 2007 (percent of GDP)0.05670.05720.0510*0.134***0.04210.06830.05550.0346
Change of Exports in Winter 2008/09 (%2007-GDP)–0.212–0.231–0.0258–0.487***–0.134–0.357**–0.124–0.184
Government Debt in 2007 (percent of GDP)–0.00845–0.008000.000485–0.00553–0.0176–0.0187–0.0459**–0.00404
Government Deficit in 2007 (percent of GDP)–0.428***–0.418**–0.262***–0.551***–0.340**–0.359**–0.228**–0.442**
KOF Economic Globalization in 2007–0.0150–0.01620.00346–0.0709**0.0112–0.02320.00726–0.0176
Under an IMF Program–1.216–1.178–0.7130.0309–1.394–1.287–2.257**–1.379
Adjusted R20.3430.3390.2340.4550.4310.4090.2870.151
Mean Dependent Variable4.1114.1273.7303.9874.1984.1523.8943.730
Standard Deviation Dependent Variable4.3184.3652.2314.2423.9794.1913.5083.201
Note: t-statistics in parentheses. Huber-White robust standard errors are used. In columns (3) to (8), the upper and lower 10 percent of the observations regarding the respective variable are removed from the sample. Dependent variable: Change in primary deficit in 2009 (%2007-GDP).***p < 0.01, **p < 0.05, *p < 0.1.
Note: t-statistics in parentheses. Huber-White robust standard errors are used. In columns (3) to (8), the upper and lower 10 percent of the observations regarding the respective variable are removed from the sample. Dependent variable: Change in primary deficit in 2009 (%2007-GDP).***p < 0.01, **p < 0.05, *p < 0.1.
Annex Table 8.1.6.List of Countries and Values for the Main Dependent Variables
CountriesPromised Discretionary Measures 2008–12Promised Stimulus 2008–09Realized Change Primary Deficit 2009Realized Change Deficit 2009Political ConstraintDemocracyCountriesPromised Discretionary Measures 2008–12Promised Stimulus 2008–09Realized Change Primary Deficit 2009Realized Change Deficit 2009Political ConstraintDemoaacy
Bulgaria2.93.783.3811New Zealand4.
Burkina Faso1.371.2600Nicaragua0.850.6511
Cabo Verde6.196.1511Panama1.681.8711
Chad13.3613.0300Papua New Guinea12.4312.1311
Costa Rica3.573.7111Poland1.
Cyprus6.957.2411Republic of Congo19.2920.5910
Czech Republic3.03.443.1811Romania2.241.8611
Cote d’lvoire1.051.2410Russia5.46.610.8110.6700
Dominican Republic0.700.3311Senegal0.280.1701
Ecuador4.1711Sierra Leone–0.88–0.6511
El Salvador2.6711Slovak Republic1.15.505.2911
Equatorial Guinea25.2900Slovenia4.994.9011
Eritrea–1.3200Solomon Islands–1.9711
Estonia–0.55–0.5211South Africa7.43.05.485.310
FYR Macedonia1.7211Sri Lanka3.781.5811
Finland3.15.76.866.5111St. Lucia2.222.3201
Guinea-Bissau–4.08–3.4311The Bahamas1.931.8001
Guyana0.120.1100The Gambia1.581.4500
Hungary–7.7–0.70.680.2611Trinidad and Tobago15.4015.0701
Israel3.023.0311United Arab Emirates21.7221.5600
Italy0.33.42.513.1811United Kingdom1.99.55.986.1201
Jamaica4.42–1.7101United States5.511.56.356.6101
Kyrgyz Republic2.1411Vietnam7.497.2300
Lao P.D.R.4.674.8700Yemen5.155.1800

    AidtToke S.FranciscoJose Veiga and LindaGoncalves Veiga. (2011). Election Results and Opportunistic Policies: A New Test of the Rational Political Business Cycle Model Public Choice. 148(1–2): 2144.

    AizenmanJoshua and YothinJinjarak. (2011). The Fiscal Stimulus of 2009-10: Trade Openness Fiscal Space and Exchange Rate Adjustment National Bureau of Economic Research Working Papers 17427.

    AlesinaAlbertoRobertoPerotti and JoséTavares. (1998). The Political Economy of Fiscal AdjustmentsBrookings Papers on Economic Activity.29(1): 197266.

    AltJames E. and DavidDreyer Lassen. (2006). Fiscal Transparency, Political Parties, and Debt in OECD CountriesEuropean Economic Review.50: 14031439.

    AlterJonathan. (2011). The Promise: President Obama Year One.London: Simon & Schuster.

    AndrikopoulosAndreasIoannisLoizides and KyprianosProdromidis. (2004). Fiscal Policy and Political Business Cycles in the EU European Journal of Political Economy.20(1): 125152.

    AngelopoulosKonstantinosGeorgeEconomides and PantelisKammas. (2012). Does Cabinet Ideology Matter for the Structure of Tax Policies?European Journal of Political Economy.28(4): 620635.

    ArmingeonKlaus. (2012). The Politics of Fiscal Responses to the Crisis of 2008–2009Governance.25(4): 543565.

    AuerbachAlan J. and YuriyGorodnichenko. (2012). Measuring the Output Responses to Fiscal PolicyAmerican Economic Journal: Economic Policy.4(2): 127.

    BartelsLarry. (2011). Ideology and Retrospection in Electoral Responses to the Great Recession. URL:

    BeckThorstenGeorgeClarkeAlbertoGroffPhilipKeefer and PatrickWalsh. (2001). New Tools in Comparative Political Economy: The Database of Political InstitutionsWorld Bank Economic Review.15(1): 165176.

    BenitoBernardinoFranciscoBastida and CristinaVicente. (2013). Creating Room for Manoeuvre: a Strategy to Generate Political Budget Cycles under Fiscal RulesKyklos. 66(4): 467496.

    BlanchardOlivierGiovanniDell’Ariccia and PaoloMauro. (2010). Rethinking Macroeconomic Policy. Journal of Money Credit and Banking.42(1): 199215.

    BlinderAlan S. (1997). Is Government Too Political?Foreign Affairs.76(6): 115126.

    BrückTilman and AndreasStephan. (2006). Do Eurozone Countries Cheat with their Budget Deficit Forecasts?, Kyklos. 59(1): 315.

    CecchettiStephen G. (2002). The Problem with Fiscal Policy. URL:

    CheibubJosé AntonioJenniferGandhi and JamesRaymond Vreeland. (2010). Democracy and Dictatorship RevisitedPublic Choice. 143(1/2): 67101.

    Chodorow-ReichGabrielLauraFeivesonZacharyLiscow and WilliamGui Woolston. (2012). Does State Fiscal Relief during Recessions Increase Employment? Evidence from the American Recovery and Reinvestment ActAmerican Economic Journal: Economic Policy.4(3): 11845.

    CoganJohn F. and John B.Taylor. (2011). What the Government Purchases Multiplier Actually Multiplied in the 2009 Stimulus PackageNBER Working Paper 16505.

    ConleyTimothy G. and BillDupor. (2013). The American Recovery and Reinvestment Act: Solely a government jobs program?Journal of Monetary Economics.60(5): 535549.

    CroweChristopher and Ellen E.Meade. (2008). Central Bank Independence and Transparency: Evolution and EffectivenessEuropean Journal of Political Economy. 24(4): 763777.

    CukiermanAlex. (1992). Central Bank Strategy Credibility and Independence.Cambridge, MA: MIT Press.

    CukiermanAlexSteven B.Webb and BilinNeyapti. (1992). Measuring the Independence of Central Banks and its Effects on Policy OutcomesWorld Bank Economic Review.6: 353398.

    CusackThomas R. (1997). Partisan Politics and Public Finance: Changes in Public Spending in the Industrialized Democracies 1955–1989Public Choice. 91(3/4): 375395.

    CusackThomas R. (1999). Partisan Politics and Fiscal PolicyComparative Political Studies.32(4): 464486.

    De HaanJakob and Jan-EgbertSturm. (1994). Political and Institutional Determinants of Fiscal Policy in the European CommunityPublic Choice. 80: 157172.

    De HaanJakob and Jan-EgbertSturm. (1997). Political and Economic Determinants of OECD Budget Deficits and Government Expenditures: A ReinvestigationEuropean Journal of Political Economy. 13: 739750.

    De HaanJakobJan-EgbertSturm and GeertBeekhuis. (1999). The Weak Government Thesis: Some New EvidencePublic Choice.101(3/4): 163167.

    DreherAxel and Jan-EgbertSturm. (2012). Do the IMF and the World Bank Influence Voting in the UN General Assembly?Public Choice. 141: 363397.

    DreherAxelJan-EgbertSturm and JakobDe Haan. (2008). Does High Inflation Cause Central Bankers to Lose their Job? Evidence Based on a New Data SetEuropean Journal of Political Economy.24(4): 778787.

    DreherAxelJan-EgbertSturm and JakobDe Haan. (2010). When is a Central Bank Governor Replaced? Evidence Based on a New Data SetJournal of Macroeconomics.32: 766781.

    DuchRaymond M. and Randolph T.Stevenson. (2008). The Economic Vote: How Political and Economic Institutions Condition Election Results.Cambridge: Cambridge University Press.

    EdinPer-Anders and HenryOhlsson. (1991). Political Determinants of Budget Deficits: Coalition Effects Versus Minority EffectsEuropean Economic Review.35: 15971603.

    Falcó-GimenoAlbert and IgnacioJurado. (2011). Minority Governments and Budget Deficits: The Role of the Opposition, European Journal of Political Economy. 27(3): 554565.

    FeyrerJames and BruceSacerdote. (2011). Did the Stimulus Stimulate? Real Time Estimates of the Effects of the American Recovery and Reinvestment ActNBER Working Papers 16759 National Bureau of Economic Research.

    FurceriDavide and Ricardo M.Sousa. (2011). The Impact of Government Spending on the Private Sector: Crowding-out versus Crowding-in Effects Kyklos. 64(4): 516533.

    G20 Information Centre. (2009) London Summit – Leader’s Statement. URL:

    GaliJordi. (1994). Government Size and Macroeconomic StabilityEuropean Economic Review.38(1): 117132.

    HeniszWitold J. (2000). The Institutional Environment for Economic GrowthEconomics & Politics.12(1): 131.

    HeniszWitold J. (2002). The Institutional Environment for Infrastructure InvestmentIndustrial and Corporate Change. 11(2): 355389.

    HerwartzHelmut and BerndTheilen. (2014). Partisan Influence on Social Spending Under Market Integration, Fiscal Pressure and Institutional ChangeEuropean Journal of Political Economy. 34: 409424.

    HibbsDouglas. (2006). Voting and the Macroeconomy in: Barry R.Weingast and DonaldWittman (eds.) The Oxford Handbook of Political Economy. Oxford: Oxford University Press: 565586.

    HortonMarkManmohanKumar and PaoloMauro. (2009). The State of Public Finances: A Cross-Country Fiscal Monitor IMF Staff Position Note SPN/09/21.

    IlzetzkiEthanEnrique G.Mendoza and Carlos A.Végh. (2013). How Big (Small?) are Fiscal Multipliers?Journal of Monetary Economics.60(2): 239254.

    International Monetary Fund (IMF). (2007). World Economic Outlook DatabaseOctober 2007 Edition. URL:

    International Monetary Fund (IMF). (2008a). World Economic Outlook Database April 2008 Edition. URL:

    International Monetary Fund (IMF). (2008b). World Economic Outlook Database October 2008 Edition. URL:

    International Monetary Fund (IMF). (2009a). The Size of the Fiscal Expansion: An Analysis for the Largest Countries. URL:

    International Monetary Fund (IMF). (2009b). World Economic Outlook; Crisis and Recovery. URL:

    International Monetary Fund (IMF). (2013). World Economic Outlook Database April 2013 Edition. URL:

    Jong-a-PinRichard M.Jan-EgbertSturm and JakobDe Haan. (2012). Using Real-Time Data to Test for Political Budget CyclesKOF Working Papers No. 313.

    Lewis-BeckMichael. S. (1988). Economics and Elections: The Major Western Democracies.Michigan: University of Michigan Press.

    Lewis-BeckMichael S. and MartinPaldam. (2000). Economic Voting: An IntroductionElectoral Studies. 19(2/3): 113121.

    LijphartArend. (1990). The Political Consequences of Electoral Laws 194585The American Political Science Review.84(2): 481496.

    LijphartArend. (1999). Patterns of Democracy: Government Forms and Performance in Thirty-Six Countries.Yale: Yale University Press.

    NordhausWilliam D. (1975). The Political Business CycleThe Review of Economic Studies.42(2): 169190.

    OECD. (2009). Economic Outlook Interim Report March 2009. URL:

    OlsonMancur. (2000). Power and Prosperity. Outgrowing Communist and Capitalist Dictatorships.New York: Basic Books.

    OrphanidesAthanasios. (2001). Monetary Policy Rules Based on Real-Time DataAmerican Economic Review.91(4): 964985.

    OrphanidesAthanasios and Simonvan Norden. (2002). The Unreliability of Output Gap Estimates in Real TimeReview of Economics and Statistics.84: 569583.

    PerottiRoberto and YianosKontopoulos. (2002). Fragmented Fiscal PolicyJournal of Public Economics.86: 191222.

    PerssonTorsten and Lars E. O.Svensson. (1989). Why a Stubborn Conservative would Run a Deficit: Policy with Time-Inconsistent PreferencesThe Quarterly Journal of Economics.104(2): 325345.

    PinaÁlvaro M. and Nuno M.Venes. (2011). The Political Economy of EDP Fiscal Forecasts: An Empirical Assessment, EuropeanJournal of Political Economy.27(3): 534546.

    PortebaJames M. (1994). State Responses to Fiscal Crises: The Effects of Budgetary Institutions and PoliticsJournal of Political Economy.102(4): 799821.

    RodrikDani. (1998). Why do More Open Economies Have Bigger Governments?Journal of Political Economy.106(5): 9971032.

    RogoffKenneth and Anne Sibert. (1988). Elections and Macroeconomic Policy CyclesThe Review of Economic Studies.55: 116.

    RomerDavid. (2012). What Have We Learned about Fiscal Policy from the Crisis? in: Olivier J.BlanchardDavidRomerMichaelSpence and Joseph E.Stiglitz (eds.) In the Wake of the Crisis: Leading Economists Reassess Economic Policy. Cambridge (Mass.): The MIT Press: 5766.

    RoubiniNouriel and JeffreySachs. (1989a). Government Spending and Budget Deficits in the Industrial CountriesEconomic Policy.8: 99132.

    RoubiniNouriel and JeffreySachs. (1989b). Political and Economic Determinants of Budget Deficits in the Industrial EconomiesEuropean Economic Review. 33: 903938.

    ShiMin and JakobSvensson. (2006). Political Budget Cycles: Do They Differ across Countries and Why?Journal of Public Economics.90(8/9): 13671389.

    SpolaoreEnrico. (2004). Adjustments in Different Government Systems, Economics & Politics. 16(2): 117146.

    StrebJorge M.DanielLema and GustavoTorrens. (2009). Checks and Balances on Political Budget Cycles: Cross-Country Evidence Kyklos. 62(3): 426447.

    SturmJan-Egbert and JakobDe Haan. (2001). Inflation in Developing Countries: Does Central Bank Independence Matter? New Evidence Based on a New Data Set CESifo Working Paper 511.

    TabelliniGuido and TorstenPersson. (2003). Do Electoral Cycles Differ Across Political Systems?IGIER Working Paper No. 232.

    UNCTAD. (2009). Trade And Development Report 2009. URL:

    Wallace-WellsBenjamin. (2001). What’s Left on the Left. Paul Krugman’s Lonely Crusade. New York Magazine. URL:

    WeiseCharles L. (2012). Political Pressures on Monetary Policy During the US Great InflationAmerican Economic Journal: Macroeconomics.4(2): 3364.

This chapter is reprinted from Kyklos, Vol. 69, Fabian Gunzinger and Jan-Egbert Sturm, “It’s Politics, Stupid! Political Constraints Determine Governments’ Reactions to the Great Recession,” ©2016, with permission from Wiley.

Fabian Gunzinger: Department of Economics, University of Oxford, United Kingdom,; Jan-Egbert Sturm: KOF Swiss Economic Institute, ETH Zurich, Leonhardstrasse 21, 8092 Zurich, Switzerland and CESifo, Munich, Germany,, phone: +41 44 632 5001, fax: +41 44 632 1150 (corresponding author). We thank Stefan Boes, Konstantin Büchel, Christian Busch, Malin Hu, Marie Proprawe, Nora Strecker, Lukas Voellmy, James R. Vreeland and participants of the 11th Carroll Round at Georgetown University (Washington, D.C., 21 April 2012), the Beyond Basic Questions Workshop (Lucerne, 13-15 June 2013), the Silvaplana Workshop in Political Economy (Pontresina, 20-24 July 2013), the ECB Public Finance Workshop (Frankfurt, 16 October 2013), the Economics Seminar at the University of Reading (10 February 2014), the 2014 Workshop on Political Economy at the Catholic University of Milan (12-13 May 2014), the Symposium on The Future of Political Economy (Freiburg, 29-31 May 2014), the International Political Economy Society (IPES) meeting (Washington, D.C., 14-15 November 2014), as well as anonymous referees and the editors for helpful suggestions and comments. All remaining errors are ours.

For an overview of stimulus sizes for all countries, see Annex Table 8.1.6.

See, for instance, OECD (2009), IMF (2009a), Aizenman and Jinjarak (2011) and Ilzetzki et al. (2013). Aizenman and Jinjarak (2011) directly test for, and confirm, their importance for the size of stimulus packages, while the findings of Ilzetzki et al. (2013) suggest that more fiscal space and less trade openness makes stimulus more effective. For a more elaborate discussion on the link between fiscal space, or “fiscal leeway,” and fiscal policy, see Blanchard et al. (2010).

For major contributions see Lewis-Beck (1988) and Duch and Stevenson (2008). For summaries of the literature see Lewis-Beck and Paldam (2000) and Hibbs (2006).

We consciously decide against differentiating between expenditure increases and bank bailouts. The reason is that the underlying political calculus for incumbent and opposition parties should be the same: if bailing out banks helps alleviate the economic shock (or prevent an even larger one), incumbents should want to do it, while the opposition should want to prevent or at least delay it. As such, the expenditures on bank bailouts are simply part of the overall fiscal package. However, taking those countries out in which substantial bank bailouts have occurred does not change our results.

For six countries where UNCTAD does not provide data, we use data from OECD (2009). The relevant countries are the Czech Republic, Denmark, Finland, Luxembourg, New Zealand and Slovak Republic.

Note, however, a country’s method for measuring its stimulus package is unlikely to be correlated with the size of that package. The consistency of our results is therefore not compromised.

Conceptually, we prefer a measure that only takes discretionary aspects into account. However, we do have to realize that it is far from obvious to disentangle cyclical and structural movements in fiscal data. Cyclically adjusted data are well-known to be heavily revised—up to the size of the actual measure (see, e.g., Orphanides [2001], Orphanides and Van Norden [2002] and Jong-a-Pin et al. [2012])—making it problematic to link it to real-time decisions. Furthermore, the sample of countries for which such data is available is very limited. As it is likely that the extent of automatic stabilizers in an economy is related to the size of the public sector, we include the latter as an explanatory variable in all of our models.

In defining democracies, we use the classification of Cheibub et al. (2010). Accordingly, the basic conditions for a regime to be coded as democratic are that i) the executive and legislative are elected and ii) multiple parties are allowed for and exist. A two-group mean-comparison test reveals that the averages of democracies and non-democracies are significantly different from each other.

During the year 2009, the only election that potentially led to a change in this variable relative to 2008 was the legislative election in June 2009 in Argentina. We, however, take values as relevant for the winter 2008/2009, which always equal those for 2008.

Henisz (2000; 2002) constructed political constraint variables that indicate whether the executive party is the largest party in the upper and lower house. As being the largest party does not necessarily imply having a majority, the correlations between the ALLHOUSE variable and those from Henisz are merely around 0.3. Given that in our line of argumentation having a majority is indispensable, we stick to using the ALLHOUSE variable.

We have also looked into using oil and gas reserves as published by British Petroleum. However, that would reduce our sample substantially.

We try to avoid issues of reverse causality by using pre-crisis data—data that is not yet influenced by the economic shock following the collapse of Lehman Brothers in September 2008.

This is in line with the findings of Rodrik (1998). He makes the point that more open economies are more likely to have larger government sectors as a form of insurance against the volatility created by openness.

The last row in Table 8.1 reports correlation coefficients between the political constraint dummy and these dependent variables. All of these are negative and mostly statistically significant indicating that also in a parsimonious regression that only includes political constraints and a constant, political constraints reduce the size of the stimulus measures. The coefficient estimates of such bivariate regressions (not shown) are comparable to those presented in Table 8.3.

When removing the IMF dummy from the first column the political constraint variable turns significant with an estimated coefficient of about −1.8. More in general, we have checked for outlying observations and, besides some countries that were under an IMF program, did not encounter such. Removing individual countries from our analysis or reducing the sample size by taking other specific groups of countries out (like oil-producing countries) does not affect our conclusions.

Only the initial deficit share explains more of the variation in the dependent variable; its removal leads to a reduction of 0.19 point of the adjusted R-squared.

Our growth forecast comparison for the year 2009 usually did not lead to an expected significant negative coefficient and is therefore not included in this baseline regression.

Although government expenditures are mathematically used in the construction of the government deficit variable, in our sample these two variables are hardly correlated (see Table 8.2). As therefore to be expected, the conclusions do not change if we include each of them separately.

Removing the KOF Economic Globalization Index does, however, increase the level of significance of the government size measure somewhat. This has no effect on our main variable of interest, political constraints.

This is in line with the findings of Streb et al. (2009) who find that political business cycles are smaller in countries where the government faces effective checks and balances, which they proxy by incomplete control of the legislative body and adherence to the law.

As is common practice in forecasting, the short-term fiscal policy assumptions used by the IMF are largely based on officially announced budgets. Hence, most if not all fiscal stimulus measures are not included in this measure, thereby alleviating the reverse causality problem.

Ideally we would have also liked to take an explicit measure for real estate crises on board. However, data availability prevents us from doing so. This is therefore indirectly taken care of via our change-in-growth-forecast variable.

We have also experimented with the long-term government bond yields, Treasury bill rates, money market rates and discount rates, as published by the IMF in its International Financial Statistics. These series are in general highly correlated. As, in contrast to these other interest rates, lending rates are available for most of the countries in our sample, we prefer using those. The results do not change qualitatively.

Based on the work of Sturm and De Haan (2001) and Dreher et al. (2008; 2010), the KOF Swiss Economic Institute published annually a database containing information on the term in office of central bank governors for almost all countries in the world starting from the year 1970. We use the 2013 vintage and calculate the average irregular turnover rate during the period 1990–2008.

We also experimented with the use of a central bank governor turnover rate that includes changes occurring after the regular term in office did end. The qualitative results are unaffected by this.

See, for instance, Lijphart (1990; 1999).

While the political fate of individual political parties is clearly tied to economic variables, this seems unlikely to be the case for the political constraints the ruling party faces. To see this, consider an exemplary case where dire economic conditions lead an incumbent party to lose both its legislative and executive powers to an opposition party. This change in political power would leave the value of the constraints dummy unchanged. However, in the case in which only legislative elections were held, it would have only lost its legislative powers, causing our constraints dummy to switch from zero to one. This stylized example illustrates that rather than depending directly on economic conditions, the political constraints variable depends on a complex mix of different factors ranging from institutional choices to economic and political conditions at the time of elections.

The high correlation between our minority government dummy and the fraction of seats held by the government and that between the coalition dummy and the probability of government members not being of the same party is expected; in both cases, the first variable is a dummy version of the second.

For recent empirical contributions that find a political business cycle, see, for instance, Aidt et al. (2011), Benito et al. (2013).

We have not only estimated models in which the election variables are interacted with the political constraint variable, but have also done this for our baseline variables and all other political-institutional variables checked in this annex. Whereas the interaction effects are basically never significant, the coefficient on our political constraints variable remains negative and almost always highly significant. Our qualitative conclusions are not affected by including such interaction effects.

The data are taken from the Database of Political Institutions and imply that the variables LIEC and EIEC of that database take on values larger than 5.

    Other Resources Citing This Publication