The slow death of politics

Broadly defined, political economy research takes two different views on the role of politics for economic reform and therefore also fiscal consolidation. An important part of the literature suggests that governments have lost their ability to design and implement policies according to their ideological preferences. One reason is that their room to manoeuvre has been shrinking, owing to the increasing international interdependence of domestic economies (Cusack Reference Cusack1999; Swank Reference Swank2001; Kittel and Obinger Reference Kittel and Obinger2003; Castles and Obinger Reference Castles and Obinger2007). This led to a “functionalist” view on the politics of reforms, policy change and party competition (Cameron Reference Cameron1978; Rodrik Reference Rodrik1998; Huber and Stephens Reference Huber and Stephens2001; Schwartz Reference Schwartz2001; Castles Reference Castles2005), which is also reflected in unusual government arrangements that have become more common in recent years. Examples are grand coalitions (in Germany) and technocratic governments (Italy, Greece), which often arise from developments in the political system that, again, can be partially traced to the societal changes triggered by globalisation.

With regard to reform politics, this view is represented by the New Politics literature (Pierson Reference Pierson1994, Reference Pierson1996, Reference Pierson2001) that sees countervailing pressures as a main source of declining political competition. On the one hand, welfare state institutions and social policies enjoy large popular support (Boeri et al. Reference Boeri, Börsch-Supan, Tabellini, Moene and Lockwood2001; Larsen Reference Larsen2006), which makes welfare states resilient to the challenges of globalisation (Pierson Reference Pierson1996; Vis Reference Vis2009a). At the same time, economic and structural problems raise the pressure to cut spending even if it is politically costly. Retrenchment events then occur out of sheer economic necessity and fiscal pressure. In addition, the extent of the retrenchment events is mostly influenced by macroeconomic factors and is not necessarily related to the preferences of the parties in government or other potential factors like elections. This leads to the following hypotheses:

Revival of politics and strategic fiscal consolidations

The empirical evidence for the “functionalist” perspective on policy change and reform, however, is mixed, and the results on the influence of partisanship contradictory. Some scholars have confirmed that conservative governments indeed implement harsher cuts (Korpi and Palme Reference Korpi and Palme2003; Allan and Scruggs Reference Allan and Scruggs2004). This is in line with the traditional view that right-wing governments promote spending less than left-wing governments. These inconclusive and contradictory findings raise doubts that political motivations indeed have become irrelevant for policy even under austerity, and suggest that the impact of politics on fiscal consolidation is more complex. I focus particularly on three factors that influence the likelihood of consolidation events, notably the ideology of the incumbent, various forms of constraints and strategic considerations related to the electoral cycle.

In the standard logic, left governments should be less supportive of fiscal consolidation than right governments, but this can manifest in different forms. Public support for redistributive policies and welfare states is particularly large during economic downturns and in times of fiscal constraints (Blekesaune Reference Blekesaune2007; Jensen Reference Jensen2007). As left governments generally represent constituencies that rely more heavily on redistributive policies, these governments are particularly vulnerable to losing political support in response to fiscal retrenchment. Left-wing governments therefore may not want to be associated with consolidation measures and hence try to avoid them. This argument implies that left-wing governments are less likely to initiate consolidation measures in the first place.

Alternatively, left-wing parties can choose a different, more strategic approach by proactively opting for consolidation, but, at the same time, design them in their interest. As the literature has shown, left governments often successfully implement consolidation measures, which means that the likelihood of a reform is not necessarily smaller than for right-wing governments (Green-Pedersen Reference Green-Pedersen2002).Footnote ¹ Left-wing governments therefore may not be less likely to pass a consolidation package. However, given that a reform occurs, the left-wing government can use the remaining room to curb spending less radically than a right-wing government would. In this way, the government can avoid being blamed for not solving an existing problem and still minimise the impact of the reform. This means we should primarily see the effect of government partisanship on the size of consolidation.

Apart from political preferences, governments face varying degrees of constraints that affect fiscal consolidation. Constraints generally reduce the ability of governments to implement their most preferred policies and are tied to specific political situations in combination with the institutional framework of a country. With respect to fiscal consolidation, a major political constraint arises from resistance among affected citizens against fiscal cuts. Citizens who suffer from these measures will oppose them and increase the pressure on governments to abandon the intended cuts or reverse previously implemented reforms. An important way in which this resistance manifests itself is through street-level politics. Although examples from the Euro-crisis are evident, there also is evidence that fiscal retrenchment is systematically and historically related to anti-government demonstrations and popular unrest (Ponticelli and Voth Reference Ponticelli and Voth2011). Such mass protests can have an important impact on reform activities in that they affect the political calculus of the government.

To what extent public opposition directly affects the policy-making process, however, depends on the institutional setting of the country. Different institutional features mediate the access of political groups to the political process (Immergut Reference Immergut1992). The electoral system, for instance, has a decisive effect on the number of political parties in government. Majoritarian electoral systems tend to produce single-party governments that are less constrained than coalition governments, which usually are associated with proportional systems. The parliamentary system is also important, because a second chamber provides the opportunity for the opposition to block government proposals. Similarly, central governments in federal systems have less autonomy over particular policy areas, such as fiscal policy, than in unitary systems. The literature agrees that the more actors involved in decision-making processes, the more difficult it is to implement highly contentious policies, such as fiscal retrenchment. The presence of political and institutional constraints therefore primarily influences the probability of consolidation. This argument is consistent with empirical evidence from other areas of fiscal policy making, such as tax policy (Hallerberg and Basinger Reference Hallerberg and Basinger1998; Basinger and Hallerberg Reference Basinger and Hallerberg2004; Hallerberg Reference Hallerberg2004).Footnote ²

Finally, there are important strategic considerations influencing governments’ reform plans. In principle, major consolidation events are considered a risky enterprise for governments. Despite the electoral riskiness of reforms, research shows that governments regularly implement unpopular reforms, taking into account potential punishment by the electorate (Vis and van Kersbergen Reference Vis and van Kersbergen2007). They do this when a deteriorating economic situation and weakening government strength manifest themselves in low public opinion polls. The combination of these factors leads governments to resort to risky reforms, with the hope that these reforms will reinstate a more favourable position (Vis Reference Vis2009b, Reference Vis2009c).

Given the political risk associated with fiscal reforms, I would expect that governments strategically time these events to maximise political gains or minimise potential losses. Following a large literature on the electoral effects on economic policy (Downs Reference Downs1957; Nordhaus Reference Nordhaus1975; Alesina Reference Alesina1987; Stimson Reference Stimson1999; Erikson et al. Reference Erikson, MacKuen and Stimson2002), it is plausible that fiscal consolidation tends to occur early in the electoral cycle for a number of reasons. First, the electoral cycle structures governments’ legislative activity, with governments’ activity being highest at the beginning of the term and declining in the second half of a legislative term (Heller Reference Heller2001; Becker and Saalfeld Reference Becker and Saalfeld2004; Martin and Vanberg Reference Martin and Vanberg2004). A second reason is the time horizon of voters, which previous research finds to be relatively short (Bartels Reference Bartels2005). If this is right, governments can hope that the negative effects of fiscal consolidation packages will be remembered less by the time of the next election. In addition, if governments cut spending at the beginning of the legislative term, the positive impact of these measures may become noticeable right before the next election, which can be used as a political asset by the incumbent and may convince the voters to support the present government in the next election. According to this logic, the probability that governments implement a substantial reform decreases the closer the next election is. I expect that elections mostly affect the timing of the reform event. If an incumbent decided that it is strategically wise to cut spending before an election, its size and content should not be affected by the fact that an election occurs in the near future.

Data

The data cover discrete consolidation events in 16 Western democracies and include a total of 159 instances of fiscal adjustments between 1978 and 2009 (Devries et al. Reference Devries, Guajardo, Leigh and Pescatori2011).Footnote ³ The selection of the consolidation events included in the data set is based on official announcements by public authorities to cut fiscal deficits in nominal terms measured as the percentage of GDP. It represents how these actions reflect on the budgetary balance of the general government and therefore avoids missing deficits, which are shifted to lower levels. Consolidation events are identified qualitatively based on the most important policy documents in each country, including Budget Reports, Budget Speeches, OECD Economic Surveys and IMF Staff Reports. These documents ‘explicitly provide evidence of what policymakers believed at the time that the decisions were taken’ (Devries et al. Reference Devries, Guajardo, Leigh and Pescatori2011, 4). For the events that are identified, the size of the respective measures is then specified by assessing the impact that the cuts have on the overall fiscal balance. The data set therefore captures the governmental decisions that can be publicly identified and were debated as consolidation events.Footnote ⁴ The consolidation events cover both consolidation through spending cuts and consolidation through changes in taxation. In additional analyses, I will distinguish between consolidation via spending and via taxes.Footnote ⁵

The raw data indicates the size of fiscal consolidation in the year when it happened. Using this information, I first construct a time-series cross-section data set with a discrete dependent variable that takes the value 1 for the year in which the data set reports that a consolidation event occurred, and the value 0 for years in which no event is reported.Footnote ⁶ This yields a balanced data set from 1978 to 2009, which can be used to estimate the probability of a consolidation event. To examine the size of consolidation, I construct another variable, which takes the value 0 when no fiscal cut took place and the amount of consolidation as listed in the original data for the year when the government implemented the reform.

The data used in this paper differ in important ways from previous studies of the determinants of fiscal consolidation. Although early work operationalised fiscal consolidation simply as the change in the budgetary deficit (Alesina et al. Reference Alesina, Perotti and Tavares1998, Reference Alesina, Ardagna and Trebbi2006), research has increasingly used the cyclically adjusted fiscal balance to identify discrete consolidation periods (Ahrend et al. Reference Ahrend, Catte and Price2006; Guichard et al. Reference Guichard, Kennedy, Wurzel and André2007; Alesina et al. Reference Alesina, Carloni and Lecce2011).Footnote ⁷ Although it is useful to isolate the structural deficit by removing the cyclical component that is due to regular business cycles, this strategy does not solve the problems mentioned above. A reduction in the structural deficit can be, but often is not necessarily, the result of a government’s policy choice. Instead, the structural deficit can fall owing to a reduction in structural unemployment, which, in turn, can be unrelated to fiscal consolidation measures.Footnote ⁸ The consolidation episodes, which are calculated based on improvements of the CAPB, hence, often do not reflect government decisions, but rather to considerable extent reflect the overall macroeconomic environment net of the cyclical component.Footnote ⁹

Assessing the impact of political and socioeconomic factors on consolidation events rather than incremental changes in public spending has the advantage that consolidation events represent pure political acts. Consolidation events therefore are based on conscious decisions of the cabinet and represent the political considerations of policy makers more properly than gradual outcome variables like spending. In addition, the effect of the key explanatory factors on reform data is straightforward and less ambiguous than for macroeconomic aggregates like spending. For instance, high levels of unemployment may be associated both positively and negatively with public spending. On the one hand, high levels of unemployment boost public expenditure because more benefits need to be paid. On the other, high levels of unemployment, in the long run, may lead to increasing pressure on governments to consolidate and cut public spending. By focusing on consolidation events, we can avoid having the ambiguous relationship between incremental changes in public spending and macroeconomic factors confound the empirical results (see also Hicks and Zorn Reference Hicks and Zorn2005). As a consequence, the correlation between the two different measures is not very high. Further analyses show that this is because consolidation episodes, as defined by the events-based measure, mostly occur in times when structural problems in the macro economy deteriorate. Consolidation episodes, as defined by the CAPB measure, often occur when structural economic problems improve. I will revisit these results in greater detail in the next section.

The key political explanatory variables are government ideology (Klingemann et al. Reference Klingemann, Volkens, Bara and Budge2006; Andersson et al. Reference Andersson, Bergmann and Ersson2012; Volkens et al. Reference Volkens, Lehmann, Merz, Regel and Werner2013), political constraints (Henisz Reference Henisz2006) and different electoral variables. Government ideology is measured using an index suggested by the study of Kim and Fording (Reference Kim and Fording1998, Reference Kim and Fording2002). The measure is based on the left-right index from the Comparative Manifesto Project and takes into account the weighted seat share in parliament of each party in the cabinet. The original index varies between −100 (most leftist position) and 100 (most rightist position). This index has been rescaled for this analysis to range between 0 and 1 (see also Table A1 included in the online appendix). Lower values still indicate more leftist party governments. To measure political constraints, I use the general constraints index by Henisz (Reference Henisz2006). Unlike other measures of political constraints, which are often time invariant, it takes into account that political constraints can vary owing to changing majorities in national decision-making arenas.Footnote ¹⁰ In addition to general constraints, I assess the impact of public discontent as another form of political constraint using a variable that captures whether anti-government protests occurred in a given year (Banks and Wilson Reference Banks and Wilson2014). The distribution of demonstrations is seriously skewed, which is why I use a binary variable distinguishing between no and at least one demonstration. The results do not change when variables with more categories are used. To assess the strategic dimensions of the politics of fiscal consolidation, I use two electoral variables, which capture various parts of the effect the electoral cycle has on government reform activity. The first is an electoral counter counting the years between two elections. The second comprises different electoral dummies for every year after the election. The main macro- and socioeconomic measures are unemployment, economic growth, demographic change and financial openness, which come from the study by Armingeon et al. (Reference Armingeon, Careja, Weisstanner, Engler, Potolidis and Gerber2013).Footnote ¹¹ In the empirical models, the socioeconomic variables are lagged by one year, as I expect the impact of these variables to take effect only during the next budgetary year (Table 1).

Table 1

Summary statistics

Source: Author’s elaborations.

Estimation strategy

Given the structure of my data, the empirical part proceeds in two steps. The first step assesses the probability that a consolidation event happens. The second step examines the size of consolidation, given that a reform occurs. Because of the binary nature of the dependent variable for the first step, I use a probit model to estimate the likelihood that a consolidation event occurs in a particular year. To account for time in a time-series cross-section setting with a binary dependent variable, I use the procedure proposed by Carter and Signorino (Reference Carter and Signorino2010), which has become a standard approach to dealing with this problem in political economy research. The variable t captures the time that has passed since the last consolidation event, i.e. it takes the value 0 after a fiscal consolidation episode and increases linearly until the next consolidation event in that country occurs. I also add t ² and t ³ to the model to capture the nonlinear effects of time on the probability of an event. These variables together capture the probability of a reform, given the number of years that have passed since the last reform. This procedure therefore is a variant of a classic duration model accounting for the hazard that an event takes place over time (see Box-Steffensmeier and Jones Reference Box-Steffensmeier and Jones2004). The models based on equation (1) assess the probability of a reform as specified by H0, H1a, H2a, H3 and H4. The results from this equation are presented in Table 2.

(1) $$\eqalignno{ Pr({\rm Consolidation}\,{\rm event}_{{i,t}} )= \ & \Phi (\alpha _{0} {\plus}\alpha _{1} {\times}{\rm Ideology}_{{i,t}} \cr & {\plus}\alpha _{2} {\times}{\rm Constraint}_{{i,t}} {\plus}\alpha _{3} {\times}{\rm Election}_{{i,t}} \cr & {\plus}\alpha _{4} {\times}{\rm Growth}_{{i,t{\minus}1}} {\plus}\alpha _{5} {\times}{\rm Unemployment}_{{i,t{\minus}1}} \cr & {\plus}\alpha _{6} {\times}{\rm Elderly}_{{i,t{\minus}{\rm 1}}} {\plus}\alpha _{7} {\times}{\rm Openness}_{{i,t{\minus}1}} \cr & {\plus}\alpha _{8} {\times}t_{{i,t}} {\plus}\alpha _{9} {\times}t_{{i,t}}^{2} {\plus}\alpha _{{10}} {\times}t_{{i,t}}^{3} {\plus}d_{i} {\plus}\epsilon ) $$

where d _i is a country dummy.

Table 2

Determinants of consolidation events

Source: Author’s elaborations.

Notes: All models include fixed effects. Robust errors are in parentheses.

*=p<0.1; **=p<0.05; ***=p<0.01.

Examining the size of consolidation raises some tricky econometric issues. First, if we examine consolidation size by looking at consolidation episodes only, we get biased estimates because of selection problems. Specifically, the set of governments that consolidates is systematically different from those that do not consolidate because they face different macroeconomic and political situations. Second, if we examine all country-years (also those when consolidation is 0), we still have the problem that governments that do not consolidate (those for which the dependent variable is 0) are all treated in the same way in a standard regression model. This is problematic because some of those governments only marginally decided against reform, whereas others were far away from implementing a reform (Bortolotti and Siniscalco Reference Bortolotti and Siniscalco2004; Hicks Reference Hicks2014). Theoretically, the problem arises because, as the previous literature has argued, the political processes that lead to consolidation differ from the political processes that lead to expansions (Pierson Reference Pierson1996, Reference Pierson1998). We therefore cannot simply expand the distribution to include both consolidations and expansions because the variables and their expected impacts differ for the two processes.

I address these problems using Tobit and Heckman models. Tobit models take into account the possibility that governments that did not reform can have different characteristics. Technically, it assumes that the underlying (latent) consolidation process is only observable when consolidation is positive, which means that the observable distribution of consolidations is truncated. I use the same basic model as before, but exclude the time variables because we do not have a binary dependent variable at this stage.

(2) $$\eqalignno{ {\rm Consolidation}\,{\rm size}_{{i,t}} =\,\,\beta _{0} {\plus}\beta _{1} {\times}{\rm Ideology}_{{i,t}} {\plus}\beta _{2} {\times}{\rm Constraint}_{{i,t}} {\plus}\beta _{3} {\times}{\rm Election}_{{i,t}} \cr &#x0026; {\plus}\beta _{4} {\times}{\rm Growth}_{{i,t{\minus}1}} {\plus}\beta _{5} {\times}{\rm Unemployment}_{{i,t{\minus}1}} \cr &#x0026; {\plus}\beta _{6} {\times}{\rm Elderly}_{{i,t{\minus}1}} {\plus}\beta _{7} {\times}{\rm Openness}_{{i,t{\minus}1}} {\plus}d_{i} {\plus}\epsilon _{{i,t}} $$

The Heckman model captures the truncation and nonrandom selection of fiscal consolidation events in a different way. It is more complex than the Tobit model because it requires the specification of two equations. One equation accounts for sample selection, and the other for the outcome of the reform. The estimation of the model is also based on the assumption that at least one variable affects the selection process, but does not affect the outcome process. The selection equation is equivalent to the probit model described in equation (1). This equation estimates the probability of observing a consolidation event, i.e. whether or not an observation is selected into the consolidation subsample. For the outcome equation, I can rely on a version of equation (2), because the dependent variable of the outcome equation is the consolidation size. To satisfy the exclusion restriction, variables that only affect the timing, but not the size of consolidation, need to be excluded in the outcome equation. An argument can be made that the time counters represent such variables. As Alesina and Drazen (Reference Alesina and Drazen1991) argue, the mere “passage of time” often increases the likelihood that a consolidation package is passed, independent of the development of the underlying structural problems (which are captured by different variables in my model). By this argument, the time variables represent the gradually increasing cost of not solving a problem, holding the size of the problem constant. In addition, according to the theoretical logic in this paper, I assume that the election cycle primarily influences the timing, but not the size of the consolidation measures. This implies that the election variable can be excluded from equation (2) for the Heckman model.Footnote ¹² Given the difficulty of establishing that a variable should truly be excluded from the selection equation, the Heckman results should be interpreted in combination with the results from the Tobit models, which rely on less restrictive assumptions.