Does Austerity Cause Polarization?

Abstract In recent decades, governments in many Western democracies have shown a remarkable consensus in pursuing fiscal austerity measures during periods of strained public finances. In this article, we show that these decisions have consequences for political polarization. Our macro-level analysis of 166 elections since 1980 finds that austerity measures increase both electoral abstention and votes for non-mainstream parties, thereby boosting party system polarization. A detailed analysis of selected austerity episodes also shows that new, small and radical parties benefit most from austerity policies. Finally, survey experiments with a total of 8,800 respondents in Germany, Portugal, Spain and the UK indicate that the effects of austerity on polarization are particularly pronounced when the mainstream right and left parties both stand for fiscal restraint. Austerity is a substantial cause of political polarization and hence political instability in industrialized democracies.

Until recently, fiscal policy in Western democracies has been characterized by a remarkable consensus among governing parties.Especially during crises, though also more generally, centre-left and centre-right governments have agreed that fiscal restraint is the appropriate policy to promote economic stability and growth (Blyth 2013).As a result, left and right political parties in many countries have embarked on a path of fiscal austerity when in government (Hopkin 2020;Mudge 2018).Key explanations for this trend turn to the constraints imposed by international integration (Konstantinidis, Matakos and Mutlu-Eren 2019), the growing appeal of technocratic solutions (Alexiadou, Spaniel and Gunaydin 2022) or the emergence of a new economic policy consensus (Blyth 2013;Mudge 2018).
In this article, we show that this fiscal policy consensus among mainstream politicians has increased party system polarization.Across Europe, party competition has radicalized due to the decline of mainstream, centrist parties and the concomitant rise of non-mainstream, radical competitors (De Vries and Hobolt 2020).In Italy, where many of these developments have been pre-empted, traditional parties have even almost vanished and been replaced by new anti-establishment parties.Similar developments can be seen in other countries, with the rise of Podemos and Vox in Spain, the Party for Freedom (PVV) and Forum for Democracy (FVD) in the Netherlands, the Sweden Democrats in Sweden, the Left Bloc and Chega in Portugal, and the Rassemblement National in France.Even ostensibly more stable countries, such as Germany, have witnessed change, with the joint vote share of the Christian Democrats and the Social Democrats falling from 76 per cent in 1998 to below 50 per cent in 2021.
Analyses of party system change and polarization generally focus on structural factors, such as economic globalization and digitalization (see, for example, Baccini and Weymouth 2021;Colantone and Stanig 2018;Gallego, Kurer and Schöll 2022), new political cleavages and changes to class politics (Halikiopoulou and Vlandas 2016;Hooghe and Marks 2018), the growing salience of identities, culture and immigration (Norris and Inglehart 2018), or economic crises (Alonso and Ruiz-Rufino 2020;Hernández and Kriesi 2016).This literature, however, has mostly ignored the role that governments and their economic decisions play in periods of economic insecurity.When economic shocks hit voters, governments can mediate their effects in different ways.As the government's main economic policy instrument, fiscal policy is decisive in helping those who suffer from economic change (Vlandas and Halikiopoulou 2022).The fiscal policies that parties propose in this context are thus important for understanding the political behaviour of voters.More broadly, elites play an important role in explaining party system stability (Tavits 2008).
If structural change creates demand for active fiscal policy, at least among a subgroup of the electorate, fiscal austerity disappoints and alienates voters (Hopkin 2020).Fiscal restraint, unopposed by mainstream competitors, creates a pool of dissatisfied voters lacking a credible antiausterity alternative among mainstream parties.Where governments opt for fiscal cutbacks, voters therefore defect from mainstream parties, as they occupy similar positions on relevant policy areas (Spoon and Klüver 2019).When defecting, some of these voters abstain from voting, while others decide to support non-mainstream, often radical, political parties that reject austerity.In the aggregate, both decisions lead to an increase in party system polarization.
Such an increase in party system polarization poses a challenge for policymaking in democracies.A more polarized political landscape makes it more difficult to construct stable government coalitions and sustainable policy solutions, both essential in times of economic insecurity (Dalton 2008;Sartori 1976).This leads to considerable policy uncertainty (Konstantinidis, Matakos and Mutlu-Eren 2019).Struggles to build stable coalitions and the associated damage to policy-making processes can be witnessed in many European countries, such as Israel, Sweden, Italy and, in particular, Spain, which held four general elections between 2015 and 2019.Our analysis suggests that solutions to these challenges need to take into account political competition over key economic policies.
We examine the impact of fiscal austerity on political outcomes in two steps.First, we use observational data from 166 elections in sixteen Organisation for Economic Co-operation and Development (OECD) countries between 1980 and 2016 to examine how austerity is associated with turnout rates, non-mainstream party vote share and party system polarization.Controlling for key economic and political factors, austerity correlates substantially with these political outcomes.If the government implements sizeable fiscal austerity packages during the legislative term, the vote share of non-mainstream parties is up to 3.5 percentage points and voter abstention up to 1.8 percentage points higher.This, in turn, increases polarization from one election to the next.A more detailed analysis of selected large fiscal adjustment episodes also shows that previously small, new and radical parties are the main beneficiaries of austerity.
The second part of the analysis uses survey experiments in Germany, Portugal, Spain and the UK to uncover the micro-level mechanism that explains these macro-level correlations.The survey experiment also indicates which parties lose or gain most after austerity.The results show that voter responses to austerity depend on the policy alternatives offered by mainstream parties.If both the mainstream left and right propose austerity, voters are more likely to turn towards smaller, non-mainstream parties or abstain from voting than if one mainstream party opposes austerity.Abstention and defection to non-mainstream parties thus occurs when none of the mainstream parties provide a credible anti-austerity alternative.

Fiscal Adjustment, Mainstream Parties and Polarization
In recent decades, fiscal adjustment has become a common policy of mainstream parties, both of the centre-right and the centre-left.These adjustment measures, often summarized under the term 'austerity', refer to a set of fiscal decisions that aim at reducing the fiscal deficit by a prespecified amount.In practice, these decisions entail a mix of spending cuts, such as a reduction in social spending, educational expenditures or public investment, and tax increases, such as higher value-added tax (VAT) or income taxes.
These austerity measures are a distributional shock that hurts a significant fraction of voters, especially those in an economically precarious situation.Fiscal adjustments often come with significant reductions in welfare provisions, especially social security transfers or unemployment benefits (Armingeon, Guthmann and Weisstanner 2016).Yet, these are precisely the instruments that governments can use to alleviate the negative impact on economically vulnerable voters of the great structural transformations of recent decades, such as globalization or automation (Baccini and Sattler 2020;Vlandas and Halikiopoulou 2022).For instance, recent research shows that social policies can effectively reduce the risk that economically vulnerable groups face in open economies (Hacker and Rehm 2022).Moreover, tax increases, especially regressive taxes like VAT, disproportionately hurt less privileged groups.As a result, fiscal adjustments have distributional implications that affect an important part of the electorate.
Nevertheless, nowadays, centre-left governments do not distinguish themselves substantially from centre-right governments in their position towards fiscal adjustments.Centre-left parties are equally likely (or, by some accounts, more likely) to pass adjustment packages compared to centre-right governments (Armingeon, Guthmann and Weisstanner 2016;Hübscher and Sattler 2017;Raess 2021). 1 More broadly, nowadays, centre-left parties tend to back economic models of which fiscal restraint is a fundamental pillar (Hopkin 2020;Mudge 2018).For example, social-democratic parties in Britain, Germany, the Netherlands and Sweden have embraced orthodox economic policies, even though their ideological foundations differ (Bremer and McDaniel 2019).
This combination of limited competition among mainstream political parties over fiscal adjustments and the distributional effects of these policies increases political polarization, which is a key characteristics of party systems (Dalton 2008;Sartori 1976).We define polarization as the degree to which political actors in a political system take distinct ideological positions and are thus ideologically dispersed along the left-right political dimension.Following Dalton (2008), we take into account both the positions of political parties and the relative popularity of these parties among voters (see also Dassonneville and Cakir 2021;Ezrow 2007). 2 Therefore, political polarization can change through two mechanisms: first, polarization increases if existing parties move further towards the edges of the political spectrum (holding vote shares constant); and, secondly, polarization increases if electoral support for (existing or new) political parties at the fringes of the political spectrum increases (holding party positions constant).This gives more weight to parties at the poles of the political system and increases the variance of ideological positions.Related to the second mechanism, polarization can increase indirectly when voters who previously supported moderate parties decide to abstain from voting.This, again, changes the relative weight of parties at the poles compared to those in the political centre.
Our article focuses on the second mechanism.3According to our argument, shifts in electoral support after fiscal adjustments increase polarization, which is our main dependent variable.This 1 We confirm these findings using our own dataset in Table A9 and Figure A8 in the Online Appendix.As Dalton (2008, 906) argues: 'a large party at the extreme would signify greater polarization than a splinter party in the same position'.We therefore conceptualize polarization as 'weighted party system dispersion' (Ezrow 2007), that is, the variance of party positions weighted by vote shares.Some studies conceptualize polarization merely as the distance between the most extreme parties or the (unweighted) variance of party positions.While vote shares may be less central when party systems with a predefined, clearly identifiable set of relevant parties are compared across countries, they matter when examining how polarization changes within countries over time because previously small parties grow or new parties emerge.This is the key mechanism increasing polarization in our study.
happens because voters defect from mainstream parties, either by abstaining or by opting for more radical parties.We, therefore, examine non-mainstream party vote and electoral abstentions as additional dependent variables.
Why Voters Leave: Austerity and Voter Defection from Mainstream Parties When voters perceive mainstream parties as similar on key policy issues, in our case, fiscal policy, some voters defect from these parties (Spoon and Klüver 2019).To understand this process, let us first look at a scenario where the mainstream left party is in government and implements fiscal austerity measures when faced with fiscal pressure.In doing so, mainstream left parties reveal that their fiscal policy position diverges from the policy preferences of an important part of their traditional electoral base.It is unlikely that the mainstream right will take a position that is more critical of austerity than the mainstream left party in government, as parties rarely leapfrog each other ideologically.Hence, voters who aim to express their disapproval of austerity should not then switch to the mainstream right party; rather, they should defect from mainstream parties altogether.
Next, take a scenario where a mainstream right party with a pro-austerity position is in government and implements austerity.The mainstream left party could benefit from taking up a contrasting stance, but their track record is likely to reduce the credibility of such stances (Horn 2021).Nowadays, centre-left parties are as likely to implement austerity as centre-right parties, and centre-left parties in most Western countries have regularly implemented major fiscal adjustment packages when in power, leading to a dilution of their party 'brand' (Bodea, Bagashka and Han 2021;Lupu 2014).As a result of the parties' past policy-making trajectory, many voters who oppose austerity will not see the mainstream left as a valid, credible alternative (Horn 2021; Karreth, Polk and Allen 2013), and the result is again defection from mainstream parties or electoral abstention. 4 The preceding logic rests on several assumptions.First, we assume that voters support parties based on the perceived proximity of the party's fiscal policy stance to their own position.The literature generally uses two dimensions along which parties and voters position themselves: an economic and a cultural dimension (see, for example, Norris and Inglehart 2018).For the sake of simplicity, we treat fiscal policy as the main policy lever for the economic dimension.This does not require that all voters privilege economic issues at all times.For some voters, fiscal policy may not be salient enough to trump other policy issues (Giger 2012).Such voters will not react to austerity measures, but instead continue to vote for the party that they supported in the past (keeping everything else constant).Our assumption is that fiscal policy is a salient issue for a significant number of voters, for instance, those who are hit by the distributional effects of austerity or who oppose fiscal adjustments for sociotropic reasons.
Secondly, we assume that government parties cannot fully spare their own base by imposing the burden of adjustment on voters of other parties.Meaningful fiscal adjustments, which often amount to deficit reductions of 2 per cent or more of gross domestic product (GDP), generally involve adjustments in most categories of the public budget.5Yet, adjustments in almost all areas of the public budget lead to a reduction in voter support, especially when social policies are concerned (Bansak, Bechtel and Margalit 2021).6This is consistent with earlier findings that social positions as fixed in the time period we study.We also do so for methodological reasons, specifically to avoid supply-side effects contaminating our analyses.

4
We present the logic in a simplified form referring to two mainstream parties, but the logic is similar for left-or right-wing coalitions.In addition, the effect of austerity on voters should magnify for coalitions across political blocs, as these make it harder for voters to distinguish the policy stance of coalition parties.
policies enjoy broad support in the population (Blekesaune and Quadagno 2003) and limits the possibilities to manipulate public support by shifting the burden of adjustment from one fiscal policy field to another.Moreover, not only left, but also right, voters negatively react to fiscal adjustments (Hübscher, Sattler and Wagner 2021). 7 Thirdly, we assume that neither mainstream right nor mainstream left parties have sufficient antiausterity credibility to attract dissatisfied voters.This does not preclude that these parties still emphasize social inequality differently, for example, in their party manifestos.However, centre-left and centre-right parties broadly agree that low fiscal deficits are a political priority, and their fiscal policy trajectories are similar when they are in government.8Such past policy decisions of parties in government strongly impact voter perceptions of party positions, more so than electoral platforms (Adams, Ezrow and Somer-Topcu 2011;Fernandez-Vazquez 2019).The fact that most centrist parties now have a history of implementing austerity, as discussed earlier, is therefore crucial.
Finally, and related to the previous point, mainstream parties implement austerity despite the risk that dissatisfied voters abandon them.International political integration, the establishment of the Eurozone and the related fiscal rules induce governments to pursue similar, low-deficit fiscal policies (Hübscher and Sattler 2022).In addition, international financial markets impose constraints on fiscal policy, especially on public deficits and debt (Mosley 2000).This pushes governments towards technocratic solutions in search of economic credibility, which again favour restrictive fiscal policy (Alexiadou, Spaniel and Gunaydin 2022).In this context, a new, pro-austerity consensus emerged that was actively promoted by powerful international institutions, such as the European Central Bank (ECB) and the International Monetary Fund (IMF) (Blyth 2013).Hence, we expect that: Hypothesis 1a (macro): Austerity measures reduce mainstream party vote share at the next election.
Hypothesis 1b (micro): Citizens are less likely to vote for mainstream parties that support austerity measures than for mainstream parties that oppose such measures.
Where Voters Go: Voter Defection and Polarization Voters who turn away from mainstream parties have two options: they can choose to vote for a non-mainstream party or they can abstain.This can happen out of protest without clear expectation of how this impacts fiscal policy.However, there are also clear policy-related reasons why voters would shift towards non-mainstream parties.First, dissatisfied voters would select a nonmainstream party if they see this party as a credible, anti-austerity alternative.Secondly, and relatedly, a voter may also expect that the electoral success of non-mainstream parties will induce a shift in the position of mainstream parties away from austerity.Both reasons require that the voter weights the economic dimension of political competition relatively highly, which is plausible when the distributional effects of austerity on the voter are large.
and public employment do not lead to a drop in support (see Bansak, Bechtel and Margalit 2021, 14, Figure 4).Governments, therefore, could only avoid a decline in public support by concentrating cuts on public employment and/or defence.Defence spending, however, generally amounts to significantly less than 2 per cent of GDP, which limits the possibility to mainly adjust via defence. 7 Left voters react more strongly to adjustments, but a majority of right voters also opposes fiscal cuts (see Hübscher, Sattler and Wagner 2021, 1755, Figure 3).Relatedly, overall public support drops after austerity, and this does not differ between left or right governments (see, for example, Bojar et al. 2022;Talving 2017).The structural changes discussed earlier likely hit not only traditional centre-left voters, but also many centre-right voters, such as conservative workers (Baccini and Weymouth 2021;Colantone and Stanig 2018).
Other voters may instead decide to abstain from voting altogether.While abstention has many motivations (Smets and van Ham 2013), disenchantment with democracy and the political establishment are a prominent determinant (Hillen and Steiner 2020).Specific policy decisions by government parties may also have a detrimental impact on political participation.For instance, centrist voters may be loath to vote for non-mainstream competitors, even when they disagree with mainstream parties' austerity policy stances, because they strongly disagree with the radical party's position on the cultural dimension.As they are indifferent between equally unpalatable choices, such voters may abstain, rather than choose between unsatisfactory alternatives.
Both defection and abstention lead to an increase in party system polarization.As discussed earlier, defection to non-mainstream, often non-centrist, parties bolsters their overall share of electoral support, increasing the (weighted) dispersion of positions in a party system (Ezrow 2007). 9Similarly, differential abstention by previous mainstream party voters also increases the (weighted) dispersion of positions, as the relative share of non-mainstream competitors increases.Ironically, then, the centripetal positioning of mainstream parties in fiscal policy leads to an overall trend towards centrifugal competition in the political system as a whole.
Again, this logic rests on a number of assumptions.First, voters need to believe that a growing vote share of non-mainstream parties reduces the likelihood of fiscal austerity.This, in turn, requires that non-mainstream parties credibly position themselves against austerity.Here, nonmainstream parties have an advantage because, unlike mainstream opposition parties, they usually do not have a history of implementing fiscal adjustments.Among the different types of nonmainstream parties, radical-left parties, for instance, Die Linke in Germany, the Dutch Socialist Party or France Insoumise, seem an obvious choice for voters who disapprove of fiscal cutbacks.Such parties take a clear stance against austerity and have strongly opposed interference by international actors promoting fiscal adjustments, such as the IMF or the European Union (EU).Some radical-left parties even emerged from anti-austerity movements, for instance, Podemos in Spain.
Radical-right parties can also be attractive to voters disappointed with fiscal policies.They strategically position themselves as an alternative to mainstream parties (De Vries and Hobolt 2020).Austerity is one area where radical-right parties regularly do so, and they often position themselves clearly on economic policies, especially when they can oppose specific austerity measures (Röth, Afonso and Spies 2018).In such cases, radical-right parties have consistently defended social policies, for example, pensions, against retrenchment proposals (Enggist and Pinggera 2022). 10Voters, therefore, can have direct policy-related motives when they select a radical-right party in response to fiscal austerity.Some voters may also have broader motives that involve other issues related to austerity.Radical-right parties effectively link austerity to Europeanization, globalization, the political establishment and immigration.Radical-right parties often advocate rolling back Europeanization and globalization to lift the constraints imposed by international integration.Examples are UKIP in Britain (Fetzer 2019) or the Rassemblement National in France, which actively courts dissatisfied voters by stressing its opposition to the political establishment and austerity policies.11Many parties, such as Lega or the Golden Dawn, also reframe austerity politics as a migration-related, cultural issue by promoting 'welfare chauvinism' (Vlandas and Halikiopoulou 2022).This shifts attention to issues where the radical right takes particularly clear positions, such as immigration and the EU.
Secondly, we assume that non-mainstream parties have a chance either to participate in government or to move the policy positions of mainstream parties.In many countries, 9 Defection to non-centrist parties can potentially decrease polarization, for example, if most voters choose one extreme alternative, creating a new consensus.Our starting point, however, is the gradual transformation of political systems with mainstream parties that still obtain a large vote share (see Figure 1).10 To give one example, candidates of the right-wing Finns Party take interventionist economic positions that are consistently more 'leftist' than those of the Finnish Social Democrats (Tähtinen 2021).
non-mainstream parties have participated in government, for example, in Austria, Finland and Italy.In line with our argument, radical parties, including those on the right, refrain from fiscal retrenchment when in government (Röth, Afonso and Spies 2018).However, even if radical parties are not in government, their electoral success can be enough to shift the positions of other parties (Abou-Chadi and Krause 2020).Therefore, it makes sense for anti-austerity voters to support radical parties even if these parties will not be part of the next government.Hence, we expect that: Hypothesis 2a (macro): Austerity measures increase polarization at the next election.
Hypothesis 2b (micro): Citizens are more likely to choose non-mainstream parties or to abstain from voting when all mainstream parties support austerity measures than when one or both mainstream parties oppose austerity.

The Macro Pattern
Empirical Design and Data The first part of our empirical analysis examines elections from sixteen advanced economies between 1980 and 2016. 12This macro approach is useful for assessing the relationship between fiscal policy and political changes at the party systems level.The country-level analysis captures the empirical relationship between fiscal policy and the patterns of national election outcomes, as well as the associated systemic changes.Given that we are assessing the variation in fiscal policy across all elections, our counterfactual is the absence of fiscal consolidation (in a given electoral cycle). 13A macro-level approach necessarily compromises on insights about causality, the underlying mechanisms and the motivations of voter behaviour.We therefore complement the macro analysis with an experimental analysis of individual-level behaviour below.We examine three different outcomes: the vote for non-mainstream parties; electoral turnout; and party system polarization.We identify non-mainstream parties based on the party family classification by the Comparative Manifestos Project (CMP) (see also Spoon and Klüver 2019). 14For our purposes, mainstream parties are those that support the current economic order and related economic policies, notably, low deficits and fiscal restraint.We classify socialdemocratic, Christian-democratic, conservative, liberal and agrarian parties as mainstream parties.Electoral abstention is measured as the share of eligible voters who do not participate in an election.Political polarization is the dispersion of positions of political parties, weighted by their vote shares.This indicator increases when more voters vote for parties at the fringes of the political spectrum.As in Dassonneville and Cakir (2021), we compute polarization as n i=1 v i (p i − p) 2 , where: p is the weighted mean of all the parties' ideological positions; p i is the ideological position of party i; v i is the share of votes that party i receives in an election; and n is the number of parties that participate in the election.To measure the positions of political parties, we use the left-right index from the ParlGov database, which integrates the results from various expert surveys. 15The ParlGov index is time-invariant, which allows us to isolate the effect of voter decisions on polarization.
12 Data on national-level parliamentary elections are from the Comparative Manifesto Project (CMP).The set of countries is limited by the availability of fiscal consolidation data and includes: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, Portugal, Spain, Sweden, the UK and the United States.The number of elections per country (166 in total) varies between eight (France and the UK) and fourteen (Australia).

13
The magnitude of the austerity impact may vary depending on the macroeconomic context, but austerity should deprive a significant group of voters across all contexts.We explore this in further analyses.
Figure 1 shows that the temporal patterns for all three outcome variables are straightforward.On average, the vote share for non-mainstream parties doubled between the 1980s and today.The non-mainstream party vote share increases sharply from the mid-1980s to 2000 and from 2009 onwards.Abstention also steadily increased, from below 20 per cent at the beginning of our period of analysis to almost 30 per cent at the end.Finally, average polarization increases considerably, with the sharpest increases between the mid-1980s and early 2000s, and again from the mid-2000s onwards.
For our independent variablefiscal consolidationwe use the events-based measure by Devries et al. (2011) and updated by Alesina, Favero and Giavazzi (2019). 16Events-based measures of fiscal consolidation qualitatively identify the timing and magnitude of fiscal consolidation packages using policy documents from governments and international organizations.This is now the standard approach of measuring fiscal consolidations because it directly captures fiscal policy decisions (Alesina, Favero and Giavazzi 2019).Out of the 166 legislative periods in our dataset, ninety-one periods include at least one consolidation event.Among these, the degree of fiscal consolidation varies considerably.A median consolidation package aims at reducing the fiscal deficit by 1.9 per cent of GDP, with a standard deviation of 2.21.
Our control variables largely follow from the literature on party system change and polarization.These include macroeconomic conditions (represented by real GDP growth and the unemployment rate), the degree of international openness and net migration.These variables enter our model as changes from one legislative period to the next.We also account for early elections, single-versus multi-party government, and the permissiveness of the electoral system by controlling for the electoral system (majoritarian versus proportional) and the magnitude of electoral districts.Some of these variables mostly account for cross-country variation, while our focus is on over-time variation within countries.Nonetheless, they account for changes after electoral systems change (Best 2012) and possibly affect how easily voters switch from one party to another. 17Table A1 in the online Appendix shows the summary statistics.
Since we are interested in gradual change in electoral behaviour over time, we use the changes in the outcome variables from one election to the next as dependent variables.This also addresses the econometric problem that the outcome variables are not stationary. 18Our analysis uses the Note: Average values across the 16 countries per year for each variable; y-axes of graphs show share of voters who support non-mainstream parties (left), share of voters who abstain from voting (middle) and the dispersion of positions of political parties, weighted by their vote shares (right) as described in section 3.1.

16
We use government policy, rather than mainstream party positions, as our key predictor.Hence, our main test concerns actual policy, rather than political rhetoric.In the following, we discuss the limitations of existing data measuring policy positions.17 Data for the control variables come from Armingeon et al. (2019) and Bormann and Golder (2013).For early elections, we follow Schleiter and Tavits (2016) and take the difference between the maximum duration of the legislative term and the actual election.We use a three-month threshold to identify elections that were held early.

18
The panel unit root tests in Table A2 in the Online Appendix reject the null hypothesis of a unit root for the differenced, but not for the undifferenced, outcomes variables.following empirical specification: where: i refers to the country and t represents an election; y i,t refers to either votes for nonmainstream parties, electoral abstentions or polarization; Austerity i,t−1 is the amount of fiscal consolidation in the legislative period prior to the election; X i,t−1 includes the control variables; d t are period fixed effects; c i are country fixed effects; and e i,t is an error term.We also use alternative empirical models, notably, models in levels, rather than differences, with a lagged dependent variable and with a time trend.As we show in the Online Appendix, the results from these models are the same.

Results
Table 1 presents our estimation results.For each outcome variable, we present three specifications: without controls; with economic and political controls; and with country and period fixed effects.The results show that the vote share for non-mainstream parties, electoral abstention and polarization increase from one election to the next when the government implements a fiscal consolidation package during the legislative term.This effect is statistically significant and robust across different specifications.
The magnitudes of the estimated effects are substantive.A median consolidation package, which aims at reducing the fiscal deficit by 1.9 per cent of GDP, increases votes for nonmainstream parties by 1.3 percentage points on average.A large package that is one standard deviation above the median, which aims at reducing the fiscal deficit by 4.1 per cent, increases non-mainstream party votes by 2.9 percentage points.The effect of austerity on electoral abstention is smaller.A median consolidation package is predicted to increase abstention by 0.6 percentage points, while a large package increases abstention by 1.2 percentage points.
These effects then translate into an important increase in the polarization of the party system.Figure 2 plots the predicted effect of fiscal consolidations on polarization across the empirically observable distribution of consolidations.For instance, a large consolidation package (one standard deviation above the median consolidation) increases polarization by about 0.1 units.This effect increases up to 0.15 for very large packages that reduce the deficit by 7-8 per cent of GDP.Compared to the empirically observable increase in polarization over recent decades, this estimated impact of austerity on polarization is quite substantial.Figure 1, for instance, shows that average polarization across countries increased by about 0.25 units between 1980 and 2015.As Table A1 in the Online Appendix shows, the average increase in polarization during an electoral period in our dataset is 0.028.A median austerity package, therefore, roughly doubles the average growth in polarization, and a large one increases it by a factor of three to four. 19mong the control variables, the variables representing macroeconomic circumstances do not have a consistent effect on the outcomes.This suggests that the impact of austerity does not simply capture the impact of the bad economic conditions that are often (albeit not always) associated with fiscal austerity packages.
A series of supplementary analyses examines the robustness of these findings.First, we examine the sensitivity of our results to alternative model specifications and estimate models in levels with a lagged dependent variable in Table A3.The results are very similar to the ones presented earlier.Secondly, we disaggregate the non-mainstream party variable and examine different nonmainstream parties separately in Figure A1 and Table A4 in the Online Appendix.The results suggest that most of the effect of austerity in non-mainstream parties works through left parties and the groups of 'other' parties without a specific family label.Thirdly, we analyse the characteristics of fiscal austerity packages in greater detail in Tables A5 and A6 in the Online Appendix.We do not find evidence that spending-or tax-based consolidations affect elections differently.We also disaggregate consolidation by timing during the legislative term.Consolidations in the year before the election year seem to have the strongest impact, but consolidations at the beginning of the term still considerably impact the election.Fourthly, Table A7 and Figure A2 in the Online Appendix examine how the effect of austerity on the outcome variables differs across subperiods of five years.While the number of observations and austerity episodes in each period is fairly small, the results for non-mainstream vote and polarization show an upward trend in the effect of austerity, with a temporary decline in the late 1990s/early 2000s.The impact of austerity is strong after 2010, but it was already substantial throughout the 1990s.As Figure A3 in the Online Appendix shows, these were also the two periods in which most austerity episodes occurred.Fifthly, Figure A4 in the Online Appendix shows that the results do not depend on the inclusion of a particular country.
Finally, our key variableconsolidation eventsreflects government policy, rather than party positions.We opted for this policy variable because it captures what parties do, rather than what they say.To address the role of policy positions more directly, we turn towards a more qualitative assessment of key cases and an experimental analysis in the following.

Who Wins? Who Loses? Evidence from Large Fiscal Adjustments
To further substantiate our macro-level results and to link the preceding analyses to our microlevel analysis in the following, we examine selected electoral cycles during which large fiscal adjustments aiming at deficit reductions of 5 per cent of GDP or more were implemented.This yields a list of ten large adjustment episodes, with five belonging to the period before, and five to the period after, the Great Recession.Looking at electoral cycles from both periods will show that the mechanisms are not unique to the extreme economic crisis that started in 2008.Table 2 lists the countries, the legislative period and the cumulative size of the adjustment packages implemented during a given electoral cycle, as well as changes in non-mainstream party vote, abstention and polarization. 20Here, we provide a synthesized account of the effects of austerity, implemented by mainstream parties and uncontested by mainstream opposition parties.

Mainstream Party Support and Positions
For all elections after substantive fiscal adjustments, the vote share of government parties declined, often considerably.The combined decline can be as little as 1 percentage point (Finland in 1995) or as high as 30 percentage points (Ireland in 2011). 21A closer look also confirms that the mainstream parties in opposition presented a credible alternative for some, but not all, dissatisfied voters.In almost all cases, the mainstream opposition did not fully absorb the loss of support for government parties.For instance, the Finnish Social Democrats or the Portuguese Socialists, in opposition in 1995 and 2015, respectively, gained voters after austerity measures, but these gains were substantially lower than the decline in the combined vote share for government parties.

Non-mainstream Party Support
The difference between government party losses and mainstream opposition gains was absorbed by non-mainstream parties.All but one austerity episode was followed by an increase in the vote share of non-mainstream parties.Prior to the Great Recession, this increase varied between 2 percentage points in Finland to slightly over 5.5 percentage points in Sweden.After the Great Recession, this increase varied between 8 percentage points in Portugal and 14.5 percentage points in Spain.The type of non-mainstream parties that benefited from austerity varied widely.While in Finland, leftist fringe parties, such as the Social Democratic Party and the Left Alliance, experienced an increase in voter support, Sweden saw an increase in the support of the conservative Christian Democrats (Kd) and the leftist/communist Vänsterpartiet.After the Great Recession, we often see parties to the left of the mainstream parties winning votes, for example, Podemos (0 per cent to 12.7 per cent) in Spain, the Portuguese Bloco de Esquerda (5.4 per cent to 10.6 per cent) and the Irish Worker's Party.

Turnout
We also see an increase in abstention in five out of ten cases.The tendency in abstention after fiscal adjustments is generally in the expected direction, though less pronounced than for nonmainstream party vote.

Survey and Experiment Design
We now turn to providing micro-level evidence of the mechanisms that create the macro-level patterns described in the previous sections.In a survey experiment fielded in four countries, we examined voter reactions to different fiscal austerity proposals from the main centre-left and centre-right parties.This means that we can test which parties voters support if both mainstream parties advocate austerity.
Importantly, this experimental approach allows us to vary the positions of the main political parties towards austerity.In an observational study, assessing how the main parties' positions on austerity matter for vote choice is challenging as many party-and country-specific factors co-vary together with the policy option pursued.Indeed, a key counterfactual, namely, that either both mainstream parties or even just the mainstream left party oppose austerity, is arguably rarely observed.Moreover, an experiment allows us to present specific, clear policy stances by mainstream parties, so we can be sure that behaviour is driven by responses to the policy dimension  Source: Data on fiscal consolidation events from Alesina, Favero and Giavazzi (2019); other data from ParlGov and the CMP.
we manipulate.Moreover, as argued earlier, it is difficult to gather data on opposition party responses to government austerity programmes.We conducted original survey experiments in the UK, Germany, Portugal and Spain, cases chosen for multiple reasons. 22First, a survey in multiple countries allows us to uncover commonalities in responses across different contexts, enhancing our confidence in the generalizability of the results.Secondly, voters in the different countries were exposed to differing degrees of austerity in the recent past.Portugal and Spain experienced a debt crisis and received international bail-out packages; the UK was hit by a short-term financial crisis and long-term austerity, but was not bailed out internationally; and Germany has proved comparatively resilient.Thirdly, the countries also have different party systems, which creates variation in the types of contestation around austerity.In Spain and Germany, party-based opposition to austerity is strongest on the radical left (Podemos and Die Linke).In Portugal, the radical left is opposed to austerity, but so is, in a more moderate way, the Socialist Party.In the UK, Labour used to take a marketfriendly approach in previous decades but moved towards a strong anti-austerity stance during the Corbyn years.
We conducted an experiment embedded in population-based surveys.The surveys were implemented by respondi, making use of different country-specific online access panels.Quotas based on age and gender were implemented.The sample is restricted to voting-age nationals.In each country, we surveyed around 2,200 individuals. 23We presented respondents with hypothetical scenarios concerning policy proposals by the two main centre-left and centre-right parties listed in Table 3.In the experiment, we also varied which party was described as the government party and which as the main opposition party.We do this to be able to examine whether the credibility of the announcement among voters differs for government and opposition parties.
To illustrate the details of the experiment for the UK, we present a possible vignette in Table 4. Before respondents saw the vignette, they were informed that their country has experienced high fiscal deficits and rising public debt for several years.The vignettes then presented respondents with the following varying information: party in government, main opposition party and the policy proposal of each party.The treatment (policy proposal in the second row) could take one of the following two values: 'Cut spending and increase taxes' or 'Keep government spending stable without tax increases'.This yields four different combinations of policy proposals: (1) both parties want to cut spending; (2) the centre-left party keeps spending as is, while the centre-right party cuts spending; (3) the centre-right keeps spending as is, while the centre-left cuts spending; and (4) both parties keep spending as is.We focus on the comparison between the 'both cut' scenario and the two counterfactual scenarios (2 and 4), where either the mainstream left or both mainstream parties keep spending as is.We disregard the third option because it is unrealistic that the centre-right keeps spending as is, while the left cuts.However, we kept this option in the design of the experiment for the sake of completeness.
Each respondent was presented with threerandomly selectedvignettes out of the eight possible vignettes (four possible proposal vignettes × two possible government-opposition assignments, as outlined above).After each vignette, the respondents had to indicate which party they would vote for given the policy scenarios presented.Specifically, we asked: 'For which party would you vote in the next election?'This became our outcome variable. 24All political parties were included as response options, not just the two main parties.We also allowed respondents to state 'Would not vote' and 'Don't know'.Respondents' reactions to these vignettes indicate how the combination of party and policy proposal influenced their vote choice.

Results
Figure 3 presents the results of a multinomial logistic regression model using vote choice as the dependent variable and the policy treatment as independent variables.The baseline category is the scenario in which both parties propose to keep spending as it is.Here, we code nonmainstream parties as in the macro analysis, namely, as radical-right, radical-left and Green competitors.We always control for simple demographics (age, education, gender and income), as well as for party choice in the prior election and economic ideology.
As noted earlier, we also varied which of the two parties was described as the government party.In our analyses, this is interacted with the austerity treatment, and the results in Figure 3 average over the government and opposition conditions.We do so because we did not find any statistically significantly different reactions to policy proposals depending on whether the party was described as governing or not.While this is consistent with our earlier claim that voter responses to austerity proposals do not depend on who is in government, we do not see this as strong evidence, as it could just be that our government/opposition treatment was somewhat abstract in the context of a hypothetical scenario.We now turn to the key results.
The results support H1b, which stated that mainstream parties that support austerity receive less support.When only the centre-right cuts, while the centre-left promises to maintain current levels of spending, the centre-left does particularly well (and the centre-right particularly badly).However, the overall support for mainstream parties is similar when both parties propose maintaining current levels or when only the left does.This is easily visible when looking at the effects of the categories capturing 'other mainstream parties', 'other non-mainstream parties' and 'abstention' in Figure 3.
The results also support H2b: what leads voters to abandon the left without rewarding the right is a context in which both parties support cuts.Thus, Figure 3 shows that the vote for other parties increases in the 'both cut' scenario in all party systems, be they mainstream or nonmainstream parties.Abstention increases as well, albeit to a lesser extent.The effects of both parties promising austerity on voting for other parties and on abstention are comparable across countries.Additional analyses show that it is voters on both the left and right political extremes that are least likely to endorse mainstream parties in a scenario where both parties endorse cuts (see Figure A15 in the Online Appendix).
In terms of polarization, we can see that the vote for non-mainstream parties increases across the board, even if not significantly in Portugal.In other words, voters' reactions to the experiment lead to more polarized party landscapes.More detail is provided in the Online Appendix, which shows full results by country.There, we can see that in Germany, the left competitor party (Die Linke) benefits most if both main parties propose to cut spending (see Figure A9).The same pattern exists in Portugal (see Figure A12), where the leftist Bloco de Esquerda benefits most from a scenario in which both main parties propose to cut.In Spain, Ciudadanos and Podemos, two relatively new competitors, benefit from voters defecting from the main parties (see Figure A11).Overall, voters also move to comparatively centrist parties, such as the Liberal Democrats in the UK or Ciudadanos in Spain.While there is movement to the extremes, voters also defect to mainstream alternatives.Hence, the situation where both parties cut is particularly conducive to engendering changes in voter decision making, at least in terms of other party vote, non-mainstream vote and abstention.This scenario essentially emulates the situation that most parties faced in the aftermath of the Eurozone crisis or during any other prolonged economic and fiscal crisis.The finding is also in line with the results from the macro-level analysis.
Finally, to link the micro analysis back to our macro-level findings, we present aggregate-level results from each proposal type in Figure 4.For this, we treat each vignette type as a hypothetical election and present the aggregate results for our three macro-level outcome variables: the change in vote share of non-mainstream parties; abstention; and left-right polarization based on ParlGov-provided expert scores for parties.For non-mainstream party vote, we can see a strong increase in all countries, especially when both parties pledge to implement an austerity package.The effects on turnout are smaller, which coincides with the macro results.Polarization increases compared to the other two experimental scenarios across all four countries, with the most pronounced effects in Germany.The reduced level of polarization when only the right proposes to cut can be explained by the increased support for the mainstream left, which lowers polarization.Overall, the aggregated effects from the experimental analysis are consistent with the findings from our macro analyses and confirm that economic policy making, in particular, fiscal consolidation, has a significant impact on the party landscape in representative democracies.

Conclusion
In this article, we find that political polarization increases after mainstream political parties implement austerity.This is the case because dissatisfied voters who do not find a credible antiausterity alternative among mainstream parties turn towards smaller, often more radical, parties.Fiscal austerity, therefore, has important political consequences even if austerity does not lead to government breakdown (Arias and Stasavage 2019).Governments may convince a large fraction of voters that austerity is necessary (Barnes and Hicks 2018) and build a majority of voters who support fiscal adjustments (Bansak, Bechtel and Margalit 2021).However, austerity still reshuffles votes within the party system towards non-mainstream parties, which consolidates their political standing over time (Bischof and Wagner 2019).
These results have important consequences for our understanding of the impact of fiscal policy.Most existing work has examined how economic conditions, such as financial crises or trade shocks, which are partially beyond the control of governments, affect parties.We highlight instead how the policy choices made by the key political parties affect the stability of the party system.Greater polarization, in turn, inhibits the ability to build viable and stable coalition governments, and leads to more difficulties in putting together a coherent government policy agenda, ultimately increasing political instability and policy gridlock (Konstantinidis, Matakos and Mutlu-Eren 2019).Ironically, these political effects then undermine the effectiveness of fiscal institutions that aim at limiting public spending and keeping fiscal deficits low (Wehner 2010).
Our findings raise a number of follow-up questions for future research.Our argument rests on assumptions about the reasoning of voters when they choose non-mainstream parties in response to austerity that we cannot test in this article.There are multiple mechanisms that lead to the results that we found.For instance, voters can select non-mainstream parties because they like their economic policy positions, because they hope to induce a policy change among centre parties or simply for reasons of protest without clear expectations of how their vote would change economic policy.It is important to understand the motivations of voters better, how they perceive the economic 2

Figure 1 .
Figure 1.Average non-mainstream party vote, abstentions and polarization over time.

Figure 2 .
Figure 2. Predicted change in polarization over empirically observable degrees of austerity.Notes: Prediction based on Model 7 in Table 1.The solid line shows predicted change in polarization (left y-axis) with 90 per cent confidence intervals; bars show the distribution of fiscal consolidations (right y-axis); std.= standard deviation.

Notes:
Austerity = cumulated consolidation during electoral term; Δ Gov support = change in vote share of parties in government; Δ NMP = change in non-mainstream party vote share; Δ Abstention = change in turnout.* In Belgium, voting is compulsory, with little fluctuation in turnout.

Figure 3 .
Figure 3. Fiscal consolidation and voter flows.Notes: Multinomial logistic regression with vote choice as DV and treatment as IV.CR = centre right; CL = centre left; ms = mainstream.Three of four policy treatments shown.Baseline category: both parties propose to keep spending.Policy treatments interacted with whether the left or right party was said to be in government.Effects shown are averaged over these two conditions.For full results with all parties, see Figures A9 to A12.Point estimates with 95% confidence intervals.

Figure 4 .
Figure 4. Aggregate effects of treatment conditions.Notes: Figure shows 95% confidence intervals.Results for non-mainstream party vote and abstention based on logistic regression with standard errors clustered by respondent and treatment group as sole predictor; results for polarization based on 1000 simulations of multinomial logit results matched with expert survey party positions.Prediction for Spain based on six largest parties only.

Table 2 .
Legislative periods with cumulative austerity > 5 per cent of GDP(pre-2008 and post-2008 crisis)

Table 3 .
Main political parties per country