Voters, economic downturn, and labor market policy

Transcription

1 Voters, economic downturn, and labor market policy Flavia Fossati and Philipp Trein 1 Abstract This paper examines how voters reward and punish major left- and right-leaning parties if they govern together in the context of an economic crisis and of policies that aim to prevent the crisis s negative effects on the labor market. In the literature on voting behavior, researchers often analyze electoral behavior in cases where either a left or a right party dominates government, and thus, such studies compare left and right parties in government and opposition context. We use a unique dataset that analyzes voting behavior, economic downturn, and labor market policy in 299 German electoral districts under the condition that a major left party and right party share power. Our results show that under the abovementioned conditions, voters reward left parties for protective labor market policies but punish them for the economic downturn. Nevertheless, right parties in government suffer less electoral loss for economic slumps than left parties do. Acknowledgments Both authors contributed equally to this work. We thank Stefanie Walter and Ruth Beckmann, whose ideas contributed in many ways to the writing of this paper. We especially want to thank Ruth for creating the dataset with the contextual information on unemployment rates and short-time work in the constituencies. We thank Patrick Emmenegger, Rafael Lalive, the participants of the Labor Market Colloquium at the University of Lausanne, and the participants of the Social Policy Panel at the Swiss Political Science Association for very helpful feedback on previous versions of this paper. All remaining errors are ours. 1 University of Lausanne, Rue de la Mouline 28, 1015 Lausanne, Flavia.Fossati@unil.ch and JosefPhilipp.Trein@unil.ch 1

2 Introduction A very important question for political science research is how economic developments and labor market policies affect voters support for government. For instance, under which conditions do voters reward which party in government for economic growth and punish it for economic downturns? What kinds of policies lead voters to support which governing party? These questions are of enduring importance across the political science literature, as they provide important insights for political parties and politicians about policy feedback from the electorate. Researchers have demonstrated that if the economy weakens prior to an election, voters tend to punish governments electorally, whereas they reward governments for a favorable economic context or a beneficial labor market policy independent of the type of party in government (e.g., Downs, 1957; Key, 1966; Tufte, 1978; Fiorina, 1981; Bechtel and Hainmueller, 2011; Lewis-Beck et al., 2013a; Stevenson and Duch, 2013). Some scholars have underlined partisan differences, arguing that left parties profit from unfavorable economic contexts because they tend to pursue policies that compensate voters for the negative impact of economic slumps (Lewis-Beck and Bellucci, 1982; Broz, 2010; Wright, 2012). Others have held that voters prefer fiscally conservative policies during times of crisis (Durr, 1993; Sihvo and Uusitalo, 1995; Stevenson, 2001). Therefore, right parties should profit electorally from economic downturns, especially from economic crises (Barnes and Hicks, 2012; Kayser, 2009; De Neve, 2014; Lindgren and Vernby, 2016). These studies, however, have often analyzed voters decisions in a context where one major party (either left or right) governs alone or dominates government and the other main party is in the opposition. Thus, an important open question is how voters reward and punish major left and right parties if they govern together. Notably, do both parties suffer similarly from economic downturns on election day if they govern together? Do voters reward both parties in the same way for protective labor market policies, or does the electorate support and punish the two parties to different extents? 2

3 In this paper, we argue that once left and right parties govern together, voters reward left parties for protective labor market policies more so than they reward right parties. In other words, it is worthwhile electorally for left parties to implement insider-oriented labor market policies, such as short-time work, when they govern together with a right party in times of crisis. Meanwhile, voters punish both parties for an economic downturn rather than solely punishing the opposition. Nevertheless, the punishment effect is more severe for left parties than for right parties in government. Although voters reward left parties for anti-crisis policies, they punish left parties more than right parties for economic downturns. The reason is that voters concerns about the economy generally outweigh their appreciation for policies addressing the labor market consequences of crises. This finding implies that even if a left party in government takes action to support voters with protective labor market policies, overall, the electorate will turn to the right because voters have more trust in the competencies of right parties regarding the economy in general. To analyze this argument empirically requires a case of a major right and major left party governing together, for example, in a grand coalition. The 2009 German federal elections are a typical case for the discussed scenario because prior to the 2009 elections, the two largest parties the center-right Christian Democratic Union (CDU) and the center-left Social Democratic Party (SPD) had been governing together in a grand coalition. When the economic and financial crisis unfolded in late 2008, the government undertook a series of measures to counteract the crisis effect on the national economy; such measures included short-time work (STW) programs to avoid layoffs (Sacchi et al., 2011). These measures lasted well into the run-up of the 2009 national elections. Given the backdrop of a grand coalition combined with the signals of a sharp decline in the economy and a policy response aiming to protect labor market insiders, this unique setting allows us to posit a more nuanced argument about voters assessment of partisan differences. For our analysis, we use a dataset that compares voting behavior, economic downturn, and labor market policy in 299 German electoral districts and combine it with high-quality survey data from the German Longitudinal Election Study (GLES). 3

4 Our results demonstrate that left parties in government indeed profit electorally in districts with higher STW rates. This finding is not surprising, since the SPD ran the employment ministry, which was responsible for these policies. By contrast, voters punished both governing parties the CDU/CSU and the SPD for rising unemployment rates. Nevertheless, our results also demonstrate that voters punish the governing left party for the economic downturn more than they punish the right, and rising unemployment affects vote decisions more directly than STW policies, to the detriment of the SPD. Our results confirm previous research and show that left parties benefit from policies such as STW that protect labor market insiders (Rueda, 2006; Emmenegger et al. 2012), but such a benefit has little worth electorally if the economy declines simultaneously. Then, voters trust right parties more concerning the economy in general. Regarding the German case, our results further underline that the labor market liberalization in the early 2000s had negative electoral consequences for the SPD (Schwander and Manow, 2016), since voters in precarious labor conditions tend to punish government for economic slumps (Marx, 2016). With fewer workers in secure employment, the potential for electoral gains through STW is low for the SPD. Literature review and theory Reward and punishment of governing parties A very prominent strand of literature regarding economic voting has referred to a reward-punishment mechanism, according to which voters reward incumbents for positive economic developments and punish them for negative ones (Downs, 1957; Key, 1966; Tufte, 1978; Fiorina, 1981; MacKuen et al., 1992; Duch and Stevenson, 2008; Lewis-Beck et al., 2013a). This argument contends that when the economy weakens, rational voters evaluate how an economic downturn affects their economic situation. If their personal outcome is negative, voters will punish government on election day. Moreover, the literature has demonstrated that voters consider prospective elements when casting 4

5 their votes. In this case, voters consider whether a potential government will be able to handle economic slumps in the future rather than basing their voting decisions on past events (MacKuen et al., 1992, 606-7; Erikson et al., 2000). Nevertheless, the link between the actual economic situation and the voting decision is rarely direct. At the micro level, voters need to perform at least two cognitive links in this situation, mediating the connection between the economy and the vote (cf. Coleman, 1986). First, voters need to evaluate the effect of the economic context on their personal situation or the national economic situation (i.e., egoor socio-tropic reasoning) (Duch and Stevenson, 2008). Second, voters need to evaluate the government s performance and decide whether they want to hold it responsible for their personal economic situation. Therefore, they need to identify an incumbent party that could be clearly responsible for the negative economic situation (Fournier et al., 2003; Hobolt et al., 2013). These assumptions imply that higher STW rates and smaller increase of unemployment in a voter s surroundings should lead to less punishment for the parties in power compared to contexts with less STW and steeper increases in unemployment. This effect should be mediated by the assessment of one s personal economic situation and the evaluation of the government s performance. Partisan differences regarding economic and labor market policies The second theory that this paper refers to accounts for partisan differences regarding labor market and economic policy. In other words, voters consider some parties to be particularly competent and responsible concerning certain policies. Thus, parties focus on specific issues to signal strength and competence to those groups of voters they want to attract (Petrocik, 1996; Bélanger and Meguid, 2008). Hence, voters decide to support a given party because of its specific policy propositions (Lewis- Beck and Nadeau, 2011; Lewis-Beck et al., 2013b) or because they infer from past experiences or beliefs that a party is particularly competent in handling a specific policy challenge (Tilley and Hobolt, 5

6 2011). According to the literature on party choice and economic policy preferences, two competing arguments are possible regarding the link between party and policy preferences of voters. 1. First, according to one strand in the partisan literature, left parties tend to represent workers or working-class issues and are therefore most intent on passing protective labor market policy since employment and social policy constitute their core competency. Conversely, conservative parties tend to cater to their electorates by showing competence regarding the stability of prices and financial markets (Hibbs, 1977; Schmidt, 1982). Therefore, voters support left parties in government in the event of an economic downturn because they expect them to resolve unemployment issues (Lewis-Beck and Bellucci, 1982; Broz, 2010; Wright, 2012). This characterization applies to voters who are labor market insiders, especially in countries such as Germany with a strong labor dualization (Rueda, 2006). Therefore, left parties in government should profit from higher STW rates more so than right parties. 2. Second, scholars have held that voters prefer fiscally conservative policies during times of crisis (Durr, 1993; Sihvo and Uusitalo, 1995; Stevenson, 2001) because they engage in prospective reasoning (MacKuen et al., 1992, 606-7; Erikson et al., 2000) and worry about the future state of the economy and the health of public finances. Therefore, right parties should profit electorally from economic downturns (Barnes and Hicks, 2012; Kayser, 2009; De Neve, 2014; Lindgren and Vernby, 2016). According to this logic, voters should support right parties in government when unemployment rises and should punish left parties. As discussed previously, this effect should be mediated by voters assessment of their personal economic situation and their evaluation of the government s performance. The argument Based on these two strands of literature, we argue that voters punish both incumbent parties for economic downturns and reward them for protective labor market policies (compared to the 6

7 opposition) because they consider government responsible for a potential economic downturn even if the main left and right parties govern together. Nevertheless, when comparing both governing parties against one another, voters tend to reward the left party in government for protective labor market policies. However, the electorate also considers the downturn of the economy and the prospect of pursuing growth-oriented policies. Thus, the more the economy weakens, the better the public tends to perceive the right party in government. Although voters reward left parties for labor market policies, they tend to support the right in cases of economic decline especially in a severe crisis. Taken together, these considerations suggest that left parties in government do not profit from economic crisis electorally in general, although they benefit from specific labor market policies on election day. Relevance of the German case During the financial and economic crisis of , policymakers around the globe undertook actions to counteract the impacts of the global economic and financial crisis on national economies. For instance, policymakers passed measures to save banks and other financial institutions from collapse and implemented stimulus policies to re-start the economy (Trein and Braun, 2016; Braun and Trein, 2014). One particularly efficient labor market-related measure is STW, consisting of payments to companies allowing them to compensate workers for reduced working hours instead of laying them off during a macro-economic demand crisis. Governments all over Europe introduced employment promotion programs aimed at supporting workers and preventing the economic downturn from being reflected in labor market statistics. Nevertheless, Germany and other Bismarckian type welfare states, such as Austria and Italy, were prominent because the governments in all three countries increased STW rates more than the governments in other OECD countries. Consequently, especially in Germany, unemployment rates 7

8 remained stable throughout the demand crisis ( ), whereas STW increased considerably and reached the OECD-27 average in 2009, when most European countries had to address a recession (Figure 1). 2 < Figure 1: about here> In Germany, STW is a working time reduction and an income replacement scheme paid for by the national unemployment insurance. Employees have a claim to STW if there is a loss of earnings due to unforeseen circumstances, such as economic crises or natural catastrophes, and if the crisis affects at least one-third of the employees in a company. 3 Legally, STW is a situation in which the employer carries the responsibility for the halting of production and the work council must agree to introduce STW. 4 In severe economic downturns, STW can be extended. During the economic crisis, the German government included temporary agency workers in the STW scheme, extended the length of STW from 18 up to 24 months; the government loosened access rules for firms, halved social security contributions for firms or paid for them entirely (Sacchi et al., 2011, 27). These decisions were made by the Ministry of Employment under the leadership of the social democrats 5 to prevent increasing unemployment rates. 6 Therefore, either both governing parties or the SPD in particular, due to their direct involvement in decisions about STW, could have profited electorally from the STW decisions. This policy context is an interesting example for political science research, as elections to the German national parliament occurred in September 2009 only nine months after STW rates began to increase sharply (Sacchi et al., 2011, 31), and thus, STW likely had an electoral impact. In the 2009 elections, the SPD lost 11% of the votes compared to the 2005 elections 7, whereas the CDU suffered only a minor 2 Japan had the highest STW rates according to the OECD definition during the recent crisis period. The presented figures include very different types of STW regimes that vary across countries. 3 Code of Social Law III (Sozialgesetzbuch (SGB) Drittes Buch (III)), articles German Civil Code (Bürgerliches Gesetzbuch), article 615. Works Constitution Act (Betriebsverfassungsgesetz), article Spiegel Online, May 20, 2009: last accessed March 7, Die Zeit Online, June 16, 2010: last accessed March 7, Refer to Figure A1 in the appendix for the distribution of electoral gains and losses in the 299 districts for each party (left graph) and for the government parties and the average for the opposition parties (right graph). 8

9 reduction in their vote share. The research on electoral behavior in the 2009 elections agrees that the economy was a very salient issue (Zohlnhöfer, 2011; Saalfeld, 2011; Beckmann et al., 2011; Goerres and Walter, 2016); however, researchers have disputed how exactly the economy affected voting behavior. On the one hand, Rattinger and Steinbrecher (2011) find an influence of economic perceptions on vote intentions but emphasize the strong role of party identification. On the other hand, Anderson and Hecht (2012) do not find a consistent influence of the economic crisis and blame this behavior on the grand coalition of the CDU and the SPD. Debus et al. (2014) find economic voting only for one party (the CDU/CSU), and Marx (2016) notes that workers in precarious situations punish the government for economic decline. A recent contribution emphasizes that economic voting primarily affected the SPD and argues that the social democrats lost votes especially in regions where unemployment increased. This effect has been particularly strong among those identifying with the SPD (Trein et al., 2017). < Figure 2: about here> Against this backdrop, the generous STW policies may have incentivized voters to punish both governing parties less severely than they would have if more people had lost their jobs. In addition, the SPD in particular could have profited from higher STW rates, as it represents the traditional working class (Fossati and Häusermann, 2014), but it lost considerable electoral support due to its reforms liberalizing the German labor market (Schwander and Manow, 2016). Empirical implications The above-discussed argument has several empirical implications. First, at the macro level, as illustrated in Figure 3 (Link 1), our argument implies that higher STW rates in a voter s district, relative to contexts with less STW, should correlate with more votes for the governing coalition. According to 9

10 the partisan difference literature, higher STW should be beneficial especially for the SPD. Conversely, the CDU should suffer less from increasing unemployment. < Figure 3: Theoretical model > Second, we derive implications for the empirical analysis at the micro level (Coleman, 1986; Walter, 2010) to link STW to voters personal economic situation, their evaluation of government performance, and vote choice (Figure 3, Links 2-3). Thereby, a context with higher STW shares should lead to a more positive assessment of the personal economic situation of voters because of the lower danger of being laid off and thus suffering from the consequences of the economic crisis. This outcome, in turn, should be accompanied with a better evaluation of government performance overall. Consequently, better government performance should result in a vote that rewards both governing parties the CDU and the SPD. However, if we differentiate the positive effect of STW, it will translate into voters more favorable evaluation of the SPD, as left parties are appreciated for being more competent in labor market policy issues. Data and methods To test the above-discussed argument, we use a broad strategy that combines different empirical tests at the macro and micro levels. At the macro level, we analyze information from the two main elements of the German electoral process in different empirical analyses. In the elections to the federal German parliament, voters are asked to cast two votes. The first vote is for a candidate (who usually belongs to a party that also presents itself on the ballot), and the second vote is for a party. According to the design of the German electoral system, half of the national parliament seats go to the candidates who win the first vote in 10

11 their electoral districts according to a first-pass-the-post logic (majority logic). The other half of the seats are distributed according to the proportion of the vote share that the parties obtain in the second vote (proportional logic). 8 At the micro level, we undertake some analyses to assess the causal chain underlying the macro relationship. To this aim, we use cross-sectional survey data from the GLES (Rattinger et al., 2011). 9 We combine the pre-and post-election surveys into a single dataset. The surveys were conducted six weeks before and after the federal election in Germany on 27 September 2009, and a total of 4288 individuals were interviewed. 10 The distributions of election results are relatively close to the actual voting results. 11 Macro level models (Link 1) Proportional vote To analyze the proportional vote at the macro level, we measure the percentage change in both coalition parties electoral outcomes and the average outcome of the opposition parties (Greens, FDP and die Linke) compared to that in the previous election, which was held in 2005, in the 299 electoral districts as a dependent variable (Figure 3, Link 1). Estimating the effect of regional labor market characteristics and policy interventions on national elections is a plausible strategy because voters refer to their local economic situation when evaluating the government performance and attempting to maximize their self-interest. Nevertheless, to ensure that the results are not distorted by individuals assessment of the national economic situation, we control for socio-tropic reasoning. 8 For details on the German electoral system. read (Falter and Schoen, 2005). 9 For more information, see 10 The results are robust to restricting the sample to the pre- and post-election samples. 11 Table S1 in the supplementary material presents a comparison of the actual voting results and the sample distribution obtained from our data. 11

12 To determine whether labor market policy intervention affects voters choices at the aggregate level, we introduced a variable capturing the share of individuals who are newly registered as working on an STW contract as a percentage of all employees in a specific region between September 2008 and September Then, municipal-level data obtained from the Bundesagentur für Arbeit (2010) was merged for the corresponding electoral region. We test our argument at the macro level by running seemingly unrelated regressions to account for the correlations among the error terms. This step is necessary because the decision to vote for one party cannot be considered independently from the choice not to vote for another party. We analyze whether the incumbents are punished for bad labor market performance, and we exploit the large variation in regional unemployment rates in the twelve months preceding the 2009 federal elections (cf. Figure 2). For this purpose, we use the change in the unemployment rate (September 2008 September 2009) in the 299 constituencies (Bundesagentur für Arbeit, 2010). The raw data were at the municipal level (Landkreise) and were merged to the constituency level using the software ArcGIS. 12 Majority vote Beyond analyzing the effect of STW and unemployment changes on the outcomes of the proportional vote, we also include two sets of analyses to clarify the determinants of the first vote (majority vote). We first estimated logit models using two dichotomous dependent variables capturing whether i) the same person or ii) the same party was elected as in Second, to test for partisan differences, we analyze whether voters who vote for a party differing from the one they voted for in 2005 move to a left or to a right party. The preference changes were captured by means of two dichotomous variables 12 A comparison with labor market data that were available directly for constituencies with our unemployment data shows a correlation coefficient of 0.98, which suggests that the re-aggregation was correct. 12

13 as well as by an alternative coding that assigns (-1) to individuals moving to a left party, (0) to individuals with unchanged voting behavior, and (1) to individuals changing to a right party. In all macro level models, we control for the level of electoral participation, for the vote share that the incumbent parties reached in 2005, and for East Germany, as the labor market situation in East Germany can still be expected to differ from that in the West. 13 Macro-micro level models (Link 2 Link 4) Following previous research (Coleman, 1986; Walter, 2010), we test the micro-foundations of our argument by running a series of regression that operationalize the different analytical linkages connecting the macro to the micro level (cf. Figure 3). The context (Link 2) connects regional unemployment and the STW rate to the respondents evaluation of their personal economic situation. In turn, this ego-tropic evaluation should trigger the individual evaluation of government performance (Link 3). The last step involves the re-aggregation of micro-level preferences that are expected to impact the macro-level outcome. We test Link 4 by regressing individual vote choice on the respondents government evaluation. If all these relationships pass the statistical tests, our argument about the micro-level foundations that link labor market policy intervention and electoral outcomes can be corroborated. We use different dependent variables to test these analytical steps. First, we assess the impact of context factors on the individual economic situation (Link 2). As suggested by MacKuen (1992), we include retrospective (vn178), current (vn179) and prospective (vn181) ego-tropic thinking and 13 We also ran some robustness checks that included controlling for the absolute level of unemployment, price developments, the share of the elderly population, the development of new job openings, and the migration balance per 100 inhabitants. Furthermore, we control for the city of Dresden, where the vote in 2005 had to be postponed due to the death of a candidate from the extreme right party NPD, as well as for the Land of Bayern, where the main conservative party is the CSU, which collaborates with the CDU in federal elections. The results also remain stable when using average or relative outcomes rather than absolute gains/losses as dependent variable and remain stable whether total or average absolute outcomes are captured. 13

14 average the scores (0=very bad to 4=very good) for all three indicators. Subsequently, Link 3 is tested by regressing the continuous variable satisfaction with the government s performance (vn112) on the ego-tropic assessment, i.e., the previous dependent variable. For these models, the empirical strategy consists of estimating hierarchical linear models that take into account the nested data structure (Steenbergen and Jones, 2002; Rabe-Hesketh and Skrondal, 2008). Finally, concerning the electoral behavior (Link 4), only voters parties that passed the 5% hurdle were considered. These parties are the SPD, the CDU/CSU, the Greens, the FDP and die Linke. Moreover, we excluded non-voters, and thus, our sample contained 2545 observations 14. To gauge the electoral choice, we focus on the second vote (proportional rule) and combine the respondents intention to vote in the pre-election sample (v254_2a) and the reported vote choice in the post-election sample (n169_2a). We recoded two versions of this dependent variable. First, a variable distinguishing between the governing parties (SPD and CDU/CSU=1) and the opposition (die Linke, Greens, FDP=0) was created. Second, we recoded a nominal variable to test for the different preferences concerning the incumbent parties by distinguishing between votes cast for the SPD, the CDU, and the opposition. The models include a number of possible confounders that account for basic sociological differences among voters (Falter and Schoen, 2005). Precisely, the models control for gender (ref. male, v1), age (in years, v436b), and age squared to account for nonlinear effects of electoral preferences. Education is modeled as continuous (5 levels, vn9a), whereas personal unemployment (vn17) and union membership (v337) are included as dichotomous variables. We also include a measure of socio-tropic reasoning by creating an index that averages the assessment of the expected past (vn182), current (vn184), and future (vn185) development of the national economy. Furthermore, we distinguish individuals with very strong or strong party identification from individuals with average or lower levels 14 We keep the number of observations constant to ensure the comparability of the results. 14

15 of party identification (vn136). Moreover, we model several political variables, including the level of political interest (5 levels, vn217), the individual position on the left-right dimension (1=left, 11=right, vn190), and the square of the left-right position. Controls are also added for East Germany and for the data sampling strategy, i.e., pre- or post-election survey wave (recoded from Erhebung ). To complete the model, changes in unemployment and the STW rate at the macro level are added. We estimate multilevel logit regressions for the decision to vote either for the governing or for the left coalition and multinomial logit models with clustered standard errors for the choice among SPD, CDU and the opposition. Results Labor market policy intervention and electoral behavior: macro-level evidence Proportional vote The analysis of the proportional vote indicates that STW boosts the electoral performance of the governing coalition, while high levels of unemployment decrease its electoral fortune (see Table 1, Model 1 and Table S2 supplementary material for alternative specifications). Conversely, the opposition loses ground in electoral districts with higher STW incidence but gains in districts with high unemployment. Therefore, we corroborate the expectation that incumbents are rewarded for labor market policy intervention and blamed for bad economic developments. <Table 1 about here> To disentangle partisan-specific preferences, we model the performance of the two governing parties separately, as shown in Table 1, Model 2 (see also Table S3 in the supplementary materials). This model, which distinguishes between left and right parties in government and the opposition parties, suggests that labor market policy intervention does not significantly influence the electoral 15

16 performance of the conservative CDU/CSU but that the party clearly suffers electoral punishment in districts with higher unemployment. This result contradicts our expectation that right parties in government should profit in situations of crisis; rather, it suggests that voters punish them for a bad economic outlook, as expected by the reward-punishment model. As a member of the governing coalition, the SPD also loses electoral ground in districts with high unemployment. Nevertheless, the SPD is rewarded for labor market policy interventions, as it loses less electoral ground in districts with high STW. If we re-run the models shown in Table 1 for all parties separately 15, we find that left parties in the opposition, e.g., the Greens and die Linke, do not profit from higher STW. In contrast, these parties significantly lose ground in districts with higher levels of labor market policy intervention, indicating that voters identify the governing SPD rather than left parties in general as managing to avoid layoffs. Moreover, these models indicate that all three opposition parties gain votes in contexts with high unemployment 16 but lose in districts with higher STW (estimates are not always significant). Majority vote The analysis concerning the first vote (majority rule, see Table A1 in the appendix) tests whether increasing STW augments the re-election chances of either the candidate or the party that won the electoral district in We also analyze whether STW incidence and increasing unemployment trigger an electoral movement to the left or to the right at all. For the analyses concerning the proportional rule, we find that higher STW correlates with an increased likelihood of choosing the same person and particularly the same party, as in the 2005 elections. The implication is that higher STW 15 Analyses are not shown but are available upon request. 16 Again, if we disaggregate the analyses by party, we see that in particular, the Greens and FDP profit significantly from high unemployment, whereas the results for die Linke are positive but not always significant. 17 Incumbent persons or parties are members of either the CDU/CSU or the SPD; only in extremely rare instances could it concern members of different parties. 16

17 rates trigger a reward for the governing coalition. In contrast to the previous findings, increasing unemployment has a positive but not always statistically significant effect on the likelihood of voting for the same person or party. In line with partisan reward theory, the probability of moving from a left to a right party decreases significantly when STW rates are high (see Table S8, Models 1-3, supplementary material and Table S9, Models 1-3). Similarly, higher STW rates decrease the likelihood of switching from a right to a left party, but this result is non-significant 18 (Table S8, Models 4-6, supplementary material). In sum, the results of our macro-level analyses corroborate our argument that the governing coalition is generally rewarded for STW and punished for bad economic development. Hence, voters blame any incumbent party independent of its political orientation for the economic downturn. However, if we disentangle the partisan dynamics, a more nuanced picture emerges, indicating that only the SPD benefits from labor market policies notably, STW measures that benefit labor market insiders (Rueda, 2006) even in times of economic crisis. Labor market policy intervention and electoral behavior: testing the micro-foundations As suggested by Coleman (1986), ecological fallacies may arise when estimating macro models. It is thus important to ensure that macro-level phenomena are well grounded and result from individual behavior following the patterns suggested by the macro-level relationship (Walter, 2010). Therefore, to link STW to the macro-level electoral outcome discussed above, we test the contextual (Model 1, Table 2), individual (Model 2, Table 2) and aggregation links (Model 1, Table 3) with micro-level data. <Table 2 about here> 18 A robustness check with the alternative coding shows the same picture, but the result for STW is significant only at the 10% level (cf. Table A9, in the supplementary material). 17

18 As Table 2 shows, increasing STW in the district affects the evaluation of a respondent s economic situation significantly in the expected direction (Model 1 19 ). In fact, on average, the higher the share of STW is, the better an individual evaluates his/her personal economic situation. Conversely, an increase in unemployment at the macro level decreases one s satisfaction with the personal economic situation, as do individual level variables that capture labor market risk exposure, i.e., personal unemployment and old age. In terms of political variables, we find that individuals positioned on the right of the political spectrum are more satisfied with their economic situation than other individuals are. We obtain positive and significant coefficients also for individuals with high education levels, strong party identification and positive judgments of the state of the national economy (socio-tropic evaluation). Model 2 in Table 2 shows the findings for Link The coefficients suggest that, the better the evaluation of one s personal economic situation is, the more satisfied the voter is with government performance. Concerning the control variables, a higher STW incidence increases satisfaction with the government (10% significance level). Conversely, a higher increase in unemployment at the regional level and older age lower one s satisfaction with government performance. Moreover, we find that women, right-oriented individuals, and respondents who evaluate their own and the national economy s situation positively tend to support the government more strongly. <Table 3 about here> Finally, we test the re-aggregation link (Table 3 21 ). The coefficients indicate that respondents sharing a positive government evaluation are more likely to vote for both the SPD and the CDU/CSU than for the opposition parties (base outcome). In these models, however, the effect of STW is no longer significantly different for the CDU, SPD and opposition parties. Only increasing unemployment affects 19 For alternative specifications, see Table S4 in the supplementary material. 20 Refer to Table S5 in the supplementary materials for alternative specifications. 21 For alternative specifications, see Tables S6 and S7 in the supplementary materials. 18

19 the SPD significantly more negatively than it affects the opposition. Apart from the macro-level analysis, this result suggests that the SPD lost more votes in case of increasing unemployment and that it could not compensate for its electoral losses directly through STW policies. In sum, our analyses at the individual level indicate that STW affects vote choice indirectly through both the ego-tropic and government evaluation. Therefore, at the individual level, we do not find a direct effect of STW on voting. However, compared to the outcomes for the opposition, the electoral fortunes decline for the SPD but not for the CDU when unemployment increases (Table 3). Overall, our results suggest that both members of the governing coalition are punished directly and indirectly for a bad economic situation and that both benefit indirectly from labor market policy intervention. Conclusions This article addresses the question of whether and how STW policy and labor market performance affected the 2009 federal German elections, which coincided with the global economic recession. This election is a particularly interesting case because the two largest German parties the conservative Christian democratic CDU and the social democratic SPD formed a grand coalition in the period prior to the election. Thus, this constellation allows us to test hypotheses of economic voting for two different parties that are in government at the same time relative to the context of an external economic shock. Notably, this case permits us to analyze whether voters punish left or right parties for economic slumps and reward them for anti-crisis policies or whether the electorate treats both parties the same way. To guide our analysis, we hold that voters punish both governing parties for bad labor market results and reward them for labor market policy interventions in this case, STW rates compared to the opposition parties. Nevertheless, when comparing the two governing parties, voters reward social 19

20 democrats for STW policy and punish them for rising unemployment rates. Conversely, they do not reward the conservative party for increasing STW but do punish this party less for an economic downturn. We analyze these questions using different empirical tests at the macro (constituency) and individual levels. Therefore, we use a unique dataset that combines unemployment and STW rates in the 299 electoral regions with information from the German pre- and post-electoral survey. At first glance, our analyses of the secondary vote results at the constituency level shows that the governing parties, if considered together, profit from higher STW rates in the constituency. However, when we disaggregate this finding, our results suggest that only the SPD gains electorally in regions with high STW rates. In other words, the governing coalition benefits from STW, but the SPD receives the merit for this result. On the other hand, voters punish both governing parties for higher unemployment rates in the constituency. The results of the first vote at the macro level confirm this finding. Further empirical analyses that combine survey data with constituency-level data reveal a particularly interesting insight. Our results reveal that STW indirectly affects individuals voting intentions. Notably, voters personal economic situation and their evaluation of the government s performance mediate the effect of STW policy on voting decisions, but there is no direct link between high STW rates and support for either governing party. Contrariwise, local unemployment rates directly affects voting intentions, which is to the disadvantage of the social democrats that lose in districts with higher unemployment rates compared to the Christian democrats and the opposition. This finding implies that voters respond to the economic context rather than to anti-crisis policies in their ballots. The main contribution of this article is certainly its analysis of economic voting under the condition that the two main parties share the incumbency (although under a CDU chancellor). We show that right parties do not lose as much as left parties in the case of rising unemployment rates if both parties are in government simultaneously. Similarly, left parties profit from STW policies, but voters did not consider these policy efforts sufficiently to compensate for their worries about (slightly) increasing 20

21 unemployment rates. Against the backdrop of a coalition government, voters seem to favor right instead of left parties in times of economic downturns. Our results are in line with those researchers who suggest that voters reward left parties for labor market insider policies (Rueda, 2006; Fossati and Häusermann, 2014) but tend to favor right parties in times of economic downturns (Barnes and Hicks, 2012; Kayser, 2009; De Neve, 2014; Lindgren and Vernby, 2016). Overall, we demonstrate that left incumbents cannot compensate for losses in labor market downturns with labor market policy interventions against a right incumbent. In the German case, the history of Agenda 2010, which alienated voters from the social democrats (Schwander and Manow, 2016), clearly reinforced this dynamic. 21

22 References Anderson, Christopher J., and Jason D. Hecht "Voting when the Economy goes bad, Everyone is in charge, and no one is to blame: The Case of the 2009 German Election." Electoral Studies 31 (1):5-19. Barnes, Lucy, and Timothy Hicks "Left Behind? Partisan Politics after the Financial Crisis." Bechtel, Michael M, and Jens Hainmueller "How Lasting Is Voter Gratitude? An Analysis of the Short-and Long-Term Electoral Returns to Beneficial Policy." American Journal of Political Science 55 (4): Beckmann, R., P. Trein, and S. Walter "Dominanz der Ökonomie: Entscheidet die Wirtschaftslage Wahlen?" In Der unbekannte Wähler? Mythen und Fakten über das Wahlverhalten der Deutschen, ed. E. Bytzek and S. Roßteutscher. Fankfurt / New York: Campus. Bélanger, Éric, and Bonnie M Meguid "Issue Salience, Issue Ownership, and Issue-based Vote Choice." Electoral Studies 27 (3): Braun, Dietmar, and Philipp Trein "Federal Dynamics in Times of Economic and Financial Crisis." European Journal of Political Research 53 (4): Broz, J. Lawrence "Partisan Financial Cycles." In Politics in the New Hard Times: The Great Recession in Comparative Perspective, ed. M. Kahler and D. A. Lake. Ithaca and London: Cornell University Press. Bundesagentur für Arbeit "Arbeitsmarkt 2009." Nürnberg: Bundesagentur für Arbeit. Coleman, James S "Social Theory, Social Research, and a Theory of Action." American journal of Sociology 91 (6): De Neve, Jan-Emmanuel "Ideological Change and the Economics of Voting Behavior in the US, " Electoral Studies 34:

23 Debus, Marc, Mary Stegmaier, and Jale Tosun "Economic voting under coalition governments: Evidence from Germany." Political Science Research and Methods 2 (01): Downs, Anthony An Economic Theory of Democracy. New York: Harper and Row. Duch, Raymond, and Randolph Stevenson The Economic Vote. How Political and Economic Institutions Condition Election Results. Cambridge MA: Cambridge University Press. Durr, Robert H "What Moves Policy Sentiment?" American Political Science Review 87 (1): Emmenegger, Patrick, Häusermann, Silja, Palier, Bruno and Martin Seeleib-Kaiser, eds The Age of Dualization: The Changing Face of Inequality in Deindustrializing Societies. New York: Oxford University Press. Erikson, Robert S, Michael B MacKuen, and James A Stimson "Bankers or Peasants revisited: Economic Expectations and Presidential Approval." Electoral Studies 19 (2): Falter, Jürgen W., and Harald Schoen, eds Handbuch Wahlforschung. Wiesbaden: VS-Verlag. Fiorina, Morris Retrospective Voting in American National Elections. New Haven: Yale University Press. Fossati, Flavia, and Silja Häusermann "Social policy preferences and party choice in the 2011 Swiss elections." Swiss Political Science Review 20 (4): Fournier, Patrick, André Blais, Richard Nadeau, Elisabeth Gidengil, and Neil Nevitte "Issue Importance and Performance Voting." Political Behavior 25 (1): Goerres, Achim, and Stefanie Walter "The Political Consequences of National Crisis Management: Micro-Level Evidence from German Voters During the 2008/9 Global Economic Crisis." German Politics 25 (1): Hibbs, Douglas A. Jr "Political Parties and Macroeconomic Policy." American Political Science Review 71 (4):

24 Hobolt, Sara, James Tilley, and Susan Banducci "Clarity of Responsibility: How Government Cohesion Conditions Performance Voting." European journal of political research 52 (2): Kayser, Mark Andreas "Partisan waves: International business cycles and electoral choice." American Journal of Political Science 53 (4): Key, V.O The Responsible Electorate. New York: Vintage Books. Kiewiet, Roderick Macroeconomics and micropolitics: the electoral effects of economic issues. Chicago: University of Chicago Press. Lewis-Beck, Michael S, and Paolo Bellucci "Economic influences on legislative elections in multiparty systems: France and Italy." Political Behavior 4 (1): Lewis-Beck, Michael S., Nicholas F. Martini, and D. Roderick Kiewiet. 2013a. "The Nature of Economic Perceptions in Mass Publics." Electoral Studies 32 (3):1-48. Lewis-Beck, Michael S., Richard Nadeau, and Martial Foucault. 2013b. "The Complete Economic Voter: New Theory and British Evidence." British Journal of Political Science 43 (02): Lewis-Beck, Michael Steven, and Richard Nadeau "Economic Voting Theory: Testing New Dimensions." Electoral Studies 30 (2): Lindgren, Karl-Oskar, and Kåre Vernby "The electoral impact of the financial crisis: Evidence using district-level data." Electoral Studies 44: MacKuen, Michael, Robert Erikson, and James Stimson "Peasants or Bankers? The American Electorate and the U.S. Economy." American Political Science Review 86 (3): Marx, Paul "The Insider-outsider Divide and Economic Voting: Testing a New Theory with German Electoral Data." Socio-Economic Review 14 (1): Petrocik, John R "Issue Ownership in Presidential Elections, with a 1980 Case Study." American Journal of Political Science 40 (3): Rabe-Hesketh, Sophia, and Anders Skrondal Multilevel and Longitudinal Modeling using Stata. College Station: STATA Press. 24

25 Rattinger, Hans, Sigrid Roßteutscher, Rüdiger Schmitt-Beck, and Bernhard Weßels Zwischen Langeweile und Extremen. Die Bundestagswahl Baden-Baden: Nomos. Rattinger, Hans, and Markus Steinbrecher "Economic Voting in Times of Economic Crisis." German Politics 20 (1): Rattinger, Hans; Roßteutscher, Sigrid; Schmitt-Beck, Rüdiger; Weßels, Bernhard; Wolf, Christof; Wagner, Aiko; Giebler, Heiko; Bieber, Ina; Scherer, Philipp (2014): Pre- and Post-election Cross Section (Cumulation) (GLES 2013). GESIS Data Archive, Cologne. ZA5702 Data file Version 2.0.0, doi: / Rueda, David "Social Democracy and Active Labour-market Policies: Insiders, Outsiders and the Politics of Employment Promotion." British Journal of Political Science 36 (03): Saalfeld, Thomas "A Seemingly Boring Election amidst Economic Turmoil." German Politics 20 (1):1-11. Sacchi, Stefano, Federico Pancaldi, and Claudia Arisi "The Economic Crisis as a Trigger of Convergence? Short-time Work in Italy, Germany and Austria." Social Policy & Administration 45 (4): Schmidt, Manfred G Wohlfahrtsstaatliche Politik unter buergerlichen und sozialdemokratischen Regierungen: ein internationaler Vergleich. Frankfurt/Main: Campus-Verlag. Schwander, Hanna, and Philip Manow " Modernize and Die? German Social Democracy and the Electoral Consequences of the Agenda 2010." Socio-Economic Review:mww011. Sihvo, Tuire, and Hannu Uusitalo "Attitudes towards the welfare state have several dimensions." Scandinavian Journal of Social Welfare 4 (4): Steenbergen, Marco R., and Bradford S. Jones "Modelling Multilevel Datastructure." American Journal of Political Science 46 (2): Stevenson, Randolph T "The economy and policy mood: a fundamental dynamic of democratic politics?" American Journal of Political Science 45 (3):

26 Stevenson, Randolph T., and Raymond Duch "The Meaning and Use of Subjective Perceptions in Studies of Economic Voting." Electoral Studies 32 (2): Tilley, James, and Sara B. Hobolt "Is the Government to Blame? An Experimental Test of how Partisanship Shapes Perceptions of Performance and Responsibility." The Journal of Politics 73 (02): Trein, Philipp, Ruth Beckmann, and Stefanie Walter "German Voters in Times of Crisis: The Impact of Perceptions and Economic Context on Electoral Behaviour." German Politics:1-26. Trein, Philipp, and Dietmar Braun "How Do Fiscally Decentralized Federations Fare in Times of Crisis? Insights from Switzerland." Regional & Federal Studies 26 (2): Tufte, Edward R Political Control of the Economy. Princeton NJ: Princeton University Press. Walter, Stefanie "Globalization and the Welfare State: Testing the Microfoundations of the Compensation Hypothesis." International Studies Quarterly 54 (2): Wright, John R "Unemployment and the Democratic Electoral Advantage." American Political Science Review 106 (04): Zohlnhöfer, R "The 2009 Federal Election and the Economic Crisis." German Politics 20 (1):

27 Tables and Figures Figure 1: Short-time work and unemployment change, Germany and OECD average Unemployment Short-time work Germany Percent of fulltime employees OECD average 27

28 Figure 2: Short-time work and unemployment change in Germany Legend: Darker shades indicate higher increases in unemployment levels between September 2008 and September 2009 and higher levels of short-time work in July