Economic Crisis, Globalization, and Partisan Bias: Evidence from Spain

Economic Crisis, Globalization, and Partisan Bias:Evidence from Spain

Jos�e Fern�andez-Albertos

Institute for Public Goods and Policies, Spanish National Research Council (CSIC)

Alexander Kuo

Cornell University

and

Laia Balcells

Duke University

Who do citizens blame for the recent European economic crisis? In this paper, we test theories about blame attributionwith respect to the economic crisis. We argue that blame for the crisis is partially conditioned by partisan bias and fra-mings of the crisis as being related to globalization. We test the argument with new survey data and a survey experimentfrom Spain. In the experiment, respondents receive different framings of the economic crisis which are endorsed by dif-ferent political parties and non-partisan organizations. We obtain the following findings: (i) blame for who is responsiblefor the economic crisis is greatly affected by partisanship; (ii) making globalization as a cause of the crisis salient exoner-ates the government of blame, but only for co-partisans of the government; and (iii) citizens are willing to blame otherglobalization-related factors for the crisis, in particular, European governments and blame the domestic government less.The results expand our understanding of public opinion dynamics during major economic recessions and also suggestconditions under which “scapegoating” globalization can occur.

The economic crisis that has hit much of Europe betweenlate 2007 and continues as of 2013 has raised many theo-retically important and policy-relevant questions regard-ing the attitudes and preferences of citizens. Who docitizens blame for the crisis? Does globalization affectwho is to blame for the crisis? Since the onset of the cri-sis, average unemployment in the eurozone area has risento above 11% by in mid-2012, while the average growthrate between 2008 and 2011 was �0.1%. Many Europeancountries have been hard hit by the crisis, most promi-nently Estonia, Greece, Ireland, Italy, Latvia, Portugal,and Spain. Descriptive overviews of this “Great Reces-sion,” which for some countries is the worst economicdownturn since the early 1970s (and for some the worstcrisis since the Great Depression), conclude that thegrowth prospects of the most crisis-afflicted states remaingrim, that the debt crisis will likely worsen, and that fur-ther bank defaults or a more prolonged economic reces-sion could possibly lead to the breakup of the Eurozone(Eichengreen 2009; Hugh 2010; Hope 2012). The crisishas forced governments to make difficult decisions in anumber of domestic and international domains, includ-ing fiscal policy (austerity vs. expansion); support forEuropean Union (EU)-funded bailouts; default on debtsfrom foreign lenders; domestic banking reform; andreforms of labor market regulations.

In this paper, we address the question of who citizensblame for the crisis, and particularly how globalizationshapes public opinion regarding the economic crisis.Among political elites and within the general public,disagreement exists about the root causes of the eco-nomic crisis, as well as the policies that governments

should pursue in response.1 For example, some econo-mists argue that adoption of the euro has significantlyworsened the crisis (in particular for Spain and Ireland),while others argue, not necessarily mutually exclusively,that lax international capital regulations allowed domesticborrowers to borrow irresponsibly, fueling asset bubbleswhich ultimately crashed. Some analysts argue that thesedynamics were worsened by liberalizing policies under-taken by conservative political parties in the late 1990sand early 2000s. Others focus on the importance ofdomestic wage rigidities and fiscal irresponsibility asbeing roadblocks to the end of the crisis. The roles ofeach of these causes and the appropriate policy responsesremain disputed.

We design a study that leverages the complexity of thecrisis and the fact that both globalization-related anddomestic-oriented causes of the crisis have been invokedin the public debate. We seek to explain who or what citi-zens think is to blame for the economic crisis, and focusprimarily on the extent to which factors linked to global-ization might affect who citizens blame. The ongoing nat-ure of the crisis and salient public debate about theappropriate government responses make survey dataabout attitudes regarding the crisis especially valuable.

We argue that existing theories about the impact ofglobalization on domestic public opinion neglect theimportant role of partisan bias. The ongoing crisis allowsus to analyze how globalization shapes public opinionand how partisan bias and competing frames affect atti-tudes and blame. We theorize that partisan bias should

1 See a summary of discussion in an EU Commission report (2009).

Fern�andez-Albertos, Jos�e, Alexander Kuo and Laia Balcells. (2013) Economic Crisis, Globalization, and Partisan Bias: Evidence from Spain. International Studies Quarterly,doi: 10.1111/isqu.12081© 2013 International Studies Association

International Studies Quarterly (2013) 57, 804–816

affect which actors citizens hold responsible for or blamefor the crisis. In particular, co-partisans of the incumbentgovernment should be more willing to blame the actorsthat the incumbent government blames for the economiccrisis. By blaming other actors, domestic governments canavoid being electorally punished for poor economic per-formance. We further argue that globalization can playan important role in the politics of blame shifting.Recent literature argues that under conditions of higheconomic internationalization, voters punish incumbentgovernments less (Kayser 2007; Duch and Stevenson2010). We argue that one important mechanism for thisresult is that supporters of the incumbent governmentare more willing to divert blame toward global factors.This intuition is also relevant to the modern economicvoting literature, which assumes that voters make certainattributions of responsibility before rewarding or punish-ing incumbent governments.

To test these conjectures, we conducted an originalsurvey of citizens in Spain in autumn 2010. Spain is a par-ticularly useful test case of public attitudes regarding thecrisis, because of the depth of the economic recession,and the intense public division over different causes ofthe crisis. The crisis in Spain began in January 2008 aftera rapid fall of asset and housing prices, a crash that hadbeen preceded by the largest period of economic growthsince the democratic transition. As of 2012, Spain isamong one the worst-performing European economies,with a projected growth rate of negative 1.5% for thenext year and an unemployment rate of more than 22%.The causes of and the policy options to exit the crisishave been the most reported issues in the Spanish mediain the last several years, as evidenced by the massive pub-lic protests in May 2011, October 2011, and September2012. In November 2011, the government led by the left-leaning Spanish Socialist Workers’ Party (Partido SocialistaObrero Espa~nol [PSOE]) suffered its worst electoral defeatin history and was replaced by the right-leaning People’sParty (Partido Popular [PP]); the grim economic situationplayed a major role in both the public protests and theelection results. Despite the policy announced by theEuropean Central Bank (ECB) in September 2012 to pur-chase government debt directly, much uncertaintyremains over the stability of the euro currency union aswell as the long-term growth prospects of Spain.

To test our hypotheses about attitudes toward the eco-nomic crisis, we implemented an experiment embeddedin an Internet survey of 1,500 Spanish adult citizens, andgathered data on their attitudes regarding who is toblame for the economic crisis. We use an experimentaldesign to evaluate whether varying the partisan endorse-ment of different framings of the crisis affects who citizensthink is to blame for the crisis. In the survey experi-ment, citizens were randomly assigned to treatmentgroups, with each group reading a different set of state-ments regarding blame for the economic crisis. Respon-dents received statements about the cause of theeconomic crisis, which either blamed globalization ordomestic factors as a cause of the crisis. Some statementswere endorsed by different political parties or a non-par-tisan source. The random assignment of specificallyworded partisan endorsements of different explanationsof the crisis allows us to isolate the causal effects ofdifferent cues, and establish baseline effects of the typesof bias on blame attribution. After receiving certainstatements, respondents were then asked which actorsthey think are most responsible for the development of

the crisis. This design allows us to evaluate whethersimple interventions highlighting factors that have beenshown to affect preferences in other domains—partisanendorsement and framings—affect citizen views of theEuropean crisis.

We find evidence that the salience of globalization as acause of the crisis can affect public opinion regardingwho is to blame for the crisis and that this effect is condi-tioned by the partisan endorsement of the framing of thecrisis. In particular, individuals who identify as beingclose to the incumbent left-wing government (PSOE) aremore likely to blame other international actors—Euro-pean governments in particular—once globalization as acause of the crisis is made salient. We find no symmetri-cal effects for PP (opposition) supporters. Further, wefind that the results are most pronounced among citizenswho identify with the PSOE, not among individuals whohave left-wing ideological beliefs. Our findings demon-strate the importance of partisan bias and salience of glo-bal issues in explaining blame for the great recession,and provide one possible explanation for why economicvoting appears to occur less in more globalized econo-mies. The results also give us better substantive under-standing of citizen preferences during difficult economictimes.

The following section summarizes recent research onthe causes of the economic crisis and presents the theo-retical expectations. The next section presents the experi-mental design and methods of analysis. The penultimatesection presents the results. We conclude in the final sec-tion.

Literature Overview and Theoretical Expectations

Globalization, Economic Crisis, and Public Opinion

Little systematic research exists regarding the impact ofglobalization on public opinion and partisan preferences,in contrast to the voluminous literature explaining prefer-ences for globalization policies (such as trade, immigra-tion, or European integration). One strand of researchfocuses on how globalization affects policy preferences,using traditional factor endowment or factor mobility the-ory. This literature tests the argument that variation ineconomic openness creates winners and losers and thosecitizens accordingly express different redistribution pref-erences to compensate for the impact of globalization. Arepresentative example of this argument that uses surveydata as an empirical test is a recent study that finds thatamong lower educated citizens in OECD states (Walter2010), globalization increases risk perceptions anddemands for income redistribution; this finding is consis-tent with the classic compensation hypotheses about howtrade openness affects redistribution (Cameron 1978;Burgoon 2001; Scheve and Slaughter 2001; Adsera andBoix 2002; Walter 2010).

An additional strand of literature complements thisapproach and focuses on the electoral consequences ofglobalization, and examines specifically how globalizationmight affect the accountability of governments to theircitizens. According to this literature, higher levels ofopenness reduce the degree of economic voting in acountry. That is, incumbents are electorally punished orbenefited less during negative economic times in moreopen countries, as opposed to incumbents in more closedeconomies. In more closed economies, stronger evidenceof economic voting exists (Hellwig 2001; Fern�andez-Al-

Jos�e Fern�andez-Albertos, Alexander Kuo and Laia Balcells 805

bertos 2006; Duch and Stevenson 2010). However, differ-ent causal mechanisms for this finding remain untested.One possible mechanism is that citizens believe that glob-alization reduces the capacity of governments to effec-tively pursue certain policies, and are more likely toattribute negative outcomes to fluctuations in the worldeconomy than to national governments.2 Another mecha-nism is that citizens who support the incumbent govern-ment and believe that economic outcomes are related toglobalization have more reason to exonerate the govern-ment for undesirable outcomes (Kayser 2007; Hellwigand Samuels 2007).

This literature suggests that globalization can affectpolitical preferences, and should be particularly relevantduring difficult economic times, as factors related toglobalization might be blamed for outcomes. Theseeffects should be expected to be particularly strong in acrisis context, in which the international aspects of theeconomic crisis (the degree of interconnectedness ofnational financial markets, global capital flows, and therole of private and public international actors) have beenespecially salient. However, few studies address theimpact of globalization on citizen blame of the economiccrisis. Little research explores whether factors related toglobalization are in fact blamed for negative outcomesand whether incumbent governments benefit from theblame of global as opposed to domestic factors. Thesegaps in the international political economy literature areimportant because public vulnerability to biases in blame(for example, partisan bias, or “scapegoating” otheractors or causes) can explain variation in subsequent sup-port for incumbent governments, political parties, or pol-icies related to the process of globalization in itself.These policies could include trade, immigration, and cap-ital mobility.

Recent literature on citizen blame in other domainssuggests that blame of the crisis should be affected by thesalience of issues as well as partisan bias. Some studiesidentify the importance of partisan bias in who citizensblame for various outcomes in both economic and non-economic contexts (Arceneaux 2003; Malhotra and Kuo2008).3 The standard operationalization of partisan biasis the effect of party cues or party endorsements on citi-zen preferences. Party cues have been found to affect awide range of political attitudes and assessments of politi-cal candidates and policies in the US context (Laymanand Carsey 2002; Cohen 2003; Achen and Bartels 2006;Ansolabehere, Hirano, Snyder, and Ueda 2006; Malhotraand Margalit 2010). But few studies assess the role ofparty cues and partisan bias on policy views in Europeanstates (Ray 2003; Hobolt 2007; Tilley and Hobolt 2011).4

Given the complex, multicausal nature of current Euro-pean economic crisis, we expect these partisan biases toplay an important role in how both international anddomestic factors are related to who citizens think isresponsible for the crisis. The political consequences ofblame during economic recessions remain overlooked in

the political science literature; few studies test for therole of partisan bias and blame for the ongoing recession.Understanding public opinion dynamics in economic cri-ses is particularly important, as attitudes formed duringrecessions can be persistent over lifetimes (Giuliano andSpilimbergo 2009). The crisis provides an opportunity tounderstand the circumstances under which globalizationis likely to be “scapegoated.”

Theoretical Expectations and Relevance of Spanish Context

We delineate some basic expectations about the connec-tion between globalization, partisan bias, and blame forthe economic crisis. We focus on a set of factors ofblame that can be classified as “domestic” causes of thecrisis and those related to “globalization.” First, wehypothesize that citizens who share the partisan affilia-tion of the incumbent government are more likely toblame factors for the crisis that the government alsoblames. Given the previous discussion, we expect that cit-izens of the same party of the incumbent governmentwill be more likely to blame globalization factors for thecrisis when they are made salient. These hypotheses arespecified more precisely after describing our estimationstrategy. In the results section, we also describe and testseveral alternative hypotheses regarding blame of theeconomic crisis.

Before describing the specific research design, we brieflydiscuss the particular relevance of the theoretical frame-work to the Spanish case, one of the worst hit countries bythe current economic crisis. Within Spain, there has beenstrong public disagreement about the cause of the crisisand what policies should be pursued.5 Globalization-related as well as domestic policy issues have been dis-cussed as reasons for the crisis. Some pundits and politicalelites have argued that under-regulated international capi-tal flows and integration with the European Union helpeddrive the housing price bubble; others have focused on theimportance of imprudent government policies and restric-tive domestic labor regulations in preventing Spain fromrecovering from the crisis (Andr�es et al., 2009; De Grauwe,2012).

An important political feature of the economic crisis inSpain is that both of the major political parties haveblamed both global (such as the actions of foreign govern-ments, supranational institutions, and international inves-tors) and domestic factors (such as labor marketregulations) as causes of the crisis.6 Representatives fromboth parties have blamed the actions of other Europeangovernments and international capital flows for facilitatingthe asset bubble which was the proximate cause of the eco-nomic crash, and have blamed European governments forinterfering in domestic banking affairs and for “unfair”banking stress tests, exacerbating investor concerns. Repre-sentatives and legislators from both of the main partieshave also blamed international ratings agencies for unfairlydowngrading Spanish debt (Mallet 2010). Similarly, both

2 See Kayser (2007) for a review. Hellwig’s account (2011) is one of thefew that systematically tests the possibility of domestic blame of globalizationas opposed to domestic governments for economic outcomes.

3 For an overview of the social–psychological literature on blame assess-ment, responsibility attribution, and scapegoating, see Shaver (1985).

4 Malhotra and Kuo (2008) argue that in the context of many non-eco-nomic policies, party cues do not necessarily provide more meaningful infor-mation to the voter; in other contexts, party cues act as a “heuristic” that canprovide meaningful information (Lupia and McCubbins 1998).

5 Division exists over whether the government should pursue austeritymeasures, liberalize labor markets, and support increased authority of EUinstitutions. As representative examples, see the debate between the two “man-ifestos” of economists advocating different labor market responses to the crisis(Andr�es, Bentolila, Dolado, and Felgueroso 2009; G�omez 2009).

6 For the PSOE, see a representative discussion in a report from theprominent PSOE-affiliated IDEAS Foundation (2009); for a representativeexample of PP views on globalization and domestic-based explanations for thecrisis, see FAES (2008)

806 Crisis, Globalization, and Partisan Bias

the PP and PSOE have pushed for labor market flexibiliza-tion as a solution to Spain’s employment woes. The salientdisagreements about the causes of the crisis, and the plausi-bility that these different causes and policy interventions(involving globalization and domestic policy issues) can bereasonably attributed to both parties, allows us to use anexperimental design that isolates each of these factors—partisan endorsement and framings of the crisis—andobserve whether partisan endorsement of different causesof the crisis affects citizen blame.

Research Design and Methods

To test the above hypotheses regarding the impact of thepartisan endorsement of different framings (global ordomestic) of the crisis on blame, we administered an In-ternet survey to 1,500 respondents in October andNovember 2010. The survey was administered by Net-quest, a Spanish survey firm. The resulting sample hassimilar demographic composition to large nationally rep-resentative surveys in Spain.7 Respondents were randomlyassigned to a treatment group; the treatment consisted ofreading a short simple statement about the causes of theeconomic crisis; this statement differed across treatmentgroups. After reading the statement, they went on toanswer a question about blame for the crisis. Demo-graphic survey data were collected at the conclusion ofthe survey.

Treatments

Respondents were randomly assigned to one of sevengroups, one control and six treatment groups. The treat-ments vary along two dimensions: the endorser of the state-ment about the crisis and the framing of the statementabout the crisis. The endorser could be from a non-parti-san source, the then-incumbent ideologically left-wingPSOE political party, or the opposition ideologicallyright-wing PP political party. The statement about the cri-sis was about the role of globalization or domestic labor mar-ket regulations affecting the crisis. The treatment statementtemplate is below, with “/” indicating different text foreach treatment group.

We would like to present to you some conclusions thathave been made about the current economic crisis.According to a report made by (non-partisan experts/an institution linked to the Socialist Party (PSOE)/aninstitution linked to the Popular Party (PP), (a seriesof international factors related to globalization/regula-tions in the Spanish labor market) have played a majorrole in the development of the economic crisis thatSpain is undergoing now.8

Table 1 displays the experimental design and classifica-tion of each treatment. In the control group, the respon-dents read no statement about the crisis and simplyanswered a set of questions about blame and policy ques-tions. In Treatments 2–4, the respondents read the state-ment about globalization as a cause of the crisis. InTreatment 2, this statement was endorsed by non-partisanexperts; in Treatment 3 by the PSOE; and in Treatment4 by the PP. In Treatments 5–7, respondents read thestatement about labor market regulations as a cause ofthe crisis. In Treatment 5, this statement was endorsed bynon-partisan experts; in Treatment 6 by the PSOE; andin Treatment 7 by the PP. As the above discussion sum-marizes, each treatment is credible in the sense that polit-ical parties in Spain have been affiliated with both typesof statements regarding the crisis.

We also code partisanship and demographic character-istics of the respondents; we code all characteristics asbinary unless otherwise indicated. Individuals whorespond as being closest to the PSOE out of a list of tenpolitical parties in Spain are coded 1, indicating he orshe is a PSOE supporter and 0 otherwise. The same pro-cedure is used for coding PP supporters. We also codepolitical ideology on a seven-point scale of “extreme left,left, center-left, center, center-right, right, extreme right.”We use a binary indicator for “left-wing ideology” if therespondent identifies as either “extreme left, left, center-left,” and 0 otherwise. Education is coded on a four-pointscale where 0 indicates non-completion of primaryschool, 1 indicates completion of primary school, 2 indi-cates completion of secondary school, 3 indicates comple-tion of tertiary schooling. Income is coded by decile.Gender is coded as 1 for female. We code respondent’s“insider” labor market status as 1 if the worker has a for-mal labor contract and 0 otherwise.

Dependent Variable

Following the above experimental treatment, we askedrespondents which factors they blame most for the crisis.All respondents had to rank three out of six factors thatthey blame the most for the development of the eco-nomic crisis. The question was:

Thinking about the current economic crisis, we wouldlike to know which are, according to you, the threemost important factors that have contributed to the cri-sis among the following six, ranking them from mostimportant to least important: The actions of the cur-rent Spanish government; the actions of Spanishbanks; the entry of Spain in the euro; the actions ofEuropean governments; the actions of internationalinvestors; the regulations in the Spanish labor market.

TABLE 1. Experimental Design

Experimentalgroup

Informationtreatment

Partisantreatment

Probability ofreceiving treatment

Control group None None .4Group 2 Globalization Non-partisan .1Group 3 Globalization PSOE endorsed .1Group 4 Globalization PP endorsed .1Group 5 Domestic Non-partisan .1Group 6 Domestic PSOE endorsed .1Group 7 Domestic PP endorsed .1

7 Appendix 1 gives an overview of Netquest’s stratification and samplingstrategy. Appendix 2 compares the samples on the relevant social and demo-graphic variables and shows no statistically significant differences in the distri-butions of these variables between the surveys.

8 The Spanish translation of the statement is: “Ahora le querr�ıamos pres-entar algunas conclusiones que se han hecho recientemente sobre la actualcrisis econ�omica en Espa~na. Seg�un un informe de “una instituci�on asociada alPartido Popular (PP)/el Partido Socialista/un grupo de expertos independi-entes”, “una serie de factores internacionales relacionados con la globalizac-i�on/las regulaciones en el mercado de trabajo espa~nol” han jugado un papelcrucial en el desarrollo de la actual crisis econ�omica que atraviesa Espa~na.”Spanish translations of the survey questions are available upon request.Appendix 3 demonstrates the randomization process was successful.


Respondents chose three out of the six to rank with 1being most to blame, 2 next, and 3 third most blameworthy;the rest of the response options are unranked.9 Three ofthe factors deal with non-domestic causes (the euro,foreign investors, and European governments), andthree factors are explicitly about domestic actors (theincumbent government, domestic banks, and labor marketregulations).

Methods

In this section, we describe the methods used for analy-sis. The appropriate model where the dependent vari-able data are ranked items is the rank-ordered logisticor “exploded logit” model (Beggs, Cardell, and Haus-man 1981; Allison and Christakis 1994).10 This modelallows us to assess which variables affect the respon-dents’ ranking of items relative to other items (forexample, blaming the government more so than domes-tic banks). We then report the likelihood that therespondent ranks a particular item higher than theother items, which are set as the baseline. Allison andChristakis (1994) and Malhotra and Kuo (2008) providea relevant application of the model; we draw on summa-ries of the model here and the latter’s notation (p.125). The rank-ordered logit model is a generalizationof conditional/multinomial logit models; the modelrelates how the characteristics of the choosers and theitems affect the likelihood of the items being selected.The process of ranking is an aggregation of these indi-vidual choices. In this study, the items are the individualitems linked to the crisis as listed above. Ranked datacan be statistically modeled by combining a set of condi-tional logit models to create the rank-ordered logitmodel. The likelihood of the rank-ordered logit modelis the product of the likelihoods of individual condi-tional logit models.

Consistent with previous applications of this model, weassume that the data are derived from a random utilitymodel where BRij represents the latent blame respondenti has for item j (out of J items). We observe Yij, which isthe blame ranking respondent i assigns to item j.11

Although BRij is unobserved, we assume that respondent iranks item j more blameworthy than item k if BRij > BRik.Each BRij is modeled as having a systematic component(lij) and a random component (eij). We estimate varia-tions of a model that account for characteristics of theblame items as well as the partisan affiliation of therespondents. We first estimate a model where we assesswhether the fact that an item is related to domesticcauses of the crisis (the government, domestic banks, andlabor market regulations) is more likely to be rankedhigher than global/non-domestic items (the actions ofother European governments; Spanish membership inthe Eurozone; and foreign investors). This equation takesthe form of:

BRij ¼ b1Dj þ b2ðDj � SiÞ þ b3ðDj � T2iÞ þ b4ðDj � Si

� T2iÞ þ b5ðDj � T3iÞ þ b6ðDj � Si � T3iÞ þ b7ðDj

� T4iÞ þ b8ðDj � Si � T4iÞ þ b9ðDj � T5iÞ þ b10ðDj � Si � T5iÞ þ b11ðDj � T6iÞ þ b12ðDj � Si � T6iÞþ b13ðDj � T7iÞ þ b14ðDj � Si � T7iÞ þ eij

ð1Þwhere Dj is a dummy variable representing whether theitem j is one of the three domestic items, Si is a dummyvariable representing whether the individual respondent iis closest to the PSOE, and T2i is the treatment variablefor Treatment 2; T3 is the treatment variable for Treat-ment 3, and so on.12 The condition variables are indexedby respondent because the treatment was administered atthe individual level and does not vary by the blame items.Dj is indexed by item because the status of being adomestic or global issue is the same for each respondent.

The substantive interpretation of the coefficients inequation (1) is as follows. b1 and b1 + b2 represent thelikelihood of PP and PSOE supporters, respectively, inthe control group ranking a domestic issue more blame-worthy than the global issues. b3 and b3 + b4 representthe likelihood of PP and PSOE supporters, respectively,in the Treatment 2 condition, ranking a domestic item asmore blameworthy than a global item, as compared tothe control group. b5 and b5 + b6 represent the likeli-hood of PP and PSOE supporters, respectively, in theTreatment 3 condition, ranking a domestic item as moreblameworthy than a global item, as compared to the con-trol group. The same interpretation of these coefficientsfollows for each treatment group: b7 and b7 + b8 forTreatment 4; b9 and b9 + b10 for Treatment 5; b11 andb11 + b12 for Treatment 6 condition; and finally b13 andb13 + b14 for Treatment 7. The substantive impact on thelikelihood of ranking an item higher than another isfound by taking the exponential of the coefficient.

In terms of the theoretical expectations laid out in theprevious section, we expect b1 to be positive and b1 + b2to be zero or negative. That is, we expect that PP support-ers should be more likely to blame domestic-related issuesthan globalization-related issues, where PSOE supportersshould either blame them equally or blame globalizationitems more. We expect b5 + b6 to be negative; that is,once presented with another set of global actors that are

9 We asked respondents only to rank the top three factors, as this proce-dure would likely provide a clearer sense of who respondents found moreresponsible.

10 We also estimated logistic regression models in which the dependentvariable is whether an item is the most blamed factor or whether it is amongthe top three most blamed factors. The results, available upon request, areequivalent to those reported here.

11 This notation is used in Malhotra and Kuo (2008:125) and follows frompreviously cited works. We coded blame to lie between 1 (least blame) and 3(most blame).

12 We do not include the terms Si, and Ti because they do not vary acrosschoices within individuals. The estimating equation must include a variablethat distinguishes within respondents; thus, some aspects of the ranked itemsmust be included for each respondent characteristic. This allows one to assessthe conditional impact of respondent characteristics (including assignment ofa treatment) on the likelihood of ranking one item higher than another(blaming one item more than another). We can also estimate equation (1)for each treatment group; the sample for each estimation is the control groupand the treatment group of interest, for respondents who identify with eitherthe PSOE or PP. We can then compare across specifications which treatmentsproduce statistically significant changes in blame rankings of domestic vs. glo-bal items. Equation (1) implies the likelihood, Li, of a respondent rankingitem k as more blameworthy than item j. Let dijk = 1 if Yik � Yij and 0 ifYik < Yij. Then,

Li ¼YJ

j¼1

expðlijÞPJk¼1

dijk expðlijÞ

26664

37775: ð3Þ


made salient, PSOE supporters should be less likely toblame domestic factors for the crisis.

We also estimate alternative specifications of the rank-ordered logit model that allow us to more directly assesswhich individual characteristics and treatments are morelikely to make respondents blame specific items more thanothers. We present an example specification where weexamine blame of each item relative to a baseline item ofblame of European governments. This means we must esti-mate a model that includes all binary indicators of theother five items, interacted with partisan affiliation, relativeto the baseline. The example specification takes the form:

BRij ¼ b1Govj þ b2ðGovj � SiÞ þ b3ðGovj � T2iÞ þ b4ðGovj� Si � T2iÞ þ b5Banj þ b6ðBanj � SiÞ þ b7ðBanj � T2iÞ þ b8ðBanj � Si � T2iÞ þ b9Eurjþ b10ðEurj � SiÞ þ b11ðEurj � T2iÞ þ b12ðEurj�Si � T2iÞ þ b13Forj þ b14ðForj � SiÞ þ b15ðForj � T2iÞþ b16ðForj � Si � T2iÞ þ b17Labj þ b18ðLabj � SiÞþ b19ðLabj � T2iÞ þ b20ðLabj � Si � T2iÞ þ eij

ð2Þwhere the variables Govj, Banj, Eurj, Forj, Labj are binary

indicator variables for the items of the domestic govern-ment actions, domestic banks, Eurozone membership,international investors, and labor market regulations,respectively. The sample for this estimation equation isrespondents in the control group and those in conditionof Treatment 2. The baseline is thus the binary indicatorof the blame item of European government actions. Theinterpretation of b1, b1 + b2, b3, and b3 + b4, so on, issimilar to that of equation (1); only instead of the impactof domestic vs. global issues on respondent rankings, wecan evaluate whether each respondent ranks each item(domestic government actions, domestic banks, Eurozonemembership, international investors, and labor marketregulations) as more blameworthy relative to the item ofthe actions of European governments. Further, as withequation (1), we can estimate the impact of being a PPvs. PSOE supporter and the impact of Treatment 2on the likelihood of ranking each item in equation (2)as more blameworthy than that of European govern-ments.13 Equation (2) can also be estimated by includingall treatment categories (T2 through T7), to evaluate theimpact of each treatment on blame. For tractability andease of reading purposes, we display the results from esti-mation of equation (2) in Table 5 where the baseline isthe item of European governments, and include all treat-ments (T2 through T7).

Results

We begin with discussion of the control group. We firstpresent descriptive statistics and then examine determi-nants of blame within the control group. After discussionof the results for the control group, we turn to the experi-mental results on how a framing of the crisis based on theroles played by globalization vs. domestic factors affectscitizen views of the crisis. We find evidence of partisan dif-ferences regarding blame of the crisis, and that makingglobalization salient helps exonerate the government for

PSOE voters, and increases blame of European govern-ments and foreign investors (with much of the blame shift-ing toward other European governments). We show thatthis result is confined mainly to PSOE supporters.

Evidence from the Control Group

Demographics of the Control GroupWe briefly summarize the descriptive statistics from thecontrol group and discuss the baseline effects and resultsfrom the control group. Table 2 presents descriptive sta-tistics of the entire sample of respondents. Consistentwith results from other nationally representative samples,a slight majority of respondents identify with one of thetwo main political parties: 30% of the control group listthemselves as closest to the Socialist Party (PSOE) and23% of the group lists themselves as closest to the conser-vatives (PP).

Blame of the Crisis OverallWhich factors did respondents rank as most to blame forthe economic crisis? Figure 1 displays the patterns ofblame of the crisis for the control group and decomposedby partisan affiliation, and demonstrates the important rolethat partisanship plays in people’s assessment of the crisis.

TABLE 2. Descriptive Statistics of Sample

Variable N Mean SD Min Max

Proximity to PP 1,501 0.25 0.43 0 1Proximity to PSOE 1,501 0.29 0.45 0 1Left Ideology 1,454 0.53 0.50 0 1Income (decile) 1,431 6.28 2.66 1 10Female 1,501 0.51 0.50 0 1Age 1,501 40.10 13.13 18 78Education level (0 no

education, 1primary, 2secondary, 3 tertiary)

1,488 2.40 0.69 0 3

Labor Market Insider 1,501 0.44 0.50 0 1

78

61

73

37

16

2733 34 34

65

37

5651

93

67

36

54

40

020

4060

8010

0

Domes

tic B

anks

EU gov

ernm

ents

Euroz

one

Foreig

n In

vesto

rs

Gover

nmen

t

Labo

r reg

ulatio

ns

PSOE partisans PP partisans All respondents

FIG 1. Who is to Blame? Percentage Ranking Factor as Top 3, byPartisan ID (Notes. Bars refer to percentages ranking factor as top

three to blame for the crisis, for the control group.)

13 Equation (2) can be estimated with a different blame item as a baselineto evaluate the likelihood of blaming one item relative to another baselineitem. The substantive results do not change.


Two-thirds of respondents (67%) blame the incumbentgovernment as a top three factor explaining the crisis. Sev-enty-three percent of respondents blame the actions ofdomestic banks as a top three factor of the respondent.14

More than half of the respondents (56%) put the actionsof foreign investors as a top three factor, whereas Spain’sadoption of the euro, the actions of other European gov-ernments, and labor regulations were less blamed (33%,26%, and 40%, respectively). In terms of the factor thatwas put as the top factor contributing to the crisis, theactions of the government, domestic banks, and foreigninvestors were the most blamed (28%, 27%, and 27%,respectively).

Blame of the Crisis by PartisanshipAs Figure 1 shows, PSOE and PP supporters differ greatlyin their views on the crisis. Partisanship greatly distin-guishes who respondents blame for the crisis for all factorsof blame, except for the factor of Spanish membership inthe Eurozone. Overall, PP supporters are more likely toblame the government and labor market regulationsmost, followed by domestic banks. PSOE partisans aremore likely to blame domestic banks, the actions of Euro-pean governments, and foreign investors. They are willingto blame the government, but far less so than PP support-ers. For PSOE supporters, the most blamed combination(domestic banks, European governments, and foreigninvestors) does not include the government.

Specifically, about half of PSOE partisans (51%) listedthe government as one of the top three factors for the cri-sis, whereas 93% of PP partisans listed the government asa top three factor (p < .001). Seventy-eight percent ofPSOE partisans blame the banks as a top three factor, asopposed to 61% of PP partisans (p < .001). Sixty-five per-cent of PSOE partisans blame the actions of foreign inves-tors as opposed to 36% of PP partisans. Thirty-sevenpercent of PSOE partisans blame the actions of Europeangovernments, as opposed to 16% of PP supporters(p < .05). Regarding domestic labor regulations, only36% of PSOE supporters list them as a top three factor,whereas more than half (54%) of PP supporters do so(p < .05). The only factor for which partisan differencesdo not exist is blame of the euro. If we examine how parti-sanship affects the top ranked factor that is blamed, thepatterns are even starker. Only 12% of PSOE supporterslist the government as most blameworthy, in comparisonwith 61% of PP supporters (p < .001). The difference inviews by partisanship on banks’ responsibilities is larger aswell: 37% of PSOE supporters put the banks as the top fac-tor to blame, in contrast to only 11% of PP supporters(p < .001). Thirty-three percent of PSOE supportersblamed the actions of foreign investors most, in contrastto 12% of PP supporters (p < .001).

These partisan differences remain when controlling forthe demographic variables of education, income, age,gender, and labor market status. As Table 3 shows, parti-san identity best explains which factor is blamed. Thetable displays logistic estimations where the dependentvariable is whether each blame item is ranked as a topthree factor for the crisis; we estimate separate equationswhere the binary party identification variable is comparedto the baseline of the rest of the respondents in the con-trol group. For each dependent variable, with the excep-

TABLE3.

Logistic

Reg

ressionsExp

lainingBlameofEachFactor,ControlGroup

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

Governmen

tDomesticban

ksEurozone

Foreigninvestors

Europeango

vernmen

tsLab

orregu

lations

Inco

me

0.00

24(0.039

)�0

.011

(0.040

)0.05

4(0.041

)0.06

1(0.041

)�0

.11(0.038

)**

�0.11(0.039

)**

0.06

6(0.037

)*0.08

1(0.038

)**

0.03

8(0.041

)0.03

9(0.041

)�0

.011

(0.037

)�0

.020

(0.037

)

Education

0.02

8(0.15)

0.03

3(0.15)

�0.11(0.16)

�0.11(0.16)

�0.38(0.15)**

�0.37(0.15)**

0.39

(0.14)**

0.40

(0.15)**

�0.097

(0.16)

�0.11(0.16)

0.09

6(0.14)

0.09

5(0.14)

Gen

der

0.02

8(0.19)

�0.11(0.19)

�0.19(0.20)

�0.14(0.20)

0.45

(0.19)**

0.45

(0.19)**

�0.27(0.18)

�0.19(0.18)

�0.057

(0.20)

�0.002

3(0.20)

0.00

56(0.17)

�0.050

(0.18)

Age

�0.003

4(0.007

3)�0

.001

3(0.007

4)�0

.005

1(0.007

6)�0

.006

3(0.007

7)�0

.014

(0.007

4)*

�0.014

(0.007

4)*

0.02

3(0.007

1)**

0.02

3(0.007

3)**

�0.005

7(0.007

7)�0

.006

4(0.007

7)�0

.002

6(0.006

9)�0

.001

7(0.006

9)

Insider

0.20

(0.19)

0.29

(0.20)

0.10

0(0.20)

0.05

8(0.21)

�0.15(0.19)

�0.16(0.20)

�0.13(0.18)

�0.22(0.19)

�0.33(0.21)

�0.34(0.21)

0.07

0(0.18)

0.12

(0.18)

Proximity

toPSO

E

�1.04(0.19)**

0.41

(0.22)*

�0.12(0.20)

0.55

(0.19)**

0.67

(0.20)**

�0.23(0.19)

Proximity

toPP

2.30

(0.36)**

�0.71(0.22)**

0.01

3(0.22)

�1.20(0.22)**

�0.68(0.26)**

0.76

(0.20)**

Constan

t1.03

(0.48)**

0.33

(0.47)

1.13

(0.50)**

1.43

(0.50)**

1.34

(0.48)**

1.29

(0.47)**

�1.99(0.47)**

�1.66(0.46)**

�0.90(0.50)*

�0.51(0.49)

�0.42(0.45)

�0.65(0.45)

N56

456

456

456

456

456

456

456

456

456

456

456

4

Pseudo-R

2.043

.095

.011

.022

.056

.055

.047

.078

.022

.017

.003

.020

Notes.Stan

darderrors

inparen

theses.

*p<.10,

**p<.05.

14 Note that these differences do not reveal whether respondents consis-tently ranked banks as more blameworthy than the government, only thatmore respondents placed the banks as a top three factors.


tion of the euro, the coefficients for the binary variables ofPSOE and PP affiliation are statistically significant. Thecoefficient for PSOE supporter is positively correlated withblame of the domestic banks, foreign investors, and otherEuropean governments. PSOE individuals are less likely toblame the incumbent government and labor market regu-lations. PP supporters, consistent with the bivariate cross-tabulations reported above, are more likely to blame thegovernment and labor market regulations and less likely toblame domestic banks, international investors, and Euro-pean governments.

Overall, partisan affiliation matters much more thanother demographic characteristics in explaining blamefor the crisis. Wealthier and more educated respondentsare less likely to blame the euro and more likely to blameforeign investors. In sum, from the control group, welearn that PSOE supporters blame the following factorsmost: banks; foreign investors; and European govern-ments. They are much less likely to blame the govern-ment (though some PSOE supporters still do). Bycontrast, PP individuals are much more likely to blamethe government and labor market regulations and lesslikely to blame the banks, international investors, andEuropean governments.15 The euro is the only issuewhere partisan identity does not affect whether it isblamed as a top three factor.

Experimental Results

We now turn to the results of the estimations of equa-tions (1) and (2), examining whether partisan identifica-tion affects the blame of some causes of the crisis overothers and whether the experimental treatments affectedblame. Table 4 (equation 1) and Table 5 (equation 2)show the results of this analysis. Model 1 in Table 4 pre-sents a baseline result to illustrate the importance of parti-san bias. If we consider the sample of respondents whosupport either the PSOE or PP within the control group,across both groups of respondents, domestic issues relativeto globalization-related issues are 1.9 times more likely tobe ranked higher (b1 = 64, p < .001). But supporters ofthe PSOE and PP individuals blame the domestic and glo-bal issues differently. PSOE respondents are 1.3 times aslikely to rank domestic issues as more blameworthy thanglobal issues (b1 + b2 = 0.25, p < .005), whereas PPrespondents are 3.3 times as likely to rank domestic issuesas more blameworthy (b1 = 1.19, p < .001). This partisangap is unsurprising, given the discussion of Figure 1 andthe higher blame that PP respondents assign to the role ofthe government and labor market regulations.

Model 2 of Table 4 presents the results from estimat-ing equation (1). We examine whether there is partisanbias in blaming domestic vs. global factors and whetherthe treatments affect such blame. We present results forjust PSOE or PP supporters. The most substantivelyimportant result is the impact of Treatment 3 on blamerankings. Treatment 3 (the PSOE endorsement of global-ization as a cause of the crisis) has a statistically signifi-cant and substantively large impact on PSOErespondents’ likelihood of ranking domestic issues asmore blameworthy than global issues, relative to the con-trol group. As discussed above, within the control group,PP partisans blame domestic issues more so than global

TABLE 4. Results of Rank-Ordered Estimations: Domestic vs. Global Items

(1) (2) (3)Control group Equation (1) Equation (1) (with ideology)

Domestic 1.19 (0.11)** 1.19 (0.11)** 0.90 (0.08)**Domestic 9 PSOE proximity �0.94 (0.14)** �0.94 (0.14)**Domestic 9 Left �0.61 (0.10)**Domestic 9 Treatment 2 0.18 (0.23) 0.33 (0.17)*Domestic 9 Treatment 2 9 PSOE �0.060 (0.30)Domestic 9 Treatment 2 9 Left �0.16 (0.23)Domestic 9 Treatment 3 0.25 (0.24) 0.21 (0.18)Domestic 9 Treatment 3 9 PSOE �0.71 (0.31)**Domestic 9 Treatment 3 9 Left �0.43 (0.23)*Domestic 9 Treatment 4 0.11 (0.24) 0.17 (0.16)Domestic 9 Treatment 4 9 PSOE �0.27 (0.31)Domestic 9 Treatment 4 9 Left �0.20 (0.23)Domestic 9 Treatment 5 �0.11 (0.26) �0.019 (0.18)Domestic 9 Treatment 5 9 PSOE 0.20 (0.33)Domestic 9 Treatment 5 9 Left 0.18 (0.24)Domestic 9 Treatment 6 0.51 (0.26)** 0.39 (0.17)**Domestic 9 Treatment 6 9 PSOE �0.40 (0.33)Domestic 9 Treatment 6 9 Left �0.21 (0.23)Domestic 9 Treatment 7 0.27 (0.26) 0.49 (0.19)**Domestic 9 Treatment 7 9 PSOE �0.028 (0.34)Domestic 9 Treatment 7 9 Left �0.37 (0.25)N 1,920 4,782 8,634Pseudo-R2 .045 .058 .042Log-likelihood �1463.72 �3594.35 �6598.78LR chi-square 136.56 (2 df)** 442.55 (14 df)** 580.85 (14 df)**

Notes. Standard errors in parentheses.*p < .10, **p < .05.

15 A small proportion of left-wing voters identify as their closest preferredparty a minor party such as United Left, ERC, BNG, or Nafarroa Bai(although such respondents may still vote for the PSOE). Given the low num-ber of such respondents in the sample, the estimates for these respondentsare less precise. Within the control group, these respondents have similarblame attitudes as those who identify with the PSOE (results available uponrequest). We focus on the contrast between PSOE and PP supporters as thoseparties constitute the largest political parties.


issues, relative to PSOE partisans. But once respondentsreceive a PSOE endorsement as globalization as the causeof the crisis, PSOE respondents are more likely to rankdomestic issues as less blameworthy than global issues, rel-ative to the control group. In fact, they are roughly 1.6times less likely to blame domestic issues (b5 + b6 =�0.47, p < .02). Consistent with our theoretical expecta-tion, there is no symmetric effect for PP partisans regard-ing the impact of Treatment 4 on PP supporters;b7 = 0.11, p < .48) (Treatment 4 is the PP endorsementof globalization as a cause of the crisis). The endorse-ment of globalization as a cause of the crisis by the PSOEhelps exonerate domestic factors for the crisis, but onlyfor those who share the party label of the government.However, citizens who affiliate with the opposition partyare not affected by this particular endorsement. Model 2also shows that PP supporters tend to blame domestic fac-tors more when labor market regulations are made sali-

ent by the PSOE (Treatment 6) (b11 = 0.51, p < .05).This effect does not exist for PSOE supporters(b11 + b12 = 0.11, p < .59). This evidence indicates thatwhen presented with intra-party endorsement of alterna-tive explanations for the crisis, PSOE supporters are will-ing to shift blame to global actors, and blame domesticfactors less. But to what specifically do these respondentsshift their blame?

Blame of Specific ItemsWe now turn to estimations that allow us to assess whichindividual characteristics and treatments are more likelyto make respondents blame specific items higher thanothers. Figure 2 displays a key experimental result, com-paring the raw percentages of PSOE and PP respondentswho include the government among the top three blamefactors in the control group, vs. those percentages underthe globalization treatment conditions. PSOE supportersare far more likely than PP supporters to respond to theglobalization treatment: Among the former, blame of thegovernment is the top three falls from 51% in the controlgroup to 32% in the treatment group.

To analyze this result more systematically, we examineblame of each factor relative to a baseline of blame ofEuropean governments, equation (2). Table 5 presentsthe results. For PSOE respondents, exposure to Treat-ment 3—PSOE endorsement of globalization as a causeof the crisis—reduces their blame of the domestic govern-ment relative to European governments. PSOE respon-dents are more likely to blame the actions of Europeangovernments when receiving the statement endorsed bythe PSOE that globalization factors are responsible forthe economic crisis. b1 is positive (2.95, p < .001), indicat-ing PP respondents are 19 times more likely to blame thegovernment more so than European governments. PSOEsupporters are far less likely to blame the government rel-ative to European governments (1.5 times as likely)(b1 + b2 = 0.38, p < .02). Treatment 3—exposure toPSOE endorsement of globalization as a cause of the cri-sis—further reduces blame of the PSOE-led governmentrelative to European governments, among PSOE support-ing respondents. PSOE respondents in this treatment

TABLE 5. Results of Rank-Ordered Estimations (Equation 2)

Item (baseline category: European governments)

Government Banks Euro Foreign investors Labor market

Item 2.95 (0.25)** 1.58 (0.24)** 0.79 (0.25)** 0.86 (0.25)** 1.32 (0.24)**Item* PSOE proximity �2.57 (0.29)** �0.29 (0.28) �0.87 (0.31)** 0.052 (0.30) �1.34 (0.30)**Item 9 Treatment 2 �0.39 (0.48) 0.23 (0.47) �0.38 (0.52) �0.13 (0.50) 0.0063 (0.47)Item 9 Treatment 2 9 PSOE 0.42 (0.61) �0.10 (0.59) 0.59 (0.66) �0.12 (0.62) 0.11 (0.61)Item 9 Treatment 3 0.44 (0.60) 0.91 (0.59) 0.056 (0.65) 0.76 (0.61) 0.59 (0.59)Item 9 Treatment 3 9 PSOE �1.25 (0.71)* �1.36 (0.68)** �0.37 (0.76) �0.60 (0.70) �1.15 (0.72)Item 9 Treatment 4 0.15 (0.54) 0.41 (0.51) �0.26 (0.58) 0.30 (0.54) 0.011 (0.52)Item 9 Treatment 4 9 PSOE �1.00 (0.65) �0.75 (0.61) �0.37 (0.70) �0.82 (0.64) �0.51 (0.64)Item 9 Treatment 5 0.31 (0.68) 0.60 (0.67) 0.56 (0.70) 0.86 (0.69) 0.39 (0.68)Item 9 Treatment 5 9 PSOE �0.14 (0.77) �0.70 (0.76) �1.00 (0.84) �0.75 (0.78) �0.35 (0.79)Item 9 Treatment 6 1.36 (0.70)* 0.74 (0.66) 0.33 (0.71) 0.66 (0.69) 1.18 (0.66)*Item 9 Treatment 6 9 PSOE �1.40 (0.78)* �0.98 (0.75) �0.54 (0.82) �1.22 (0.77) �1.36 (0.76)*Item 9 Treatment 7 0.048 (0.57) 0.43 (0.55) �0.0099 (0.60) �0.087 (0.60) 0.15 (0.56)Item 9 Treatment 7 9 PSOE 0.37 (0.69) �0.28 (0.68) 0.17 (0.75) �0.14 (0.72) �0.12 (0.71)N 4,782Pseudo-R2 .14Log-likelihood �3273.862LR chi-square (70 df) 1083.54**

Notes. Standard errors in parentheses.*p < .10, **p < .05.

32

51

9093

020

4060

8010

0

PSOE partisans PP partisans

PSOE/glob treat Control group PP/glob treat Control group

FIG 2. Effect of Partisan Globalization Treatment on GovernmentBlame, by Partisan ID. Bars refer to percentages ranking the govern-ment as a top three factor to blame for the crisis, for the control

group and partisan blame of globalization treatments.


condition are 1.6 times less likely than in the controlgroup of blaming the government over European govern-ments (b5 + b6 = �0.47, p < .03). We do not find sub-stantive effects of the treatment on changing the blameof the other items relative to European governments.

We now examine whether the effects of the treatmentare driven by political ideology, as opposed to party affili-ation. Some literature suggests that more left-wing ideo-logical citizens are more skeptical of globalization; thiscould be because of the belief that globalization harmsworkers or that it constrains left-wing governments’ policyagendas (Kayser 2004; Fern�andez-Albertos 2006). We testthis account by estimating equations (1) and (2) andinteracting the treatments with left-wing ideology of therespondent (as opposed to party affiliation). The resultsare displayed in the third column of Tables 4 and 6. Inboth models, the results demonstrate that the PSOEendorsement of framings of the crisis has a greaterimpact on blame for PSOE respondents than for justideologically left-wing respondents. Left-wing respondentsare less likely to blame domestic issues when exposed toTreatment 3, but the effect is much smaller than forPSOE respondents.16 This indicates that the impact ofthe treatment more likely operates through partisan affili-ation and less through ideological beliefs about the roleof globalization.17

Conclusions

Who do citizens blame for the recent European eco-nomic crisis? What impact does globalization have on citi-zen views of the economic crisis? This paper addressesthese questions and provides evidence on how globaliza-tion can be politically used to shift blame during harsheconomic times. The recent substantial economic hard-ships already suffered in many European economies havetheir roots in domestic as well as global factors, and thereremains much debate about the causes as well as the bestpolicy responses. Understanding variation in blame forthe crisis is an important foundational task for explainingcross-national variation in responses to the crisis.

In this paper, we test theories about the impact of global-ization on blame for the economic crisis, with evidencefrom the important case of Spain. We argue that blame forthe crisis is partially conditioned by partisan bias and fra-mings of the crisis as being related to globalization, and wefind that different framings about causes of the crisis,endorsed by different political parties, do affect who peo-ple blame for the crisis. Consistent with recent literaturethat documents the impact of partisan bias on policy pref-erences, we show that partisan attachments help explainpeople’s views about the economic crisis. More interestingly,these attachments matter more if globalization-related expla-nations of the crisis are made salient. Our findings contrib-ute to the literature on the political consequences of glob-alization by showing how the salience of the issue can affectviews of economic crises, and how partisan bias might con-dition the use of globalization as a blame-shifting device.

Specifically, we find that partisan attachments explain agreat deal of who people think is responsible for the crisis:While supporters of the conservative party in the oppositiontend to blame the government, those who identify with theincumbent party tend to blame foreign investors anddomestic banks instead. The incumbent party endorsementof globalization as a cause of the crisis affects the party’s sup-porters’ views of the crisis: They help exonerate the govern-ment, and blame other European governments more. Thissuggests that it is not difficult, given the large magnitude ofthe effects and the minimal nature of the treatment, to

TABLE 6. Effect of Treatments on Blame of Individual Items by Political Ideology

Item (baseline category: European governments)

Government Banks Euro Foreign investors Labor market

Item 2.33 (0.17)** 1.61 (0.17)** 0.78 (0.18)** 1.12 (0.17)** 1.02 (0.17)**Item* Left �1.85 (0.21)** �0.20 (0.21) �0.80 (0.23)** �0.097 (0.21) �0.97 (0.22)**Item 9 Treatment 2 0.046 (0.37) 0.30 (0.37) �0.18 (0.40) �0.17 (0.38) 0.27 (0.37)Item 9 Treatment 2 9 Left 0.34 (0.48) �0.22 (0.47) 0.57 (0.52) 0.078 (0.49) �0.019 (0.49)Item 9 Treatment 3 �0.26 (0.36) 0.12 (0.35) �0.54 (0.40) �0.22 (0.37) 0.0036 (0.36)Item 9 Treatment 3 9 Left �0.16 (0.45) �0.44 (0.43) 0.22 (0.50) 0.35 (0.45) �0.26 (0.46)Item 9 Treatment 4 �0.17 (0.34) 0.21 (0.34) �0.23 (0.37) �0.24 (0.36) �0.079 (0.35)Item 9 Treatment 4 9 Left �0.22 (0.44) �0.38 (0.43) �0.29 (0.49) �0.063 (0.45) �0.31 (0.46)Item 9 Treatment 5 0.083 (0.40) 0.12 (0.40) 0.040 (0.42) 0.092 (0.41) �0.18 (0.42)Item 9 Treatment 5 9 Left 0.14 (0.49) �0.21 (0.48) �0.31 (0.54) �0.089 (0.50) 0.37 (0.52)Item 9 Treatment 6 0.60 (0.42) 0.70 (0.42)* �0.0012 (0.47) 0.50 (0.43) �1.08 (0.50)**Item 9 Treatment 6 9 Left �0.56 (0.50) �1.08 (0.49)** �0.31 (0.55) �1.08 (0.50)** 0.73 (0.42)*Item 9 Treatment 7 0.58 (0.43) 0.63 (0.42) 0.047 (0.47) 0.13 (0.44) 0.60 (0.43)Item 9 Treatment 7 9 Left �0.61 (0.52) �0.73 (0.51) �0.15 (0.57) �0.51 (0.53) �0.56 (0.53)N 8,634Pseudo-R2 .11Log-likelihood �6111.44LR chi-square (70 df) 1552.52**

Notes.. Standard errors in parentheses.* p < .10, ** p < .05.

16 Across the entire sample, 49% of ideologically left-wing respondentsidentify with the PSOE. The results do not substantively change if we consideralternate codings of partisan identification; nor do they changed based onideology if we use alternative binary classifications of “left-wing” to include the“center” category.

17 Additional suggestive evidence of the relative importance of partisan-ship over ideology comes from the analysis of the treatment effects for non-PSOE left-wing partisans, that is, respondents whose closest party is a smallleft-wing party such as United Left, or ERC, BNG or Nafarroa Bai (resultsavailable upon request). The estimated effect of the treatment for suchrespondents is similarly signed to that of PSOE identifiers, but the standarderrors are much larger (this could be due to smaller numbers of such respon-dents). The “exonerating” effect that the PSOE-endorsed globalization treat-ment has for PSOE partisans is not statistically significant for non-PSOE left-party supporters.


induce a specific subset of citizens to “scapegoat” globaliza-tion in difficult economic times. These results are drivenmore by partisan affiliation of citizens than ideological views.

Additionally, these findings have implications for under-standing the burgeoning literature on how globalizationattenuates the economic vote. This study suggests one plau-sible mechanism for this finding: Partisan bias leads votersto blame globalization-related factors more for economiccrises than the actions of incumbent governments. We findthis effect mainly for co-partisans of the incumbent govern-ment. Our findings suggest that for some citizens, incitingprotectionist views may not be difficult and that such viewscan help reduce the culpability of the government.

The theory, design, and results raise additional ques-tions for further research. Extrapolation of the resultsbeyond Spain is necessarily speculative, but the blame ofglobalization effect found in this study should exist inother countries and may persist as the crisis worsens. Oneof the scope conditions could be absolute exposure of astate to globalization. In smaller and more open econo-mies, the potential for blaming globalization (and subse-quent electoral support for the incumbent government)could be expected to be even higher than in a relativelylarger economy such as Spain.

The experimental results could also depend on the ideo-logical orientation or partisan affiliation of the incumbentgovernment. More moderate or right-leaning governmentscould plausibly benefit from shifting blame to globaliza-tion-related factors and shifting blame away from incum-bent government action (at least among co-partisancitizens). The specific aspects of globalization to blamemight differ depending on the ideological orientation ofthe political party. More Euro-skeptic parties may blamespecific European countries (such as Germany) or Euro-pean institutions (such as the ECB); other parties mightblame international financial actors, capital flows, or immi-gration. Our conjecture is further supported by the factthat in Europe, since the onset of the recession, many right-wing parties and governments have attempted to shiftblame for the crisis toward other actors, very much as thePSOE government did in Spain. The crisis has been themost important cause of government turnover in Europe(either by elections or by the collapse of parliamentary sup-port). In Great Britain in 2010 and Portugal in 2011, previ-ously moderate to left-wing governments were replaced,while in Ireland the long-ruling centrist Fianna Fail partysuffered its worst electoral defeat. In Greece, the NewDemocracy party won the elections in June 2012, replacinga previous coalition government that was formed becauseof turmoil related to austerity demands; the left-wingincumbent party PASOK suffered its worst defeat in theseelections. In Italy, the long-serving Prime Minister Silvio Ber-lusconi was ousted after failure to resolve the sovereign debtcrisis. In each of these cases, incumbent political parties—even those considered more right-wing—partially blamedglobal actors for the dire economic consequences.18 As the

economic crisis continues and further unpopular austeritymeasures in countries are debated, blame of global forcescould increase. Recent empirical analysis of elections in thecrisis (through mid-2011) indicates that incumbent govern-ments were electorally punished for lower economic out-comes, regardless of government ideology or which partywas in control (Bartels 2011). This evidence suggests that asthe European economic crisis deepens, we may observeboth left-wing and right-wing parties seeking to shift blameto factors related to globalization (such as European gov-ernments, international investors, or the Eurozone gener-ally) to avoid electoral punishment for the crisis. Ourfindings support the contention that globalization may be aconvenient target for such governments.

Our results also indicate that replicating the experi-mental design in other countries with incumbent govern-ments of other ideologies would be a useful extension.Another set of studies should examine why some globalactors are targeted for blame and not others (such asEuropean governments as opposed to foreign investors),and whether more specific framings of the crisis affectblame. Regardless of the ultimate outcome of the Euro-pean crisis, the blame of globalization (fairly or unfairly),as well as the consequences of such blame and backlash,will become more important questions for scholars ofinternational political economy.

Acknowledgments

The authors thank the Spanish National Research Coun-cil (CSIC) (project number: 200910I070), and the Span-ish Ministry of Science and Innovation (projectsCSO2009-12620 and ECO2011-25293) for financial sup-port of this project. The authors thank Laura Cozma atthe Institute for Economic Analysis (CSIC), for valuableresearch assistance. Previous versions of this paper werepresented at the annual meetings of the Midwest PoliticalScience Association, European Political Science Associa-tion, Council for European Studies, as well as workshopsat the Center for Advanced Studies of the Social Sciences(CEACS) at the Juan March Institute, Institute for Eco-nomic Analysis (IAE-CSIC), Center for Political andConstitutional Studies (CEPC), and the Autonomous Uni-versity of Barcelona. Finally the authors thank StephenCeccoli, Luis de la Calle, Sebastian Lavezzolo, SandraLeon, Yotam Margalit, Pedro Riera, Ruben Ruiz-Rufino,Thomas Scotto, two anonymous reviewers, and the editor,for valuable feedback. The usual disclaimer applies.

References

Achen, Christopher H., and Larry M. Bartels. (2006) It Feels LikeWe’re Thinking: The Rationalizing Voter and Electoral Democracy.Paper presented at the Annual Meeting of the APSA, Philadelphia,PA.

Adsera, Al�ıcia, and Carles Boix. (2002) Trade, Democracy, and theSize of the Public Sector: The Political Underpinnings ofOpenness. International Organization 56 (02): 229–262.

Allison, Paul D., and Nicholas A. Christakis. (1994) Logit Modelsfor Sets of Ranked Items. Sociological Methodology 24 (1994): 199–228.

Andr�es, Javier, Samuel Bentolila, Jos�e J. Dolado, and Florentino

Felgueroso. (2009) Propuesta para la reactivaci�on laboral enEspa~na. El Pais. Available at http://www.elpais.com/articulo/primer/plano/Propuesta/reactivacion/laboral/Espana/elpepueconeg/20090426elpneglse_16/Tes. Accessed February 2011.

Ansolabehere, Stephen, Shigeo Hirano, James Snyder, and

Michiko Ueda. (2006) Party and Incumbency Cues in Voting:

18 Right-wing governments have made statements blaming globalizationfor the economic crisis, in some cases to justify unpopular policies. For exam-ple, former Italian Prime Minister Berlusconi blamed “speculation on interna-tional markets” and the euro for Italy’s economic situation (Discourse toParliament, August 3, 2011). Former Greek Prime Minister Karamanlis arguedin September 2009 that the austerity measures were intended to “defend thecountry from the effects of the global economic crisis” (Speech at the 72ndInternational Trade Fair in Thessaloniki, September 7, 2009). Former IrishPrime Minister Brian Cowen linked Irish economic woes to external factorsand the necessity of selling to world markets (Speech at the Global Irish Eco-nomic Forum in Farmleigh House, September 18, 2009).


Are They Substitutes? Quarterly Journal of Political Science 1: 119–137.

Arceneaux, Kevin. (2003) The Conditional Impact of Blame Attributionon the Relationship between Economic Adversity and Turnout.Political Research Quarterly 56 (1): 67–75.

Bartels, Larry M. (2011) “Ideology and Retrospection in ElectoralResponses to the Great Recession.” Unpublished manuscript..

Beggs, Steven, N. Scott Cardell, and Jerry A. Hausman. (1981)Assessing the Potential Demand for Electric Cars. Journal ofEconometrics 17 (1): 1–19.

Burgoon, Brian. (2001) Globalization and Welfare Compensation:Disentangling the Ties That Bind. International Organization 55 (3):509–551.

Cameron, David. (1978) The Expansion of the Public Economy.American Political Science Review 72 (4): 1243–1261.

Cohen, Geoffrey L. (2003) Party over Policy: The Dominating Impactof Group Influence on Political Beliefs. Journal of Personality andSocial Psychology 85 (5): 807–822.

Cole, Richard L., and John Kincaid. (2006) Public Opinion on USFederal and Intergovernmental Issues in 2006: Continuity andChange. Publius: The Journal of Federalism 36 (3): 443–459.

Cole, Richard L., John Kincaid, and Alejandro Rodriguez. (2004)Public Opinion on Federalism and Federal Political Culture inCanada, Mexico, and the United States, 2004. Publius: The Journal ofFederalism 34 (3): 201–221.

De Grauwe, Paul. (2012) In Search of Symmetry in the Eurozone. PolicyBrief (268). CEPS.

Duch, Raymond M., and Randy Stevenson. (2010) The GlobalEconomy, Competency, and the Economic Vote. Journal of Politics72 (1): 105–123.

Eichengreen, Barry. (2009) The Crisis and the Euro. Berkeley, CA:University of California, April, 1-09.

European Commission. (2009) Economic Crisis in Europe: Causes,Consequences and Response. Luxembourg: European Commission.Available at http://ec.europa.eu/economy_finance/publications/publication15887_en.pdf. Accessed February 2011.

FAES. (2008) La Crisis Financiera. Or�ıgenes y Soluciones.Fern�andez-Albertos, Jos�e. (2006) Does Internationalization Blur

Responsibility? Economic Voting and Economic Openness in 15European Countries. West European Politics 29 (1): 28–46.

Giuliano, Paula, and Antonio Spilimbergo. (2009) Growing up in aRecession: Beliefs and the Macroeconomy. Cambridge, MA, USA:National Bureau of Economic Research.

G�omez, Manuel V. (2009) Los sindicatos movilizan a 700 expertoscontra la rebaja del despido. El Pais, June 20.

Hansen, John M. (1998) Individuals, Institutions, and PublicPreferences Over Public Finance. American Political Science Review92 (3): 513–531.

Hellwig, Timothy. (2001) Interdependence, Government Constraints,and Economic Voting. Journal of Politics 63 (4): 1141–1162.

Hellwig, Timothy. (2011) The World Economy, Political Control, andResponsibility for Economic Performance. Unpublished manuscript.Paper presented at European Responses to the Economic Crisis:Lessons for the United States Conference, April 8, 2011, IndianaUniversity Bloomington.

Hellwig, Timothy, and David Samuels. (2007) Voting in OpenEconomies: The Electoral Consequences of Globalization.Comparative Political Studies 40 (3): 283–306.

Hobolt, Sara B. (2007) Taking Cues on Europe? Voter Competenceand Party Endorsements in Referendums on European Integration.European Journal of Political Research 46 (2): 151–182.

Hope, Katie. (2012) Firms Urged to Prepare for Eurozone Break-Up,BBC online, August 9.

Hugh, Edward. (2010) Spain on the Verge of a Nervous Breakdown.Foreign Policy, December 3.

IDEAS Fundaci�on. (2009) Nuevas ideas para mejorar el funcionamientode los mercados financieros y la econom�ıa mundial.

Kayser, Mark A. (2007) How Domestic Is Domestic Politics?Globalization and Elections. Annual Review of Political Science 10:341–362.

Layman, Geoffrey C., and Thomas M. Carsey. (2002) PartyPolarization and ‘Conflict Extension’ in the American Electorate.American Journal of Political Science 46 (4): 786–802.

Lupia, Arthur, and Matthew D. McCubbins. (1998) The DemocraticDilemma. Cambridge: Cambridge University Press.

Malhotra, Neil, and Alexander G. Kuo. (2008) Attributing Blame:The public’s Response to Hurricane Katrina. The Journal of Politics70 (01): 120–135.

Malhotra, Neil, and Yotam Margalit. (2010) Short-TermCommunication Effects or Longstanding Dispositions? The Public’sResponse to the Financial Crisis of 2008. The Journal of Politics 72(03): 852–867.

Mallet, Victor. (2010) Spain’s Credit Rating Cut Over Growth Fears,Financial Times.

Ray, Leonard. (2003) When Parties Matter: The Conditional Influenceof Party Positions on Voter Opinions About European Integration.Journal of Politics 65 (4): 978–994.

Scheve, Kenneth F., and Matthew J. Slaughter. (2001) WhatDetermines Individual Trade-Policy Preferences. Journal ofInternational Economics 542: 267–292.

Shaver, Kelly G. (1985) The Attribution of Blame: Causality, Responsibility,and Blameworthiness. New York: Springer-Verlag.

Tilley, James, and Sara B. Hobolt. (2011) Is the Government toBlame? An Experimental Test of How Partisanship ShapesPerceptions of Performance and Responsibility. Journal of Politics 72(2): 1–15.

Walter, Stefanie. (2010) Globalization and the Welfare State: Testingthe Microfoundations of the Compensation Hypothesis.International Studies Quarterly 54 (2): 403–426.


Appendix 1: Netquest Survey Protocol

The Netquest survey uses opt-in panels, based on existing databases of nationally representative samples of residents ofSpain. The panel is constrained to individuals at least 18 years of age. The sample is stratified with representative quotasof the Spanish population by geographical area (seven geographical areas), age group, and gender. Netquest compen-sates economically all participants with vouchers that can be used later to purchase goods at Netquest’s online store.Full documentation on sample compilation is available upon request.

Appendix 2: Comparison of Samples

VariableNetquest survey

Nationalrepresentative survey

Survey*

Household monthly income in euros <900 8.8 <900 euros 17.5 European Social Survey, 4th Wave, April 2010901–1,400 20.8 901–1,400 31.4

1,401–1,900 17.3 1,401–1,900 26.51,901–2,700 27.6 1,901–2,700 15.6

>2,700 25.6 >2,700 11.1Gender Women: 51.0 Women: 51.0 CIS 2553. November 2010Age 18–24: 13.7 18–24: 9.2 CIS 2853. November 2010

25–34: 25.2 25–34: 19.935–44: 24.6 35–44: 20.845–55: 21.2 45–54: 16.755+: 15.4 55+: 33.6

Vote recall PSOE 38.8 PSOE 38.2 CIS 2853. November 2010PP 26.5 PP 26.4

Ideology (1–10) 4.6 (SD: 2.1) 4.8 (SD: 1.9) CIS 2853. November 2010Vote intention PSOE 19.8 PSOE 18.7 CIS 2847. October 2010

PP 24.8 PP 24.8Closest party PSOE 26.3 PSOE 28.8 CIS 2847. October 2010

PP 22.9 PP 24.6

Notes. *The closest survey available was the November 2010 Barometer (CIS 2853). However, this survey did not include questions on vote intention and partycloseness, so for these questions we use the next closest survey, CIS 2847, conducted in October 2010.

Appendix 3: Randomization Check

Treatment 2 Treatment 3 Treatment 4 Treatment 5 Treatment 6 Treatment 7

Education �0.016 (0.18) �0.16 (0.18) �0.27 (0.18) 0.16 (0.19) 0.25 (0.18) �0.080 (0.19)Income �0.020 (0.031) 0.042 (0.031) 0.039 (0.031) 0.037 (0.033) �0.054 (0.030) 0.037 (0.033)Gender �0.099 (0.17) �0.16 (0.17) �0.23 (0.17) 0.23 (0.18) 0.12 (0.17) 0.25 (0.18)Age �0.0067 (0.0067) 0.016 (0.0064) * 0.0039 (0.0064) �0.011 (0.0071) �0.0028 (0.0065) �0.0099 (0.0071)PSOE partisan 0.20 (0.21) �0.21 (0.21) �0.064 (0.21) 0.017 (0.21) �0.080 (0.20) �0.055 (0.22)PP partisan 0.40 (0.26) 0.28 (0.26) �0.21 (0.25) �0.21 (0.28) �0.29 (0.25) �0.037 (0.28)Ideology (L-R) 0.0019 (0.052) �0.088 (0.050) 0.088 (0.050) 0.0057 (0.054) 0.081 (0.049) �0.0054 (0.054)Constant �1.92 (0.38) ** �2.50 (0.38) ** �2.63 (0.39) ** �2.29 (0.40) ** �2.17 (0.37) ** �2.19 (0.40) **N 1,501 1,501 1,501 1,501 1,501 1,501Pseudo-R2 .006 .012 .010 .007 .008 .005

Notes. Estimations are logistic regressions predicting respondent placement in each of the seven experimental groups. Standard errors in parentheses.*p < .05, **p < .01.


Documents

Economic Crisis, Globalization, and Partisan Bias: Evidence from Spain