Decentralization and electoral accountability

1 IntroductionThe critical importance of assigning political responsibility for government action isnot a new topic. In the Federalist Papers, Hamilton (1961 [1788], page 424) warned that:

` It often becomes impossible, amidst mutual accusations, to determine on whom theblame or the punishment of a pernicious measure, or series of pernicious measures,ought really to fall. It is shifted from one to another with so much dexterity andunder such plausible appearances, that the public opinion is left in suspense aboutthe real author. The circumstances which may have led to any national miscarriageor misfortune are sometimes so complicated where there are a number of actors whomay have different degrees and kinds of agency, though we may clearly see upon thewhole that there has been mismanagement, yet it may be impracticable to pronounceto whose account the evil which may have been incurred is truly chargeable.''

More recently, authors have begun to study how clarity of responsibility within nationalgoverning institutions shapes citizens' ability to assign responsibility to incumbents formanaging the economy (eg C J Anderson, 2000; Powell and Whitten, 1993; Rudolph,2003a; 2003b). Nevertheless, as C D Anderson (2006, page 450) suggests, much less isknown about the effects of multilevel governance on economic voting or about citizens'ability to effectively attribute political responsibility at different levels of government.

Clarity of responsibility consists of two dimensions: ` horizontal clarity'' measuresthe degree of clarity of responsibility within a national government, while ` verticalclarity'' examines how multiple levels of government and/or significant decentralizationclouds responsibility attributions (C D Anderson, 2006, page 451). Previous studieshave favored cross-national comparisons aimed at maximizing the variance of institu-tional variables to explain horizontal clarity. As a result, case studies of vertical clarityare very unusual [C D Anderson (2008) is an exception], and predictors of verticalclarity have scarcely been analyzed, despite their centrality in understanding differencesbetween countries.

Multilevel institutions create new opportunities for local actors,(1) improve thequality of public policy, and reduce administrative costs by increasing the proximity

Decentralization and electoral accountability

Ignacio Lago-Pen¬ asDepartment of Political and Social Sciences, Universitat Pompeu Fabra of Barcelona,Ramon Tr|as Fargas 25-27, 08005 Barcelona, Spain; e-mail: [email protected]

Santiago Lago-Pen¬ asREDE, IEB and the University of Vigo, Department of Applied Economics, Campus Universitario,32004 Ourense, Spain; e-mail: [email protected] 18 May 2009; in revised form 11 December 2009; published online 22 March 2010

Environment and Planning C: Government and Policy 2010, volume 28, pages 318 ^ 334

Abstract. On the basis of aggregated and individual-level survey data of national and regional electionsin Spain, this paper analyzes how economic voting is impacted by vertical and horizontal dimensions ofclarity of responsibility. Our findings suggest that economic voting is enhanced when mechanismsof accountability are simple.

doi:10.1068/c0981

(1) Perkmann (2007) shows that multilevel governance patterns in the EU generate opportunities forentrepreneurial policy organizations to attract policy tasks and resources.

between government and citizen as well as governmental understanding of citizenpreferences (among others see Escobar-Lemmon, 2003; Kyriacou and Roca-Sagales,2008; Weingast, 1995). However, if the electoral control of incumbents falters as a resultof an inability to assign responsibility in multilevel governance, politicians are no longerforced to act in the best interest of the public (eg C D Anderson, 2006; Cutler, 2004;Gelineau and Remmer, 2003; 2005; Wibbels, 2003). Clarity of responsibility might bethen a challenge for decentralization processes.

In contrast to C D Anderson's (2006) influential work based on a cross-nationalcomparison of vertical clarity, the present study focuses on a single country, Spain.Anderson analyzes institutional variability along two dimensions of clarity of respon-sibilityöhorizontal and verticalöthrough a comparison of several countries; we havedone so by comparing national and regional governments (Autonomous Communities)in Spain. Furthermore, we maximize control over the research question by handlingfewer variables and avoiding the black box problem of fixed effects or country dummiesin the model.

The case of Spain is particularly well suited for exploring the impact of multilevelinstitutions on clarity of responsibility. In Spain voters can express their preferencesthrough different levels of government, and party elites have strong incentives to assurea contented constituency despite often competing interests. Moreover, the Spanishpublic sector has undergone extensive decentralization in the past thirty years. (2)

Currently Spain boasts one of the highest rates of subnational governmental spendingautonomy (specifically regional or autonomous governments). In 2006, regional gov-ernments were responsible for the 35.8% of total public expenditures,(3) and the localgovernments accounted for 13.3%. In total, half of all governmental spending (49.1%)was managed directly by subnational governments (Garc|a-D|az, 2007). Nevertheless,decentralization in Spain has not proceeded along a linear trajectory, specifically withregard to the redistribution of fiscal powers.

The coexistence of central control alongside largely autonomous regional spendingcontrol allows researchers to contrast the speed with which institutional changes firsttranslate into citizen perceptions and subsequently citizens' capacity to manipulategovernmental decision making. Additionally, the central administration retains controlover much of the basic regional legislation, like that of public health care and educa-tion. Finally, the decentralization that has taken place in Spain over the past threedecades allows researchers to estimate the whole impact of decentralization unlike inthe analyses of more traditional decentralized countries such as Canada where theimpact can only partially be measured due to the unavailability of data (C D Anderson,2008).

Section 2 presents the research question and a general model for the assignmentof political responsibility. Our empirical analysis is based on individual and aggregatedata and is divided into two parts. In section 3 we discuss the assignment of politicalresponsibilities from an individual perspective. We assess whether voters know whichlevel of government is responsible for various public policies and the effects ofthis knowledge on the electoral control of governments. In section 4 we study theextent to which we can explain differences in the degree of electoral control ofgovernments among Autonomous Communities. We introduce an index to measureeconomic voting based on the percentage of individuals voting in congruencewith their evaluation of the government. This indicator is designed to facilitate

(2) See Castells (2001) and S Lago-Pen¬ as (2005) for an overview of Spanish fiscal federalism at theregional level. For the local level, see Suarez-Pandiello (1996).(3) In 1986, the figure was 13.8% (DGCHT, 2001).

Decentralization and electoral accountability 319

cross-national and cross-time comparisons. In section 5 we will discuss researchfindings and offer directions for future research.

Three major findings result from our analysis. First, the attribution of politicalresponsibility depends on institutional arrangements and the `history' of decentraliza-tion.When political competences are in the hands of the national government, citizensmore accurately assign political responsibility than when the regional governmentsare in charge. Meanwhile, we observe that a longer history of decentralization changesthe default assignment of responsibility and makes citizens more likely to assignresponsibility at the local level. Our second finding suggests that electoral controlis significantly weaker in a coalition government. Finally, when the same party rulesat both the regional and the national levels, it is easier for voters to attribute respon-sibility for economic conditions given that parties tend not to engage in blame shiftingand credit taking. This in turn leads to stronger economic voting (and accountabilityin general).

2 Theoretical argumentsThe ability of voters to correctly assign responsibility is conditio sine qua non to holdgovernments accountable for actions and outcomes (Hamilton, 1961 [1788]). If votersdo not correctly assign responsibility for economic conditions, retrospective votingdoes not guarantee that bad governments are punished and good governmentsrewarded. The decision criteria of voters who are unable to assign responsibilities arerandom, like tossing a coin. The problem is that random action is not enough to ensurethat incumbents feel threatened into acting in the best interest of citizens. Indeed,electorate control of governments and imperfect information are at odds.

As is well known, information on government responsibility is not randomlydistributed among the electorate. When one analyzes differences in citizens' capacityto correctly assign political responsibility, systematic differences emerge betweenindividuals in a country or political system, as well as between countries or politicalsystems. In the case of individuals, all else equal, some citizens are more capablethan others of determining the role a government has played. At the aggregate level,the assignment of responsibility differs between countries or political systems, despiteoften similar demographic characteristics.

Individual differences are related to the oft-demonstrated theory of voter ignoranceand lack of sophistication.(4) Knowledge about the economy, unemployment levels,inflation rates, economic growth, or public debt is limited among voters, as is knowl-edge of how economic systems function (Aidt, 2000, pages 357 ^ 359). In general, thosewho are older, who have higher levels of education and above-average incomes arebetter informed (Delli Carpini and Keeter, 1996).(5)

According to Rudolph (2003a; 2003b), the assignment of responsibility depends onvoters' economic ideologies and party identification. In an analysis of AutonomousCommunities in Spain, Leon-Alfonso (2007) argues that the attribution of politicalresponsibilities is correlated with voters' political information. These findings supportDuch and Palmer (2002) and Duch et al (2000) who suggest that the subjectiveheterogeneity in evaluation of the economy depends upon proximity to the governingparty, one's own personal economic situation, and voter sophistication or level ofinformation.

(4) One can find thorough reviews of this literature in Kinder (1983) and Sniderman (1993).(5) In any case, we already know that a lack of information does not mean that citizens areincapable of making reasoned political decisions. Through the use of heuristics or informationshort cuts, they can compensate for their limited political information.

320 I Lago-Pen¬ as, S Lago-Pen¬ as

Literature on electoral and political behavior has turned towards institutionalismin recent years (C J Anderson, 2007a; 2007b). Theory suggests that the connectionbetween voters' perception of the economy and their vote depends on the extent towhich structural features of polities can inhibit voter access to information aboutrepresentatives' activities. The general argument, based largely on Powell and Whitten(1993), is that institutions increase or reduce citizens' capacity to reward or punishincumbents. Once voters have formed their opinion about the state of the economy,the translation to a vote for or against the incumbent is based on what they perceiveto be the incumbent's responsibility. Institutional features can cloud this perceptionby diminishing the government's clarity of responsibility'.Voters' ability to rely on theirown evaluation of the economy is undermined, leading to greater difficulty in decidingwhom to reward or punish. Thus, clearer political responsibility leads to a stronger linkbetween the state of the economy and electoral support of incumbents.

According to Powell and Whitten, there are five institutional characteristics thatweaken clarity of responsibility: (i) a multiparty cabinet; (ii) a minority government;(iii) weak party cohesion; (iv) opposing-party control of key policy development institu-tions; and (v) strong legislative committees presided over by members of opposingparties. As suggested by C J Anderson (2000), we also include two variables thatare specific to the mechanism of reward/punishment: (vi) the existence of a crediblealternative to the governing party, thus making electoral politics a zero-sum game;and (vii) the type of majority that the governing party has in the government orlegislature.

The capacity of voters to assign political responsibility is a function of the ease withwhich they can identify a competent candidate. ` The greater the perceived unifiedcontrol of policymaking by the incumbent, the more likely is the citizen to assignresponsibility for economic and political outcomes to the incumbents'' (Powell andWhitten, 1993, page 398). If a credible alternative to the governing party exists, poorresults attributed to the governing party will result in votes for the opposing party.

Empirical evidence largely supports the argument (eg C J Anderson, 2000; Lewis-Beck, 1986; Lewis-Beck and Stegmaier, 2007; Powell, 2000; Samuels, 2000), althoughthere are some who disagree (eg Royed et al, 2000). C J Anderson (2000, page 168)summarized the evidence as follows:

` these findings suggest that voters' ability to express discontent with economicperformance is enhanced when accountability is simple. Voters' economic assess-ments have stronger effects on government support when it is clear who the targetis, when the target is sizable, and when voters have only a limited number of viablealternatives to throw their support to.''Nevertheless, this discussion of clarity of responsibility has heretofore been limited

to its horizontal dimension, or the sharing of responsibility within a single level ofgovernment. In multilevel governance, political authority is disbursed among multiplelevels of government (Hooghe and Marks, 2001). Political authority may be dividedbetween a national government and a supranational organization, as in the case ofthe European Union, or subnational governments as in the case of the United Statesof America. In such cases, there is a vertical dimension of political responsibility(C D Anderson, 2006).

According to C D Anderson (2006, page 450; 2008, page 330) and Cutler (2004,pages 19 ^ 22), multilevel governments diminish the clarity of responsibility for tworeasons. First, roles are less clear in multilevel governments, and voters require additionalinformation in order to correctly assign responsibility. In single-level governments(national), virtually all the responsibility is assigned to that level of government. However,when subnational governments are present, the assignment of political responsibility is no

Decentralization and electoral accountability 321

longer direct: voters must now know whether a particular task is shared or exclusive. Asa result, retrospective voting becomes more demanding (Downs, 1999).

Second, in multilevel governments the executive branch has great incentive toassume responsibility for good political or economic outcomes and to evade responsi-bility for poor outcomes. We refer to this as the squid strategy of governmentsöhidingpolitical responsibilities in the ink of the distribution of functions. As mentioned above,in addition to using horizontal diffusion of responsibilities to avoid responsibility, thereis the possibility of vertical diffusion. Responsibility for an unpopular decision maybe transferred to an institution or actors situated at the same level of governmentöasin the voluminous literature on divided or multi-party governmentsöor at anothersuperior level (McGraw, 1990). National governments may also follow similar strategiesby blaming previous administrations or the international context for unpopular decisions.Voter confusion is the best ally of bad governments.

In addition to these two mechanisms, the multilevel distribution of political author-ity also diminishes the effects of economic evaluation on retrospective voting. Allother things being equal, economic voting is more likely to occur in a highly central-ized state than in a highly decentralized one. Meanwhile, within decentralized states,economic voting at the national level should be weaker than economic voting at thesubnational level, given subnational governments' inability to reassign responsibilityfor poor economic outcomes.

3 The assignment of political responsibility in SpainThe aim of this section is to test whether voters know which level of government isresponsible for which public policy. As discussed above, adequate electoral controldepends on how effectively citizens assign responsibility and hold different levels ofgovernment accountable within a country. In decentralized countries like Spain, wherenational governments dominate subnational branches, we expect the assignment ofpolitical responsibility to be better among national than among subnational functionsand policies.(6) Thus, our hypothesis follows:

Hypothesis H1: Voters have more difficulty assigning political responsibility at theautonomous level than at the national level.

Furthermore, due to Spain's two-tier decentralization process, it is possible to analyzethe impact of the degree of decentralization on the assignment of political responsi-bility. According to the Spanish Constitution, certain Autonomous Communities weredesignated `fast-track'öAndaluc|a, Canarias, Catalun¬ a, Galicia, Navarra, the Pa|sVasco, and the Comunidad Valencianaösince they had distinct earlier timetables toassume functions and higher competency levels than `slow-track' Autonomous Com-munities (Aragon, Asturias, Baleares, Cantabria, Castilla-La Mancha, Castilla y Leon,Extremadura, La Rioja, Madrid, Murcia). A longer history of self-government in theformer should involve a better knowledge of political processes and assignment ofresponsibility. For instance, while in the case of slow-track Autonomous Communitieshealth-care services were devolved in 2002, they had already been devolved to Catalun¬ ain 1981, to Andaluc|a in 1984, to the Pais Vasco and the Comunidad Valenciana in 1987,to Navarra and Galicia in 1990, and to Canarias in 1994.(7) The analysis of separateregions within a single country is especially interesting in order to estimate the causal

(6) Contamination effects are present when the behavior of a voter, a party, a candidate, or alegislator in one tier or arena is demonstrably affected by the institutional rules employed in theother tier or arena (Ferrara et al, 2005, page 8).(7) Regarding primary and secondary education, the devolution to fast-track regions took placebetween 1980 and 1983 (Navarra in 1990), while in low-track regions it was between 1997 and 1999.

322 I Lago-Pen¬ as, S Lago-Pen¬ as

effect of decentralization, given that we can control for more potential confoundingvariables. Our hypothesis is as follows:

Hypothesis H2: The assignment of political responsibility for competences in the handsof the regional government should be more accurate in fast-track than in slow-trackAutonomous Communities.

Nevertheless, this same logic applied to national government competences, would leadus to believe that in (slow-track) Autonomous Communities where fewer resourcesare controlled by autonomous governments, correct assignment of political responsi-bility will be more likely than in (fast-track) Autonomous Communities where moreresources are controlled by autonomous governments. Our hypothesis follows:

Hypothesis H3: Assignment of responsibility for competences in the hands of the nationalgovernment should be more accurate in slow-track Autonomous Communities than infast-track Autonomous Communities.

Finally, after more than twenty-five years since the start of the decentralization pro-cess, it would be interesting to see whether citizens have learned how to maneuvercomplex institutional networks or whether, on the contrary, the greater resources of theAutonomous Communities have complicated citizens' decision making. According toC D Anderson (2006, page 459), greater fiscal decentralization results in a weakereconomic voting. Unfortunately, we cannot study this issue in Spain, due to the lackof survey data from the beginning of the decentralization process.

In order to test the hypotheses, we have used data from the survey on ` EconomicOpinions and Attitudes of Spaniards, 2005'' collected by the Institute of Fiscal Studies(IEF).(8) The questionnaire included a battery of questions on political responsibilityfor different public services and social benefits. We selected five especially relevantquestions which we expected respondents to answer correctly. As such, we are playingin favor of the alternative hypothesisöthat citizens are able to distinguish between thepowers of each level of government.

Health and education budgets have been in the hands of Autonomous Communi-ties since 2002 and account for more than half of the regional budget. Nevertheless,basic legislation still depends on the national government. Accordingly, when the IEFsurvey asks which level of administration is responsible for education spending, weconsider correct those answers that assign responsibility to the communities or indicateshared responsibility between the communities and the state.

As shown in table 1, more than half of the respondents incorrectly assigned res-ponsibility for public health and education administration (53% and 54% incorrectresponses, respectively). Respondents tended to incorrectly assign responsibility to thenational government. Conversely, centrally administered competences such as socialsecurity, are much less likely to be incorrectly attributed to regional governments: 30%for retirement pensions, 26% for unemployment insurance, and 32% for medical leaveor disability. In other words, as predicted by our first hypothesis, when competencesare administered by the national government, citizens are more likely to correctly

(8) Survey information: Geographic scope: national territory. Population: all individuals aged 8 yearsor older, economically active and inactive. Quantitative methodology: a structured questionnaireadministered in the home. Sample design: multistage sampling through stratification and substra-tification of units of the first stage (census sections). Sample size: 1410 personal home interviews.Margin of error: �2.7 for global data with an assumption of simple random sampling andequiprobable categories. ( p � q � 50%) at a confidence level of 95%. Fieldwork: 28 Novemberto 22 December 2005. Survey firm: CUANTER, SA.

Decentralization and electoral accountability 323

assign political responsibility. We have checked the robustness of the results usingsurveys for years 2006 and 2007 also carried out by the IEF.(9)

Given that the allocation of expenditure responsibilities has changed over time inSpain, it is possible that fast-track Autonomous Communities would show a greaterpercentage of correct responses. These communities gained control over education andhealth systems expenditures in the 1980s and the early 1990s. Data largely confirm thesecond hypothesis. As seen in table 2, errors in the assignment of political responsibil-ities reach 43% in the fast-track regions, compared with 69% in the slow-track com-munitiesöa net difference of 26 percentage points. The same phenomenon occurs witheducation, where errors reach 67% and 46%öa difference of 21 percentage points. Inboth cases we performed a t-test to determine significance of error differences, assum-ing unpaired data. Insofar as p-values are very low ( p 5 0.00001), the null hypothesisof equality can be discarded.

As hypothesis H3 suggests, errors in the assignment of political responsibility arereduced in the slow-track Autonomous Communities when dealing with centralizedcompetences. As demonstrated by table 3, only 19% of respondents incorrectly attrib-uted responsibility for retirement pensions and unemployment insurance, while only22% incorrectly attributed responsibility for temporary or permanent disability insur-ance. The corresponding percentages for the `fast-track' communities are between 38%

Table 1. Percentage answers to the question: ` Who do you think is responsible for the followingsocial services and benefits? For example to whom would you submit a formal complaint: thecentral state, the autonomous community, the city council, or is responsibility shared?''

Services and benefits Administration currently responsiblea

state autonomous local shared DN/NAb errorcommunity

Public health 44 39 3 8 6 53Education 42 36 6 10 6 54Retirement pensions 70 16 2 5 7 30Unemployment insurance 64 20 3 6 7 36Medical or disability leave 68 16 3 6 7 32

aWith the exception of the final column, the percentages have been calculated according torow totals. The error column sums up the percentage of respondents who answer incorrectly.bDN/NAÐdon't know/no answer.

(9) In all cases errors were slightly greater than those in table 1: public health (55% in 2006 and 61%in 2007), education (54% and 58%, respectively), retirement pensions (36% and 35%), unemploy-ment insurance (42% and 43%), and medical or disability leave (39% and 39%). Main resultsfrom both surveys are available at http://www.ief.es/Publicaciones/Documentos/Doc 15 08.pdf andhttp://www.ief.es/Publicaciones/Documentos/Doc 21 07.pdf

Table 2. Percentage errors in the assignment of responsibility in education and health care.

Slow-track autonomous Fast-track autonomouscommunities (N � 569) communities (N � 839)

Health 69 43Education 67 46

Note: number of observations is given in parentheses.

324 I Lago-Pen¬ as, S Lago-Pen¬ as

and 44%. As in table 2, t-tests of equality were performed. Differences are statisticallysignificant in all cases ( p 5 0.00001).

Finally, misinformation on allocations of tax revenue is even more accentuated.(10)

According to table 4, between 78% and 81% of respondents were unaware that per-sonal income tax, VAT, and excise duty were partially allocated to the AutonomousCommunities in 2001. The percentage of error decreases to 55% for the fully centralizedbusiness tax.(11)

4 Electoral control of regional governments in SpainThis section outlines the consequences of assigning responsibility and holdinggovernments accountable for their actions. Not all clarity-of-responsibility predic-tor variables offer insight into differences in electoral control between national andregional governments in Spain. For example, party cohesion remains constant betweennational and regional levels of government, and therefore cannot be used to explainregional differences. Also, the analysis does not focus on the characteristics of the

(10) In what follows, results for the Pa|s Vasco and Navarra are discarded because of the special taxagreements concerning tax collection.(11) As happened with public expenditures, surveys for 2006 and 2007 are in line with those for 2005.Incorrect answers concerning the personal income tax were 83% in both 2006 and 2007. For theVAT percentages were 86% in both years. For business taxes: 50% in 2006 and 51% in 2007.Finally, errors in the case of excise duty were 86% in 2006 and 87% in 2007.

Table 3. Percentage errors in the assignment of responsibilities in the matters of retirementpensions, unemployment insurance, and temporary or permanent disability.

Slow-track autonomous Fast-track autonomouscommunities (N � 569) communities (N � 839)

Retirement pensions 19 38Unemployment insurance 19 44Temporary or permanent disability 22 38

Note: number of observations is given in parentheses.

Table 4. Percentage answer to the question: ` With respect to the following taxes, to whichadministrative branch do you think the money is directed, the central state, the AutonomousCommunity, or is the money divided among the two?''

Tax Level of governmenta

everything everything to part to the state, DN/NAb errorto the state the autonomous and part to the

community autonomouscommunity

Personal income tax 54 9 23 15 78VAT 54 11 19 16 81Business taxes 45 15 19 22 55Excise duty 44 14 20 22 80

Note: total number of observations � 1325.aObservations for Navarra and the Pa|s Vasco (N � 83) are excluded. With the exception of thefinal column, the percentages have been calculated according to the row totals. The error columnsums up the percentage of respondents who answered incorrectly.bDN/NAÐdon't know/no answer.

Decentralization and electoral accountability 325

legislative commissions, even if such characteristics seem especially relevant forSpain. Instead, empirical analysis of clarity of responsibility will focus on threevariables: government composition, viability of alternative administrations, and thepresence of nationalist sentiment.

Coalition governments tend to obscure responsibility for the consequences ofgovernment action: the greater the complexity of the government, the weaker theeconomic voting (C J Anderson, 1995; 2000). As Lewis-Beck (1986, page 341) sum-marizes, ``the more parties in a coalition government, greater is the diffusion ofgovernment responsibility.'' According to empirical evidence provided by Urquizu(2008) in his analysis of twenty-two OECD parliamentary democracies from 1946to 2006, the electoral outcomes of parties which control the executive branch incoalition governments are clearly affected by the state of the economy: no differencewas found between coalition and single-party governments. Conversely, electoral pay-offs for the remaining parties in coalition cabinets do not depend on the economy.In other words, voters focus on the most visible party and assign blame to it based onits performance. Hence, the following hypothesis:

Hypothesis H4 : In the presence of a coalition government, electoral control of thegovernment will be weaker.

Second, according to C J Anderson (2000), Powell and Whitten's (1993) argument onclarity of responsibility assumes the existence of alternative parties or candidates whenvoters want to unseat the incumbent. The less the fragmentation of the party system,the easier it will to rally voters behind a single party and hence the stronger theeconomic voting. In short, alternatives to the current administration must be clearlydefined if the reward ^ punishment mechanism is to give rise to a new administration.According to the example provided by C J Anderson (2000, page 156), an electoralcompetition between an incumbent with 51% of the seats and ten opposition partieswith the remaining 49%, is not the same as one where only one single party stands inthe opposition with 49% of the seats. Our hypothesis is as follows:

Hypothesis H5: The more fragmented the party system, the weaker the electoral controlof the government.

Finally, according to recent research by Aguilar and Sanchez-Cuenca (2008) andAlonso (2008), nationalist voters show less voting elasticity than do nonnationalistvoters. The reason is that they cast their ballots based not on what the governmentdoes, but on whom they represent (Aguilar and Sanchez-Cuenca, 2008, page 113). Thisrelationship would suggest that economic voting is more likely among nonnationalistvoters. Our hypothesis is as follows:

Hypothesis H6: The greater the nationalism (the regional cleavage) in an autonomouscommunity, the weaker the electoral control of the government.

In addition to these variables with demonstrated links to clarity of responsibility, weincorporate three others into the analysis. The first is the impact of time on assigningresponsibility and ultimately holding governments accountable for their actions. We donot have a defined hypothesis. On the one hand, it is possible to expect that voterslearn to evaluate the governments based on their performanceöif this is the case,votes would become increasingly elastic with each election. However, we also knowthat, at the subnational level, advances in fiscal decentralization make the division ofresponsibilities more blurry and, therefore, attenuate economic voting (C D Anderson,2006).

326 I Lago-Pen¬ as, S Lago-Pen¬ as

The remaining variables are included to measure electoral contamination. As inpresidential systems (Golder, 2006), the effect of national elections on subnationalelections may depend on how closely together the elections are held. Thus, the nation-alization of regional elections will be greater when national and regional electionsare held simultaneously, or when several Autonomous Communities hold elections atthe same time. And nationalization is synonymous with weaker electoral control of thesubnational governments. Our hypothesis is as follows:

Hypothesis H7:When regional elections coincide with national elections or when variousregional elections are held simultaneously, there will be weaker electoral control of theautonomous governments.

Similarly, all other things being equal, when the same party governs at both thenational and regional levels, we can expect accountability to be greater. This is basedon the assumption that, when the same party governs at both levels it will be easier forvoters to hold governments accountable for their actions. Similarly, having the sameruling party in both levels should discourage governments from confusing voters andthus government responsibility becomes more transparent. Our hypothesis is as follows:

Hypothesis H8: When the same party rules at both the national and regional levels,electoral control of the regional government should be greater.

The dependent variable (Control) is an index constructed to measure the probability thatan individual will vote in accordance with his or her appraisal of incumbent performance.Empirical research on electoral control of governments and economic voting is based onestablishing statistically significant correlations between economic conditions and theincumbent vote. This is a legitimate objective, since these correlations are essential tomodels of electoral control. Nevertheless, voters' utility functions are likely to changebetween elections and political arenas. We therefore cannot use such correlations todeduce the degree of electoral control, nor can we determine changes in electoralcontrol over time or political arenas of greatest electoral control. This is a limitationthat contrasts with other studies of electoral behavior that have attempted to measurethe intensity of the relationship between voting and its determinants. For example,in the analysis of class voting, we refer to Alford's index and its successive improvements(eg, Manza et al, 1995).

Our measure of electoral control of governments centers on the reward ^ punishmentmechanism behind retrospective voting. It seeks to determine the degree of votingelasticity with relation to government action or, conversely, whether variables includedin the voter utility function decrease voter sensitivity to incumbent performance. Datacollected from postelection surveys, report the percentage of respondents whose voteis in keeping with their assessment of government performance. A vote is consideredto be consistent when an individual assesses government performance as either goodor very good and casts a vote in favor of the governing party/parties or assesses theperformance as poor or very poor and casts a vote for the opposing party/parties.Conversely, a vote is considered inconsistent when an individual assigns governmentperformance a ranking of poor or very poor yet still casts their vote in favor ofthe governing party/parties or when an individual assigns government performancea rating of good or very good, yet casts their vote in favor of the opposing party.The greater the percentage of voters whose vote is in keeping with their assessmentof government performance, the greater the electoral control of the government. Theindex can range between 100, indicating maximum control, with all individuals votingconsistently with their evaluation of the government, and 0, indicating minimumcontrol, where no individual votes consistently with their evaluation of the government.

Decentralization and electoral accountability 327

In order to clarify how we calculate this index, table 5 presents the levels ofelectoral control for the La Rioja region in 1987. Data come from a postelectionsurvey. As shown in the table, 155 individuals assess government performance asgood or very good and vote in favor of the incumbent party, the PSOE, comparedwith 53 individuals who favorably assess government performance, but vote for theopposing parties. Two individuals assess the performance of the La Rioja govern-ment as either poor or very poor but vote cast ballots in favor of the currentadministration, whereas 43 who similarly assess government performance and votefor the opposition. In the table, consistent voters are those situated on the maindiagonal, in the shaded squares (155 + 43 = 198). Using the total of 253 respondentsas our denominator (155 + 53 + 2 + 43), the value of the index of electoral controlis 100(198/253) = 78%.

In the analyses of electoral control, it is important that surveys be conductedpostelection or at least close to the date of the elections in order to maximize thereliability of vote reporting. The data used for the present study were collected onlya few days after the elections, and are therefore especially appropriate for analysis. Inaddition, government assessment is likely to change over time so is best reported at theend of the term. Due to these restrictions, we have only been able to find data forthe second election from the beginning of the establishment of Autonomous Commu-nities and the two most recent autonomous elections. The data collected in 2002 comefrom a cross-sectional survey carried out approximately one year prior to the election.

Table 6 displays the levels of electoral control of the regional and national govern-ments. There are notable differences between Autonomous Communities. There isnearly 20 percentage points difference between Murcia and Canarias in the secondelection, 24 between Galicia and Aragon in 2002, and 25 between Galicia and theCanarias in the most recent election. Although we will explain the causes of thisvariability later, one possible explanation presented earlier is that of weaker electoralcontrol in the presence of electoral governments. This is corroborated by the fact thatthe five lowest values in the most recent election, Canarias, Catalun¬ a, Aragon, thePa|s Vasco, and Cantabria, correspond to Autonomous Communities with coalitiongovernments. Navarra is the only case that does not follow this pattern.

The set of independent variables measures the mechanisms behind the clarity ofresponsibility and electoral contamination:Government: This is a dichotomous variable in which a value of 1 indicates a coalitiongovernment, and 0 indicates a single-party government.ENPP: The clarity of available alternatives or party-system fragmentation is measuredby the effective number of parliamentary parties in the previous election (Laakso andTaagepera, 1979). This is calculated as follows: for n parties which obtain votes,

N ��Xni

i� 1

p 2i

�ÿ1,

where the term pi is the proportion of votes obtained by party i in the election.The variable ranges between 1.9 for Castilla-La Mancha and 4.0 for the Pa|s Vasco.

Table 5. Electoral control in La Rioja, 1987.

Vote Performance appraisal

``good'' or ``very good'' ``poor'' or ``very poor''

Incumbent 155 2Opposition 53 43

328 I Lago-Pen¬ as, S Lago-Pen¬ as

The source of the data is Ocan¬ a and On¬ ate (1999) for the second election, and Lagoand Montero (2009) for 2002 and the two most recent elections.Nationalism: In public opinion surveys, Spaniards were asked to report whetherthey considered themselves to be ``only [regional]'', ``more [regional] than Spanish'',``as [regional] as Spanish'', ``more Spanish than [regional]''or ``only Spanish''. Theintensity of the regional cleavage is measured by the proportion of individuals ineach category. We have created an aggregate measure of the intensity of the regionalcleavage in each autonomous community based on the simple addition of two scalecategories: ``only [regional]'' or ``more [regional] than Spanish''. The data come fromLago and Montero (2009).Time: Time is a dichotomous variable in which a value of 1 corresponds to the secondelection and a value of 0 corresponds to the 2002 and most recent elections.(12)

Simultaneousness: Simultaneousness is a dichotomous variable in which a value of 1corresponds to regional elections that are held together with or at the same time asnational elections. A 0 represents nonsimultaneous elections.SRP (same ruling party): This is a dichotomous variable in which the value 1 representselections in which the same party governs at the national and regional level. A valueof 0 represents different parties governing at the national and regional levels.

According to `Klein's rule' (Greene, 2003), there are problems of multicollinearitywhen Government, ENPP, and Nationalism are included in the same model. Therefore,the variables were initially included in different models, and subsequently analyzedwithin the same model.

(12) The results do not change appreciably if the variable has three categories instead of two.

Table 6. Index of electoral control in regional elections (source: own results based on CIS surveys).

Region Electiona

second 2002 most recent

Andaluc|a 87 74 89Aragon 86 68 77Asturias na 86 88Baleares na 69 79Canarias 72 71 67Cantabria na 89 78Castilla-La Mancha 79 79 84Castilla y Leon 82 88 87Catalun¬ a 82 73 75Extremadura na 84 88Galicia 89 92 92Madrid 85 81 87Murcia 91 91 87Navarra 80 87 89Pa|s Vasco 88 82 77La Rioja 78 85 82Comunidad Valenciana 89 77 82

Average 84 81 83

Note: naönot availablea The second election in Andaluc|a was held in 1986 and the most recent in 2008; the dates forCatalun¬ a are 1984 and 2006; for Galicia, 1985 and 2005; for the Pa|s Vasco, 1984 and 2005; andfor the remaining Autonomous Communities, 1987 and 2007.

Decentralization and electoral accountability 329

We use a pooled cross-sectional analysis to estimate the effect of the independentvariables and to compare their predictive performance across elections. Given thesmall number of observations (seventeen or even fewer) for each election, we havemerged three different surveys at different points in time. A pooled cross-sectionalanalysis is the proper method for an empirical test designed to demonstrate temporaldifferences in the effect of an independent variable (eg voters' economic expectations)on the probability of rewarding the incumbent. Additionally, all standard errors areclustered by region to account for nonindependence in the data structure. Finally, wehave also checked the reliability of the results using the quasi-maximum likelihoodestimator (QMLE) designed to deal with constrained fractional response variables(Papke and Wooldridge, 1996). By definition, the endogenous variable is constrainedon the interval [0,100]. However, ordinary least squares (OLS) does not guarantee thatthe predicted values of the dependent variable lie in the unit interval. The QMLE is notsubject to this limitation. In order to compare the OLS and the QMLE estimates,marginal effects computed with all variables held at their mean values are reportedin lieu of coefficients. This comparison is only shown for model 2 in table 9 due tospace constraints.

The main results are displayed in tables 7 ^ 9. In table 7 the degree of electoralcontrol is explained by the clarity of the responsibility index. Government in the firstmodel and ENPP in the second are the only statistically significant variables atthe 0.05 and 0.1 level, respectively. Both relationships are in the expected negativedirection: demonstrating that coalition governments and more fragmented partysystems lead to weaker accountability. Thus, coalition governments in comparisonwith single-party governments reduce electoral control by nearly 7 percentage points,and each additional parliamentary party attenuates the electoral control score byapproximately 4 percentage points. The effect of Time is unclear; while Nationalismshows the expected negative relationship, but is not statistically significant. None ofthe interactions between Time and the remaining independent variables is statisticallysignificant.

In table 8 clarity of responsibility is replaced with electoral contamination asthe determinant of electoral control of governments. The first model includessimultaneousness, Time, and SRP. Only SRP is statistically significant, at the 0.1 level:indicating that the same governing party at both the national and the regional levelsincreases economic voting. In the second model the irrelevant variables in the previousmodel are omitted and a spline function (Greene, 2003) is used to test whether the

Table 7. Clarity of responsibility and electoral control.

Variables Model 1 Model 2 Model 3

Government ÿ6.77** (2.70)Time ÿ0.58 (1.75) 2.09 (1.70) 1.87 (1.75)ENPP ÿ4.27* (2.25)Nationalism ÿ0.09 (0.10)Constant 83.69*** (1.53) 95.22*** (5.92) 85.74*** (2.06)R2 0.19 0.15 0.06N 47 47 47

*** p < 0:01; **p < 0:05; *p < 0:1.Note: Observations are clustered by region. The estimation is by ordinary least squares.Standard errors are given in parentheses.

330 I Lago-Pen¬ as, S Lago-Pen¬ as

effect of SRP is homogenous over time.(13) The variable coefficient is somewhat greater in2002 and the most recent election (2.42 for all observations versus 3.18 for the most recentelections) as is statistical significance ( p 5 0.10 for all observations versus p 5 0.05 for themost recent elections). That is, this indicates that government processes are clearer whenthe same governing party is in power at both the regional and the national level.

Table 8. Electoral contamination and electoral control.

Variables Model 1 Model 2

SRP 2.42* (1.40) 3.18** (1.46)Simultaneousness ÿ0.05 (1.79)Time 1.03 (1.76)SRP6Time ÿ1.23 (2.06)Constant 81.86*** (2.29) 80.95*** (1.70)R2 0.05 0.05N 47 47

*** p < 0:01; **p < 0:05; *p < 0:1.Note: Observations clustered by region. The estimation is by ordinary least squares. Standarderrors are given in parentheses.

(13) Therefore, the estimated specification in the second model is: Controli � a� b SRPi�gTimei6SRPi � ei . The effect of SRP is b� g for observations corresponding to the secondelection and b for the most recent election. The inclusion of Time would involve a lessconstrained econometric specification, with all the constitutive terms of the interaction:Controli � a� b SRPi � dTimei � gTimei6SRPi � ei . However, multicollinearity was very strongin this second case. Moreover, the statistical significance of the variable Time is extremely low inmodel 1 of table 8 (p � 0:56). Imposing d � 0 in model 2 of table 8 then seems reasonable.

Table 9. Clarity of responsibility, electoral contamination and electoral control of governments.

Variables Model 1 Model 2 Model 2 Model 4 Model 5 Model 6 Model 2(QMLE)

Government ÿ6.69** ÿ7.07** ÿ5.08* ÿ6.69** ÿ46.53***(2.79) (2.77) (2.63) (2.85) (16.32)

Time ÿ1.28 1.82 1.39 ÿ0.33 ÿ1.29(1.76) (2.00) (1.92) (1.76) (1.71)

ENPP ÿ4.17 ÿ2.05(2.41) (2.18)

Nationalism ÿ0.09 0.00(0.11) (0.11)

SRP 2.23 3.29* 1.36 1.88 1.76 2.23 23.01**(1.68) (1.70) (1.60) (1.69) (1.56) (1.71) (11.38)

Simultaneousness 0.00 ÿ0.93 ÿ0.69 ÿ0.44 0.00(1.74) (1.98) (2.10) (1.84) (2.06)

SRP6Time ÿ3.59 ÿ26.13(2.29) 16.27)

Constant 81.97*** 83.20*** 94.48*** 84.58*** 88.14*** 81.96***(2.51) (1.54) (7.34) (3.88) (6.13) (3.51)

R2 0.21 0.24 0.17 0.08 0.17 0.21N 47 47 47 47 47 47 47

*** p < 0:01; **p < 0:05; *p < 0:1.Note: Observations clustered by region. Models 1 to 6 are estimated using ordinary leastsquares. Model 2 in the last column is also estimated using quasi-maximum likelihood estimates.Marginal effects computed with all variables held at their mean values are reported. Standarderrors are given in parentheses.

Decentralization and electoral accountability 331

Finally, table 9 displays results for six regressions that combine clarity of responsibilityand electoral contamination variables. The results are very robust to different modelspecifications. Government is statistically significant in all models, even when theENPP and Nationalism are added (models 5 and 6). A coalition government reducescontrol by five to six percentage points compared with a singe-party government. Noneof the remaining clarity of responsibility variables is statistically significant. Concerning thecontamination model, the effect of SRP is again stronger and more significant in 2002and the most recent election. Estimates using QMLE instead of OLS (reported in the lastcolumn of table 9) strongly support these results.

5 ConclusionsThe purpose of this study was to examine how accountability is affected by clarityof responsibility within a government. According to aggregated and individual-levelsurvey data on national and regional elections in Spain, clarity in the assignment ofpolitical responsibilities, and the electoral control of governments depend on govern-mental transparency. Simply put, accountability is enhanced when mechanisms ofaccountability are simple.

The paper presents three main results. First, it is easier for citizens to correctlyassign responsibility for government action when the political competences are inthe hands of the national government. However, the longer subnational governmentshave been decentralized, the more likely citizens are to correctly assign responsibility.Second, electoral control is significantly weaker in the presence of a coalition govern-ment. Finally, accountability increases when the same party governs at the national andthe regional level.

In sum, while multilevel governance has been shown to improve the quality ofpublic policy it may also diminish incentives for politicians to act in the best interestof the public by confusing responsibility assignment. As shown by Rodr|guez-Pose andSandall (2008), both democracy and good governance are key issues surrounding thecurrent wave of decentralizations worldwide. Nevertheless, the benefits provided bygiving political and fiscal authority to different levels of government are far fromautomatic. In a best-case scenario, the benefits arise years after starting the processof decentralization. Decentralization is not simply a matter of changing institutionalframeworks, devolving expenditure and tax powers, and setting up regional parlia-ments, but rather, depends on citizens. If citizens do not have sufficient informationto evaluate incumbent governments, then governments do not have incentives to act intheir best interest.While devolving tax powers is a better solution than merely grantingrevenues,(14) maximum benefit from fiscal decentralization requires that tax assign-ments remain transparent to the average voter. Moreover, shared expenditure powersalso have a cost in terms of transparency and control.

A decentralized framework is more complex than a centralized framework. Thusadditional efforts to clarify and understand political processes and outcomes areneeded. Some of them refer to institutional design, but others to political culture andinformation flows.

Acknowledgements. Financial support from the Centro de Investigaciones Sociologicas (CIS) isgratefully acknowledged.

(14) Rodr|guez-Pose and Gill (2003) provide valuable insights into the issue of the imbalancebetween the devolution of powers and resources and its effects on choices by all fiscal tiers.

332 I Lago-Pen¬ as, S Lago-Pen¬ as

ReferencesAguilar P, Sanchez-Cuenca I, 2008, ` Performance or representation? The determinants of voting

in complex political contexts'', inVoters, Institutions and Accountability Eds J M Maravall,I Sanchez-Cuenca (Cambridge University Press, Cambridge) pp 105ö130

Aidt T S, 2000, ` Economic voting and information'' Electoral Studies 19 349 ^ 362Alonso S, 2008, ` Enduring ethnicity: the political survival of incumbent ethnic parties in western

democracies'', inVoters, Institutions and Accountability Eds J M Maravall, I Sanchez-Cuenca(Cambridge University Press, Cambridge) pp 82 ^ 104

Anderson C D, 2006, ` Economic voting and multilevel governance: a comparative individual-levelanalysis''American Journal of Political Science 50 449 ^ 463

Anderson C D, 2008, ` Economic voting, multilevel governance and information in Canada''Canadian Journal of Political Science 41 329 ^ 354

Anderson C J, 1995, ` The dynamics of public support for coalition governments'' ComparativePolitical Studies 28 350 ^ 383

Anderson C J, 2000, ` Economic voting and political context: a comparative perspective'' ElectoralStudies 19 151 ^ 170

Anderson C J, 2007a, ` The end of economic voting? Contingency dilemmas and the limits ofdemocratic accountability''Annual Review of Political Science 10 271 ^ 296

Anderson C J, 2007b, ` Interaction of structures and voter behavior'', in The Oxford Handbookof Political Behavior Eds R J Dalton, H D Klingemann (Oxford University Press, Oxford)pp 589 ^ 609

Castells A, 2001, ` The role of intergovernmental finance in achieving diversity and cohesion:the case of Spain'' Environment and Planning C: Government and Policy 19 189 ^ 206

Cutler F, 2004, ``Government responsibility and electoral accountability in federations'' Publius:The Journal of Federalism 34 19 ^ 38

Delli-Carpini M X, Keeter S, 1996 What Americans Know about Politics and Why It Matters(Yale University Press, New Haven, CT)

DGCHT, 2001, La descentralizacion del gasto publico en Espan¬ a. Per|odo 1984 ^ 1998 Directorate-General for Financial Coordinationwith the SubcentralGovernments,Ministerio de Econom|ay Hacienda, Madrid

DownsW M, 1999, `Accountability payoffs in federal systems? Competing logics and evidencefrom Europe's newest federation'' Publius: The Journal of Federalism 29 87 ^ 110

Duch, R M, Palmer H D, 2002, ` Heterogeneous perceptions of economic conditions'', in cross-national perspective'' Economic Voting Eds H Dorussen, M Taylor (Routledge, London)pp 139 ^ 172

Duch RM, Palmer H D, Anderson C J, 2000, ` Heterogeneity in perceptions of national economicconditions''American Journal of Political Science 44 635 ^ 652

Escobar-Lemmon M, 2003, ` Political support for decentralization: an analysis of the Colombianand Venezuelan legislatures''American Journal of Political Science 47 683 ^ 697

Ferrara F, Herron E S, Nishikawa M, 2005 Mixed Electoral Systems: Contamination and itsConsequences (Palgrave Macmillan, NewYork)

Garc|a-D|az M.A, 2007, ``Los presupuestos de las Comunidades Autonomas en 2007'',WP 15/07,Instituto de Estudios Fiscales, Madrid

Gelineau F, Remmer K L, 2003, ` Subnational electoral choice: economic and referendum votingin Argentina, 1983 ^ 1999'' Comparative Political Studies 36 801 ^ 821

Gelineau F, Remmer K L, 2005, ` Political decentralization and electoral accountability: theArgentine experience, 1983 ^ 2001'' British Journal of Political Science 36 133 ^ 157

Golder M, 2006, ` Presidential coattails and legislative fragmentation''American Journal ofPolitical Science 50 34 ^ 48

GreeneW H, 2003 Econometric Analysis (Prentice Hall, Englewood Cliffs, NJ)Hamilton A, 1961 [1788] The Federalist Papers (Mentor, NewYork)Hooghe L,Marks G, 2001Multi-level Governance and European Integration (Rowan and Littlefield,

Lanham, MD)Kinder D R, 1983, ` Diversity and complexity in American public opinion'', in Political Science:

The State of the Discipline Ed. A Finifter (American Political Science Association,Washington,DC) pp 391 ^ 401

Kyriacou A, Roca-Sagales O, 2008, ``Fiscal decentralization and the quality of government:evidence from panel data'',WP 3/08, Instituto de Estudios Fiscales, Madrid,http://www.ief.es/Publicaciones/PapelesDeTrabajo/pt2008 03.pdf

Decentralization and electoral accountability 333

Laakso M, Taagepera R, 1979, ` Effective number of parties: a measure with applications toWestEurope'' Comparative Political Studies 12 3 ^ 27

Lago I, Montero J R, 2009, ` Coordination between electoral arenas in multi-level countries''European Journal of Political Research 48 176 ^ 203

Lago-Pen¬ as S, 2005, ` Evolving federations and regional public deficits: testing the bailouthypothesis in the Spanish case'' Environment and Planning C: Government and Policy 23437 ^ 454

Leon-Alfonso S, 2007 The Political Economy of Fiscal Decentralization: Bringing Politics to theStudy of Intergovernmental Transfers (Institut d'Estudis Autonomics, Barcelone)

Lewis-Beck M S, 1986, ``Comparative economic voting: Britain, France, Germany, Italy''American Journal of Political Science 30 315 ^ 346

Lewis-Beck M S, Stegmaier M, 2007, ` Economic models of voting,'', in The Oxford Handbookof Political Behavior Eds R Dalton, H D Klingemann (Oxford University Press, Oxford)pp 518 ^ 537

McGraw K M, 1990, `Avoiding blame: an experimental investigation of political excuses andjustifications'' British Journal of Political Science 20 119 ^ 131

Manza, J, Hout M, Brools C, 1995, ` Class voting in capitalist democracies sinceWorld War II''Annual Review of Sociology 21 137 ^ 162

Ocan¬ a FA, On¬ ate P, 1999 Analisis de Datos Electorales (Centro de Investigaciones Sociologicas,Madrid)

Papke L E,Wooldridge J M, 1996, ` Econometric methods for fractional response variables withan application to 401(k) plan participation rates'' Journal of Applied Econometrics 11 619 ^ 632

Perkmann M, 2007, ` Policy entrepreneurship and multilevel governance: a comparative study ofEuropean cross-border regions'' Environment and Planning C: Government and Policy 25861 ^ 879

Powell, G B Jr, 2000, Elections as Instruments of Democracy:Majoritarian and Proportional Visions(Yale University Press, New Haven, CT)

Powell G B Jr,Whitten G D, 1993, `A cross-national analysis of economic voting: taking accountof the political context''American Journal of Political Science 37 391 ^ 414

Rodr|guez-Pose A, Gill N, 2003, ` The global trend towards devolution and its implications''Environment and Planning C: Government and Policy 21 333 ^ 351

Rodr|guez-Pose A, Sandall R, 2008, ` From identity to the economy: analysing the evolution ofthe decentralisation discourse''Environment and Planning C: Government and Policy 26 54 ^ 72

Royed T, Leyden K M, Borrelli S A, 2000, ` Is clarity of responsibility' important for economicvoting? Revisiting Powell and Whitten's hypothesis'' British Journal of Political Science 30669 ^ 668

Rudolph T J, 2003a, ` Institutional context and the assignment of political responsibility'' Journalof Politics 65 190 ^ 215

Rudolph T J, 2003b, ` Who's responsible for the economy? The formation and consequences ofresponsibility attributions''American Journal of Political Science 47 698 ^ 713

Samuels D, 2000, ` The gubernatorial coattails effect: federalism and congressional elections inBrazil'' Journal of Politics 62 240 ^ 253

Sniderman P M, 1993, ` The new look in public opinion research'', in Political Science: The Stateof the Discipline II Ed. A Finifter (American Political Science Association,Washington, DC)pp 219 ^ 245

Suarez-Pandiello J, 1996, ` Financing local governments in Spain: new solutions to old problems''Environment and Planning C: Government and Policy 14 411 ^ 430

Urquizu I, 2008 The Political Consequences of Coalition Governments: Multiparty Cabinets andAccountability (Instituto Juan March, Madrid)

Weingast B R, 1995, ` The economic role of political institutions: market-preserving federalismand economic development'' Journal of Law, Economics and Organizations 11 1 ^ 31

Wibbels E, 2003, ` Bailouts, budget constraints, and Leviathans: comparative federalism and lessonsfrom the early United States'' Comparative Political Studies 36 475 ^ 508

ß 2010 Pion Ltd and its Licensors

334 I Lago-Pen¬ as, S Lago-Pen¬ as

Conditions of use. This article may be downloaded from the E&P website for personal researchby members of subscribing organisations. This PDF may not be placed on any website (or otheronline distribution system) without permission of the publisher.