ESTUDIOS SOBRE LA ECONOMÍA ESPAÑOLA

Joan Costa-Font Ricard Ramon –Sumoy

EEE 216

November 2006

FEDEA Fundación de Estudios de Economia Aplicada

ISSN 1696-6384

Las opiniones contenidas en los Documentos de la Serie EEE, reflejan exclusivamente las de los autores y no necesariamente las de FEDEA.

The opinions in the EEE Series are the responsibility of the authors an therefore, do not necessarily coincide with those of the FEDEA.

JOAN COSTA-FONT^* Universitat de Barcelona and London School of Economics

RICARD RAMON –SUMOY Universitat Autònoma de Barcelona

Contact address: Dr Joan Costa -i-Font, London School of Economics, Cowdray House, J7, Houghton Street, WC2A 2AE, London. E-mail: j.costa-font@lse.ac.uk Acknowledgements: This research was supported by the Ministry of Education (Grant number: CICYT SEJ2005-06270) and the Generalitat de Catalunya (Grant number: 2005-SGR-460).

Abstract. Rational voting models assume turnout results from net benefits being positive. However, low turnout questions the legitimacy of democracies, especially when it affects newly created institutional structures such as those resulting from decentralisation processes. In this paper we empirically estimate whether turnout rates of “less decisive” elections can be increased by the coincidence of several elections. We use a unique data from Spanish regional elections 1980-1999 where low turn out would influence institutional legitimacy of newly created regional state in a novel democracy. We employ controls for economic voting namely: inflation, unemployment and regional income. Our empirical findings illustrate that simultaneity systematically increases individuals average turnout regardless of the elections considered and that regional income reduces turnout. .

Key words: rational voting, turnout, economic voting, decentralisation, Spain. JEL: D72, F22, H73

1. INTRODUCTION

A cause for alarm for the subsistence of modern democracies is ‘low’ voter turnout. Although a careful definition of what low turn out is, systematic low turnout - relative to previous times of other settings - might well compromise the legitimacy of elections as grounding tenets to elicit political support for competing political projects. Accordingly, the feature poses a question to analysts for a better underlying explanation of the phenomenon. Well known theories on rationality being abstention depart form Downs’ (1957) seminal contribution on the rational choice of voting. Accordingly, voting takes place when it portrays non-negative expected payoffs to individuals¹. In this case, voting can be viewed as consumption actions (Riker and Ordershook’s, 1968), or a positional good leading to some subjective satisfaction and self-affirmation of one’s views with regard to the polity. This can be compared supporting sports competitions (Brennan and Buchanan, 1984). In turn, electoral abstention is a form of “free riding”, given the collective nature of electoral benefits of other’s “expressing” their vote along with the limited sanctions resulting from nonparticipation. This is especially the case of elections with limited decisiveness (e.g., regional elections) or those with null chances of being a pivotal voter (Barry, 1982).

Yet, the likelihood of an individual to participate in an election is still open to theoretical and empirical scrutiny. An array of theoretical and empirical contributions has dealt with the influence of a set of features affecting an individual’s perceived costs (e.g., nature of election, electoral day and mechanism and so on). However, limited emphasis has been given to institutional determinants behind an individual’s perceived benefits from voting. and the final voting action.

In dealing with the free riding problem, one might claim for individual’s social responsibility and social inclusiveness (Downs, 1957, Riker and Ordershook’s, 1968) should suffice. Furthermore, so long as voting takes the characteristics of positional goods, one might invoke prescriptive reasons such as civic values and moral obligations (Opp, 2001, Riker and Ordershook’s, 1968 Blais et al, 2000)². Conversely, there is broad agreement that voting is a low cost - low benefit activity (Aldrich, 1993), so that the final outcome might be tightly dependent on the institutional determinants affecting the opportunity costs of voting, as well as any expected benefits from casing a vote³. Furthermore, evidence indicates that people who vote tend to overestimate their influence in electoral outcomes (Opp, 2001), which suggests that the act of voting might still be rational though subject cognitive limitations.

¹ That is, when its (expected) utility offsets a wide range of (expected) individual’s specific costs, including time, information update and other transaction costs as well as psychological deliberation efforts and emotional distress.

Beside voter cognitive limitations and indifference attitudes towards politics (Colomer, 1991), non-pivotal voters might perceive that their contribution is not decisive or would make very little difference in the results⁴. This is especially the case of regional elections, where the benefits of the expected outcomes might not be perceived as being high enough to overcome the range of the information and procedural costs of casting a vote. Some individuals are strategically able to differentiate different electoral contests and to select those calls for which there is perceived participation is worthwhile - and opt-out form those where their vote might not be perceived as marginally decisive - . In other words, whilst national politics might be envisaged as ‘high politics’ and worth contributing to, regional politics might arguably be conceived as “less influential”⁵. On the other hand, regional elections provide higher incentives to free-ride. A counter argument might indicate that in regional elections, because the potential electorate is smaller than that at the national level, one might well argue that the probability of influencing the final electoral outcome is larger, and thus instrumental conceptions of voting might be more meaningful.

² However, it does not seem plausible to solve the well sustained problem of voluntary contributions to collective goods (e.g., the formation of a parliament and democracy itself) claiming ethical or moral principles, especially when individuals disagree on the composition of such a collective good (e.g., the parliament composition).

³ Even anecdotal voting circumstances, including weekend day, sunlight, and the efficacy in casting the vote seem to influence the probability of electoral turnout.

Anecdotal evidence form Spanish regional elections suggests that the most decisive electoral call which could have determined the ‘nationalist majority’ in the Basque Parliament in 2001 is the one that has lead to the highest turn out ever in a regional election in Spain

⁵ Whilst some political activity was held in the Spanish region-states in during the dictatorship top oppose the system , with the democratization of Spain and the expansion of the public sector. High political activity has been shifted to the central state, s whilst only regional specific of local issues are discussed at a lower level.

The effect of simultaneous voting in increasing participation has received only anecdotal support. One exception refers to a study on European elections which identifies simultaneity as rising turnout by as much as 20 percentage points (Smith, 1999). However, this has not been examined further for other key election, such as regional and local elections which could condition the legitimacy of the electoral system and representative democracy. Indeed, prior studies do not control for other relevant factors. Arguably, simultaneous elections reduce the costs of voting if voting for several elections is conveyed as a “single civic duty”. Another explanation for regional turnout is the role of economic factors, especially regional income, inflation and unemployment. This has two main explanations, on the one hand, Spanish regional governments are entitled decide on some economic and social policy issues, so that participation could be viewed as an underlying expression of a views on such areas . On the other hand, in a decentralized setting one might believe that regional governments might lobby to shift policies at a state level through some form of vertical competition (Costa-Font and Rico, 2006) . Finally, in poorer regions, regional governments might supply a large share of the employment in the area, so that a change in the regional government is likely to affect the economic activity and foster people’s electoral participation.

The examination of Spanish regional elections has the added advantage of allowing for different “types” of simultaneous elections, e.g., regional elections coinciding with local, national, and/or European elections. All of them can take place together to reduce the costs of an election and its associated campaign. This allows the authors to compare the effect of simultaneously holding different types of elections on turnout. Furthermore, incumbent politicians might have some special interest in deciding the electoral calendar in a way that does or does not coincided with another election, so that coincidence might become a decision variable⁷. Some evidence from the US suggests that ultimately, the purpose would be to ensure that the incumbent candidate is ultimately elected (Kayser, 2005). Accordingly, turnout cannot be envisaged as exogenous but rather as an endogenous variable whenever incumbents can decide the electoral calendar.

⁶ Important areas of public expenditure have been regionally decentralized such as health and education, and policies are decided often at the region-state - or so called autonomous communities (AC)- level.

⁷ Consequently, one might argue that depending on the political environment and whether the incumbent is of the same party as the ruling party at the national level he/she might have an interest in either fostering or discouraging simultaneous elections.

This paper empirically examines the influence of the coincidence of several elections in the final electoral turnout. We employ evidence from Spanish regional elections using a panel from 1980 to 1999⁸. This country which allows for different types of elections was, up until 1978, not governed by a democratic regime. The institutional framework that emerged after democratisation has lead to the creation of a multilevel government, which effectively implied the coexistence of multiple elections. This stands as an empirical test of a commonly believed feature, namely the voting enhancing effect of simultaneous voting in federal or asymmetrically federal settings such as within Spanish institutional design. Indeed, whilst the parliaments of four of region-states (so-called “autonomous communities”, AC) in the period examined (Catalonia, Basque Country, Galicia and Valencia) are entitled to decide the electoral calendar we can distinguish region-states with the capacity of influencing the incumbent’s agenda in these ACs , and in turn shifting the effect of simultaneous elections to influence turnout on their benefit. Therefore, the second hypothesis to test is whether there are any differences between those regions that are empowered to modify the elections calendar and those that are not.

Our empirical analysis includes a set of controls for economic voting which have not yet been tested using Spanish data, namely the unemployment rate, inflation and regional income (Lewis-Beck and Paldam, 2000). Interestingly, previous studies point out that unemployment and inflation, are the two main variables that exhibit associations with the probability of individuals casting a vote. However, evidence on elections in one country might not be applicable to other countries which call for the role of institutional factors behind an individual’s vote, and specifically the implicit rules governing an individual’s voting.

⁸ We examine this period because there have been important changes in the decentralization processes from 2002 onwards as the remaining regional parliaments have been empowered with more responsibilities, which we believe deserves separate treatment. .

The structure of the paper is the following. Section two is devoted to the conceptual framework underpinning electoral turnout. Section three describes the institutional environment in Spain. Section four is devoted to the data and methods. Section five reports the results and the last section concludes.

2. RATIONALITY AND ELECTORAL TURNOUT

2.1 Rational Voting and Extensions

Rational voting models explicitly attempt to describe the rationality of voting by identifying the underlying outcomes of different courses of action, namely the benefits and costs of individuals casting a vote. Accordingly, rational individuals would be expected to participate in an electoral contest only if their perceived expected benefits from voting outweigh its costs, else the expected strategically optimal alternative would be abstention (Downs, 1957). The costs of electoral participation include tangible sacrifices (e.g., distance, physical impediments etc) and/or intangible and individual specific loses (e.g., day lost, prospects of undesired policies etc) as well as ample measures of expected benefits (e.g., desired policy change) and beliefs even if they are irrational which sustain so-called rational irrationality hypothesis (Caplan, 2001). Therefore, despite some authors having defined voting as being an irrational act , some might argue that by extending the classification of costs and benefits of voting so as to include civic duty (Mueller, 1989).

Theoretically the expected pay-off of voting results from the different expected net benefits in each of the J future policy scenarios that derive from the success of each one of the candidates, where measures the probability of each J benefit scenario at time t+1. Furthermore, individuals take into account a specific and expressive benefit at time t which would materialise from casting their vote for a certain candidate along with some time-specific costs form casting a vote which give rise to the following aggregate expression for voting turn out at time t :

\[T _ {i, t} = \sum_ {J, t + 1} \pi_ {i} B _ {i} + D _ {i t} - C _ {i t}\tag{1}\]

Accordingly, in a specific setting it might well be that given that certain individuals perceive that their chances of their voice give rise to a winner results are small, or that which implies that the probability of their vote being decisive is very small. Other alternative scenario might well be that of being significantly small and therefore the resulting benefits from the voting action would be expected to result purely on expressive grounds. Despite conflicting evidence, rational voting models provide us with broad theoretical grounds to examine voter turnout. Decision making theories such as prospect theory (Kanheman and Tversky, 1979) argue that individuals weight benefits and costs differently. In particular, one can argue that both might be weighted so that although benefits might appear to overcome the costs, individuals’ biased perceptions might act in a different way so that the costs are overestimated or the benefits underestimated. However, even accounting for possible cognitive biases, the general rational choice predictions (Downs, 1957), whereby voters are more likely to vote, the lower the costs of voting and the higher the expected benefits would still apply. If voting is a low costs-low benefit act (Aldrich, 1993), one might expect individuals not to participate in the collective decision making process in the first place. That is they exhibit higher participation uncertainty and lower incentives to support a specific option on purely rational grounds. However, they could potentially benefit for the resulting decision later . Palfrey and Rosenthal (1983), drawing from a game theoretic approach, argue that if one knows that it is rational for everyone to abstain, it becomes rational to vote and voting and decisiveness (Brenan and Buchanan, 1984) become simultaneously determined. Therefore, if incumbents have in their hand the possibility of influencing the individual’s voting decision, they will be expected to do some if they perceive the marginal vote to be decisive.

Individuals’ decision to, or not to participate in electoral contests result from a set of conditions (Schram, 1991) . First, the probability of being decisive is somehow negligible. Second, significant uncertainty in the capacity to benefit from the winner as opposed to the preferred alternative is sometimes very small .Finally, even if the preferred alternative coincides with the winner, the benefits enhanced from this feature might be collectively spread throughout the entire population rather than being individually enjoyed, regardless of their prior preference or vote . However, Riker and Ordeshook (1968) argue in favor of the individual’s commitment to the so-called “civic duty “as a reason behind the individual’s act of voting. However, this latter argument has been qualified as somehow axiomatic though not necessarily backed by empirical evidence. Therefore, it is important to find out empirical evidence behind competing explanation of individuals vote.

⁹ Accordingly one might argue that there is an incentive to “free ride”, especially if they see themselves not different form the “median voter”. The latter implies, that if individuals position on a certain important policy dimension is likely to differ from that of their perceptions of the majority or the “median voter” then they would be expected to participate.

Alternative explanations to purely instrumental conceptions of voting conceive the act of voting as being purely expressive, by giving utility to voters supporting regional specific interests, or and an act to avoid individuals’ regret in not having cast a vote (Ferejohn and Fiorina, 1976; Brenan and Lomansky 1993). As several authors have observed, voting is also both an "expressive" and "signalling" act, and these aspects have to be accounted for in the computation of the respective costs and benefits of voting and not voting Thus, the paradox of voting asserts that self-interested individuals will not vote is in part solved by Brenan and Lomansky’s (1993) assumption on irrationality whereby voting is an act of “expressive behaviour” as well as similar explanations based on regret minimization (Ferejohn and Fiorina, 1974). Finally, model assumptions based on individuals’ bounded rationality and game theoretic approaches based on strategic interactions seem to agree that some expressive factor influences high levels of voter turnout (Dhillon and Peralta, 2002).

Electoral participation, it has been argued, depends on the social and economic development of certain areas such as access to information, higher levels of education and income levels (Powell, 1982) as well as the regional diversity of interests (Blais and Dobrzynska, 1998). In addition, small countries and those with higher population density tend to have higher incentives to vote as elections take place in political environments where community relations are closer and more direct. Other important variables refer to the voting system. Countries with a proportional voting system exhibit a higher turn out (Blais and Dobrzynska, 1998; Blais et al , 2003). Finally, other relevant variables are the voting day and the ease of voting.

Some other explanations for voter turnout are the following. First, several studies demonstrate that compulsory voting stands as the most significant variable in explaining turn out. Indeed, compulsory voting increases voting turnout by over 30% (Smith, 1999). Other important variables are demographic factors, (Smith, 1999), the election days - (e.g., an election on a weekend increases turnout); the voting system e.g., open lists are more attractive and the electoral system (Lijphartm 1997), Other studies highlight the importance of public expenditure in certain countries as explaining voter turnout (Colomer, 1991). If voting indirectly implies voting for several public policy programs, those countries where the public sector size is relatively more important relative to total economic activity (e.g., most EU countries) should be expected to exhibit higher levels or voter turnout. Opp (2001) finds those perceptions of being influencing final political outcomes does determine voting turnout. These effects are of remarkable importance in explaining regional voting. Those countries that have devolved some of their powers to regions as is the case of Spain might be concerned with the lower levels of voter turnout. This might result from the novelty of the decentralization process as well as individuals’ misperceptions of the benefits that provide regional governments in their actions.

Previous work has shown that the economy does affect turnout (Radcliff 1992), however results are disputed and recent research does identify the significance of such variables although the do not provide an explanation for such an effect, in particular they find a negative effect of inflation and unemployment (Bengstsson, 2004). Under a decentralized structure, regional governments might act as potential lobbies with the central state as has been identified is some studies examining key policy areas such as health care ( Costa-Font and Rico, 2006). Therefore, the evolution of economic activity and employment might be seen as forces providing incentives to put some additional pressure on central governments through regional counterparts. Therefore, the role of the significance of certain economic variables is grounded on the idea that expected benefits of certain elections might not be the same for different individuals. That is those individuals who happen to be unemployed or in a weaker socioeconomic position would be more likely to participate in influencing the implementation of certain social programs.

3. THE INSTITUTIONAL SETTING

3.1 Decentralisation

This study takes advantage of the fact that the Spanish State has a decentralised government structure following the development of the devolution process after the inception of democracy. This has lead to the progressive creation of a set of 17 regionstates ( so called “autonomous communities”, AC) with their own regional parliaments progressively invested with more public policy responsibilities. However, different AC exhibit differences over time in the extent of government responsibilities and whether they can make use of the prerogatives of parliament dissolution and setting the timing of new elections. Furthermore, interactions between the social (reference) groups individuals belong to – which are likely to be regionally specificare argued to affect turnout (Schram and Sonnemans, 1996). Therefore, regional specific effects will be expected to measure institutional variations as well as differences in values and social interactions that take place at the regional level.

Institutional determinants are important. Evidence from the US elections suggest that congressional and state elections produce higher turnout when held together than on their own. Furthermore, evidence from Spanish elections suggests as exhibited in Table 1 that turnout in National elections is systematically higher than turn out in other elections. However, there are wide differences in the variability of elections across regions. Regional calls in some region-states are on average similar to local calls though there is wide variability among elections as the variance coefficient suggests. As expected, European elections seem to be the ones with lowest turnout. On the other hand, inter-regional variability – measured by the coefficient of variation- is similar regardless of the electoral call which suggest that patterns of participation follows similar regional patterns regardless of the elections. Yet, by attempting to explain the forces driving the turn out variability, Figure 1 plots the variance of turnout depending on the regional per capital GDP. Interestingly, we find that variability in the turn out is explained by the GDP which would suggest that richer regions are more likely to exhibit a largest “turnout gap” between national and regional elections, arguably suggesting that individuals in those regions are more likely to rely less on the policies at the regional level and distinguish “high” politics for “low level” politics.

[Insert Table 1 and Figure 1 about here]

3.2 Electoral Calls

Spanish citizens are entitled to vote to choose the members of their autonomous regional parliaments, the parliament in Madrid, and their local authorities as well as to select MEPs for the European elections. Therefore, as in other countries there is some multilevel political competition that influences electoral behaviour. This has lead to the differentiation between first order elections ( national) versus second order elections, although in Spain the similar patterns in electoral behaviour do not allow to envisage a different level of political competition (Pallares and Keating, 2003). Indeed, in Spain it is possible to distinguish national and regional cleavages. Elections within each region have promoted the “regionalisation” of party politician strategy too, so that most candidates at the regional level act as “double agents” subject to their own regional constituents and to party discipline.

The electoral system--with the exception of the Senate, which uses a majority system--is set forth in the electoral law of 1977, updated in 1985, which is based on the d'Hondt system of proportional representation. Although the Constitution prescribed proportional representation, the electoral law finally stipulated a system of party lists based on provincial constituencies, with a minimum of two MPs per province. Four regions Catalonia, the Basque Country, Galicia and Andalucía, as established in the article 151 of the Spanish Constitution can set the timing of their own elections while the other 13 ACs used to hold the elections on the same day. This issue has important implications for voter turn out ( Pallares and Keating, 2003). Traditionally, abstention has been less prevalent in national elections (20-30%) as compared to regional elections (30-40%). Some argue that regional elections lack the mobilization of individuals, political competition which, arguably foster the emergence of a ‘dual voting’ for national and regional parties.

In characterizing abstention, there is agreement within the literature in pointing out that electoral abstention in Spain in rather voluntary and, thus determined by individual rather than compulsory behaviour. On the one hand, elections take place normally on weekends. The institutional organisation of electoral calls minimises

possible barriers for individuals to cast their vote (e.g., early postal voting, inclusion of non-residents etc). Evidence suggests that so called technical or compulsory abstention has declined progressively and in 1991 was only 35% of total abstention (Justel, 1998). In Spain the questioning of the democratic system has exerted some influence on electoral behaviour. This might have lead to the so-called “chronic abstentionists” and rebel abstentionists (Astorika, 1998). There are geographic patterns of abstention whereby turn out is higher in the middle of Spain as compared to the periphery with the expectation of Madrid so called “ abstentions island” (Justel, 1998), although Galicia is the “ abstentionists region” .

3.3 The role of Economic controls

The influence of economic variables has traditionally been dismissed as explaining abstention in Spain with few exceptions. Colomer (1996) finds no association between turn out and the share of public expenditure. Interestingly, in 1983 only 14.4% of public expenditure was decentralized and turnout in regional election was 66%, , in 1991 where 25.4% of total public expenditure was decentralised to regions turn out was 61% and in 1995 with decentralized expenditure being 27% turn out increased moderately to 69%. Therefore, there is no evidence of higher decentralization as influencing voting turn out. However, one might argue that other economic variables might be influential; for example employment and occupation rates, economic activity and inflation. Indeed, there is ample evidence that points towards the existence of a general individual’s response to macroeconomic performance (Kramer, 1971, Paldam, 1991, Powell and Whitten, 1993). The influence of economic voting on turnout is complex to disentangle, as far as individuals might participate to support or penalise a certain incumbents on its assessment of government outcomes (Chappel and Gançalves Veiga, 2000). Therefore, extreme economic conditions such as large inflation rates, large unemployment or a reduction in regional income could be expected to influence turnout in an election. However, given that regional elections are not normally perceived as decisive as nationwide elections one might expect a smothering influence highly dependent on the nature of political responsibilities, which in the Spanish case mainly refers to social policy.

4. DATA AND METHODS

4.1Data

The data has been collected from the publications of the body charged with managing the elections in Spain (Junta Electoral Central, Ministerio del Interior) for several years. From this data we computed the turnout in 100% by each AC. On the other hand economic control variables were obtained from the National Institute of Statistics (Contabilidad Regional de España¹⁰) . The definition of the summary statistics and sources of all variables employed is provided in Table 2.

The regional average turn out in regional elections is about 67%. In those elections, 15% exhibited Type 1 simultaneity, 10% Type two and in some case we find Type 3 and Type 4. Table 2 includes three economic variables to look at. First the GDP in logatuthm (logs) nd employment rate to capture whether there is a socioeconomic vector as explaining electoral turn out. Second, whether, the inflation rate - as in other countries - as an index of consumer capacity places any role in determining the decision to participate in the Spanish regional elections.

[Insert Table 2 about here]

Simultaneity or coincidence of electoral calls with regional calls has been measured as a single variable and besides, it has been splited in all four possible combinations o electoral calls possible. We hypothesise that the heterogeneous effect in different electoral calls would capture the different costs and benefits that individuals might obtain from the act of voting in each electoral call, and hypothetically guide their behaviour. Furthermore, by combining two elections in one, individual’s regret for not having voted might be likely to increase which should encourage people to cast their vote. We distinguish between four different types of simultaneity: Simult I (coincidence of regional and local elections), Simult II (coincidence of European, regional and local elections), Simult III (coincidence of national and regional elections) and Simult IV (coincidence regional and European elections). Moreover, by aggregating all possible form of simultaneity we add up to a measure of total simultaneity between regional calls and other calls designated as Simult-Total .

¹⁰ This data can be retrieved from www.ine.es

4.2 The empirical strategy

The Spanish data allows us to, on the one hand pool all regions and time periods together, as well as build up a panel of regions in order to account for potential time and region specific effects, this should ensure that the results are robust to some regional unobserved heterogeneity. The former is of great importance given that we do not observe the political context of each regional election which is likely to be region specific. Accordingly, we build up a panel of the 17 regions for the period 1980 to 1999. Moreover, as discussed before we distinguish total simultaneity from each specific type of simultaneity separately.

Yet, when performing panel data analysis we draw upon an aggregate measure of simultaneity given that the relatively small number of years that we still have recommends pooling the data together. Overall Table 3 contains a description of the variable measuring electoral coincidence (simultaneity) for all the sample as well as a sub sample of those region-states that do not have the capacity to determine the electoral calendar for comparative reasons. We did not include interaction effects given that they might cancel out with region-specific fixed effects. Interestingly, table 3 indicates that about 27% of all electoral calls experience some form of electoral simultaneity. The most frequent source of simultaneity was between local and regional (54-555%), followed by local regional and European (41%) whilst other forms of simultaneity were considered though they appear to be marginal . It is important to note that in the Spanish setting, given that the set of new region-states were set up after the inception of democracy, electoral simultaneity was an institutional response to avoid that a low turnout could have been put the decentralisation process into question.

[Insert Table 3 about here]

In examining the results we compare the estimates retrieved from ordinary least squares (OLS) or pool data from that of generalised least squared (GLS)¹¹ using both fixed and random effects so that we take advantage of panel data techniques so as to examine the robustness of our results to the inclusion of unobservable effects. One additional concern might be that of enodgeneity of the variable “simultaneity” when regional incumbets could determine the electoral calendar The endogegeneity hypothesis is tested for the total sample and the sub sample of those region-states whose parliaments can call for elections. In both cases we reject the null hypothesis of endoegeity and (Davidson and McKinnon, 1993).

To capture other control determinants of voter turnout we examine the role of economic factors hypothetically affecting the costs and benefits of voting The equation estimated is the following:

\[y _ {i} = \beta_ {0} + Z _ {i} \beta_ {j} + \sum_ {k = 1} ^ {k} \beta_ {k} ^ {\prime} X _ {k i} + v _ {i t} + \varepsilon_ {i}\tag{2}\]

refers to the turnout in the regional elections in region i in year t, The coefficient is expected to indicate the effect of simultaneity in hypothetically boosting turnout if the predicted effects take place. refer to a set of ‘control variables’ that following the theoretical predictions would indicate the presence of the so-called economically driven turnout.

Different specifications differ in whether we control for specific cross-section times series effects or not. This is important because electoral calls are not phenomena that take place every other day, and accordingly they are likely to be influenced by context effects that are region and time specific. Furthermore, a key issue is the structure of the idiosyncratic disturbance and in particular whether random of fixed effects should be employed. The difference relies in that a fixed effect model assumes differences in intercepts across groups or time periods, whereas a random effect model explores differences in error variances ( Balaghi, 1993). We provide several estimates and

¹¹ Given that under the presence of dummy variables OLS models might not produce consistent estimates we employ Generalised Least Squares (GLS) methods as is common practice in order to avoid any

finally on the basis of the Lagrange Multiplier Breusch –Pagan Test we choose the fixed effect model.

5. RESULTS

We report empirical evidence in the following way. First, we report the coefficient of the relevant variable under analysis, namely the effect of electoral simultaneity and then we examine specifically the effect of economic controls. Furthermore, some specification include as well regional specific effects for comparative purposes. Tables 4 and 5 provide the OLS estimates for all the sample and that of region-states subject to the general regime, namely those that could not manipulate the calendar. On the other hand, Table 6 provides panel data estimates.

Table 4 reveals that when splitting simultaneity across different calls the effect of coincidence does not necessarily come up as significance with the expunction of Simultaneity III. This results because simultaneity in each case is compared with all the other electoral calls that contained simultaneous calls too ( estimates (1)-(4). Yet, when we group then (5)-(6) we find that the effect is robust and exhibits a coefficient of 6.7- 6.8 suggesting that simultaneity increases turnout about 7%. Interestingly, we find that after controlling for electoral coincidence, richer regions seem to exhibit lower turnout. Therefore, consistent with the hypothesis that richer regions are more affected by the composition of state level parliaments than those at the regional level.

[Insert Table 4 about here]

As noticed our strategy has been to examine the full sample and the sample of those regions without agenda setting responsibilities. The reason for this rather than to add region specific dummies results from the use of cross-section specific fixed effects that could cancel out the dummy specific effect. Furthermore, a dummy effect in the Spanish setting does not necessarily capture strategic interactions to manipulate turnout as interactions might to in opposite direction which ultimately cancel out. Indeed, incumbents are presumably more interested in maximizing their share of the votes, rather than the turnout rate.

problem concerning the existence of time and fixed effects.

Our model suggests that whereas the predicted average turn out level is 64.2%, simultaneity increases the turn out to 68.7% whereas the lack of simultaneity will lead to an expected of participation rate of 62.7%. When particular sorts of simultaneity are examined, we find that the coincidence of local and regional implies an expected participation rate of 68.2 whereas the lack of coincidence leads to a rate of 65.6% . in the case where in addition to the last three the European elections were added, then participation rate goes up to 69.5%. However, by far the largest increase in the participation rate happens when national and regional elections coincide, in this case the expected participation rate goes up to 74.4% . Finally, when local and general elections coincide the expected participation rate is 69.6%.

When regions that are subject to specific regimes are excluded, we find that interestingly the effect of electoral coincide is even larger so that coincidence increases turnout by 10.8% ( see (10)-(11)) . This results suggests that in those region-states where lower turnout could have questioned the legitimacy of the decentralisation process, electoral simultaneity seem to have expanded turnout more intensively. Another interesting results is that in regions subject without the capapcity to determine the calendar ( general regime) the role of income does not come up as a significant influence of turnout. Indeed, the effect of income was only significant when certain forms of electoral simultaneity were excluded.

[Insert Table 5 about here]

The fact that the coefficient is different amongst electoral calls leads to conclude that the effect is election and context specific. Yet, given the small number of observations, in the following panel data approach analysis we proceed with an aggregate measure of simultaneous elections only . Table 6 contains several specifications for the fixed and random effect and finally we have included a time trend to control for any time specific effect correlated with unobservable variables. Specification (13)-(14) contain the models estimated by random effect which suggests similar results as that of OLS models though now the income specific effects disappear. However, the Breusch Pagan test suggests that a fixed effect specification would be employed which now similar effect for the coincidence variable though interestingly that for general regions now income suggests lower turn out as expected from Figure 1 ( see (15)-(16)). Interestingly, the results holds as when a time trend in introduced. The effect of including fixed effect changes slightly up the coefficients for the model that employs the total sample and the opposite for the restricted sample of region-states that are subject to centrally determine electoral calendar. Finally, the inclusion of the time trend ((17)-(18)) reduced slightly the effect of simultaneity though essentially the OLS results hold , suggesting an expansion of turnout due to coincidence of electoral calls. Furthermore, the poorer the region-state is, the more likely it is to vote in regional electoral calls.

[Insert Table 6 about here]

6. DISCUSSION

This paper has attempted to empirically test the effect of a potentially influential institutional determinant of voter turnout, namely that of holding simultaneous elections calls. Evidence of the effect of electoral simultaneity is in line with rational voting¹² assumptions and indicates that by accumulating several electoral calls it is possible to raise the general turn out. This is especially relevant given the encompassing effects on the legitimacy of the system. Tests for simultaneous elections require often several governmental layers that are subject to political representation in regional parliaments as well as local councils. For countries that are taking part in regional integration processes such as the European Union, this implies electing members of parliaments at the European level.

This study has examined evidence from Spanish regional elections. Given the decentralized institutional structure (1980-1999) adopted by Spain after the introduction of democracy in the late 1970s, a set of new entities so called autonomous communities were formed and progressively regional parliaments have been increasingly relevant political actors in certain policy areas. Yet, potentially low turnout could have questioned the legitimacy of regional elections when the AC where fist created. Our empirical findings, suggest some compelling evidence indicating that on average, turnout rates in regional elections held concurrently with other elections (especially national elections) were higher than in regional elections held separately. The magnitude of this effect is higher (about 9.5-10.8%) for region-states that could not determine the electoral calendar whilst, the average effect was about 6.4% . Moreover, we control for the influence of economic voting, namely some economic indicators which indicate that especially in newly created regions where the legitimacy of regional parliament could be called into question, regional income is a significant determinants of lower voter turn out, suggesting in turn some anecdotal evidence of economic voting that deserves further examination in future studies. The latter is found when we use panel data models and we control for fixed effects, possibly picking up the influence of unobservable variables.

The main contribution of this paper is the finding of a robust influence of simultaneity or electoral call coincidence in increasing turnout. Simultaneity, it is claimed, increases the expected benefits of a “probabilistic influence” of voting as in one single shot individual can participate in more than one call, thus declining the “costs per call”. Furthermore, these results suggest that from an institutional perceptive, governments can actively influence an increase in voter turnout, especially when electoral costs do not seem to stimulate participation. Therefore, attempts to promote voter turn-out should be directed to reducing individuals’ costs. The simpler and easier it is to participate, then the more likely it will be for individuals would cast their vote. In fact, one might argue that simultaneous elections often take place for reasons of efficiency, as far as public expenditures of an electoral contest are minimized. However, one potential caveat is that simultaneity might affect individuals’ reasoning of each electoral call separately and might expose individuals to more intensively to the effect of short sighted emotional voting, because electoral programs of two different elections might be perceived as mixed up together. Frequently individuals’ preferred candidates for local, regional and national elections are from different parties.

¹² Coincident elections might hypothetically affect the subsequent net benefits – both direct or indirectly through expressive - from voting are added whilst tangible costs remain constant – though psychological costs to decide the most suitable candidate might well expand-

It might be the case that people are more prone to vote for the same party in two elections arranged at the same time, however this information is not observed. In the Spanish setting regional governments can influence the electoral calendar to make it coincide with other elections – or prevent them to coincide- then we might expect that politicians might use this feature to stimulate individuals’ votes. Therefore, the coincidence of several elections can become a mechanism on the part of regional incumbents to influence on their interest turnout. This study highlights that in decentralized countries; individuals are still able to distinguish central–state politics which is identified as ‘high politics’ from regional politics. We find evidence of some influence of regional income in determining turnout which arguably ciould be explained by the influence of public sector employment (e.g., in influencing rents to individuals depending on the party in government in regional governments in poorer regions. The only exception in the Spanish setting, is that of two AC – namely the Basque Country and Navarra- collect taxes on their own, therefore in those AC regional calls are more crucial relative to the rest.

References

1. Aldrich, J.H., 1993. Rational choice and turnout. American Journal of Political Science 37, pp. 246–278.
2. Astorika, M (1998). Evolución de la abstención electoral en España: 1976-1991. Del Castillo, P (ed). Comportamiento Político y electoral. Centro de Investigaciones Sociológicas Madrid.
3. Barry, B (1970). Sociologists, Economists and Democracy. Collier-MacMillan, London.
4. Bengtsson, A (2004). Economic voting: The effect of political context, volatility and turnout on voters’ assignment of responsibility. European Journal of Political Research 43: 749–767, 2004
5. Blais, A , Massicotte, L and Dobrzynska, A (2003). Why is Turnout Higher in Some Countries than in Others? Elections Canada,, www.elections.ca.
6. Blais, A and Dobrzynska, A (1998). Turnout in Electoral Democracies. European Journal of Political Research, 33: 239-261.
7. Blais, A, Young, R and Lapp, M (2000). The Calculus of Voting : An Empirical Test-European Journal of Political Research, 37: 181-201.
8. Brenan, H.G. and Lomansky, E (1993). Democracy and decision. Cambridge: Cambridge University Press.
9. Caplan, B. 2001 ‘Rational Ignorance versus Rational Irrationality,’ KYKLOS,. 54: 3– 26.
10. Chappell, H.W and Gonçalves Veiga, L (2000). Economics and Elections in Western Europe 1960-1997. Electoral Studies, 19: 183-197.
11. Colomer, J.M (1991). Benefits and costs of voting. Electoral Saudíes, 10:313-325.
12. Colomer, JM (1996). La Politica en Europa. Ariel, Barcelona.
13. Costa-Font, J and Rico, A (2006).Vertical Competition in the Spanish National Health System (NHS), Public Choice, 128( 3-4): 477-498.
14. Dhillon, A and Prelata, S (2002). Economic Theories of Voter Turnout. Economic Journal, 112: 332-352.
15. Downs, A (1957). An Economic Theory of Democracy. , Harper and Row, New York (1957).
16. Ferejohn, J and Fiorina, M (1974). The Paradox of Not Voting: A Decision Theoretical Analysis. American Political Science Review 68 (1974), pp. 525–536.
17. Font, J., A. Bosch, J. Contreras, and G. Rico. 1999. Participació i sistema electoral: propostes per al debat. Barcelona: Fundació Jaume Bofill.

Justel, M (1998). Composición y dinámica de la abstención electoral en España. Del Castillo, P (ed). Comportamiento Político y electoral. Centro de Investigaciones Sociológicas Madrid.

Kayser, Mark A. 2005: Who Surfs, Who Manipulates? The Determinants of Opportunistic Election Timing and Electorally Motivated Economic Intervention. American Political Science Review 99(1): 17-28

1. Kanheman, D and Tversky, A (1979). Prospect Theory: a decision making under risk. Econometrica, 47: 263-292.

Lijphart, A., 1997. Unequal Participation: Democracy Unresolved Dilemma. American Political Science Review 91 1, pp. 1–14.

1. Lewis-Beck, Michael S. and Martin Paldam. 2000. ‘Economic Voting: An

Introduction.’ Electoral Studies 19: 113-121.

1. Mueller, D.C (1989). Public Choice II. Cambridge, Cambridge University Press.
2. Opp, K-D (2001). Why do people vote? The cognitive-illusion Proposition and Its Test .Kyklos, 54:355-378.
3. Pallares, F and Keating, M (2003). Multi-level electoral competition. Regional and party systems in Spain. European Urban and Regional Studies, 10(3): 239-255.
4. Powell, B (1982). Contemporary Democracies: Participation, Stability and Violence. Harvard University Press, Cambdridge.

Radcliff, B. (1992). The welfare state, turnout and the economy: A comparative analysis. American Political Science Review 86(2): 444–454.

1. Riker, W and Ordeshook, P.C (1968). The Theory of Calculus of Voting. American Political Science Review, 62: 25-43.

Schram, A and J. Sonnemans (1996), Why people vote: experimental evidence, Journal of Economic Psychology 17:417–442.

Schram, A.J.H.C., 1991. Voter Behavior in Economic Perspective. , Springer-Verlag, Heidelberg

Smith, J., 1999. Europe's Elected Parliament, Sheffield Academic Press, Sheffield.

Table 1. Average turn out in each Spanish region electoral calls (per cent) 1980-1999

Source: Complied from the reports from the Central Electoral Authority, Ministry of Internal Affairs (Junta Electoral Central, Ministerio del Interior).

Figure 1. Variability in Regional Turn Out and GDP per capita

Table 2. Variable Definition, descriptive statistics, Expected Sign and Sources

Notes: A. Complied from the reports from the Central Electoral Authority, Ministry of Internal Affairs (Junta Electoral Central, Ministerio del Interior). B. Instituto Nacional de Estadística, Contabilidad Regional de España ( www.ine.es)

Table 3. Percentage of simultaneous turnout by type of AC

Table 4. Turn out in all regional electoral calls 1980-1999 (Ordinary Least Squares (OLS) Estimates)

Note: ^a 1% significance. ^b5% Significance.

Table 5. Turn out in regional electoral calls 1980-1999 (Ordinary Least Squares (OLS) Estimates)

Note: ^a 1% significance. ^b5% Significance.

Table 6. Turn out in regional electoral calls (both total and general) 1980-1999 (Generalised Least Squares (GLS) Estimates)

Note: ^a 1% significance. ^b5% Significance