Economic Conditions and Electoral fractionalization. The "Hostage Voter" e ect.

Transcription

1 Economic Conditions and Electoral fractionalization. The "Hostage Voter" e ect. Dimitrios Xefteris University of Cyprus Konstantinos Matakos University of Warwick May 3, 2010

2 Abstract We develop a simple model of opportunistic voting in a multi-party political environment with two dominant parties, where voting is expressive. That is, given that the electorate is large each individual vote does not a ect the nal outcome. Hence, the only incentive that citizens have, to publicly express their political preferences, is to receive the rents when they side with the winner and avoid the cost of supporting the loser. In such a set up, where a fraction of agents is employed and the rest are unemployed, we show how macroeconomic variables such as debt, unemployment rate and social expenditures a ect the voting decision of the agents. In particular, we show that fractionalization in the parliament moves in the same direction with debt and social expenditures and in the opposite one with unemployment. In the empirical section of the paper we test the theoretical predictions of the model and we nd the results to be statistically signi cant and qualitatively important in magnitude. Keywords: expressive voting, fractionalization, debt, unemployment, social expenditures

3 1 Introduction-Related literature A brief overview of elections results in representative democracies around the world indicates a wide dispersion of parties political power fractionalization. That is, in several countries just a small number of political parties (usually two) receive the lion s share of votes, while in other political environments a lot of parties receive a respectable vote share. Political science has always been interested in this phenomenon and has adopted various approaches to explain it. Duverger (1951), rst noted that this divergence of electoral fractionalization among countries may be attributed to the particular political institutions that characterize the political environment of each state. This idea suggests that if, for example, a majoritarian electoral rule is applied, then we should expect a low electoral fractionalization; the two larger parties shall be receiving the vast majority of votes, and when a proportional electoral law is in place, electoral fractionalization should be high. Individual voters are "driven" to vote for the larger parties in the rst case, as a vote for a small party in a majoritarian system is nothing else than a lost vote. In a sense, this accounts to citizens voting strategically, since voting for a minor party that has minimal chances of wining may alter the election result in their disfavor. In the second case though, voting for small parties is not so "strategical". Colomer (2005), Ergun (2007), 1

4 Xefteris and Matakos(2009) point out that the causality of the phenomenon could be very well completely reverse. Political institutions in an environment of representative democracy do not appear "out of the blue" and are constantly subject to alterations by the political parties. That is, if a political opportunity arrises, certain parties may have incentives to change the institutions for their bene t in order to solidify the status quo. Imagine, for instance, that under proportional representation two parties receive a very large share of votes which would allow them to change the electoral rule. In this case we should expect that the two dominant parties will attempt to distort the proportionality of the electoral rule and they would try to implement a more majoritarian one, in order to reinforce their position in the forthcoming elections. From a purely economic perspective several studies have been conducted that relate electoral outcomes with macroeconomic variables. Persson and Tabelini (2003) and Barro (1996) for example, have extensively analyzed the impact of institutions and other political indicators on the economic performance of a country.. What we intend to do in this paper is, once more, to try to walk the same path in the opposite direction. That is, we shall attempt to investigate how macroeconomic variables, such as the public debts, unemployment and social expenditures, in uence the political environment; and more speci cally electoral fractionalization. 2

5 The in uence of economic variables on electoral outcomes has been studied by a large variety of scholars. Both individual (Fiorina 1981; Kinder and Kiewiet 1979,1981; Kiewiet 1983) and country-level (Kramer 1971; Arcelus and Meltzer 1975; Bloom and Price 1975; Tufte 1978; Kinder Adams, Gronke 1989; Campbell 2000; Holbrook 2008) economic conditions are found to have a severe impact on voting behavior. The main focus of this literature is to examine if economy is a determinant factor of the elections winner. By the means of the following study we attempt to shed some light on a di erent element of the electoral outcome; the fractionalization of parties vote shares. We build and solve a simple theoretical model and then test the results using a panel data analysis (few words about data structure). We consider that voters are both ideological and rent seeking at the same time and that ballots are public; each voter publicly reveals her vote. The theoretical result indicates that electoral fractionalization depends positively on public debts and social expenditures and negatively on unemployment. The intuition behind these results is relatively simple. Voters expect high rents if they support a party that has a good chance of winning the elections when public nances are healthy (low debts). Voters with low personal income (unemployed population) are severely less risk averse than other, richer, voters, and thus, choose to support a large party even if this party is not the one that o ers the best ideological platform for them, as they expect high rents. These low income (unemployed) part of the population, has 3

6 less incentives to abandon their preferred party in ideological terms when social expenditures are high; a respectable level of living could be preserved even without the rents and, thus, they would generically support the party that they ideologically prefer. The empirical analysis is in complete accordance with these ndings. We observe that electoral fractionalization is positively associated with public debt and social expenditures and negatively with the unemployment rate. More speci cally, we show that an increase in social expenditures or deterioration in public nances, depicted through higher public debt levels, are associated with an increase in the level of fractionalization. That is, public vote is more dispersed among the parliamentary parties. On the other hand, we also show that an increase in the level of unemployment is negatively associated with the fractionalization index, that is, public vote is more concentrated in the parties that have higher probability of wining in the forthcoming elections. Furthermore we test our empirical model under various econometric speci cation using both xed and random e ects estimators. We also use a number of controls in order to account for biases. To our reassurance the model is very robust to all the above speci cation variations. Moreover, results are statistically signi cant under both speci cations. 4

7 2 Theoretical Model 2.1 Assumptions As said, we shall rst solve a simple theoretical model and then test its implications by the means of an empirical study. We consider that the electoral competition is conducted between four parties: = fl; L; R; rg where l is a party with an extreme-left ideology, L is a party with a moderate-left ideology, R is a party with a moderate-right ideology and r is a party with an extreme-right ideology. We shall assume that a party s index! 2 coincides with its ideological platform; that is party L s ideological platform is L. The party that receive the larger share of votes wins the election and forms a single party government. Parties are solely policy motivated in this framework; once elected in o ce they implement their preferred policy. We also consider a large electorate (in nite number of voters) which is composed of individuals that receive utility from the policy that shall be implemented and from their personal income. The utility function of a voter i when party! 2 is in o ce will be the following: U i (!) = i (!) + u(w i ) 5

8 where i (!) is the utility that individual i receives from announcing policy! 2 and u(w i ) is the utility that the same individual receives from her personal income w i. We assume that the political preferences of the voters are single-peaked in the political axis. For example voter who receives the highest utility from an extreme-left policy, prefers moderate-left to moderate-right and moderate-right to extreme right. Practically speaking, the rst term i (!) of the utility function measures the ideological proximity of a voter i with respect to some party! 2 : Given that voters preferences are single-peaked, it might be useful to think of i (!) as the utility function induced by the Euclidean metric (known as Euclidean preferences) with its domain being the real line. Nevertheless, this extra assumption is not a necessary condition to yield the results in the following sections. Hence, we abstain from making this assumption and we just note that Euclidean preferences, or even quadratic preferences for that matter, are just a subset of the set of preference relations that satisfy the single-peakedness property. As far as the income part of the utility is concerned, we assume that u 0 (w i ) > 0, u 00 (w i ) < 0 and that an individual with an initial wealth w 0 i 2 fw(s); W g has an income: w i = w 0 i + r(d) if she has voted for the party that won the elections and 6

9 w i = w 0 i c if she has voted for any of the other parties that are in opposition. Individuals with high initial wealth W are considered to be receiving their income from their activities in the free market, and individuals with low initial wealth w(s) are unemployed individuals whose income depends on public expenditure s (unemployment remuneration, public services, etc.). Further assume that a fraction q of the society is unemployed, that is w 0 i = w(s); and the remaining (1 q) has initial wealth w 0 i = W. Parameter r(d) represents the rents that the supporters of the winning party receive and parameter c > 0 is the cost that voters of the opposition receive. The rents depend on the state s nancial health d; the more healthy the public economics are (that is, the lower the level of debt d is) the more dividends a government can distribute to its supporters. In other words, r(d) is strictly decreasing in d: Ballots, in this model, are open. That is, each individual has to publicly announce its vote. Actually, this is all that matters. Since there is an in nite amount of voters no one is pivotal and hence, the election result does not depend on any individual s i vote. The elections result just depends on the probability distribution p! 2 [0; 1];! 2 and all that citizens do, is to just announce publicly their choice of party! 2. As a result, by construction of the model, there 7

10 is no strategical dimension in the voting decision of an individual i. This is so, because, one of the aims of this analysis is to investigate the rent seeking behavior of voters. If we consider secret ballots, then no voter can claim rents from the winning party (as it cannot be proved that this voter actually supported the party) and no voter can receive costs for opposing to the winner during the electoral campaign period. The voters have exogenous information about the winning probability of each party. This information is summarized by the probability distribution: p! :! [0; 1]; such that P!2p! = 1 where p! is the probability of party! 2 to win the elections. During the elections each party receives a vote share, denote by: v! 2 [0; 1] such that P!2v! = 1 and following Rae (1968) and Laakso and Taagepera (1979) we can de ne the electoral fractionalization of the described procedure by the following formula. F = P 1 v! 2!2 8

11 2.2 Analysis-Results In this section, we will proceed with our theoretical analysis and provide our main result with the aid of an illustrating example. Without any loss in generality, simply assume that society we are examining is characterized by the following probability distribution: p L = p R = 1 2 and p l = p r = 0: That is, we have a bipartisan political system, with two dominant parties that have a positive probability of winning in the forthcoming elections. The aim of this example is to study how the vote share of each party, and as a consequence the Rae fractionalization index as well, varies with changes in the economic policy variables such as the unemployment rate, public debt and social expenditures. In order to do so, we have to examine the voting behavior of three categories of agents, by means of comparing their expected utilities of announcing one of the four parties. More speci cally, we compare those who have to decide whether to announce l or L, those between L and R and nally those that have to decide among R and r. It is obvious that, given the above probability distribution and the fact that preferences satisfy single-peakedness, any other comparison between alternatives is pointless. 9

12 Let us rst focus on those agents that have to decide whether to announce L or R. Given that both dominant parties have equal probability of winning in the forthcoming elections the income component u(w i ) of the expected utility function EU i (!), plays no role regardless of the initial wealth endowment w 0 i being w(s) or W: Hence, for both employed and unemployed agents in this economy, their decision on whether to publicly announce their support for either party L or R; solely depends their ideological proximity to each of the two parties. That is to say, it solely depends on i (!): If, for instance, parties L and R were symmetrically positioned around some point x, and if a continuum of voters was uniformly distributed on that interval, then the two parties would split the public vote based on the equidistant criterion: the voter "sitting on point x" would have been indi erent between announcing support for party L or R:In our case, since we abstain from making those explicit assumptions, what matters is that the decision on whether to announce L or R does not depend on income and thus by the economic policy parameters of the model. In turn, this implies that the share of votes that each of those two parties receives does not depend on the policy parameters as well. Hence, the Rae fractionalization index is not a ected. The interesting part of the analysis occurs when we examine the choice between parties l and L. The results obtained here can also be extended without any loss in generality in the case where citizens choose between parties R and r due to the symmetry of the problem. Here in this case, 10

13 when we examine the income component u(w i ) of the expected utility function, we observe that in order for both employed (with initial wealth W ) and unemployed (with w(s)) to prefer the smallest party (to the left or to the right) they have to be "closer" to that party, in terms of ideological proximity, measured by i (!): That is, in order for an individual i to be indi erent, it must be the case that for both types of agents the following conditions hold: i (l) > i (L) and i (r) > i (R) The above condition simply says that given that the two dominant parties have strictly positive probability of winning the election, announcing support for them has a strictly positive monetary e ect, namely r(d). Hence, if we are to equalize expected utilities, such that EU i (l) = EU i (L), it must be the case that the monetary loss associated with announcing the support for a minor party, captured by u(w i ), will be o set by higher ideological proximity with that party. In other words, this agent must be "closer" to that party. This trade-o reveals that the share of votes that is split between a minor and a dominant party is not invariant to the economic policy parameters, such as public debt d, social expenditures s and the unemployment rate q. Therefore, the fractionalization index must also vary with those parameters. 11

14 Yet, even though the analysis above holds true for both types of agents regardless of their initial wealth, the magnitude or the degree of ideological proximity and "closeness" to the minor party varies with w 0 i, since the lower the initial wealth endowment is the more risk averse the individuals are. Technically speaking, this is due to the concavity of u(w i ). This implies that, for an unemployed person, with w 0 i = w(s); to be indi erent between supporting a dominant or a minor party, she has to be relatively more close to the minor party compared to an employed person with w 0 i = W: To sum up the discussion so far, we have argued that in these two cases economic policy parameters do a ect the share of votes that the parties receive and as such they do have an impact to the fractionalization index. Our next step then is to examine the direction and the magnitude of this e ect to the above mentioned index. That is we will perform a comparative statics exercise. Although we will formally and analytically present those results in the section that follows this discussion, we will take the opportunity now to comment on the qualitative aspect, in order to clarify the underlying intuition. From the analysis above, it is clear that the share of votes that is distributed among parties l and L (and by analogy among r and R) is a ected by unemployment rate q: As the number of agents in the economy with lower initial wealth endowment w(s) increases the vote share of the dominant parties L and R will also increase, resulting in lower value for the fractionalization index. 12

15 That is unemployment and fractionalization are negatively associated. The intuition behind this result is simple. As the fraction of unemployed agents rises, the fraction of people who can a ord supporting the minor party because they feel ideologically closer to it shrinks. An unemployed person has to be relatively more partisan, ceteris paribus, to an employed one in order to announce support for a minor party. As a result, an increase in the unemployment rate reduces the support and, of course, the vote share for the two minor parties in favor of the two dominant ones. Hence, fractionalization decreases. Coming to our second key economic parameter, social expenditures s, rst we note that behavior of the employed agents with initial wealth W is not a ected by changes in s, since their income is independent of social expenditures. In analyzing the behavior of agents with initial wealth w(s) < W, we note that as social expenditures are rising so does the initial income of the unemployed individuals. Keeping their ratio q in the economy xed, we observe that an increase in s makes them relatively more well o. As a result, some of those are willing to sacri ce the bene t that they would have otherwise received by announcing the support of one of the two dominant parties and vote for the party that is "closer" to them. Hence, as s increases, the fraction of those unemployed agents that are willing to sacri ce the rents of supporting the winner and behave more ideologically also increases. That is, the vote shares of the two dominant parties are shrinking whereas that 13

16 of the two minor ones increases with social expenditures. This implies that fractionalization and social expenditures move in the same direction. Intuitively, it is as if the state indirectly subsidizes the smaller parties through the means of providing more "ideological autonomy" to the least wello citizens of the society, by liberating them from the tyranny of "voting for money". Thus, social expenditures can be viewed as an indirect mechanism that supports pluralism and fractionalization. Finally our third key policy parameter of interest is the level of public debt d. It is clear that as public nances deteriorate (increase in d) the size of rents r(d) is decreasing. Hence for both types of agents, either employed or unemployed, the incentive to announce support for the dominant parties is decreasing. That is, voters become more ideologically driven, as the size of rents that they can extract by announcing support for one of tho dominant parties is shrinking, due to the deterioration of public nances. As a result, the vote share of minor parties l and r (and in turn fractionalization) increases as public debt is increasing. The Proposition that follows, summarizes the comparative statics results that were extensively discussed above. But, before presenting the main theoretical result of this paper, we nd it useful to rst establish the following lemma. Lemma 1 Let assumptions outlined above hold. Given the following probability distribution p! : 14

17 (p l ; p L ; p R ; p r ) = (0; 1 2 ; 1 2 ; 0), regardless of initial wealth endowment (that is w0 i 2 fw(s); W g ) and the size of rents r(d) and cost c; EU i (L) = EU i (R) if and only if i (L) = i (R): The importance of this lemma is the following: it just states that the share of votes distributed among the two dominant parties is xed and it does not vary with the economic policy parameters. Hence, any e ect that those parameters have on the fractionalization index come through the changes they cause to the share of votes that is to be distributed between the minor parties l and r and the dominant ones L and R; respectively. Proposition 1 Given the assumptions underlined above w.r.t the probability distribution p! 2 [0; 1];! 2 we can state the following: (i) As the fraction of unemployed people in the economy increases (that is q! 1), ceteris paribus, the electoral fractionalization index F is decreasing. (ii) As public (social) expenditures s are increasing, ceteris paribus, the electoral fractionalization index F is increasing. (iii) As public debt d is increasing (and hence distributed rents r(d) are decreasing), ceteris paribus, the electoral fractionalization index F is increasing 1. 3 Empirical Analysis 3.1 The Data 1 Proof in the Appendix A. 15

18 We use the observations of "The Comparative Political Data Set " which is a collection of political and institutional data which have been assembled in the context of the research projects Die Handlungsspielräume des Nationalstaates and Critical junctures. An international comparison directed by Klaus Armingeon and funded by the Swiss National Science Foundation. It consists of (mostly) annual data for 23 democratic OECD countries for the period of 1960 to In the cases of Greece, Spain and Portugal, political data were collected only for the democratic periods 2. The data set is suited for cross national, longitudinal and pooled time series analyses. In what follows we will use this data set to conduct a cross-national panel data analysis, using a 2SLS xed e ects estimator. The data set contains some additional demographic, socio- and economic variables. A few variables have been copied from a data set collected by E. Huber, Ch. Ragin, J. Stephens, D. Brady and J. Beck eld (2004), as well as from a data set collected by D. Quinn. 3.2 The model We estimate the following model using a 2SLS xed e ects estimator, where we instrument debt, unemployment and social expenditures with their lagged variables over a period of four years, which is the average electoral cycle in the OECD countries. The reason we have decided to use a 2 Data for Greece are missing during the period Data for Portugal are missing until 1975, and for Spain until

19 2SLS xed e ects estimator is related with problems of endogeneity that a standard OLS estimator would have had. Since recent literature in political economy has documented that political cycles and changes of parties that hold o ce can indeed have an e ect on macroeconomic variables such as debt, unemployment rate, de cit, social expenditures etc., it would have been di cult to estimate the direction of causality. To put it more simply, would it be the case that those macroeconomic variables (debt, unemployment, social expenditures) a ect the voting behavior of the agents and as a result, the fractionalization in the vote and the nal outcome of the electoral process, or is it the other way around. That is, the outcome of the elections (left vs. right or single-party vs. coalition governments) a ects in turn the macroeconomic variable themselves. In order to deal with this endogeneity problem, we instrument the macroeconomic variables (debt, unemployment and social expenditures) with their lagged values over a period of four years. Hence, by doing so, we can escape from this problem. We also perform the relevant Sargan test which rules in favour of the validity of our instruments. Furthermore recent evidence in political science and economics have document that the electoral results and hence, the fractionalization of the public vote can critically depend on individual country speci c characteristics such as irregularities in the electoral rules, country speci c demographics (minority parties etc.), urbanization, the structure of the political system itself, constitutional 17

20 arrangements, the political history of each country and many others. In order to deal with this problem as well we use a xed e ects (FE) estimator which is suggested by the Hausman test as opposed to a random e ects (RE) estimator. Formally we estimate the following model F i;t = a 0 + b 1 (Debt) i;t + b 2 (Unemployment) i;t + b 3 (Soc:Expenditures) i;t + b 4 (Disproportionality) + X i;t + Controls + u i;t where the dependent variable F i;t is the Rae fractionalization index for each country over the period and the independent variables are all measured in percentage terms. More specifically debt and social expenditures are measured as a percentage of GDP, whereas unemployment is measured as a percentage of the active labor force. For a more detailed description of the those variables, the instruments and some other key variables of the data set we have included a more detailed list in the second appendix of the paper. 3.3 Results The estimated results of the above 2SLS model are presented in Table 1. The results in Columns 2 and 3 will be the main focus of this section. Although the results in all 5 columns are qualitatively equivalent (since they have the same sign across columns and they are numerically close) we have decided to focus on those two since they are the two simplest ones. 18

21 At this point, we can take the opportunity to report that we have found our model to be very robust in various changes of speci cations. Given the richness of the data set in terms of the control variables we have tried many possible speci cation using di erent measures for each control variable (that is absolute, relevant, and Gallagher disproportionality measures). We have also done the same for the instrumental variables. The results were qualitatively equivalent and statistically signi cant in all those speci cations. Therefore, since the changes in magnitude were not statistically signi cant from the ones reported here in Table 1, those estimates are not reported in this paper. 3.4 Discussion of the Results e par-ele e par-ele-2 e par-ele-3 e par-ele-4 e par-ele-5 (1) (2) (3) (4) (5) debt (.003) (.003) (.003) (.004) (.004) unempl (.018) (.018) (.018) (.018) (.018) socexp-t (.044) (.044) (.042) (.044) (.047) dis-rel (.629) (.678) dis-gal (.014) (.015) instcons (.074) (.078) cons (.789) (.753) (.743) (.836) (.868) e(n) e(r2) e(f) Table 1: 2-SLS xed e ects estimator. Dependent variable electoral fractionalization (e par_ele). Robust standard errors in parantheses. (***) statistically signi cant at the 1% level, (**) statistically signi cant at the 5% level, (*) statistically signi cant at the 10% level. 19

22 4 Concluding remarks To sum up, in this paper we have attempted to give an alternative explanation to the phenomenon of varying electoral fractionalization in western representative democracies. Rather than arguing in the traditional way, that the degree of electoral fractionalization depends upon the existing electoral rule, we have developed an analytical model that relates electoral fractionalization with economic conditions. In speci c, we have attempted to show how opportunistic voting can a ect electoral fractionalization in a democracy. To this aim, we have built a model of bipartisan political competition where voting is expressive (open ballots). Of course, this is not to say that opportunistic voting is the sole determinant of fractionalization of public vote in a democracy. There might be other factors, such as ideology, that a ect citizens voting decisions, which in turn determine the level of fractionalization. Yet, as our theoretical model suggests, there might be a signi cant relationship between economic conditions, such as public debt, unemployment and social expenditures, and fractionalization. In addition to the theoretical predictions of our model, we set out to test empirically the above mentioned hypothesis with respect to the determinants of electoral fractionalization. Our empirical ndings, as presented in the paper, conform with the predictions of the model. In fact, we nd 20

23 a statistically signi cant relationship among electoral fractionalization and three main economic indicators: public debt, unemployment and social expenditures. To be more precise, an increase in public debt and social expenditures is positively associated with an increase in fractionalization, whereas an increase in unemployment rates is found to be negatively associated with a decrease in the fractionalization index. Finally, we conclude this study by conducting robustness checks to our empirical ndings that con rm once more the statistically signi cant and important in magnitude relationship between macroeconomic indicators and electoral fractionalization. 5 References References [1] Arcelus, F., and A. H. Meltzer The e ects of Aggregate Economic Variables on Congressional Elections. American Political Science Review 69: [2] Armingeon Klaus, Panajotis Potolidis, Marlène Gerber, and Philipp Leimgruber. "Comparative Political Data Set I ". Institute of Political Science. University of Berne, [3] Barro, Robert J. "Institutions and Growth, an Introductory Essay." Journal of Economic Growth, June 1996, 1(2), pp

24 [4] Bloom, Harold, and D. Price Voter Response to Short-run Economic Conditions: The Asymmetric E ect of Prosperity and Recession. American Political Science Review 69: [5] Campbell, James E., and James Garand Before the Vote: Forecasting American National Elections. Thousand Oaks, CA: Sage Publications. [6] Colomer J.M., It s Parties that Choose Electoral Systems (or, Duverger s Laws Upside Down), Political Studies, vol. 53, 2005, pp [7] Cox, Gary W., and J. Morgan Kousser, "Turnout and Rural Corruption: New York as a Test Case", American Journal of Political Science 25(4): , 1981 [8] Duverger M., Political Parties: Their Organization and Activity in the Modern State, Wiley Press, New York, [9] Ergun S.J., From Plurality Rule to Proportional Representation, unpublished paper (mimeo), Universitat Autonoma de Barcelona, [10] Fiorina, Morris RetrospectiveVoting in American National Elections. New Haven, CT: Yale University Press. 22

25 [11] Holbrook, Thomas M Incumbency, National Conditions, and the 2008 Presidential Election. PS: Political Science and Politics 41: [12] Kiewiet, D. Roderick Macroeconomics and Micropolitics: The Electoral E ects of Economic Issues. Chicago: University of Chicago Press. [13] Kinder, Donald, Gordon S. Adams, and PaulW. Gronke Economics and Politics in the 1984 American Presidential Election. American Journal of Political Science 33: [14] Kinder, Donald, and D. Roderick Kiewiet Economic Discontent and Political Behavior: The Role of Personal Grievances and Collective Economic Judgments on Congressional Voting. American Journal of Political Science 23: [15] Kinder, Donald, and D. Roderick Kiewiet Sociotropic Voting: The American Case. British Journal of Political Science 11: [16] Kramer, Gerald Short-Term Fluctuations in U.S. Voting Behavior, American Political Science Review 65: [17] Laakso, Markku and Rein Taagepera (1979): E ective number of Parties: A measure with Application to West Europe. Comparative Political Studies 12, No.1:

26 [18] Persson T. and G. Tabellini, The Economic E ects of Constitutions, MIT Press, Cambridge, MA, [19] Rae, Douglas W. (1968): A note on the Fractionalization of Some European Party Systems. In: Comparative Political Studies, 1, [20] Tufte, Edward Political Control of the Economy. Princeton, NJ: Princeton University Press. 6 Appendix 6.1 Proofs 6.2 Data Description VARIABLE LIST General variables year Year of observation country Country names countryn Country code: 1 Australia; 2 Austria; 3 Belgium; 4 Canada; 5 Denmark; 6 Finland; 7 France; 8 Germany; 9 Greece; 24

27 10 Iceland; 11 Ireland; 12 Italy; 13 Japan; 14 Luxembourg; 15 Netherlands; 16 New Zealand; 17 Norway; 18 Portugal; 19 Spain; 20 Sweden; 21 Switzerland; 22 United Kingdom; 23 USA. Notes: Data up to the end of 1990 are for the Federal Republic of Germany before reuni cation only (West Germany); unless otherwise indicated, they are for the whole of Germany from 1991 onwards Party-system rae_ele Index of electoral fractionalization of the party-system according to the formula [F ] proposed by Rae (1968). rae_ele = 1 mx (v i ) 2, where v i is the share of votes for party i and m i=1 is the number of parties. Period covered: Missing: none. Source: own calculations. rae_leg Index of legislative fractionalization of the party-system according to the formula [F ] proposed by Rae (1968). 25

28 rae_leg = 1 mx (s i ) 2, where s i is the share of seats for party i and m i=1 the number of parliamentary parties. Period covered: Missing: none. Source: own calculations. e par_ele E ective number of parties on the votes level according to the formula [N 2 ] proposed by Laakso and Taagepera (1979). The e ective number of parties carries the same information as the Rae-Index and is calculated from this index as follows: ef f par_ele = 1=(1 rae_ele) Period covered: Missing: none. Source: own calculations. e par_leg E ective number of parties on the seats level according to the formula [N 2 ] proposed by Laakso and Taagepera (1979). The e ective number of parties carries the same information as the Rae-Index and is calculated from this index as follows: ef f par_leg = 1=(1 rae_leg) 26

29 Period covered: Missing: none. Source: own calculations. dis_abso Index of absolute disproportionality. A score of 1 means, for example, that on the seats level there is, on average, one (e ective) party less than on the votes level. Absolute refers to the calculation of the di erences between the e ective number of parties in elections and parliaments without taking into account the fractionalization of the party system. The stronger the fractionalization of the party system the more likely are, ceteris paribus, high values of dis_abso. The index is calculated as follows: dis_abso = effpar_ele effpar_leg Period covered: Missing: none. Source: own calculations. dis_rel Index of relative disproportionality. This index is weighted on the total number of parties on the votes level. It is calculated as follows: dis_rel = (ef f par_ele ef f par_leg)=ef f par_ele 27

30 Period covered: Missing: none. Source: own calculations. dis_gall Index of disproportionality according to the formula [least squares] proposed by Gallagher (1991). Calculated as follows: dis_gall = s 1 2 mx (v i s i ) 2, where v i is the share of votes for party i, i=1 s i is the share of seats for party i and m the number of parties. Period covered: Missing: none. Source: own calculations Macroeconomic data gdpgr Growth of real GDP, percentage change from previous year. Period covered: Missing: Australia: 1991; Austria: 1984; Germany: 2002; Greece: 1990; Italy: 2003; New Zealand: 1990; Spain:

31 Source: Until 1970: OECD Economic Outlook, various years; 1971 onwards: OECD Factbook 2009: Economic, Environmental and Social Statistics - Online Version, (Download: ). debt Gross government debt ( nancial liabilities) as a percentage of GDP. Period covered: For most countries from 1970 onwards or earlier, except for Australia (1988), Denmark and Iceland (1980), Finland (1975), Ireland (1974), Luxembourg, Spain and Switzerland (1990), New Zealand (1993) and Portugal (1995). Missing: see note below. Source: SourceOECD, Economic Outlook, No. 77, (Download: ); SourceOECD, OECD Economic Outlook Database, Economic Outlook: Annual and quarterly data, Vol. 2008, release 01, (Download: ). de cit Annual de cit (government primary balance) as a percentage of GDP. Period covered:

32 Missing: Denmark: 1971; Luxembourg and Switzerland: ; Netherlands: ; New Zealand ; Norway: ; Portugal: Source: SourceOECD, OECD Economic Outlook Database, Economic Outlook: Annual and quarterly data, Vol. 2008, release 02, (Download: ). openc Openness of the economy in current prices, measured as total trade (sum of import and export) as a percentage of GDP. Period covered: Missing: Germany: Source: Penn World Table 6.2, (Download: ) Labour force data ttl_labf Total labour force, in thousands. Period covered: Missing: USA:

33 Source: SourceOECD, Employment and Labour Market Statistics, Labour force statistics - Summary tables. Vol.2008, release 01, (Download: ). emp_un Unemployed, in thousands. Period covered: Missing: Iceland: 1960; Luxembourg: Source: see ttl_labf. unempl Unemployment rate as a percentage of civilian labour force. Period covered: Missing: see emp_un. Source: see ttl_labf. st_unemp Standardised unemployment rates (series adjusted for seasonal variations). Period covered: For most countries from the early 1980s onwards or earlier, except for Austria (1993), Iceland (1991) and Switzerland (1992). Source: SourceOECD, Main Economic Indicators Comparative 31

34 subject tables, Vol. 2009, release 02, (Download: ). Note: Germany : Data is taken from Main Economic Indicators, Vol. 2008, release 01 (Download: ) Public social expenditure and revenue data sstran Social security transfers as a percentage of GDP. Social assistance grants and welfare bene ts paid by general government (bene ts for sickness, old-age, family allowances, etc.). Period covered: Missing: Canada: 1999/2000; Greece, Iceland, Portugal and Spain: ; Luxembourg: ; New Zealand: ; Switzerland: 2000; USA: Source: Until 1969: OECD Historical Statistics, various years; 1970 onwards: SourceOECD, National Accounts Statistics, Historical Statistics - Structure or composition of certain economic aggregates, Vol. 2002, release 01, 32

35 (Download: ). socexp_t Total public social expenditure as a percentage of GDP. Period covered: Missing: Austria: , ; Iceland: ; Norway: , Source: OECD (2007), Social Expenditure Database (SOCX), , (Download: ). socexp_c Public social expenditure in cash as a percentage of GDP. Period covered: Missing: see socexp_t. Source: see socexp_t. socexp_k Public social expenditure in kind as a percentage of GDP. Period covered: Missing: see socexp_t. Source: see socexp_t. health Total public expenditure on health as a percentage of GDP. Period covered:

36 Missing: see socexp_t. Source: see socexp_t. family Total public expenditure which supports families as a percentage of GDP. Period covered: Missing: see socexp_t. Source: see socexp_t. unemp Cash expenditure compensating for unemployment as a percentage of GDP. Period covered: Missing: Austria and Norway: , ; Iceland: ; Ireland: Source: see socexp_t Demographic data pop Total population, in thousands. Period covered:

37 Missing: none. Source: SourceOECD. Employment and Labour Market Statistics. Labour force statistics - Summary tables. Vol. 2008, release 01, (Download: ). pop15_64 Population 15-64, in thousands. Period covered: Missing: none. Source: see pop. Notes: 1. France (2007): Estimated value. pop65 Population over 65, in thousands. Period covered: Missing: Italy: ; Portugal: , Source: SourceOECD. Employment and Labour Market Statistics. Population and labour force statistics. Vol. 2008, release 01, (Download: ). Notes: 35

38 1. France (2007): Estimated value. elderly Population 65 and over as a percentage of population. Period covered: Missing: see pop65. Source: see pop65. Notes: 1. France (2007): Estimated value. 36