11 October 2014

Veto Player and the Greek Constitution, Part 4

For previous parts of this series: part 1, part 2, part 3.

3. How we can shape things up through the electoral system.

I showed you in the first part of my talk that it is difficult to make changes, and in the second that most people who want to make them are looking not just in the wrong place, but also in the opposite direction: they try to increase the number of veto players, when the only way we will get out of the wrong equilibrium is if we decrease them (or bring them closer together). Unlocking the constitution would be one step in the right direction, but it will not help very much if we make modifications like the ones discussed above (if we increase the number of veto players we will reach the same “locking” with different means).

The Greek political system is in serious trouble. This is neither an original nor a novel assessment. The statement can generate unanimous consent, but does not advance our understanding of causes of the identification of necessary cures. I think we would advance our analysis if we focused on the party system. In each of the statements I will be making I may have disagreements, but I hope that by the end of the analysis I will have a large majority in favor of the diagnosis and the proposed cure.

The problem with the Greek party system is that it is centrifugal. Despite the fact that a large majority of people desire a solution to the serious problems I outlined above, the party system responds in the opposite direction: polarization. If we want to analyze this polarization more closely, we will see two features: first, that extreme parties have unusual strength (in the Greek case, particularly on the right), and second, that the other (more moderate) parties are being influenced by these extreme parties (or their own internal forces that feel close to these parties). So, SYRIZA is very concerned about its left side and New Democracy is about its right.

Let me use one example from German policymaking borrowed from Der Spiegel (Spiegle.de 2014). The article states that  “Merkel is afraid that, for the EP election on May 25, the Euro-skeptic party Alternative for Germany will profit from a discussion about new aid for Greece” on the other hand Schaeuble “sees the danger that – without the prospect of new aid – radical parties in Greece will gain electoral support at the election.” What is interesting to note here is that both German politicians are concerned about the electoral consequences not the policy consequences of their choices. You can recognize the phenomenon of policy impact of extremist parties although official statements will ignore such parties and offer different justifications.

If we want such phenomena to change, we must do so through the electoral system: in a democracy, ultimately what happens is consistent with what the people select in elections. And an electoral system is the primary institution that contains incentives both for the people who elect and the people who get elected. Since at least 1951, with Duverger’s law, political science has been studying electoral systems.

I will start with one actual modification of an electoral system that was designed to reduce polarization and has had successful results. In the United States, the electoral system involves a primary election within each party, after which the winners of these primary elections (along with candidates from other parties, as determined by the electoral law of each state) confront each other in the final election. The candidate with a plurality of votes is the winner. This electoral system had produced many extreme candidates with the following mechanism: within each of the two major parties it was possible that the extremist side was more powerful than the centrist, so that one party selected an extremist representative in the primaries. In the final election, voters had to choose between two extreme candidates, which lead to a divided Congress where representatives mainly confronted each other along party lines rather than compromise. In 2013, some 80 Republican Representatives sent a letter to the (Republican) Speaker of the House asking him to shut down the government rather than compromise with the President on budgetary issues. The result has been successive blockings of US policymaking. President Obama complained because the shutdown was caused by a fraction of one party among the two political parties, in one of the two Houses in Congress, in one of the two branches of Government. This remarkably small subset of politicians wanted to impose its will upon all the rest.

To address the polarization and reduce the number of extreme candidates it produced, two states of the US decided to change their electoral system in the 2000s. The process took more than a decade to be completed, and it involved referendums and Supreme Court decisions. The modification was called a “blanket primary” and it involved two steps: first, the primaries took place together—rather than separately by party—where voters select candidates of their choice; second, that only the first two candidates (regardless of party) would participate in the election. The outcome of this modification is that if, in a constituency, two candidates of the same party are selected for the final confrontation, then the more moderate will prevail, because the voters of the other party will select him (along with his initial supporters). We can evaluate the success of this gap by examining how many of the more extreme representatives signing the letter to the Speaker of the House (mentioned above) come from these two states: only 3 were from California (the most populated state) and 0 were from Washington (Wilson 2013). In the 2014 election a tea Party Republican (one of the three singators of the letter to the Speaker I discussed above) is now challenged by a moderate in the blanket primary (for the first time ever).

The attempts to introduce blanket primaries in Washington go back to the 1930s, the attempts in California led to a successful referendum in 1996 that was rejected by the American Supreme Court. A modified blanket primary was introduced in Washington and upheld by the Court, and California passed its own (identical) provision by referendum in 2010.

A similar (in terms of the political effects) electoral system of alternative vote was introduced by the Liberal party in the UK in a referendum agreed as part of the government agreement with the Conservative party (BBC.com 2011). The electoral system was rejected by British voters.

The US and UK have plurality electoral systems, so, their examples are not applicable in Greece. A successful modification needs to change both the political system and mentalities: the expectations of voters and parties with respect to who is selected. Secondly, the change in parties needs to produce candidates who are more moderate than those currently selected. Greece has been an important influence in the design of electoral systems worldwide. Let me explain first the positive features of the current Greek system, analyse the criticisms and proposed alternations and then present the modifications I claim would help shape representation and politics in the desired dimension of discussing and problem solving.

An electoral system is expected to satisfy two requirements: representation of the population, and government stability. Because these two requirements are contradictory, as we saw in the first portion of the paper, different countries elect their parliament with different electoral systems placing different weights in these requirements: from a purely proportional electoral system in Israel and the Netherlands where the whole country is an electoral constituency and there is no required threshold, leading to multiparty governments, to the plurality electoral system of the UK, where the Liberals never get seats proportional to their share of the vote and which usually produces single party governments. In this continuum, the electoral system of Greece has an exceptional placement, since it is proportional and gives small parties (over 3%) proportional representation, but, because of the bonus of 50 seats, was producing single party governments from 1981 until 2012. In the seven party parliament of today, a two party coalition has the majority and can form a stable government.

However, people now have found different problems with the current electoral system and several proposals have been made to try to address special problems that manifest themselves. Some people see that the personal vote (the cross of preference) in large constituencies is correlated with corruption and therefore propose a “German” system with small constituencies. Others consider the 50-seat bonus to the largest party as promoting polarization and impoverishing the political debate. Still others consider the 3% threshold as eliminating interesting voices that should be heard in the Parliament. Several voices have been raised in favor of reduction of the size of Parliament. Recently, there was a competition between SYRIZA and DHMAR to determine who favored the proportional system the most, and supporters of these parties wanted to include a “pure” proportional representation system to the constitution in order to resolve the issue once and for all.

All these criticisms (except for the last one) make valid points. The last one shows complete lack of understanding of politics in two fundamental dimensions: first, the electoral system should be satisfying two constraints, not just one. Pure proportional representation would not produce stable governments (a simple proportional allocation of seats one the basis of votes in the 2012 election is sufficient to make this point), and therefore is not the eternal ideal system. As such, it should not be included in the constitution, unless we consider the restrictions this constitution already imposes as insufficient!

The other criticisms however, have serious factual basis, and a reduction of the size of large constituencies, and/or the size of the bonus could improve features of the current system. The most founded (and unquestionably beneficial) would be the reduction of the size of Parliament. However, they do not address the fundamental problems that we face: first that the electoral system produces very significant centrifugal forces, and second the need to have a government that will coordinate our movement away from the current (bad) equilibrium.

There is a very simple solution that will fundamentally transform the electoral system, and the political system that it will generate. I have proposed that every voter receives multiple votes instead of one (Tsebelis 2013). My initial proposal involved up to three votes, but since then Potami has emerged as a significant party, so, the number of votes has to be increased to up to four: however, it is not the number of votes that matters as much as the way the votes are cast. Voters can use these votes to select up to four parties of their preference (cumulation, that is use of two or more votes for the same party is not allowed). None of the other features of the electoral system needs to be changed, although if a consensus about additional changes emerges the revised system can accommodate them.

The simple institutional change of multiple votes has profound consequences on the meaning and the effects of voting. With respect to the meaning, instead of the voter deciding with a single choice which party (s)he identifies with, the multiple choice indicates which parties have characteristics that the voter appreciates (their political positions, their ability to form coalitions, the personal qualities of their representatives). It possible, but not likely that one single party will satisfy a voter with respect to all these dimensions. So, the perspective changes, and instead of the voter trying to fit the best possible in the existing party system, he becomes the center of the act, and decides which parties express better his preferences and values.

Let us now turn to the consequences. How will such a system affect voters, parties and government?

Voters: Under the current system with 1 vote, each voter has the possibility of electing 1 out of 8 parties (let us ignore the smaller ones for now, the larger parties play such a large role that we will first focus on them), or abstaining, which gives him 9 choices. Under the system I propose, each voter will be able to vote for one, two, three, or four parties, or abstain. That will produce 163 different choices (8 single party choices, 28 two party choices, 56 three party choices, 70 four party choices, and 1 abstention). So, each voter will have many more choices to express him/herself. For example, people who like the center left, will be able to vote for PASOK, Potami, and DHMAR, as well (if they so wish) for a major party; voters who want a government of national unity may vote for ND and SYRIZA (and maybe also PASOK, Potami, or DHMAR), people who want a government of ND or SYRIZA will be able to vote in favor of other parties too, in order to assure their ability to form a government coalition in (the all but certain) case of lack of 151 votes. Given that voters will have many more choices, they will pay more attention to political debates, trying to decide whether they will exercise their right to vote for four parties, or select three, two, or even one. They have an incentive to vote for more than one party because they can only allocate one vote per party and using more of their votes allows them to maximize their influence on electoral outcomes.

Parties: The parties will adopt strategies that will maximize their votes. Given that they can gain votes not only from their members or traditional supporters but also from supporters of other parties, they will shift their discourse from ideological to pragmatic, and from aggressive to conciliatory, so that they will be able to attract the additional votes (under the current system a second or third-ranked party would not receive a vote, but are likely to receive a vote under the new system). In other words, parties will be actively focusing on centrist voters because voters have the possibility of rewarding four parties each.

Party system: The party systems that will emerge will have a series of desirable characteristics. Centrist parties will be stronger than they currently are because they will receive additional votes. Extremist parties will be less influential than under the current system. As we observed in Figure 2, this change will lead to greater potential for policy change and yield policies more reflective of the people’s wishes. Centrist parties will emphasize solutions over ideology, and the political debate will be centered around problem solving and differences in the solutions proposed as opposed to matters of principle.

Government: The government will be formed by the centrist parties, because they will receive a majority of votes, with most agreement on the solutions proposed. Therefore it will be more able to coordinate the plan for modification of the status quo (the veto players will have smaller ideological distance among them and therefore will be less policy stability as I demonstrated in Figure 2). In particular, it will make feasible a serious constitutional revision (not a minimal one like the upcoming).

Moving to this new system may raise some questions, particularly regarding the number of votes per voter, the mechanisms behind its success, and whether such a system has ever been ‘proven’ or used in practice. These are very reasonable questions and we turn to each one of them at a time.

In his seminal article on ideological positions generated by electoral systems Gary Cox proves that electoral systems with a number of parties more than double the number of votes per voter will produce party systems that are ideologically dispersed (Cox 1990, 903-935). Fundamental assumptions for Cox’s argument are single peaked preferences of voters[1], one-dimensional politics, and mandatory use of the number of votes. These assumptions are necessary in order to prove a theorem of existence and uniqueness, but are more or less plausible and more or less required for a likely outcome.

While these are plausible in general, they are not all equally relevant when we consider political systems. For example, while single peakedness of preferences is the most important factor driving Cox’s results, it is also a very reasonable and realistic assumption. Another assumption, one dimensionality is less reasonable: there are multiple issues that are addressed by a party system. Not only the left-right continuum, but also immigration, the EU, globalization, healthcare, social security, environment, etc. Closeness in one of these dimensions does not guarantee proximity in the others. Furthermore, mandatory use of all votes can indeed guarantee convergence of parties, but (in my opinion) is too restrictive to the freedom of choice of voters. Why should a voter be forced to use all four of his votes?

To explore what the application of my proposal would look like and assure the reader that it accomplishes what I have argued, I have created a model of the electoral system[2] where the reader can explore its properties. By allocating their vote to parties, and deciding how many additional votes they will use and in favor of whom (this is the most consequential choice), we can compare its results to those of the current system.

Number of Permitted Votes 4
Election share under one-vote system (%) 7 6 30 8 10 2 30 7 0 0 100
Prob of a voter from party GD voting for other parties 1 0.2 0.4 0 0 0 0 0 1.6
Prob of a voter from party ANEL voting for other parties 0.2 1 0.4 0 0 0 0.3 0 1.9
Prob of a voter from party ND voting for other parties 0.1 0.3 1 0.7 0.5 0.3 0 0 2.9
Prob of a voter from party PASOK voting for other parties 0 0 0.8 1 0.8 0.6 0.2 0 3.4
Prob of a voter from party POTAMI voting for other parties 0 0 0.6 0.9 1 0.7 0 0 3.2
Prob of a voter from party DHMAR voting for other parties 0 0 0.7 0.8 0.8 1 0.3 0 3.6
Prob of a voter from party SYR voting for other parties 0 0.2 0 0 0 0.2 1 0.2 1.6
Prob of a voter from party KKE voting for other parties 0 0 0 0 0 0 0.2 1 1.2
Prob of a voter from party I voting for other parties 0.0
Prob of a voter from party J voting for other parties 0.0
Percentage (%) 5 10 21 17 14 12 15 6 0 0 100


Here are some of the major features:

  • Centrist voters will be more comfortable casting more votes than extremist ones (this will show up at the margin of the table, where parties like PASOK, DHMAR will have voters who are more willing to vote for other parties).
  • Extremist voters will oscillate between voting for their own party and expanding towards parties more to the center (at the margin of the table they will be using less votes than centrist voters)
  • Centrist parties will see their percentage increased compared to the current system (comparison of the first and last lines of the table will corroborate this statement).

The table will persuade the reader of the effectiveness of this electoral system to solve the current problems of the Greek party system. In addition, it will demonstrate that a higher number of votes would increase the centripetal forces (the reader can venture with a 3 or 4 or 5 vote system). However, the most interesting change cannot appear in the percentages of parties. It concerns the political discourse, because it will be the most problem solving oriented parties with civil discourse that will be receiving the most votes.

The system is effective because it has two different effects: one mechanical and the other strategic. The mechanical effect is that because of single peakedness of the voters’ preferences, centrist parties get higher representation than in the current one vote system. The strategic effect is that the leaders of the parties, knowing the mechanical effect will result in more centrist votes, will themselves move toward the center and away from ideologies. This is because the center holds the largest reservoir of multiple votes. To obtain some of these votes, party leaders will need to focus on problem solving instead of confrontational and ideological statements that will appeal to a smaller number of voters.

One very reasonable strategic objection is that this proliferation of voter’s choices will lead to proliferation of parties. What if Golden Dawn splits in four parties so that its supporters will vote for all four of them and exhaust their four votes when say some of PASOK supporters will not use all four of their votes because they do not want to vote for DHMAR (or vice versa)? Such behavior would inflate the votes of Golden Dawn. The possibility of party proliferation can be dealt with a series of auxiliary measures, like the absence of financing for new parties until after they reach some percentage in the election, necessary number of signatures for creation of a party (like in Italy), and prohibition the creation of new parties six months before an election. This particular deadline is necessary because the number of votes should be at least half the number of parties.

Although it may be tempting to enshrine this solution by locking it constitutionally, doing so would limit adjustments down the road. The current system is designed to solve today’s problems: how to reduce the weight and pressure from extremist parties, how to facilitate coalitions among centrist parties; how to make these parties focus on problem solving instead of position taking. I expect that we will be facing problems like immigration, globalization, social security, health care for many more years to come. Each one of these problems is dependent on values that we hold dear: equality, justice, efficiency. Some of them affect intergenerational transfers and tradeoffs. A democratic debate on these issues has not happened not only in Greece but in any advanced industrialized country. It is long overdue, and we have to welcome and facilitate these upcoming debates.

On the other hand, this electoral system may not help alternation of different parties in power. If this happens, the political system becomes corrupt like the political systems of Italy and Japan in the past. If signs of government coalition stability and corruption emerge, we should be able modify the electoral law and not be prevented by the constitution.

The system I propose is a mixture of multiple vote with proportional representation. Multiple vote has been used with plurality electoral systems (approval voting, and transferable vote to a certain extent), but never with proportional representation. Approval voting has been the most stable electoral system in the history of our country (1864-1920). It was imported from Eptanisa, which got it from the Venetian Empire. Actually, the Doges of Venice were elected by this system, and during the Middle Ages this is how the Catholic Church elected its Popes (Colomer and McLean 1998, 1—22). If we go further back in time, we can find a more generic form of voting called “range” voting where voters can give to candidates a grade instead of a binary “approve”, “disapprove”. We can find this “range” electoral system in the Olympic Games today (in composite evaluations like skating on ice, or diving), and its origins are in the voice voting (δια βοης) of ancient Sparta. So, while the components of the system I propose have been used in the past (some of them going back to older or even ancient Greece), this particular combination has never been used. But then again, neither have the problems we’re facing been so profound and the need for solutions so urgent.

In conclusion, I discussed in the first two parts that we should enable ourselves to decide both in terms of what is included in the constitution and in terms of the things that should not make their way into it. And I suggested a way of voting (collectively deciding) which will make our life easier in the tough times we’re going through now, and are likely to come in the foreseeable future. I hope that I did these things thoroughly and scientifically: επισταμενως.

[1] That voters can somehow rank or order their choices.

[2] In http://sitemaker.umich.edu/tsebelis/data the interested reader can click at “multiple vote electoral system”