Title Fair representation and the concept of proportionality

The idea of proportionality, which compares votes cast to seats won, is a common test of fair representation. But fair representation is a relation between electoral aims and electoral outcomes. The proportionality test falsely presupposes that each voter aims to support a political party. It therefore tells us nothing about fair representation. We need to construct a deeper criterion of fair representation which takes account of citizens' multiple political concerns. What is the best criterion for deciding whether one electoral system is fairer than another? One widely employed criterion -what I shall call the 'proportionality test' -is to compare the proportion of votes cast for each party with the proportion of seats won by each party. One system is more proportional than another just in case there is a better fit between these two sets of proportions. Quite what counts as a better fit is of course a matter of further specification, but that is not my concern here. My aim here is to show that the proportionality test is normatively indefensible; that is, that it cannot be used as a criterion of fair representation. This is because electoral fairness is a relationship between the electoral aims of voters and electoral outcomes. In some electoral systems, the proportionality test does capture the electoral aims of voters. But in other systems, voters can have other aims, and the proportionality test simply does not apply (sections 1 and 2). We need a deeper criterion of electoral fairness based on voters' multiple political concerns (section 3). By way of caveat, I should say that nothing in this paper presumes that fair representation is the only virtue of an electoral system, or even the most important one. If fairness matters at all, then we need to consider how to assess it. The proportionality test gives us an easy, but indefensible, way of doing so. 1. Why does proportionality matter? What is the normative point of proportionality? Why should anyone care whether electoral systems are proportional or not? As ever in political theory, this question is unlikely to have a univocal answer. But the most obvious answer, and the one I will assume to be central for what follows, involves the relationship between proportionality and fair representation, where fair representation is itself related to the idea of political equality. The point might be put as follows. In an election, each voter acts within the rules defined by the electoral system to achieve certain electoral aims. The overall outcome is more or less fair according to whether it gives more or less equal weight to each voter's electoral aims. Proportionality measures the degree to which voters' aims are given equal weight. Proportionality is by no means the only criterion of fair representation. Another widely discussed criterion is whether the legislature reflects the electorate in terms of gender, race, ethnicity, class, age, and other politically important characteristics; according to this criterion, all contemporary legislatures are highly unrepresentative. A related question is whether the options placed before the electorate reflect the full range of their political concerns; in this respect, an electoral system in which some significant political objectives are excluded from the ballot, whether directly or as an effect of the way the system operates, is bound to produce unrepresentative results, even if the outcome is proportional to the votes for each party. Again, given that every electoral system is in principle open to strategic voting, we might want to ask whether the results are representative of the electorate's sincere political priorities as distinct from the votes actually cast. Yet another aspect of fair representation concerns the relationship between votes and legislation. For example, the Westminster legislative system might be considered unfair because voters for the government party end up having far more influence on policy than those who vote for the opposition, even if the electoral results are proportional. All of these criteria of fair representation are concerned in one way or another with the degree to which electoral or legislative outcomes give equal weight to the political concerns of all citizens. By comparison, the idea of proportionality is a more superficial criterion of fair representation, because it does not look beyond the actual votes cast in a given election or electoral system. Despite this superficiality, proportionality can be considered normatively important because it focuses on a central and ineliminable stage in the process by which the political concerns of citizens influence legislative outcomes, and therefore is central to the very idea of democracy as self rule. Proportionality may not measure the degree to which each citizen's overall political concerns are equally influential over legislative outcomes, but at least it seems to measure the degree to which each citizen's electoral aims are equally influential over electoral outcomes. It may therefore be thought to capture a significant if partial feature of political equality, and to do so in an empirically straightforward way. 2. The proportionality test Proportionality, then, gets its importance as a measure of the degree to which citizens' electoral aims are equally influential over electoral outcomes. It follows that the following questions are central to any test for proportionality: (1) What are the voters' electoral aims? (2) What is the electoral outcome? Now the standard proportionality test operates as follows. First, count the votes cast for each party as a proportion of the total valid poll. Second, count the seats won by each party as a proportion of the total number of seats filled. Third, compare the degree to which these proportions diverge. What answers to the two questions listed above are presupposed by this procedure? Clearly, they are: (A1) Each voter aims to support a party. (A2) Each party wins a certain number of seats. It is only by making support for a party and seats won by a party the key concepts here that it is possible to aggregate all the votes cast for each party's candidates and compare them with all the seats each party's candidates win. Otherwise, there is no reason for thinking that a Democrat winning a seat in New York contributes to the fair representation of a Democratic voter in Iowa. Parties are thus the essential vessels for making the proportionality test relevant to fair representation. I want to focus on presupposition A1. In some electoral systems, namely pure list systems, it is perfectly reasonable to describe the electoral aims of voters as support for a party. The system does not allow them to do anything else. But what are we to say about the single-member constituency plurality (SMP) system, or about systems, whether based on single-member or multi-member districts, which use the single transferable vote (STV)? The fact that technically speaking voters in these systems vote for individuals, not for parties, means that the system does not require them to have party support as their aim. It might indeed be their aim, but their aim might rather be to vote for a neighbour or a co-religionist or a woman. Moreover, in STV systems, voters vote not for a single individual but for an ordered set of individuals. Thus the only general way of describing a voter's aim in an STV system is something like this: 'my highest aim is to get A elected; my next highest aim is to get B elected; my third highest aim is to get C elected; etc.', where 'getting A elected' may be most accurately filled in as 'getting a Labour candidate elected' but may also be filled in in a number of other ways. Empirical studies of voting behaviour in SMP and STV systems confirm that party support is far from being the exclusive aim of all voters. Since the account of voters' aims presupposed by the proportionality test does not necessarily or in fact correspond to the aims actually pursued in these electoral systems, the test cannot be applied. And since a test which cannot be applied cannot be normatively defended, I conclude that the test is normatively indefensible. In other words, the proportionality test tells us nothing at all about fair representation in systems other than pure list systems. 3. Judging fair representation The proportionality test is normatively indefensible. Although it has a straightforward application to pure list systems (where it might reasonably be used, for example, to test the effect of district size on electoral fairness), it fails to apply elsewhere. How, then, should we judge the relative fairness of different electoral systems? From what I have said in this paper, it should be clear that no test based on votes and outcomes will work, because fairness concerns the relation of voters' aims to electoral outcomes, and it is only in the simplest, pure list system, that we can identify aims from votes. A voter in a pure list system has no choice but to aim to support a single party. A voter in an SMP system can aim to support a single party, but can also aim to elect a neighbour or an advocate of a particular policy or a person with integrity. A voter in an STV system can aim to support a single party or individual, but can also aim to give graded support to other parties and individuals. It follows that any criterion for fair representation has to be less superficial than the proportionality test, extending beyond voting to the aims of voters. I do not pretend that there is an easy or even a possible procedure for cranking out fairness indices in these terms. On the contrary, I presume that assessments of fairness must inevitably be qualitative and based on arguments rather than calculations. At the moment, however, the deceptive ease with which the proportionality test can be applied distracts us from the fundamental question here, which is how to formulate a criterion for electoral fairness. A good place to start is the criterion given by John Stuart Mill when he called for an electoral system which: secures a representation, in proportion to numbers, of every division of the electoral body: not two great parties alone, with perhaps a few large sectional minorities in particular places, but every minority in the whole nation, consisting of a sufficiently large number to be, on principles of equal justice, entitled to a representative. Mill thought that the criterion was best satisfied by Thomas Hare's electoral system, which employed STV with respect to one country-wide district. What is interesting in the present context is to ask what Mill's argument presupposes about the aims of voters. The presupposition seems to be: (B1) Each voter aims to elect a representative of the political group she or he belongs to. Put this way, we can see that Mill's criterion, though an improvement on the proportionality test, nevertheless continues to think of voters' aims by reference to a single group, organised as a party or otherwise. In modern, pluralist societies, it is more plausible to suppose that many voters would like to express their affiliation to a plurality of overlapping groups based on party, locality, gender, single issues, or other considerations. If we really want the legislature to reflect these complex aims, what we need is a modification of Mill's criterion of the following form: a system which reflects, in proportion to numbers, every division of the electoral body, recognising that these divisions overlap and that voters have multiple identifications. Which electoral system best satisfies this criterion is an open question. The point is that we cannot even begin to address that issue until we have developed a criterion of fair representation which takes account of citizens' multiple political concerns. In doing so, we inevitably engage with some of the deeper conceptions of fair representation with which I contrasted the proportionality test in section 1, because the criterion will be sensitive to gender, ethnicity, and so on, because it allows us to critique electoral systems in which these concerns are excluded by a focus on party support or by the power to control political agendas, and because it encompasses the incentives for and consequences of strategic voting. I have little doubt that any such criterion would support systems we think of as 'proportional' as against SMP systems. But the arguments between 'proportional' systems are another story. The proportionality test gives us no reason for thinking that one 'proportional' system is fairer than another.