How many electoral candidates are voters able meaningfully to choose between? This is an important question for anyone thinking about designing or evaluating different electoral systems.
Voters often say that they want more choice: ‘voter choice’ was identified as a key desideratum by citizens’ assemblies that investigated electoral reform options in British Columbia and Ontario in 2004 and 2007; it was also highlighted in focus groups conducted by David Farrell and Michael Gallagher in the UK in the late 1990s.
On the other hand, too much choice may be a bad thing: it may become overwhelming, causing voters to use shortcuts such as party allegiance and position on the ballot paper, rather than thinking about the candidates themselves. The Jenkins Commission, which investigated electoral reform in the UK in 1997–8, put the point colourfully:
it should be stated that the Commission sees the extension of voter choice as highly desirable up to the point at which the average voter is able and eager meaningfully to exercise choice, both between and within parties. But that where the choice offered resembles a caricature of an over-zealous American breakfast waiter going on posing an indefinite number of unwanted options, it becomes both an exasperation and an incitement to the giving of random answers. In voting rather than in breakfast terms exasperation may discourage going to the polls at all and randomness lead to the casting of perverse or at least meaningless votes.
But what do we know about the amount of choice that voters can actually cope with? I’m writing a paper at the moment for which it would be useful to have a clear answer to this. So far as I can tell, however, we don’t have a clear answer. So I’m writing this post partly to run through the evidence that I have found so far and partly as a plea to all of you to let me know if you think there is some relevant evidence that I have missed.
One way of looking at the issue of how much choice voters can cope with is to consider how many choices voters make – or, more accurately, how many preferences voters express – when given the opportunity. We can consider this by investigating elections using the Single Transferable Vote (STV) electoral system, under which voters can rank all the candidates. Most Australian elections use versions of STV under which voters must rank a specified number of candidates – in some cases, all the candidates – in order to cast a valid vote. These versions are not relevant here. But Ireland, Malta, Northern Ireland, and Scotland all use forms of STV where voters can express as many or as few preferences as they want.
Unfortunately, in most of these elections we can’t see how many preferences voters actually express. Lower preferences count towards the result only if the candidates receiving higher preferences have been eliminated, which means that many of the preferences expressed are completely ignored. In a few elections, however, the relevant data are available. Three Irish constituencies experimented with electronic voting in the 2002 Dáil election, and complete anonymized voting records were made available. In Scottish local elections, votes are counted electronically and a few local authorities publish data on all preferences. Michael Laver has analysed the Irish data and found that the mean number of preferences expressed by voters was 4.98 in Dublin North, 4.43 in Dublin West, and 4.65 in Meath. A quick analysis at the data available for the Glasgow City Council elections held earlier this month shows that the mean number of preferences expressed across the whole city was 3.18, ranging from 2.82 in Pollockshields to 3.56 in Hillhead. Similarly Shaun Bowler and David Farrell analyse mock ballots presented to survey respondents in Ireland in Galway West in 1987 and in the context of the European Parliament elections of 1989. Their data show that voters expressed on average 3.73 and 3.35 preferences respectively. In each of these cases, the modal number of preferences expressed was 3.
We cannot simply conclude from this, however, that most voters are able to make a meaningful choice only of three or four options. As Laver emphasizes when discussing the Irish results, voters in an STV election might express fewer preferences than they actually have, either because they realize that lower preferences are unlikely to be counted or because they stop when they reach candidates among whom they are indifferent, even if they have clear dislikes further down their preference ranking. Furthermore, it is reasonable to suppose that many of those who express only three preferences value the presence of many more candidates on the ballot paper: they may, for example, want to support candidates from only one party but still want a choice of candidates within that party.
Support for the idea that having as few as three candidates on the ballot paper is optimal does, however, come from other sources. Analysis of elections in the United States by Lau, Anderson, and Redlawsk finds that voters are more likely to make ‘mistakes’ in their vote if there are more than two candidates. In a recent study conducted in Brazil, Saul Cunow uses mock ballot papers to assess abstention rates among respondents who are asked to make a choice among 2, 3, 6, or 12 candidates. He finds the lowest abstention rates where there are just 3 candidates. Related to Cunow’s findings, many studies into the determinants of electoral turnout have found that a greater number of parties is associated with lower turnout (André Blais provides a useful summary) – though whether this can be accounted for in terms of greater complexity or other factors such as weakened accountability is hard to say.
These studies provide hints that many voters struggle if they are offered more than a very few options, but they are hardly conclusive. So far as I have been able to see so far, they provide pretty much all the evidence we have at the moment on this question. As John Carey and Simon Hix observe in an important recent paper, ‘to our knowledge cognitive capacity has attracted no serious attention in research on electoral system design’.
In order to get beyond this, Carey and Hix turn to psychology. They cite a famous paper, published in 1956, by George Miller, according to which voters are able to make meaningful choices among seven options, but not much more than that. Drawing on this, they posit that the voters should be equally able to deal with the choice available to them in districts electing anything from one to around six members, but that voters will struggle in districts electing more than about ten members.
Carey and Hix are certainly right that Miller’s analysis has been corroborated by numerous studies across many fields in the decades since it was published. Still, it would be valuable to have more concrete evidence that relates specifically to the electoral context. So I close with a question: Is anyone aware of more evidence on elections that could offer useful guidance here? I shall be very interested to find out what I’ve missed.