Extending the Rational Voter Theory of Tactical Voting

Whilst tactical (or strategic) voting is the result of voters responding to the strategic situation they face, rational voter theory may not be a full description of what happens in practice. This paper tests a number of additional hypotheses using survey data on voters in England from 1987 to 1997. Effects of political interest, education, strength of party identification and being against the incumbent party are identified, along with changes over time in the propensity to vote tactically. However, local campaign spending and specifically anti-Conservative sentiment did not seem to be related to the level of tactical voting. Also the extensions to the rational voter theory are generally quite weak. The rational voter model may not be sufficient to account for the pattern of tactical voting we observe, but it is by far the most important set of explanatory factors. Extending the Rational Voter Theory of Tactical Voting by Stephen D Fisher 2 It is well known that electoral laws not only affect the composition of legislatures given the votes cast, but also structure the vote choice itself. Electors must translate their feelings for the parties and candidates into a voting strategy. In most cases this is straightforward, but when voters decide it is optimal to abandon their first preference party and vote for another it is said that they voted tactically (or strategically). Various theoretical works have attempted to describe the micro-logic behind tactical voting (e.g. Cain 1978, Cox 1997, Myatt 2000). Although the approach of each is very different the components are essentially the same. Each assumes that voters are instrumentally rational and respond to the strategic situation in their constituency. These rational voter models are have met with varying degrees of success in describing the pattern of tactical voting in England (Fisher 2000). But it would be surprising if any of them can be sufficient when they are based on such a parsimonious framework. There are a number of obvious possible points of departure. First, rational voter models ignore parties as possible relevant actors when tactical voting could well be influenced by constituency level campaigns. Secondly, whilst there are incentives for some electors to vote tactically, it is not clear that voters respond to these incentives in a similar manner. More specifically, the level of sophistication and knowledge required may be so great that tactical voting is the preserve of an elite. Thirdly, since the level of tactical voting has changed over time, it is important to ask whether this can be explained from within the rational voter model. Fourthly, in Britain tactical voting is often considered to be about, ‘getting the Tories out’. As a result tactical voting may be more prominent among those who hate the Conservatives than it is for others facing the same strategic situation. Finally, tactical voting campaigns have been about defeating the Conservatives in seats they won previously. It may well be much easier to mobilise tactical voting against the incumbent party than it is to protect the incumbent party. All these hypotheses are tested in an effort to address the central question of whether the rational voter models of tactical voting need to be extended to account for the pattern of tactical voting we observe in reality. Extending the Rational Voter Theory of Tactical Voting by Stephen D Fisher 3 Rational voter models The rational voter approach is highly suited to the study of tactical voting. In fact, it is the only framework which predicts tactical voting and through which it can really be understood. Rational choice models assume voters to be utility maximisers, or short term instrumentally rational. Clearly no one fulfils this abstract conception of a voter perfectly, but tactical voters in real life can be identified by having instrumental reasons for their vote choice. Also, it is convention that tactical voters are those who vote for a party that is not their first preference. So a tactical voter can be defined as someone who votes for a party they believe is more likely to win than their preferred party, in order to vote effectively. How much tactical voting there is and when to expect it is the subject of substantial debate. Fisher (2000) shows that the Myatt (2000) theory of tactical voting is the only rational voter theory that fits the pattern of tactical voting for England from 1987 to 1997. In the Myatt model the level of tactical voting depends on a number of factors, for voters who prefer the party they believe is most likely to come third of three parties in their constituency. Firstly, tactical voting should decrease with the relative strength of preference for the favourite party over the second preference party. Secondly, tactical voting should increase with the relative strength of preference for the second favourite party over the least preferred party. Thirdly, tactical voting should increase with a specially derived strategic incentive variable, which depends on the shares of the votes for the different parties in the constituency. The first two of the Myatt propositions are intuitive and occur elsewhere in the literature (see especially Heath et al. 1991 and Evans 1994). The tactical incentive variable is highly positively correlated with the distance from contention, which is the share of the vote for the second placed party in the constituency minus that for the favourite party. After controlling for this association, the tactical incentives in the Myatt model are weakly positively correlated with the margin of victory, which is the gap in support between the winner and the runner-up. The standard intuition would tell us that tactical voting should increase with the closeness of the race, so the predictions of the Myatt model are surprising. Extending the Rational Voter Theory of Tactical Voting by Stephen D Fisher 4 The intuition behind the Myatt model starts with the observation that the only circumstance in which a voter can influence the result of an election is when there is a tie for the lead, i.e. when the voter is pivotal. People need to know who they will be pivotal between if they are to be pivotal. Therefore it is the conditional probability that different pairs of parties are tied for the lead, given that there is a tie, that becomes important. Now when the margin of victory in a constituency widens the absolute probability of a tie for the lead may decrease, but the conditional probability that, if there is a tie, it is between the top two placed parties can actually increase. So, supporters of the third placed party have more incentive to vote tactically, because the probability that any tie for the lead involves their party has gone down. The aim here is to investigate whether the effects predicted by rational voter theory are sufficient to account for the pattern of tactical voting observed in practice. In particular, whether there are effects of local campaigning, political interest, education, strength of party identification, time, anti-Conservative feelings or anti-incumbent feelings, after controlling for the factors within the Myatt model. Although the model is not the only rational voter model, it does have a better fit to the data used here than the standard intuition. However, the results are essentially the same if variables from the standard intuition are used instead. Data and methodology The British Election Studies (BES) for 1987, 1992 and 1997 provide high quality postelection survey data with sufficient questions to study tactical voting in depth (Heath et al. 1991, 1994, Evans and Norris 1999). Scotland and Wales were excluded because both have very strong nationalist parties and therefore genuine four party competition. In England the same three parties stood at all three elections and took first, second and third 1 Note that the Cox(1997) model would not be helpful for this project because it has no comparative statics for the level of tactical voting. Extending the Rational Voter Theory of Tactical Voting by Stephen D Fisher 5 places in all but a couple of cases. The theories we are testing assume that people do vote, so non-voters were also excluded. Tactical voting is measured using responses to the following BES question. A. Which one of the reasons on this card comes closest to the main reason you voted for the party you chose? 1. I always vote that way 2. I thought it was the best party 3. I really preferred another party but it had no chance of winning in this constituency 4. Other (write in) 5. None of these/Don’t know Tactical voters identified by response option 3 in question A were asked a follow up question. B Which was the party you really preferred? Some respondents who gave tactical reasons for their vote in answer 4 were also coded as tactical so long as this was consistent with their declared voting behaviour and other questions relating to their order of preference for the parties. By definition tactical voters do not vote for their preferred party, so if there was any indication that they did so then the respondent was not coded as tactical. This was done using the ‘strength-of-feeling scores’ (or ‘approval ratings’) for the parties. The strength-of-feeling score for a party is the response coding (1 to 5) from the following question about the party. 2 These cases were excluded from the analysis. Extending the Rational Voter Theory of Tactical Voting by Stephen D Fisher 6 C. Please choose a phrase from this card to say how you feel about the (Conservative Party/Labour Party/Liberal Democrats/...)? 1. Strongly in favour 2. In favour 3. Neither in favour nor against 4. Against 5. Strongly against When the respondent is a tactical voter the preferred party is provided by the response to question B above, or is imputed from the strength-of-feeling scores. For non-tactical voters the party voted for is the first preference party, unless there is a clear indication otherwise on the strength-of-feeling scores. The second preference party is defined, for all respondents, as the party with the best strength-of-feeling score that is not the first preference party. The third preference party is similarly defined. Sometimes there is a tie for second preference on the strength-of-feeling scores. This is either decided according to who the respondent said they would vote for if they had a second vote, or, in a small minority of cases, the identity of the second choice is chosen so as to prejudice the test against the Myatt theory, but results under other coding schemes are similar (Fisher 2000). The level of tactical voting in England according to this measure was 5.0 per cent in 1987, 7.7 per cent in 1992 and 8.5 per cent in 1997. However, tactical voting as a proportion of the total population of voters is a somewhat artificial measure. It is more illuminating to look at tactical voting as a proportion of those voters who actually faced a tactical decision. Measuring tactical voting as a proportion of a population at risk of tactical voting is helpful because it adds context. Blais and Nadeau (1996) also suggested the idea of a risk population to aid identification of tactical voting. But, the risk population is most essential when analysing the determinants of tactical voting. Failure to limit analysis to the risk population can drastically change the meaning of any effects found in a logistic regression. For example, a negative association between the winning Extending the Rational Voter Theory of Tactical Voting by Stephen D Fisher 7 party share and the probability of tactical voting is expected for the population of voters as a whole. This is because the winning party share will be negatively correlated with the level of support for the third placed party and hence with the size of the population facing a strategic decision. This problem disappears when the analysis is restricted to the population at risk of tactical voting. Identifying members of the risk population within the BES is not trivial however. The risk population is pragmatically defined as all those voters whose preferred party came third or lower in the constituency at the election under investigation, at the previous election or in a poll estimate of the election result. Although the Liberals came third nationally at each election, they came first or second in roughly half of the constituencies. The risk population is composed of 4.5 per cent Conservative, 49.1 per cent Liberal, 38.1 per cent Labour and 8.3 per cent minor party supporters. So all parties are represented, not just third and minor parties nationally. Since the risk population includes only a quarter of all voters, the BES cross-section surveys for 1987, 1992 and 1997 are pooled for the analysis to avoid the problems associated with small numbers of cases. Since the cross-section surveys vary considerably in size, the pooled data set is weighted so that each election is equally represented. However, the wider sample of voters for each election remain weighted to the share of the vote. Analyses on an election-by-election basis produce similar coefficients but the p-values of the significance tests vary.