Published on The Plot blog on 18 July 2016
The outcome of the Brexit referendum has sparked a wave of comments to understand who the Leave voters are. The evidence so far indicates that those who decided to leave the EU tend to be older, less educated and poorer than those who decided to remain in the EU. However, there is another factor that seems to explain support for Brexit, i.e. voters’ inclination to take risks. Despite presenting two apparently equivalent options, referendums require voters to choose between the uncertainty of a change – such as a constitutional change or the introduction of a new policy – and the relative certainty of maintaining the status quo. In most cases, the ‘change option’ is intrinsically riskier than keeping the status quo, and the outcome of Britain’s EU referendum clearly shows that “Brexit is riskier than Bremain”, as Timothy Garton Ash wrote.
Given these differences between referendum options in terms of risks and uncertainties, we can expect risk preferences to influence not only voting decisions, but also how voters react to campaign arguments from both sides of the referendum campaign. Were risk takers more likely to vote Leave in the Brexit referendum, and how did information influence their voting decisions?
Risk propensity and Leave votes
Preliminary analysis of the newly released Wave 7 of the British Election Study Internet Panel (Fieldhouse et al., 2015) confirms that respondents’ risk attitudes are significantly correlated with the probability to vote Leave. Figure 1 plots regression coefficients for a logistic regression model in which the dependent variable corresponds to the intention to vote Leave, as measured in the period April-May 2016. As the graph shows, those who are very willing to take risks are five percentage-points more likely to vote Leave than those who are very unwilling to take risks (reference category). This result is statistically significant after controlling for several key factors that are correlated with voting Leave and risk attitudes, such as gender, age, education, party identification, trust in MPs, British identity, and approval of EU decisions.
Figure 1. Probability to vote Leave by risk attitudes
Note. Probability to vote Leave by risk attitudes (reference category = ‘very unwilling to take risks’), controlling for gender, age, education, trust in MPs, British identity, party identification and approval of EU decision. 95% confidence intervals. Source: Author’s calculation on BES Panel data, Wave 7, April-May 2016 (Fieldhouse et al., 2015).
Balancing voting intentions
Given this correlation between risk propensity and support for Brexit, it is worth questioning whether voters responded differently to campaign arguments depending on their inclination to take risks. To explore this question, at the beginning of June I conducted an online experiment financed by the European Union Democracy Observatory (EUDO) at the European University Institute and by the Centre for Experimental Social Sciences (CESS) at Nuffield College, Oxford. In the experiment, a pool of participants (N=259) read a balanced set of pro/anti EU arguments that were carefully selected from publicly available sources of information. The study was conducted online with a diverse pool of British participants (median age=35, SD=15.4) provided by CESS.
Figure 2 presents preliminary findings from the analysis of experimental data. The values on the Y-axis correspond to changes in pro-leave attitudes (left-hand graph) and intentions to vote Leave (right-hand graph) in the treatment condition (information) compared a control condition of no information. Positive values indicate that exposure to information increased support for leaving the EU compared to the control group (value 0), while negative values indicate a decrease in support for leaving the EU compared to control.
Figure 2. Effects of information on pro-leave attitudes and intentions to vote Leave
Note. Y-axis: changes in pro-leave attitudes (left-hand graph) and changes in the intentions to vote Leave (right-hand graph) compared to control group (value 0), with 90% confidence intervals.
As the graph bars show, participants reacted differently to the same type of information stimuli, depending on their inclination to take risks, as measured on a standard risk-attitude question. After reading a balanced set of pro/anti EU arguments, support for leaving the EU increased among those who were more risk averse compared to a condition of no information. On the other hand, the opposite pattern occurred among risk takers, since information reduced support for leaving the EU within this group of participants.
Interestingly, if we measure participants’ risk propensity with a behavioural indicator – i.e. the decision to either play or not to play in a risky lottery – the results are similar to those presented in figure 2: information increased support for Leave among those who decided not to play the lottery, while it reduced it among those who chose to play.
A puzzling result?
This somewhat puzzling finding can be explained as the ‘balancing effect’ of information. For those who were less inclined to take risks, reading a set of convincing arguments might have reduced the uncertainties related to leaving the EU, thus information increased their support for that side of the referendum campaign. On the other hand, the same set of carefully selected arguments might have induced risk takers to reconsider both the risks of a Leave option that they were initially more inclined to support, and the benefits of remaining in the EU. In other words, provision of balanced information contributed to balancing voting intentions by reducing the gap between risk-averse and risk-taker voters.
These results suggest not only that further attention should be devoted to how risk attitudes influence voting behaviour in referendum campaigns, but also that referendum campaigners should keep in mind that different voters might react very differently to the same campaign arguments. Reducing the uncertainties related to changing the status quo can prove crucial to attract risk-averse voters to the ‘change side’ of a referendum campaign, and ultimately secure their support on voting day.