Cognitive ability and voting behaviour in the 2016 UK referendum on European Union membership

Abstract

On June 23rd 2016 the UK voted to leave the European Union. The period leading up to the referendum was characterized by a significant volume of misinformation and disinformation. Existing literature has established the importance of cognitive ability in processing and discounting (mis/dis) information in decision making. We use a dataset of couples within households from a nationally representative UK survey to investigate the relationship between cognitive ability and the propensity to vote Leave / Remain in the 2016 UK referendum on European Union membership. We find that a one standard deviation increase in cognitive ability, all else being equal, increases the likelihood of a Remain vote by 9.7%. Similarly, we find that an increase in partner's cognitive ability further increases the respondent's likelihood of a Remain vote (7.6%). In a final test, restricting our analysis to couples who voted in a conflicting manner, we find that having a cognitive ability advantage over one's partner increases the likelihood of voting Remain (10.9%). An important question then becomes how to improve individual and household decision making in the face of increasing amounts of (mis/dis) information.

Fig 1. Cognitive ability and a vote to remain. Bar chart of the relationship between deciles of our composite measure of cognitive function, ‘Cognitive Ability’, and the frequency of a Remain vote.

Horizontal dashed line at 50%. Sample of 6,366 individuals from 3,183 households.

Fig 2. Cognitive ability and a vote to remain or leave. Distributions of our standardized composite measure of cognitive function, ‘Cognitive Ability’, for Remain and Leave voters.

Sample of 6,366 individuals from 3,183 households.

Fig 3. Within household differences in cognitive ability. Histogram of the variation of our composite measure of cognitive function, ‘Cognitive Ability’, within households where the actor and partner voted in a conflicting manner.

Based on 926 individual observations from 463 households.

Fig 4. Cognitive ability and the propensity to vote remain.

Estimates from regressions where our composite measure of cognitive function, ‘Cognitive Ability’, is entered in deciles. The omitted category is the l^st decile. Point estimates and 95% and 90% confidence intervals are shown. Panels A, B, C, D relate to the econometric specifications in Column’s 1, 2, 3, 4 of Table 2, respectively.