Analysing vote-choice in a multinational state: national identity and territorial differentiation in the 2016 Brexit vote

ABSTRACT

Striking territorial variations in the 2016 Brexit referendum are neglected in the explanatory literature, a gap our analysis of the British Election Study (BES) helps to fill. Rather than modelling Britain as one political system, we present parallel models for England, Scotland and Wales. Typical in other multinational states, this approach is innovative for ‘British politics’. The analysis shows different variables help to explain Leave support in England, Scotland and Wales. To analyse complex multilevel national identities, we develop a relative territorial identity (RTI) measure. The measure shows that those who prioritize their territorial identity in England tended to vote Leave, while the opposite is true in Wales and Scotland. The performance of this RTI measure helps to explain territorial differentiation in the 2016 Brexit vote.

KEYWORDS:

• vote-choice
• Brexit
• territorial differentiation
• national identity
• relative territorial identity (RTI)
• survey data analysis

JEL:

• D72
• H10
• R50
• Y8

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1. A lack of comparable data from Northern Ireland, which might otherwise allow us to test various explanations of vote-choice, precludes a cross-UK analysis; a key (if unavoidable) limitation of this study. For detailed analysis of the Brexit vote and its aftermath in Northern Ireland, see Garry et al. (2018, 2020).

2. All data are available from the BES website (http://britishelectionstudy.com).

3. The BES does not include other widely used measures of identity (Bechhofer & McCrone, 2015; Mendelsohn, 2002), including the Linz-Moreno scale (Scottish not British, More Scottish than British, etc). Others, including Moreno (2006), have compared the relationship of different measurement techniques for national identity. Consistent with their claims, Henderson et al. (2014) find that a relative territorial identity manages to capture both strength of attachment and primary political community in a more nuanced way than Moreno.

4. The state and sub-state identities relate to each other in different ways, more entangled in England, and more zero-sum in Scotland. This is reflected in the distribution of the relative measure (available from the authors upon request) and in the eventual findings.

5. We have excluded occupational and subjective social class from our analysis for three reasons. First, the BES occupational class variable captures professional job codes and over-represents middle-class respondents. Second, once a respondent’s education level is included in the model, the effect of class is no longer significant. Third, subjective social class is highly correlated with national identity and residence: respondents in Scotland and Wales are more likely to claim they are working class when, by objective measures, they are not. Since neither the occupational nor the subjective social class measures suit our purposes, we use a measure of personal income and education level.

6. This is obviously a proxy measure designed to capture the aggregate-level explanations raised by others with the individual-level data that best approximate the core idea.

7. The correlation between Wave 9’s immigSelf (with ‘don’t knows’ removed) and Wave 8’s immigCultural (with ‘don’t knows’ removed) is 0.725; and the correlation between Wave 9’s immigSelf (with ‘don’t knows’ removed) and Wave 8’s immigEcon (with ‘don’t knows’ removed) is 0.729.

8. For each model we have employed weighted data using the full weight for Wave 9 (the wave which contains our dependent variable).

9. The direction and significance of the Low Efficacy coefficient depends on whether the conceptually similar anti-expert measure is in the model (both variables appear in the same BES question battery). It is included here for completeness.

10. Anyone interested in regional variation within Britain might reasonably query regional variation within England. We have run the five models in each of the regions of England and find that RTI has a consistent and positive effect on Leave support across all regions of England, including London. Englishness seems to operate as a consistent national identity across England. The sole exception is when (in model 5) we add attitudes to migrants, which causes the identity effect to disappear in those regions with the smallest samples. Full results are available from the authors upon request.

11. As a robustness check, we tested whether the RTI model results were affected by respondents who were born in one nation but settled in another. Although < 3% of the English sample and < 10% of Scottish respondents were born elsewhere in the UK, such intra-UK migration is potentially more significant in Wales (one in five Welsh residents were born in England at the 2011 Census) (Office of National Statistics (ONS), 2011). However, of the total Wave 9 Welsh sample of 1079, only 46 respondents in Wales expressed a stronger English identity than a British or a Welsh identity. Excluding these individuals from the model does not change the significance or direction of the coefficients.

12. BES weights increase the regression models’ standard errors. Our results are more robust without them: model fit improves, the RTI coefficients are as anticipated (positive for England, negative for Scotland and Wales) and significant everywhere at p < 0.01.

Additional information

Funding

This work was supported by the Economic and Social Research Council (ESRC). Specifically, Daniel Wincott's time on this project was funded by grant numbers ES/P009441/1 and ES/R007500/1. Jac Larner's time was funded by grant number ES/R007500/1.