Working title: Intergroup relations as predictors of vote choice and election results in Great Britain from 2015 to 2019

Introduction

A. Context of study

  • The United Kingdom has experienced a turbulent electoral period between 2015 and 2019. This timeframe has included both the unexpected result of the Brexit referendum in 2016 as well as three General Elections, held over the span of these five years, two more than originally expected following the 2015 election. During these politically turbulent and uncertain years, key social psychological concepts from intergroup relations provide potential explanation and predictive ability, both in terms of individual vote choice and parliamentary constituency election results.

B. Lit review summary

Intergroup relations, voting, and elections

  • Previous piecemeal applications of intergroup relations to British politics have produced an incomplete picture of how this theoretical framework can provide insight to individual vote choice as well as election results.

  • This study aims to provide a more cohesive application of the dual process model of intergroup relations to illustrate the social psychological trends in the British public. Our orientations towards group membership, within group organization, and between group hierarchies have been shown to influence individual voting decisions. Relational orientations refer to how we relate to each other within and between groups as well as how we think relationships should be structured on a societal level.
    • (Group identities -> in-group membership and identification)
    • Authoritarianism -> within group structures and norms
    • Egalitarianism and social dominance -> between group hierarchies

  • Election prediction often prioritizes demographic factors as predictors of individual and constituency electoral outcomes rather than social psychological factors, such as these relational orientations. Intergroup and relational theorists suggest this only provides an incomplete illustration of the forces which shape our electoral decision-making and outcomes. Application of relational orientation to individual vote choice and constituency-level election results enables to not only predict the outcomes but also shed light on the influences of electoral decision-making. While the prioritization of demographics is often a result of data availability, we aim to illustrate the utility of such relational orientations and measurements in election prediction.

Electoral consequences of ideological and geographic sorting

  • Evidence from the US suggests that people are migrating to places with ideologically similar people.

  • This kind of sorting suggests that key social psychological influences on political attitudes and behaviours, specifically relational orientations, should be predictive of election results for electorally relevant areas, such as parliamentary constituencies.

  • In addition to moving to areas with ideologically similar people, do our relational orientations shift closer to those proximally near us? Especially in terms of politically-relevant geographic areas, such as our parliamentary constituency?

D. Research questions

  • RQ1: Do relational orientations predict individual vote choices above and beyond demographic factors?

  • RQ2: Do aggregate relational orientations of parliamentary constituencies predict election outcomes above and beyond demographic factors?

  • RQ3: Does the ambient relational orientation of one’s parliamentary constituency predict individual vote choices?
    Need to refine the RQ for Study 3

E. Methods summary

Study 1 investigates the predictive power of relational orientations in individual vote choice as compared with demographic variables.

  • Should I use logistic regression? Or a multi-level logistic regression for this?

  • Comparison of logistic regression models using BES data
    • Model 1: y = vote choice (1 = for right-wing parties, 0 = left-wing parties), x = demographic variables
    • Model 2: y = vote choice, x = relational variables (English, British, European identifications, authoritarianism, support for social equality, and support for wealth equality)
    • Model 3: y = vote totals for conservative parties, x = both demographic and relational variables

  • Repeat this for 2015 GE, 2016 referendum, 2017 GE, and 2019 GE.

Study 2 uses constituency-level estimates of relational orientations to predict election results and compares those results with the predictive ability of constituency-level measures of demographic varibables.

  • Constituency-level estimates of relational orientations are generated with BES data and used to fit linear regression models for constituency-level election results.
    • Model 1: y = vote totals for conservative parties, x = demographic variables
    • Model 2: y = vote totals for conservative parties, x = relational variables
    • Model 3: y = vote totals for conservative parties, x = both demographic and relational variables
  • Repeat this for 2015 GE, 2016 referendum, 2017 GE, and 2019 GE

Study 3 adds the constituency-level estimates of relational orientations to the models from Study 1 to assess if the ambiant relational oreination of one’s constituency has an association with individual vote choice.

  • Multi-level regression model comparisons

  • I could also see if people shift closer to the ambient relational orientations of the constituencies during this period. I’d limit the sample to those voters who stay in the same constituency and respond to the BES during these years, and calculate the mean difference between individuals and their constituency for each year to see if it changes. I could also do this for people who differ from their constituency (i.e. those who are relatively high in authoritarianism in a seat that’s relatively low in authoritarianism).

F. Results summary

  • This study finds that relational orientations are stronger predictors of vote choice and constituency-level election results than demographics in recent British elections. In fact, it appears that British electorate between 2015 and 2019 increasingly behaved as the theoretical framework of intergroup relations would predict.

Literature Review

Methodos

Results

Study 1

  voted Con 2015 voted Con 2015
Predictors Odds Ratios CI Odds Ratios CI
(Intercept) 0.14 *** 0.10 – 0.19 0.04 *** 0.03 – 0.05
englishness 2015 1.18 *** 1.16 – 1.20
britishness 2015 1.19 *** 1.16 – 1.23
europeanness 2015 16 0.95 *** 0.93 – 0.97
authoritarianism 2015 1.36 *** 1.30 – 1.44
wealthequal 2015 0.76 *** 0.75 – 0.77
socialequal 2015 0.87 *** 0.83 – 0.91
Age indexed to February
2014 		1.02 *** 1.02 – 1.02
Education level 0.96 ** 0.94 – 0.99
Gender 1.17 *** 1.10 – 1.24
white 1.66 *** 1.43 – 1.94
income Personal 1.08 *** 1.07 – 1.09
Observations 18061 27592
R2 Tjur 0.218 0.029
  • p<0.05   ** p<0.01   *** p<0.001
  voted Leave voted Leave
Predictors Odds Ratios CI Odds Ratios CI
(Intercept) 1.96 *** 1.38 – 2.77 1.13 0.89 – 1.44
englishness 2015 1.17 *** 1.14 – 1.19
britishness 2015 1.08 *** 1.05 – 1.11
europeanness 2015 16 0.53 *** 0.51 – 0.54
authoritarianism 2015 1.75 *** 1.65 – 1.85
wealthequal 2015 0.99 0.98 – 1.01
socialequal 2015 0.55 *** 0.52 – 0.59
Age indexed to February
2014 		1.02 *** 1.02 – 1.02
Education level 0.65 *** 0.63 – 0.67
Gender 0.96 0.89 – 1.02
white 1.25 ** 1.06 – 1.48
income Personal 0.98 *** 0.97 – 0.99
Observations 16108 16209
R2 Tjur 0.420 0.114
  • p<0.05   ** p<0.01   *** p<0.001

  • I know I need to narrow these samples down to the same sample (/observations which have all of the necessary variables). Is it okay if they’re different samples for each year? I don’t need to just narrow it down to those who have responsed to the necessary items for 2015, 2016, 2017, and 2019, right?

  • Also, I know these variable names need to be cleaned up. Are these the right stats to report?

Study 2

Study 3

Discussion

Conclusion