Geographic Mobility, Social Connections and Voter Turnout

Abstract

It is known that those recently moved are less likely to vote. If the decision to vote is a cost–benefit calculation, then those intending to move may also be less likely to vote as they will not be affected by the local result. This self-interested calculation might be mitigated if people vote sociotropically. We show that the effect of intending to move is conditional on levels of social capital: those who have invested in their community are less affected by their intention to move in their decision whether to vote. Using survey data from the UK we show that (a) those intending to move are less likely to vote, (b) higher social capital increases the likelihood of voting and (c) the negative effect of intending to move diminishes when people have more social investments in their community.

Acknowledgement

The research reported in this paper was supported by an Economic and Social Research Council grant number RES-166-25-0012 “Public Services: Exit and Voice as a Means of Enhancing Service Delivery”.

Notes

Stephenson and Crête's analysis suggests that in models of turnout, differences in survey mode do not produce significantly different results (2008: 11).

Responses of “very likely” and “likely” are coded as 1; “neither likely nor unlikely”, “fairly unlikely” and “very unlikely” are coded as 0.

The options are: children's education/schools; youth/children's activities (outside school); education for adults; sports/exercise (taking part, coaching or going to watch); religious (e.g. church); political (e.g. a group fighting for a particular cause or issue, political party); trade union; the environment, animals; justice and human rights; citizens' groups; local community or neighbourhood groups; the elderly; health, disability and social welfare; safety, first aid; hobbies/recreation/arts/social clubs; and Internet group or chat room.

The social trust question is the standard, “Which of these statements do you agree with: ‘Generally speaking most people can be trusted’ or ‘You can't be too careful when dealing with people’?”

For excellent discussions of interaction effects, see Brambor et al. (2006) and Gill (2001).

Other variables are fixed at their mean for continuous variables or median for categorical variables. Race is set as white and trust is fixed at “people can be trusted” in both figures.

We have also estimated each of the models above, controlling for how far respondents intend to move and whether they intend to move across a jurisdictional boundary such as an electoral district or borough council. In none of the alternative specifications do the results change appreciably. Accordingly, we are confident that our findings are not driven by the distance of intended moves.