| What’s the impact on voter turnout of distance? More generally, how do the practical considerations of personal convenience affect the decision to vote? |
| Grebner :: Technical Politics – The effect of distance between home and polling place on turnout. |
| Taking the Michigan voter file as it existed in September 2006, I conducted a logistic regression analysis, with voting as the dependent variable, and independent variables including previous voter record, various demographic variable, and a transformed measure of distance between the registration address and the coordinates of the polling place.What I found was a small but consistent effect: being further away makes voting slightly less likely. For a person who would normally have a 50% chance of voting, adding one mile between home and polling place reduced the turnout to about 48%. I used the actual distance, rather than the distance along the driving route, because it was easier, and I was examining millions of points.
The distance variable was coded as the quartile of distances among voters in each precinct. That is, comparing people who lived within a given precinct, I compared the 25% living closest to the poll, to the 25% living farthest away. This was necessary because raw distance to the poll is correlated to urbanization and socio-economic status – cities have geographically small precincts, while cities and rural areas are spread out. So I compared turnouts among people who live in precincts of similar sizes. The effect was quite robust, and the coefficients were consistent in size and sign for precincts of different sizes. I was worried that even within a precinct, people who live near the polling place may live in different kinds of neighborhoods than people who live farther away. Examining distance to the NEXT nearest polling place showed no statistical relation to turnout, which makes me confident the main result I found was not a statistical artifact. The effect was greatly reduced for people who ordinarily use absentee ballots when they vote, which makes sense. The research I performed doesn’t distinguish between literal distance and the lack of familiarity with the polling place. That is, I don’t know how much of the effect is due to voters’ reluctance to TRAVEL, as opposed to their reluctance to go somewhere UNFAMILIAR. Maybe it would be possible to code polling places by how visible they are from major thoroughfares – that’s beyond my interest. I also didn’t look at the effect of MOVING a polling place, which I would expect would have a separate effect of reducing voter turnout, which would presumably be transient and fade within a few years. |