Indeed an interesting question. I was forwarded a link to the following paper, “Can Where People Vote Influence How They Vote? The Influence of Polling Location Type on Voting Behavior”, by Jonah Berger, Marc Meridith and S. Christian Wheeler of Stanford University. The abstract of their paper summarizes their work:

Can the type of polling place in which people vote (e.g. church, school, or firehouse) influence how they cast their ballot? Results of two studies suggest it can. A field study using Arizona’s 2000 general election found that voters were more likely to support raising the state sales tax to support education if they voted in schools, as opposed to other types of polling locations. This effect persisted even when controlling for voters’ political views, demographics, and unobservable characteristics of those individuals living near schools. A voting experiment extended these findings to other initiatives (i.e. stem cells) and a case in which people were randomly assigned to different environmental primes (i.e. church-related, school-related or generic building images). The present studies reveal that even in noisy, real-world environments, subtle environmental cues can influence decisions on issues of real consequence.

The authors develop an interesting theoretical argument, that voters may be affected by subtle (or sometimes not so subtle) cues in their behavioral environment. In the behavioral literature, this is typically called “priming”, and there is much experimental research that shows how “priming” (environmental cues) can alter behavior. They test this theory, as relating to polling place type, by looking at data from a recent election in Arizona as well as a real laboratory experiment. Both of these studies point to the conclusion that (in the Arizona case) those who voted in schools were more likely to support raising their taxes to pay for education.

I’ve got little doubt that these sort of subtle or not so subtle cues often do influence voting behavior. An example that I wrote about in 2005 was in Riverside County. Here’s a snip from the essay I wrote then about this type of phenomenon, here in a Catholic Church:

… in a vast parking lot behind the polling place located in a parish hall at the Corpus Christi Catholic Church (3750 Magnolia, Riverside County), one small “Yes on 73” sign was located in the parking lot, as was one large “Yes on 73” sign nearby. For those of you not closely following the California special election, Proposition 73 was the “Waiting period and parental notification before termination of minor’s pregnancy” measure (this wording is from the official voter guide). There were no other political signs located in the vast church parking lot, and there is no indication that church officials or anyone actually associated with the church put the signs in the lot. However, the relationship of such signs in a catholic church parking lot, on election day, may raise some concerns. When organizations take positions on certain issues, and those issues are prominent features of an election like the special election, that may mean that election officials might exercise caution about using the facilities of such organizations for polling places.

As to the Berger et al. research, it is intriguing and hopefully will get others thinking about this same question. It would be interesting to see if the Berger et al. result from the Arizona case holds up with stronger statistical methodologies (for example, applying some of the more modern ecological inference tools like King’s “EI” or causal inference models [see Jas Sekhon’s tools here]). It would also be fascinating to see if this result holds in other settings, though the difficulty will be getting good data on what type of location a particular precinct used in any election.