Voting and the Environment… But Not What You Think

Just read a fascinating report from Christian Wheeler’s group at Stanford’s Graduate School of Business.  The report tackles an interesting question: Does how you vote depend on where you vote?
As we’ve noted before, subtle features can (and sometimes do) have significant effects on how people behave.  The Stanford team hypothesized that voting at polling places in schools could gently nudge people toward voting for ballot initiatives supporting increased school funding.

Specifically, they looked at support for Proposition 301, a school-funding initiative from the state of Arizona in 2000.  They found that support for Prop  301 was higher among folks voting in schools than those voting elsewhere (55.3% vs. 53.4%).

Of course, this is not a fair comparison because people who vote in schools may differ systematically from those who vote elsewhere.  The researchers worked to account for these differences by controlling for voting on 13 other propositions and for president, as well as a number of other factors (e.g., comparing folks who voted in non-school locations but lived within 0.15 miles of a school to those who voted in school locations).

When the dust settled, the researchers found that voting in a school still had an effect (albeit smaller) on how people voted on the school-funding proposition — about a 1% premium for those who voted at school polling locations.

The big shortcoming, of course, is that people aren’t randomized to voting location.  So the researchers moved from the field into the lab and undertook a “priming” study.  In such studies, researchers expose some subjects to the prime (usually some subtle feature); the rest form a comparison (or control) group.

Priming works best when the subjects are unaware that it’s taking place.  So Wheeler and colleagues asked subjects to rate the brightness of 15 pictures.  In the school group, 10 of the 15 pictures were of well-maintained schools; the remaining 5 were of nondescript buildings.  In the control group, all 15 pictures were of nondescript buildings.  And in the church group, 10 of the 15 buildings were images from churches (e.g., pews).

They then asked subjects to “vote” on a number of different initiatives: Prop 301 from the Arizona ballet, California’s 2004 stem-cell funding initiative, and five other ballot initiatives from the 2000 Arizona ballot.

The results are very interesting.  Note that because the study design used randomization, confounding is not an issue; assignment to the priming condition was accomplished randomly (and was therefore disconnected from any underlying inclinations to vote one way or another).  The chart below shows the results, which are quite interesting:

Subjects in the church prime group were significantly less likely to support the stem-cell research initiative than those in the control or school prime groups (p = 0.02), while those in the school prime group were significantly more likely to support the education funding initiative than those in the control or church prime group (p = 0.02).

The real-world effects are much smaller than those seen in the lab (differences of around 1% versus differences of about 20%).  Nonetheless, very subtle features can in fact measurably affect how people respond and behave.

These features are so subtle that they may escape our notice as we develop messages to healthcare consumers.  We need to keep on our toes about the possibility that we’re inadvertently nudging people one way or the other when we communicate with patients.

(Note: this entry originally appeared at

About the author


View all posts