In new research with Dr. Sara Cordes, Dr. Liane Young and RA Mackenna Woodring, I investigated the role of voters’ beliefs and desires in the mathematical processing of election-related information just before the 2012 U.S. Presidential Election.

We presented voters with mental math problems based on fictional polling results, and asked participants who they intended to vote for, who they expected to win, how emotionally invested they were in the preferred candidate winning the election, and how much they liked and agreed with the preferred candidate.

We found that both committed Obama and Romney voters arrived at solutions that symmetrically underestimated support for the opponent and overestimated support for the preferred candidate. An identical experiment conducted 10 months after the election revealed the absence of this symmetrical bias: Obama and Romney voters alike produced estimates that aligned with the Obama win (i.e., they underestimated Romney leads). This result, and a finding in the first experiment that a subset of voters who didn’t expect the preferred candidate to win did not underestimate the opponent’s lead suggest that estimates were largely constrained by participants’ expectations about the likely or actual state of the world. Moreover, within the constraints of expectations, estimates tracked with the value voters attached to the candidates, i.e., how much participants liked and agreed with them.

This work suggest that a necessary condition for the biased processing of quantitative information in favor of one’s preferred candidate may be a true belief that they will actually win. A simple desire that they will win may be insufficient.