In noncompetitive, low-profile races like the first and third districts, voters take a bit longer to decide than they do in competitive, highly publicized races like the fourth district.
This analysis was performed by Robert Richards, a student research fellow at BYU’s Center for the Study of Elections and Democracy (“like” CSED on Facebook), in collaboration with CSED faculty. The writing is mostly his. Inquiries about this research should come to Adam Brown or Quin Monson.
In many elections, the press tends to make a big deal about the number of undecided voters in any given race and the potential of these undecideds to change the outcome of the election. So what kinds of people wait until a few days before the election to make up their minds? One way to examine undecided voters is to look at what kinds of voters decide earlier in the election cycle rather than later. We asked questions on the Utah Colleges Exit Poll about when they decided who they would vote for in their U.S. House election.
Who decides early
It turns out that early deciders and late deciders tend to be different on certain demographic factors. Those who decided earlier who to vote for in their U.S. House elections were older, wealthier, more educated, more Democratic, and more likely to be male. The early deciders also tend to be more strongly ideological (whether in the conservative or liberal direction) and more strongly partisan (whether Republican or Democrat). Even when controlling for strength of partisanship, Republicans tend to take longer to make up their minds.1
Comparison across Congressional districts
In addition to differences among voters, there were interesting differences across districts in when voters decided who to vote for. The graph below shows some of these differences, controlling for various political and demographic variables, as well as the party and incumbency status of the candidate the person eventually voted for.2 The solid lines show what percentage of voters remained undecided by the specified time period. The dotted vertical lines mark the point at which 50% of voters in each district had made their decisions.
The first and third districts were virtually the same in this regard. Both of these districts had incumbents running without exceptionally strong challengers. These two races largely went unnoticed by the rest of the state, and the incumbents ultimately took home at least 65% of the vote in their respective districts.
The second district battle between Chris Stewart and Jay Seegmiller also went largely unnoticed statewide, but the seat was open due to Jim Matheson’s decision to run in the new fourth district. Without a well-known incumbent running, and without a high level of media attention to help them out, voters took slightly longer to decide in the second district than in other districts.
Meanwhile, voters in the fourth district decided far earlier than voters elsewhere. This was the intense, highly publicized campaign between Jim Matheson and Mia Love, which turned out to be a very close race. Despite the large amounts of money spent late in the race, this race’s intense media coverage apparently helped voters in this race make their decisions earlier than voters in the other districts.
Based on this analysis, it seems that the number of undecided voters in a Congressional race is affected by the kind of race going on in their district. In noncompetitive, low-profile races like the first and third districts, voters take a bit longer to decide than they do in competitive, highly publicized races like the fourth district. In competitive races without a lot of outside attention, voters take slightly longer to decide. We only have four cases to compare here, but the story seems reasonable based on the data.
The Utah Colleges Exit Poll was conducted on Election Day 2012 by student volunteers from various colleges and universities in the state of Utah. In addition to the Election Day polling, the UCEP also administered an online survey to early and absentee voters. One of the Election Day forms of the statewide questionnaire and the early voter survey both included the question about vote timing used in this analysis, which appeared as follows:
When did you finally decide who to vote for in the race between [name of Republican] and [name of Democrat]?
- 2-3 days ago
- Sometime last week
- Earlier in October
- In September
- Before September
Data for the modeling in this analysis was left unweighted, but was consistent with bivariate analysis using the weights.