Who’s on first?
Published 8:39 am Saturday, August 15, 2009
Every year, about this time, I hear from people who have watched their state set the order of candidate appearance on the fall ballot. Some states put candidates in chronological order of their official filing with the elections board, while others choose candidates or parties by lot. Regardless of the method, inquiring political minds want to know, does it make any difference?
The theory has always been that the most desirable position on the ballot for any candidate is to be listed first. Dozens of published articles and conference papers have been written on the subject.
While not all the research is in agreement — no surprise there —my own interpretation of the bulk of the findings suggests the following:
1. There is an advantage to being listed first on the ballot. Voters who do not have well-defined choices prior to voting appear to latch onto the first name on the ballot for each office, a phenomenon we call “first-listing bias.”
2. The advantage for first-listed candidates varies widely. In some elections a first-listing produces just a handful of votes, though they can make the difference in an extremely close election. In other elections a first-listing can generate extra votes up to about 5 percent of the overall tally, according to some studies.
3. Offices at the top of the ballot, for president, governor and senator, produce the fewest additional votes for a first-listed candidate. That is because the candidates for these high-visibility offices tend to be well-known.
4. Offices in the middle and bottom of the ballot are especially susceptible to the first-listing bias. Many candidates for lower statewide elected office (such as lieutenant governor, labor commissioner, etc.) and other localized offices (state legislators, city councilors) are surprisingly little known by many voters.
5. Partisan elections have a lower first-listing bias than nonpartisan elections. The party identification listed next to candidates serves as a major voting cue — an inducement to cast a ballot for a candidate — for two-thirds of Americans who have a partisan affiliation.
6. Elections without well-known incumbents are more susceptible to first-listing bias than those with such incumbents. Incumbency can substitute for a party label, in that less attentive voters may use name identification as a vote prompt where party identification is not available.
7. Primary elections are more susceptible to first-listing bias than general elections. By definition, party primaries do not contain a party identification prompt. All the candidates are either Democrats or Republicans, and so party voters lack a key voting cue.
8. The first-listing bias can be just as helpful to minor-party candidates and independents who gain the top ballot position. In other words, the first-listing bias doesn’t merely assist Democrats and Republicans.
9. There is some evidence that, in a long listing of candidates for a particular office, being listed last is almost as good as being listed first. This is somewhat biblical “the first shall be last and the last shall be first”— but essentially, the suggestion is that the voter’s eyes assess a large, multi-candidate field by focusing on the first listed candidate and then the last.
10. Of all these principles that govern the first-listing bias, the most important are the degree of information held by individual voters and the position of the office on the ballot. Elections that draw a disproportionate number of well-informed voters have lower first-listing bias effects.
One of the key 2009 elections, the Virginia gubernatorial battle, is an echo of an earlier contest that may have involved first-listing bias. In 2005 Republican Bob McDonnell ran against Democrat Creigh Deeds for the office of state attorney general in Virginia. Attorney general is the third of three statewide offices listed on the ballot. As determined by lottery the prior June, McDonnell was put first on the ballot, Deeds second, and there were no other candidates. Party labels were listed on the ballot. After a recount, McDonnell won with 970,981 votes to 970,621 for Deeds, an astonishingly small difference of 360 votes — the closest statewide election in Virginia’s history.
About 40,000 people who had voted for governor did not vote for attorney general, and disproportionately, these voters were found in heavily African-American precincts (where Democrat Deeds received almost 90 percent of the vote). This racially-tilted ballot drop-off is the single best structural explanation for Deeds’s narrow loss.
However, the McDonnell-Deeds contest was not highly publicized, certainly compared to the race for governor. Few would argue that a sizeable percentage of the voters on Election Day knew little about either candidate. Given our principles of first-listing bias, it is highly probable that first-listed candidate McDonnell gained considerably more than his 360-vote margin from the luck of the June drawing.
This November, McDonnell and Deeds are facing off once more, this time for the top-of-the-ballot office of governor. McDonnell and the Republicans won the first listing again. The probability is great that this year’s election won’t be nearly as tight as the one four years ago. Deeds certainly hopes so, or he could be foiled a second time by first-listing bias.