First Among Equals?

The research, by Yuval Salant, assistant professor of managerial economics & decision sciences at Kellogg, and Marc Meredith, an assistant professor at the University of Pennsylvania, also examined why voters show a preference for candidates in the first position on the ballot. The pair have not found a complete answer to that question yet. But they conclude that the reason most commonly given for voters’ preference—choosing the first name that appeals to the voter as he or she reads down the ballot—does not fully account for the result.

Salant’s research focuses on framing effects—how presentation can influence a person’s choice in such situations as the order of products shown to online shoppers and the precise placement of items in supermarkets. “There’s a large experimental and empirical literature on that, but my focus is mainly on the theory behind framing effects,” Salant explains.

A Natural Experiment
Salant and Meredith sought to measure the impact of framing on winning office rather than vote share, and in the process “collected an entirely new dataset,” Salant recounts. He and Meredith based it on the results of elections to city councils and school boards in California. Why California? “We were looking for a natural experiment: a situation in which the order of candidates on the ballot is determined randomly,” Salant explains. The Golden State satisfied that requirement, because it has a unique method of assigning ballot positions to ensure that candidates whose surnames start with the letter “A” do not dominate the top of the ballot.

“After the California local election entry-deadline has passed, the California Secretary of State draws a random order of the alphabet according to which candidates are listed on all ballots,” the researchers’ paper states. Because candidates positions’ on the ballot are quasi-random, Salant and Meredith expected the distribution of candidates’ characteristics and the number of winners to be similar for all ballot positions. They then “counted how many winners came from the first position and calculated the expected number from that position, and compared them using statistical analysis,” Salant says.

Clear-cut Results
The results were clear-cut. “In one out of ten elections, the candidate listed first won just because he was listed first,” Salant recalls. “The first candidate advantage,” the paper notes, “comes primarily at the expense of candidates listed in the median ballot position who are 2.5 percentage points less likely to win office than expected absent order effects” (Figure 1). The first candidate advantage was “similar in city council and in school board elections, in races with and without an open seat, and in races consolidated and not consolidated with statewide general elections.” In addition, the percentages of winners from specific positions remained similar whether the elections were designed to produce one or more winners.

In more important elections such as those for governors, senators, and the presidency, ballot position may not have as much of an impact. “I would expect that the effect is smaller the more important the elections are,” Salant says. “But the results will still be there, particularly in close races. However, we don’t have enough data to confirm that.”

Nor did the project produce enough information to determine just why ballot order has such a significant effect on winning office. A popular explanation is that voters indulge in “satisficing”: they evaluate the candidates as they scroll down the ballot and choose the first one that meets their basic criteria, instead of selecting the best candidate from the entire list. However, Salant and Meredith caution against over simplified explanations. Careful analysis of the California results indicates that satisficing does not account for the complete first candidate advantage. “There is something else going on here, but we don’t have enough data to determine what it is,” Salant says.

Practical Considerations
Overseas authorities have already taken practical advantage of ballot order. Salant and Meredith quote the example from Russia’s regional parliamentary elections in March 2007. In a supposedly random allocation of parties to ballot positions in these elections, the then-President Vladimir Putin’s Unified Russia party appeared in the first ballot position in eight of the fourteen regions, a full six regions more than expected under a random allocation. “This finding underscores the importance of ensuring that any policies done to mitigate order effects, like randomization or rotation, are done so in a fair manner,” they conclude.

Nevertheless, their research shows that authentic randomizing of ballot order can improve the fairness of elections. Yet is the California system fair enough? An ideal system, Salant says, would show each individual voter a randomized ballot order. But that, “would be very costly,” he adds. Electronic voting could help reduce the financial cost of such a system, but with randomization “voters may still get confused, which is the other source of cost,” he says.

Voters are not the only ones who can benefit from this research. “Details that seem irrelevant may affect the way we behave in other settings, like the supermarket, buying online, and choosing a retirement plan,” Salant says. “As individuals, we should be alert to order and other framing effects when making decisions.”

Related reading on Kellogg Insight:

Engineered Electioneering: The “when” and “what” of candidates’ messages

Voters Love Winners (And So Do Endorsers): How endorsers can gain the upper hand

Predicting Politics: Prediction markets out-predict political pollsters