PROJECT SUMMARY

Reviewing qualitative information developed by earlier researchers and conducting their own research and outreach to election authorities, a team of Harris students working with the Center for Survey Methodology has determined that, in partisan elections, only 12 of the nation's 50 states rotate or otherwise randomize candidate names across precincts, counties, legislative districts or other jurisdictions within the state. Earlier studies have found a statistically significant candidate name order effect on election outcomes. The absence of randomization procedures in partisan elections in 38 states suggests that an alarming number of election contests may be affected by prejudicial name order methodologies. Election data from states that rotate or otherwise randomize candidate names in partisan elections will enable researchers to replicate and expand upon the earlier studies. So, the Harris team has begun to meticulously inventory and collect the needed available data, which includes candidate name order by unit of rotation, election results, and voter registration and partisan identification data. (Data from 10 of the 12 states is being collected; in the remaining two states, the rotation or randomization occurs across too few jurisdictions to make statistical analysis useful.)

BACKGROUND

In June of 2017, the Center for Survey Methodology and the Project on Political Reform hosted a conference on proposals for improving the integrity of the electoral system in the United States. Conference participants included nationally-prominent Republican and Democratic political practitioners and scholars from around the country who study democracy and governance.

During the conference, Stanford University Professor Jon Krosnick presented the results of his investigation into the effects of candidate name order on election ballots. Professor Krosnick also discussed at length related studies by other scholars. Based on analysis of the data, Professor Krosnick concluded that, in jurisdictions in which candidate name order is not rotated or otherwise randomized, name order can have a statistically significant impact on election results, sometimes determining which candidate is deemed the winner of the election.

At the conclusion of Professor Krosnick's presentation, political practitioners of both parties voiced concerns about the possibility that name order effects could be unfairly disadvantaging candidates whose names are systematically listed beneath the first position on a ballot due to requirements of, or deficiencies in, state election law. For example, in some states, name order is determined by which party's candidate(s) won the previous election. In other states, name order is alphabetical. In still other states, a single random order is adopted and mandated for use throughout the state. These and other methods used to establish non-randomized name orders may be, according to the studies conducted by Professor Krosnick and other scholars, distorting election outcomes in close races.

As a result of the concern expressed by the practitioners attending the conference, the Center for Survey Methodology has initiated a comprehensive investigation of the effects of candidate name order on election ballots. The investigators have determined that, in partisan contests -- that is, contests in which candidates may be identified as affiliated with a political party -- only 12 of the nation's 50 states rotate or otherwise randomize candidate names across precincts, counties, legislative districts or other jurisdictions within the state.

Here is a spreadsheet summarizing the current status of data identification and collection efforts.