Ballot Design and Unrecorded Votes

University Of Missouri-Saint Louis, Saint Louis MO

Investigators

Abstract

The Problem. Recent elections indicate that a significant number of voting errors occur because of malfunctioning equipment or confusing voting procedures. For example, roughly two million voters in the United States failed to cast a valid vote for president in the 2000 election. The recount of votes in Florida in 2000 was complicated by unrecorded votes (the difference between the number of people who went to the polls and the number of valid votes cast for president). In the wake of the 2000 election, state and local governments are implementing several reforms to improve the voting process, and much of the reform effort has focused on technology. Hundreds of counties have bought new voting equipment in the last three years. However, even though the main source of voting errors in Florida in 2000 was a ballot design flaw, election officials and the academic community have not given enough attention to ballot design as a source of voting errors. Proposed Research. This investigation tests whether ballot features are associated with unrecorded votes for several contests in the 2004 election. The investigators use a theoretical framework based on research in survey methodology, psychology, and other disciplines. Their framework for evaluating ballot design was developed in a pilot study of 250 counties from five states that had gubernatorial elections in 2002. For the 2004 election, they obtain ballots from roughly 1,750 counties that use hand-counted paper or optical scan ballots. They assess each county's ballot features and collect voting returns, demographic data, and information on voting technology for each county. The researchers conduct a statistical analysis to assess the impact of ballot features on unrecorded votes in several contests. Scientific impact. This investigation assesses the impact of ballot design on voting errors. The pilot study in 2002 found that poor ballot design had a stronger effect on unrecorded votes for governor than any other factor (including voting equipment). The current investigation of the 2004 election examines additional ballot features in a larger and more diverse sample of counties. The researchers test the impact of ballot features in the presidential contest as well as other down-ballot contests. Finally, the investigators maintain an inventory of ballots collected from the 2004, 2002, and 2000 elections that will be available to researchers and election officials. Broader impacts: This study will demonstrate the degree to which ballot design influences a voter's ability to cast a ballot accurately. State and local election officials who buy new voting equipment without paying attention to ballot design may not achieve the desired result of reducing voting errors. Similarly, election officials who resist buying new electronic voting machines because of cost and security concerns can still reduce voting errors by incorporating the ballot design recommendations from our study. The investigators use the results of this study to create educational materials on best practices in ballot design for state and county election officials.