Last month, the Government Accounting Office (GAO) issued a congressional report on the problem of long lines and extended waits experienced by voters on Election Day in the November 2012 election.
This is a fallow and inviting field for operations research, as noted in a 2008 Caltech field study of 30 California polling places conducted by Douglas Spencer and Zachary Markovitz. Spencer, Douglas., and Z. Markovitz, Long Lines at Polling Stations? Observations of an Election Day Field Study, Election Law Journal Vol. 9, No. 1 (March 1, 2010).
There is an existing literature on queuing theory developed for proprietary businesses that studies the experience of standing in line for services, there are deeper and more thorough surveys of local election officers being produced, and there is a burgeoning amount of scholarly literature on wait times and the voting experience (see, e.g., Stewart, Charles III, and S. Andsolabehere, Waiting In Line to Vote, Caltech/MIT Voting Technology Project Working Paper No. 114 (July 28, 2013)).
In general, the studies of voting have shown that long lines and slow service at polling places disproportionately affects minority voters, with particularly strong effects on African-American voters, who (all other things being equal) tend to wait in line for an average of around 30 minutes, compared to an average wait of around 14 minutes for white voters. Interestingly, African-American voters (across all jurisdictions) are more likely to be asked to show a photo I.D. than whites. Stewart, Charles III, Racial Differences in Election Administration, Caltech/MIT Voting Technology Project Working Paper No. 82 (July 2009).
Compared to the more comprehensive and detailed surveys and research papers produced by academics, the GAO report is much more limited in scope. The value of the GAO report is that it describes some broad organizing factors that should be considered when studying long lines at the polls. These are factors that academic researchers may want to study more closely when designing future election studies.
1. There is currently no uniform definition of what constitutes a long wait – whether a wait is too long or not is a subjective judgment by polling place workers, and varies widely. Some election workers would call a 10-minute wait to vote “long” while others would only flag a wait as long if a voter’s time in line waiting to vote exceeded 30 minutes or more.
2. In addition to a lack of any objective standard for “long” versus “short” waits, there is no such thing as a comprehensive database of actual waiting times at any jurisdictional level.
3. Wait times are influenced by nine major factors, including:
- The extent of early voting opportunities in the jurisdiction;
- The number of polling places;
- The number of poll workers;
- The extent of poll worker training;
- The number of voting stations at each polling place;
- The format and organization of the poll book used to look up voters;
- The complexity and number of steps associated with determining voter eligibility;
- The length and format of the ballot; and
- The availability of voter information and voter education.
With so many potentially interdependent variables affecting waiting times, any broad comparative survey of voter waiting times would need to consider and address these factors.
It’s not that such a survey would be impossible (one could either conduct a broad range of field studies of different elections, or run multiple series of time-motion studies on mock elections, while incrementally altering one condition at a time, thereby establishing some general observations based on those time-motion studies), but every level of complexity would dramatically increase the expense of designing and executing such a survey.
So the real value of the GAO study lies not in what it uncovered, but in the care with which its authors considered its limitations. In identifying the weaknesses of their own study, the GAO staff have produced a roadmap for better academic surveys of the polling place experience in the future.
For those of you using web readers, here are the links. As you probably expect, many academic papers are posted as .pdfs, which may cause problems for web reader software; in that case the publishing institution often has an accessible version of the paper available on request.