http://gao.gov/products/GAO-12-69. From the abstract:
For the 2010 general election, 35 states and the District provided voters at least one alternative to casting their ballot on Election Day through in-person early voting, no-excuse absentee voting, or voting by mail. Specifically, 33 states and the District provided in-person early voting, 29 states and the District provided no-excuse absentee voting, and 2 states provided voting by mail to all or most voters. Of the 9 states and the District where GAO conducted interviews, all but 2 states provided voters the option of in-person early voting in the 2010 general election, and 5 states and the District offered both early voting and no-excuse absentee voting. Implementation and characteristics of in-person early voting varied among the 7 states and, in some cases, among the jurisdictions within a state. For example, 5 states and the District required local jurisdictions to include at least one Saturday, and 2 states allowed for some jurisdiction discretion to include weekend days.
State and local election officials GAO interviewed identified challenges they would anticipate facing in planning and conducting Election Day activities on weekends—specifically, finding poll workers and polling places, and securing ballots and voting equipment—and expected cost increases. Officials in all 17 jurisdictions and the District we contacted said they expected the number of poll workers needed for a 2-day weekend election would increase. Further, officials in 13 jurisdictions said that some poll workers would be less willing to work on the weekend because of other priorities, such as family obligations or attending religious services. Officials in 14 of the 17 jurisdictions and the District expected that at least some of the polling places they used in past elections—such as churches—would not be available for a weekend election, and anticipated difficulty finding replacements. Officials in all 9 states, the District, and 15 of the 17 local jurisdictions said ensuring the security of ballots and voting equipment over the Saturday night of a weekend election would be both challenging and expensive. Officials in 5 of the 7 states and the District that conducted early voting and provided security over multiple days explained that the level of planning needed for overnight security for a weekend election would far surpass that of early voting due to the greater number and variety of Election Day polling places. For example, officials in one state said that for the 2010 general election, the state had fewer than 300 early voting sites—which were selected to ensure security—compared to more than 2,750 polling places on Election Day, which are generally selected based on availability and proximity to voters. In addition, officials in all 9 states, the District, and 15 of the 17 local jurisdictions said they expected overnight security costs to increase.
Weekend elections have not been studied, but studies of other voting alternatives determined that voter turnout is not strongly affected by them. Since nationwide federal elections have never been held on a weekend, it is difficult to draw valid conclusions about how moving federal elections to a weekend would affect voter turnout. GAO’s review of 24 studies found that, with the exception of vote by mail, each of the alternative voting methods was estimated to change turnout by no more than 4 percentage points. GAO’s analysis of early voter turnout data in Maryland found that 1.5 percent of voters we analyzed cast ballots on the weekend during the 2010 general election.