Polls have faced growing challenges in recent years, including declining response rates, differential participation by Republicans and Democrats, overrepresentation of highly-engaged voters and difficulties obtaining reliable samples of young voters.
The poll, conducted by The Washington Post and Schar School of Policy and Government at George Mason University, sought to address these challenges and to reach voters who don’t usually participate in election polls, with the help of SSRS, the research firm that carried out the survey.
To encourage participation, SSRS mailed every sampled voter a one-dollar bill with their survey invitation and offered another $10 if they completed the survey, a common method to boost response rates in surveys. Over a six-week period, live interviewers made additional contact attempts, and sampled voters were sent emails, text messages and postcards when contact information was available. The overall response rate for the survey was 8.5 percent, compared with 1 to 2 percent response rates that are typical in telephone surveys with short field periods.
To mitigate nonresponse bias — which happens when people with different opinions respond to surveys at systematically different rates — SSRS tracked voter participation by partisanship based on party registration, past primary turnout and other predictors of party affiliation. The survey was also conducted in two phases: A smaller first phase helped gauge response patterns by groups, then a larger second phase where SSRS could adjust which types of voters it contacted to maximize the final sample’s representativeness.
To ensure less-engaged voters were adequately represented, SSRS oversampled voters who voted in one to three of the last four general elections and those aged 18 to 25, who were therefore too young to vote in the 2016 election. In total, the survey interviewed 1,645 voters who turned out in one to three of the last four general elections and 454 who were 18 to 25 years old. The final sample was weighted to ensure these groups were represented in proportion to their share of all registered voters.
On average, the poll’s response rate was 9.5 percent among Republicans, 8.6 percent among Democrats and 7.7 percent among voters unaffiliated with a party. After data collection was finished, the poll was weighted to accurately reflect the partisan makeup of each state according to the L2 file and support for Trump and Biden in the 2020 election, in addition to standard demographics like age, race and education. (Full technical details are available on the last page of the poll results page).
In addition to surveying a representative sample of voters in these states, a key goal of the Post-Schar School poll was to estimate how many voters are up for grabs in key states, dubbed “Deciders” in this poll.
There are two big questions about any voter in key states: whether they will turn out and who they will vote for. The Post and the Schar School identified voters uncertain of either.
First, voters were classified as sporadic if they voted in just one of the last two presidential elections, were registered to vote but too young to participate in the 2016 election or if they registered for the first time since the 2022 midterm elections. This information came from the L2 voter database, which relies on public records of voter behavior.
Second, voters were classified as uncommitted if they did not say they would “definitely” vote for Biden or Trump in the survey. A small slice of voters who reported voting for Biden or Trump in 2020 but a different party in 2016 were also classified as uncommitted.
The poll finds 61 percent of registered voters in these key states are either sporadic voters (33 percent) or uncommitted (44 percent), with 17 percent fitting both categories.
While the size of potentially persuadable voters might seem high, it’s echoed by other data. The Pew Research Center found that 25 percent of 2020 voters had not voted in the 2016 presidential election. And a June Monmouth University national poll found that 63 percent of registered voters said they would definitely vote for Trump or Biden, leaving nearly 4 in 10 who would “probably” vote for a candidate or were unlikely to support either.
The Post and Schar School plan to continue interviewing Decider voters in key states to see how opinions change through the campaign.
Emily Guskin contributed to this report.
