Overwhelming majorities of both Republican and Democratic voters have retained their party affiliation over the past two years, a tumultuous period marked by a global pandemic, mass protests against racial injustice and a presidential impeachment.

Since 2018, comparably small shares of registered voters in both parties have changed parties. About one-in-ten voters (9%) who affiliated with the Republican Party or leaned Republican in September 2018 now identify as Democrats or lean Democratic. An identical share of voters (9%) who two years ago identified as Democrats or leaned Democratic now align with the GOP.

A new study, conducted on Pew Research Center’s nationally representative American Trends Panel, is based on interviews with the same set of 11,077 registered voters on five occasions over the past two years, from September 2018 to July 2020.

While individual-level change has not resulted in a significant net shift in the overall balance of party identification in the electorate, that is not the case within demographic groups. For instance, among white voters without a college degree, a larger share of 2018 Democrats now tilt to the GOP than vice versa (12% vs. 6%). The reverse is true among white college graduates: 4% of 2018 Democrats in this group now associate with the GOP, while 8% of 2018 Republicans now associate with the Democratic Party. Among nonwhite voters, 10% of 2018 Democrats have moved to the GOP, while roughly twice that share of 2018 Republicans (21%) have moved to the Democratic Party.

These patterns are similar to those seen in prior years and are consistent with the long-term shifts in the composition of Republican and Democratic voters. A recent Pew Research Center examination of trends in partisan identification since 1994, based on telephone surveys, illustrated that white college-educated voters have moved in a Democratic direction over time, while white voters without a college degree have become more Republican.

For the most part, the partisan leanings of voters do not change over short periods. The current study finds that nearly nine-in-ten voters who leaned toward or identified with a given party in 2018 currently report identifying with or leaning toward that same party – including about eight-in-ten who have consistently reported that affiliation or leaning in five surveys conducted over the past two years.

Overall, about one-in-five voters made at least some change in their partisanship over the past two years, but these changes have not favored either party. Among all registered voters, 4% identified with or leaned to the GOP in 2018 and now call themselves Democrats or Democratic leaners, while a similar share of all voters (5%) now call themselves Republicans or Republican leaners but had called themselves Democrats or Democratic leaners in the fall of 2018.

Party switchers are less politically engaged than consistent partisans

Voters who have not wavered in their party loyalty are more engaged in politics and demographically different than those who have changed their party affiliation one or more times over the past few years.

Majorities of both Republican (61%) and Democratic registered voters (63%) who consistently identified with or leaned toward their parties in five surveys between September 2018 and July 2020 say they follow what is going on in government and politics most of the time.

That compares with 36% of voters who currently align with the Republican Party but had shifted their party identification at some point in the past two years and 41% who are currently Democrats but had changed their party previously.

Consistent partisans are also substantially more likely than others to say they talk about politics at least a few times a week. About half of consistent Republicans (49%) and Democrats (52%) say they talk about politics every day or a few times a week. That compares with only about third of Republicans (29%) and Democrats (35%) who have shifted their party affiliation or leaning over the past two years.

Voters who have been inconsistent partisans also tend to be younger and have less formal education than those who consistently identify with or lean to the same party.

While changes in partisanship, particularly over the short term, are not common, the patterns of switching among the electorate in the past few years are largely continuations of long-term trends.

For example, white voters with no college degree have been moving steadily toward the Republican Party over the past 10 years, and in this analysis white voters with no college degree who were Republicans in 2018 are significantly more likely to have remained consistent Republicans throughout this time period than white Democrats with no college degree were to have remained Democrats (84% vs. 78%).

At the same time, white voters with a college degree have trended toward the Democratic Party. White voters with a college degree who were Republicans in 2018 are about twice as likely to have left their party than white voters with a college degree who were Democrats in 2018 (8% of white, college-educated voters who were Republicans in 2018 are now Democrats, versus 4% of white college-educated voters who were Democrats in 2018 but are now Republicans).

Nonwhite voters who were Republicans in 2018 were much less likely to stick with the Republican Party than were nonwhite voters who were Democrats.

This report is a collaborative effort based on the input and analysis of the following individuals:

Principal Researchers

Bradley Jones, Research Associate Jocelyn Kiley, Associate Director, Political Research Carroll Doherty, Director, Political Research

Research team

Andrew Daniller, Research Associate Hannah Hartig, Research Associate Amina Dunn, Research Analyst Hannah Gilberstadt, Research Assistant Ted Van Green, Research Assistant Vianney Gomez, Research Assistant

Communications and editorial

Nida Asheer, Communications Manager Calvin Jordan, Communications Associate David Kent, Senior Copy Editor

Graphic design and web publishing

Alissa Scheller, Information Graphics Designer Sara Atske, Associate Digital Producer

Methodology

Andrew Mercer, Senior Research Methodologist Arnold Lau, Research Analyst

We also would like to thank Scott Keeter and Claudia Deane for their valuable feedback on this project.

The American Trends Panel (ATP), created by Pew Research Center, is a nationally representative panel of randomly selected U.S. adults. Panelists participate via self-administered web surveys. Panelists who do not have internet access at home are provided with a tablet and wireless internet connection. The panel is being managed by Ipsos.

The data in this report comes from a longitudinal dataset comprised of the 13,570 ATP members who completed the panel’s 2018 annual profile survey. In particular, it focuses on the 11,077 of these panelists who indicated they were registered to vote.

Panelists who were newly recruited in 2018 and those who joined prior to 2018 were profiled separately. Of the 13,570 in total, 4,792 were preexisting members who joined the panel prior to 2018. The profile survey for preexisting panelists was conducted from Aug. 20 to Oct. 28, 2018, and had a response rate of 88%. The cumulative response rate accounting for nonresponse to the recruitment surveys was 2.5%.

The remaining 8,778 panelists were newly recruited in 2018. These panelists were both profiled and invited to join the panel in a survey fielded between Aug. 8 and Oct. 31, 2018. This survey employed an address-based sample (ABS). Respondents were sent a survey invitation via mail and instructed to go online to complete the survey. This survey had a response rate of 11.7%. Although a total of 9,396 individuals completed this survey, only the 8,778 who agreed to join the ATP are included in this analysis.

For these 13,570 panelists, their 2018 profile data was combined with their answers to questions asked on 10 subsequent surveys conducted up through August 2020. On average, panelists who completed the 2018 profile survey responded to seven of the 10 follow-up surveys. Only about 41% of the panelists in this dataset responded to all 10.

There is some evidence that those who are most likely to participate consistently in the panel are more interested and knowledgeable about politics than those who only periodically respond. Omitting the individuals who did not participate in every wave of the survey might overstate the amount of stability in individuals’ partisanship.

Rather than rely only on data from panelists with perfect survey participation, this analysis uses a technique known as multiple imputation which uses statistical models to fill in missing values for panelists who did not respond to a given wave. The particular missing data imputation algorithm we used is a method known as multiple imputation by chained equations, or MICE. The MICE algorithm is designed for situations where there are several variables with missing data that need to be imputed at the same time. MICE takes the full survey dataset and iteratively fills in missing data for each question using a statistical model that more closely approximates the overall distribution with each iteration. The process is repeated many times until the distribution of imputed data no longer changes. Although many kinds of statistical models can be used with MICE, this project used a machine learning method called classification and regression trees (CART).¹

The American Trends Panel survey methodology

The ATP was created in 2014, with the first cohort of panelists invited to join the panel at the end of a large, national, landline and cellphone random-digit-dial survey that was conducted in both English and Spanish. Two additional recruitments were conducted using the same method in 2015 and 2017, respectively. Across these three surveys, a total of 19,718 adults were invited to join the ATP, of whom 9,942 agreed to participate.

In August 2018, the ATP switched from telephone to address-based recruitment. Invitations were sent to a random, address-based sample (ABS) of households selected from the U.S. Postal Service’s Delivery Sequence File. In each household, the adult with the next birthday was asked to go online to complete a survey, at the end of which they were invited to join the panel. For a random half-sample of invitations, households without internet access were instructed to return a postcard. These households were contacted by telephone and sent a tablet if they agreed to participate. A total of 9,396 were invited to join the panel, and 8,778 agreed to join the panel and completed an initial profile survey. The same recruitment procedure was carried out on August 19, 2019, from which a total of 5,900 were invited to join the panel and 4,720 agreed to join the panel and completed an initial profile survey (although it should be noted that no data from panelists recruited in 2019 was used in this report). Of the 23,440 individuals who have ever joined the ATP, 15,410 remained active panelists and continued to receive survey invitations at the time the most recent survey in this analysis was conducted.

The U.S. Postal Service’s Delivery Sequence File has been estimated to cover as much as 98% of the population, although some studies suggest that the coverage could be in the low 90% range.² The American Trends Panel never uses breakout routers or chains that direct respondents to additional surveys.

Weighting

The ATP data was weighted in a multistep process that begins with a base weight incorporating the respondents’ original selection probability. The next step in the weighting uses an iterative technique that aligns the sample to population benchmarks on the dimensions listed in the accompanying table.

The use of multiple imputation means that it is not possible to reliably report a single margin of sampling error for the sample as a whole or for particular subgroups. The variability of multiply imputed estimates differs substantially from question to question, and questions asked on later waves with larger amounts of missing data can have much more variability than earlier waves.

Sampling errors and test of statistical significance take into account both the effect of weighting and the imputation of missing data. Interviews are conducted in both English and Spanish.

In addition to sampling error, one should bear in mind that question wording and practical difficulties in conducting surveys can introduce error or bias into the findings of opinion polls.

© Pew Research Center, 2020