Two decades of partisanship in the Cooperative Election Study

Executive summary

• Over the past 20 years, the Democrats’ advantage in party affiliation among the general public has been steadily shrinking
• Over the same period, the share of the public that identifies or leans toward the Republican Party has stayed relatively stable
• The share who say they have no political affiliation has roughly doubled from 10% to 20% during this time
• Nearly every major demographic group has moved away from the Democratic Party during this period
  • Voters without a college degree (especially White women and Black men) have made the biggest shifts
  • Younger White men with a college degree are among the only groups to shift substantially toward the Democratic Party during this period
  • For many groups, the move away from the Democratic Party has not meant a positive affiliation with the Republican Party, but rather a greater share who have no partisan affinity
  • Over the past 10 years, people in the most rural parts of the country have made large shifts away from the Democratic Party

Introduction

Party identification is one of the most important factors in American politics. Due to a variety of structural factors, the Democratic and Republican Parties dominate the political landscape to the exclusion of any others, and the vast majority of Americans identify themselves with, or at least say they lean toward, one of the two major parties. Since 2006 (the first wave of the Cooperative Election Study), nearly eight-in-ten Americans say they identify or lean toward either the Democrats or Republicans politically. Less than one-in-four say they are politically unaffiliated.¹

Over the past 20 years, the United States has undergone dramatic political changes. Partisan control of the presidency has switched four times during this period. The same is true for both the House of Representatives and the U.S. Senate. During these tumultuous years, the partisan complexion of American political institutions has shifted from unified Republican control, to divided government with Republican control of the executive and Democratic control of the legislative branches, to united Democratic control, to Democratic control of the presidency and Republican control of the legislature and back again.

Over the past 20 years, Americans have witnessed large swings in domestic policy from the passage of the Affordable Care Act in 2010 to the Trump tax cuts in 2016 and then Biden’s infrastructure investments in 2022 followed by the second Trump administration’s sweeping crackdowns on immigrant communities beginning in 2025. During this time, America has undergone some other major changes from the financial collapse of 2008, to the legalization of gay marriage in 2015, the impeachment of Donald Trump in 2019, to a once-in-a-century global pandemic in 2020, to the January 6, 2021 insurrection and the overturning of Roe v. Wade in 2022 to name just a few.

This has also been an era of unprecedented government dysfunction. Since 1980, the federal government has had to reduce services due to budget impasses within Congress for a total of 137 days – 101 of those days (about three-quarters) during the last 20 years (the period covered by the CES). This total does not include the ongoing (as of this publication) shutdown of DHS.

Through all of this change, partisanship has shifted only slowly. Political scientists have long argued that partisanship is a stable orientation, and this assertion is borne out by the CES data. Over the past 20 years, the share of the general public that identifies or leans toward the Republican Party has stayed relatively constant, while the share who identify or lean toward the Democratic Party has declined somewhat from nearly half of the general public in the mid-aughts to about four-in-ten in 2025.

These shifts away from the Democratic Party are seen across nearly every demographic group, and they are especially pronounced among Americans without a college degree. The so-called “diploma divide” in partisanship has widened during this period – especially among White Americans.

The shift in the partisan balance among the general population impacts electoral politics in nuanced ways. Over the past 20 years, Democrats have become increasingly geographically concentrated in ways that disadvantage them structurally. On the other hand, the well-documented associations between turnout in elections and both education and gender tend to favor Democrats. Republicans’ relative fortunes have increased somewhat in terms of partisanship in the general public, and there is some evidence that the increasing share of those who say they do not identify with either party have been important to Trump’s electoral coalition (though they turn out to vote at substantially lower rates than those who have some political affiliation).

Data

The data for this report is drawn from the Cooperative Election Study (formerly the Cooperative Congressional Election Study). This study has been fielded every year since 2006, and in even numbered years, the CES typically includes interviews with 60,000 Americans (in odd numbered years, the sample size is substantially smaller – ranging from 10,000 to 25,000). The study design is modeled on the American National Election Study with pre-election interviews typically conducted in October and post-election interviews in the weeks immediately following the election (the latest round was completed in late 2025; see the methodological appendix for details of the sample size and field dates). The “Cooperative” in CES refers to the way in which the survey brings together teams of researchers from dozens of universities to field their own survey modules with a core set of common content that is shared across the teams. The common content includes core demographic items, vote intent (pre-election), vote choice (post-election), and a set of issue positions and policy preferences. Due to the evolving methodology of the study, this report focuses on data from 2007 onward.

Throughout this report when we talk about Republicans and Democrats, we refer to the set of Americans who either identify or lean toward each party. A great deal of political science research shows that so-called independent leaners look like partisan identifiers in their behaviors and attitudes.² In many cases, we are concerned primarily with the partisan balance of a group. In these cases, partisan balance is defined as the share who identify or lean toward the Democratic Party minus the share that identify or lean toward the Republican Party. The results shown here are substantively similar if we were to consider partisan identifiers only (e.g. including those who “lean” toward one of the major parties with independents).

To read more about the CES methodology, see the methodological appendix.

National trend

Over the past 20 years, the share of Americans who identify with or lean toward the Democratic Party has decreased at a steady pace. Two decades ago, about half of the public identified or leaned toward the Democratic Party; that share has decreased by about seven percentage points. The share of the public that completely rejects any partisan association has increased from just over 10 percent to about 20 percent today. The share who identifies with or leans toward the Republican Party has remained relatively stable between 35 and 40 percent during this period.

Shrinking Democratic advantage in the general public

Today the share of Americans who identify or lean toward the Republican Party is nearly the same as the share who identify or lean toward the Democratic Party. The Democratic advantage in party identification and leaning among the general public–which was a defining characteristic of U.S. politics for at least a century–is now almost entirely gone.

The chart below shows the party identification advantage among the general public going back to 1952 in the American National Election Study timeseries. In the 1950s and 1960s, the average Democratic advantage in the general public was around 20 percentage points, and today it has almost disappeared. The breakup of the so-called Solid South³ has been well-documented and accounts for a great deal of the decline in the Democratic advantage throughout the 20th century. The CES trend lines up closely with the ANES trend.

Major demographic groups

Generation

For the past two decades, politics in the United States has been divided by age. The oldest generations have been consistently more Republican on balance than the younger generations. With the exception of the youngest generation in the electorate – Gen Z, the oldest of whom have only been eligible to vote in federal elections since 2016 – all generational groupings have trended away from the Democratic Party.

These generational groups have taken somewhat different trajectories. For Millennials, the biggest part of their shift away from the Democratic Party came in the early part of the time series. Part of this change came as members of this generation came of age. In 2006, the oldest Millennials were only 25 and much of their generation was still in middle- or high school. It wouldn’t be until 2014 that the youngest Millennials aged into the electorate. Overall, partisan balance among Millennials mirrors the national trend. The modest trend toward the Republican Party among Millennials is less about an affirmative identification with the GOP and more due to increasing rates of political disaffection.

For Gen X and older generations, the shift away from the Democratic Party began toward the end of the Obama presidency and with the election of Donald Trump. Both the Baby Boomers and Gen X were on balance Democratic in 2007. Today, Gen X is about evenly divided between the Democrats and Republicans and Boomers are on balance Republican in their orientation. Gen Xers and Boomers, in contrast to their younger counterparts, have been increasingly likely to either identify with or lean toward the Republican Party in addition to decreasing their rates of affiliation with the Democrats.

Over the past 20 years, the Silent and Greatest generations (those born prior to 1945) have shrunk significantly. The youngest members of these generations were 81 in 2025, and these Americans now make up less than 5 percent of the adult public. As their numbers have declined, this group has remained the most significantly Republican-leaning generational grouping, and it has only become more so since the early 2010s.

Race/Ethnicity

Racial and ethnic divides in American politics have deep roots. For the history of the Cooperative Election Survey, White Americans have been Republican on balance. They have shifted substantially in the Republican direction over the past 20 years. This change among White Americans comes from a modest increasing likelihood of identifying with or leaning toward the Republican Party as well as a steady decline in Democratic affiliation.

Black Americans have long been among the most reliably Democratic groups in the population. That is still the case, but the overall balance has shifted from D+70 or so in the early part of the series to about D+60 today. For Black Americans the biggest driver of their movement away from the Democratic Party has been the much greater share who have been declining to identify with either party rather than a positive move toward the Republican Party.

Asian Americans have remained remarkably stable in their overall partisan balance over the past 20 years. There has been very little change in the share of Asian Americans who identify or lean toward either major party. During this period, the share of Asian Americans who have no political affiliation has consistently hovered around 25 percent.

Over the past 20 years, Hispanic Americans have shown a steady increase in the share who say they have no partisan affiliation. Over the last 5 years or so, the share that identifies or leans toward the Republican Party has ticked up in a way that was not true of the first 15 years of the CES trend.⁴

Education

Americans with higher levels of educational attainment are substantially more likely to identify or lean toward the Democratic Party compared to those with less formal education. Among those who have never attended college, 30 percent identify or lean toward the Democratic Party. Among those with a postgraduate degree, that figure rises to 54 percent.

Much of the change in partisan balance between different levels of educational attainment comes from changes in the relative shares of those who say they have no partisan affiliation and those who identify or lean toward the Democratic Party. The share who identify or lean toward the Republican Party between those with a 4-year college degree and those with a HS education or less varies within a much narrower range.

Over the past 20 years, Americans with no college experience are the broad educational grouping that has shown the most significant shifts. Twenty years ago, about 45% of Americans with a high school degree or less identified or leaned toward the Democratic Party. That share has declined to about 30% today. At the same time, the share who identify or lean toward the Republican Party has been increasing steadily since the mid-2010s.

Gender

In the early part of the CES timeseries, the gender gap in partisan identification was significantly wider than it is today. Women have steadily been moving away from the Democratic Party since the earliest part of the CES, but there was a period that roughly corresponds with the Obama presidency when men were moving in a Democratic direction overall. Around 2014, the trend among men began moving in the Republican direction and the trends for both men and women have been roughly parallel for the past 10 years or more.

The share of Americans who claim no partisan affiliation does not differ much by gender. The trends for both men and women very closely match the national average. Men’s rates of association with the Democratic Party held stable overall for the first part of the timeseries before starting to tick down around 2016. Among women, there has been a steeper trend away from affiliation with the Democratic Party. Twenty years ago, somewhat less than 35% of women identified or leaned toward the Republican Party (a figure that has not changed much in the intervening years).

Groupings by Race / Gender / Education

Political swings in the U.S. almost always are seen across groups roughly in parallel, and this is indeed what we see for nearly all demographic groupings defined by gender, race/ethnicity, and education. With one exception, every major demographic group has become more Republican on balance in its partisan composition since the mid-aughts. In many cases, these moves are not due to the group increasing its share that identifies or leans toward the Republican Party but by increasing shares saying they have no partisan affinity and fewer claiming affiliation with the Democratic Party. The only exception to this trend among these groups is White men with a college degree or postgraduate degree.

White men with a college degree by generation

The countertrend among White men with a college degree has been driven primarily by the changing composition of Millennial men as well as the entry of Gen Z college graduates in the last few years. In 2007, a substantial proportion of the Millennial generation had not yet entered adulthood. The oldest Millennials in 2007 were only 26 years old, and they made up about 5% of all White men with college degrees in the United States. Today, Millennials account for a plurality of White men with college degrees, and they are a reliably Democratic cohort. Older generations of White men with college degrees are more Republican than Democratic on balance.

The shifts among White Millennial men with college degrees has been both a shift toward greater Democratic affiliation as well as a shift away from Republican affiliation. Over this period the shares of this group that claims no partisan affinity has only increased modestly.

Among White men with college degrees, Gen X is the only to make positive movements toward greater Republican affiliation during this period. Prior to about 2020, this group of Gen X men was about evenly divided in its partisanship. After 2020, the share who identify or lean toward the Republican Party increased markedly from about four-in-ten in 2020 to nearly half today. The share identifying or leaning toward the Democratic Party dropped from about 40% in 2020 to about one-in-three today.

White women without a college degree by generation

White women without a college degree are among the groups that have made the most substantial moves away from the Democratic Party over the past 20 years. Prior to 2010, all generational groupings of White women without a college degree leaned toward the Democratic Party. Today, White Gen Z women without a college degree are about evenly split between the Republican and Democratic parties (though, as is true among all groups of younger Americans and those without college degrees, a substantial share of this group has no partisan leaning).

White Millennial women with no college degree have been steadily moving away from affiliation with the Democratic Party. Their affiliation with the Republican Party has been increasing in recent years but so has the share who have no partisan affiliation. In 2007, over half of this group identified or leaned toward the Democratic Party. Today that share has decreased to less than one-in-four.

White Gen X and Boomer women have made much more steady movement in terms of positive Republican affinity. In 2007, less than 40% of this group identified or leaned toward the Republican Party. Today more than half of women in this group claim Republican affinity. At the same time, the share who identify or lean toward the Democratic Party has decreased from about 45% in 2007 to less than 30% today.

Geography

Over the past 20 years, the urban/rural divide in the country has only deepened. While the move away from the Democratic Party is seen in the most developed⁵ areas of the country, the move has been most dramatic in the least developed places. The biggest shift toward the Republican Party during the first Trump administration occurred among the most rural parts of the country.

By state politics and development level

High development areas in blue states are more Democratic than high development areas in red states. The same is true for low development areas in blue states compared to red states. The differences aren’t dramatic, suggesting that the greatest part of the difference between blue and red states is accounted for by the different composition of those states. In blue states, about 30% of the population lives in high development areas compared to less than 15% in red states; in red states about 20% of the population lives in low development areas compared to about 10% in blue states.

Religion

As we saw with the demographic groupings, nearly every major religious grouping has become more Republican on balance over the past 20 years. As with the shifts in partisan balance among demographic groups, the greatest part of this change is the result of individuals becoming less likely to identify or lean toward the Democratic Party and having no political affiliation rather than an increasing share who identify or lean toward the Republican Party.

The two groups that have moved to become more Democratic in their partisan balance are Mormons and atheists (both groups that are Whiter and more likely to have a college degree than the average American). In the case of Mormons, the large shift in the Democratic direction was only enough to move them from the most Republican leaning religious grouping in the mid-2000s to the second most today. White evangelical Christians are now the most solidly Republican leaning major religious grouping in the country.

The other group that shifted in the Democratic direction was atheists. This group is now in competition for the most solidly Democratic group (the difference between atheists and Black protestants in terms of their partisan composition is difficult to precisely measure).

Methodology

The analysis in this report is based on the Cooperative Election Survey (CES) cumulative file (2007-2025). During this period, the CES conducted 682,534 interviews. In most years, the surveys were conducted online with a combination of YouGov’s panel as well as respondents provided by trusted partners. In even-numbered years, the CES is fielded in two stages. The preelection survey is fielded several weeks prior to the election. The postelection survey is fielded immediately after the election. The data in this report is based on party identification measures collected prior to the election (in even-numbered years) with the preelection weights applied. For odd-numbered years, there is only one round of data collection (after off-year elections are held).

In addition to sampling error, surveys are subject to other kinds of error. Care should be taken in the interpretation of the results.

YouGov panel methodology

The YouGov panel is a nonprobability panel. Panelists are recruited via a range of methods (primarily through different forms of online advertising), or they come to the website and sign up to be a part of the panel of their own initiative.

YouGov employs a variety of methods to ensure the authenticity and quality of panelists. When panelists first join, they are invited to an introductory survey where they must successfully complete a CAPTCHA test as well as other quality control measures. New panelists are also subjected to a probationary period where their qualification for the panel is determined. Once new panelists have completed the probation period, surveys regularly include quality measures that are reviewed by automated methods as well as project managers. Additionally, YouGov controls incentive disbursement, and suspicious patterns of incentive redemption can also disqualify panelists from participation in future surveys.

Methodology statements and codebooks for each year of data collection are available at the CES Dataverse.

For a comparison between some of the partisan estimates reported here and comparable estimates from a series of large-n probability based samples, see this document.

Weighting and sample matching

YouGov’s weighting and sample matching procedure is designed to produce samples that closely match the characteristics of a randomly selected sample. The procedure is described in detail by Rivers (2007), but in brief the process begins by drawing a sample from a modeled frame which constitutes the “target frame” (in this case, the American Community Survey microdata with modeled political variables appended). Panelists are then matched to cases in the target frame via a propensity score matching process. Any remaining differences between the sample and the target population are corrected with weighting.

Detailed descriptions of the weighting procedure for each wave of the CES can be found at the Dataverse. For most years, the weighting targets were drawn from the most recent available ACS 5-year estimates using interlocking frames for age, education, race and gender.

Supplementary data

The analysis in this report is primarily based on the publicly available Cooperative Election Study data files, but for the analysis by geography, ZIP-code tabulation area-level measures of development intensity were merged into the file. This data was originally collected by the US Geological Survey and includes measures of development intensity based on satellite imagery. The USGS data has been aggregated to the ZCTA by the National Neighborhood Data Archive.

To classify ZCTAs into development categories, we calculated the total share of the land area (the total area less the area that is classified as “open water” or “permanent ice and snow”) that was classified as developed. The different categories used in this report correspond to the following ranges:

When the data was reported by state politics, we used the following definitions (based on 2024 vote returns):

• Red states: AL, AK, AR, FL, ID, IN, IA, KS, KY, LA, MS, MO, MT, NE, ND, OH, OK, SC, SD, TN, TX, UT, WV, WY
• Swing states: AZ, GA, MI, NV, NC, PA, WI
• Blue states: CA, CO, CT, DE, DC, HI, IL, ME, MD, MA, MN, NH, NJ, NM, NY, OR, RI, VT, VA, WA

Funding

The Cooperative Election Survey is funded by teams of academic researchers from a range of universities and colleges. The study also receives funding from the National Science Foundation.

Question wordings

For the full questionnaires see the CES dataverse. The partisanship question is asked in two stages:

party_id Generally speaking, do you think of yourself as a . . . ?

1 Democrat
2 Republican
3 Independent
4 Other
5 Not sure

[if party_id = 1,2]
party_str Would you call yourself …

1 A strong [Democrat/Republican]
2 Not a very strong [Democrat/Republican]

[if party_id= 3,4,5]
party_ln Do you think of yourself as closer to the Democratic or Republican Party?

1 The Democratic Party
2 The Republican Party
3 Neither
4 Not sure⁶

Acknowledgements

This report benefited greatly from thoughtful comments and consultation with many individuals around YouGov. Carl Bialik provided thorough editorial guidance. David Montgomery contributed his expertise in visualization. Alissa Stollwerk, Marissa Shih and Reagan Bijou provided helpful comments and edits that greatly improved the final product. Any residual errors that escaped the watchful eyes of the many generous individuals who provided feedback are mine.

Image: Getty

¹ Surveys that ask about partisan identification can vary dramatically in their estimates of the share of the U.S. that is classified as politically unaffiliated (sometimes called “pure independents”). Divergent estimates can be traced to a variety of factors. One important factor is the mode of the interview. The original survey questions that were designed to measure partisanship were administered in in-person interviews. The partisanship measure is typically delivered in two parts. First respondents are asked if they think of themselves as Republicans, Democrats or Independents. For those who indicate one of the two major parties, they are then asked about the strength of their attachment. For those who indicate that they think of themselves as independent, they are asked if they lean more toward one party or the other. In a survey that includes an interviewer (face-to-face or over the phone), the interviewer can record so-called “voluntary” responses, and historically about 10 percent of the public would ‘volunteer’ that they identified with neither party. In the move to online surveys (that are most typically collected with a self-administered questionnaire), researchers must choose whether or not to display an explicit “neither” option in that follow-up question. Pew Research Center’s online surveys, for example, do not show this option and typically receive very low shares of any sample refusing to answer the question (which is coded by Pew to mean that the person is unaffiliated politically). The CES has always included the “neither” option and, consequently, has a larger estimate. For more on mode effects in the measurement of partisan identification see Dyck and Santucci (2026).

² A classic political science study on the topic is the Myth of the Independent Voter.

³ The Solid South refers to a period of Democratic dominance in the former Confederate states. This was the result of historical animus against the Republican Party dating back to the Civil War as well as legal and extra-legal intimidation of the Black population in the South that prevented them from participating in traditional forms of political participation. There were a range of factors that contributed to the demise of this system, but its legacy can be seen in party identification trends long after its formal end.

⁴ With the exception of 2016 when the CES included Spanish interviews, the study does not do interviews in Spanish or any Asian languages, so the results here reflect the English-speaking populations for these groups.

⁵ “Development” in this section of the report has been drawn from the National Land Cover Database collected by the US Geological Survey and aggregated up to the ZIP code tabulation area by the National Neighborhood Data Archive. The development scores here are based on processing satellite imagery to measure the intensity of development in a given area. Self-identified measures of community type (urban/suburban/rural) correlate strongly with these measures, but the community type question was not included consistently on the CES surveys.

⁶ “Not sure” was not included in the party_ln followup in 2007.