The Fault Lines of America

New Ipsos research finds that identifying as either a Democrat or Republican is a real and more meaningful source of difference than race or religion in the minds of many Americans today — particularly for Democrats.

This is evident by the results of a novel study in which Ipsos designed a poll that presents a nationally-representative sample of Americans with a variety of profiles of other people—two at a time. Respondents were then asked to choose the profile with whom they felt more in common. Each profile features different attributes related to race, religion, country of origin, education level, urban or rural residence, and political party—forcing respondents to decide between coexisting pieces of their identity. Because these attributes are shuffled randomly, over about 35,000 choices, researchers can use these polling data to calculate the importance of each attribute in driving how Americans identify with others.

The data are compelling and relevant to brands across all industries.

Download our detailed paper now or visit TIME for more highlights and interactive QUIZ.

PARTY FIRST?

New research finds that identifying as either a Democrat or Republican is a real and more meaningful source of difference than race or religion in the minds of many Americans today —particularly for Democrats. In this novel study, Ipsos designed a poll that presents a nationally-representative sample of Americans with a variety of profiles of other people — two at a time. Respondents were then asked to choose the profile with whom they felt more in common. Each profile features different attributes related to race, religion, country of origin, education level, urban or rural residence, and political party — forcing respondents to decide between coexisting pieces of their identity. Because these attributes are shuffled randomly, over about 35,000 choices, researchers can use these polling data to calculate the importance of each attribute in driving how Americans identify with others. Some of the most compelling data came from comparing self-identified Democrats and Republicans, and then rank ordering how much they identified (or didn’t) with the descriptions of various types of people. Holding all other factors constant, Democrats are 41% more likely to select a Democrat over a non-Democrat. Conversely, Democrats are 29% less likely to select a Republican than a Democrat and 12% less likely to select an Independent than a Democrat. In comparison, Republicans are 29% less likely to select a Democrat over a non-Democrat, 21% more likely to pick a Republican over a Democrat, and 8% more likely to pick an Independent over a Democrat. The “distance” that American partisans apply to profiles featuring the opposing party identities is the greatest observed between any two subgroups—the most significant fault line in America today.

In fact, political party has become Democrats’ primary source of identity, and it isn’t even close. The experiment also allowed researchers to measure what attributes drive profile choices, and political preferences drive 38 percent of Democrats’ decisions about with whom they identify. Religious backgrounds drive 27 percent of Democrats’ decisions, while racial backgrounds only drive 12 percent of these decisions. Similarly, Republicans feel they have less in common with Democrats than any other group in American society. But the primary drivers of Republicans’ affinity are religion and race. Religious similarities drive 35 percent of Republican identification, while race drives 25 percent and political preferences drive 22 percent of perceived commonality. Viewed from the other angle, Republicans are 57 percent more likely to identify with Christians and 43 percent more likely to identify with White people, but they are only 21 percent more likely to identify with fellow Republicans. In other words, Republicans’ ratings show them to be more driven by an ethno-religious sense of solidarity. The emphasis on partisanship, however, should not overshadow the influence of race and religion as enduring American fault lines. White Americans are 16 percent less likely to identify with people of Asian backgrounds, 18 percent less likely to identify with Blacks, and 19 percent less likely to identify with Hispanics. Christians are 17 percent less likely to identify with people of Jewish backgrounds, and 23 and 24 percent less likely to identify with atheists and Muslims, respectively.

RACE

The Ipsos poll also reveals complexities inside America’s lingering racial politics. The social distance between White Americans and the country’s racial and ethnic minorities certainly persists. About 36 percent of White people’s decisions about commonality are driven by racial attributes, an effect that is on par with the impact of religion for Whites. As noted above, Whites are 16 percent less likely to identify with people of Asian backgrounds, 18 percent less likely to identify with Blacks, and 19 percent less likely to identify with Hispanics. However, this sense of distance is generally not symmetric, or even mutual. Asians (23 percent), Blacks (23 percent) and Hispanics (18 percent) are far less influenced by racial attributes in determining a sense of commonality. Instead, their choices were more driven by religious background. Further, while White Americans are significantly less likely to select the profiles of racial minorities, only Blacks are equally less likely to select the profiles of White Americans (18 percent) as Whites are to select the profiles of Black Americans. Among both Hispanics and Asians, the extent to which they are more or less likely to identify with a White person (6 percent less for Hispanics, 3 percent more for Asians) is too small to be statistically significant. The same applies to Hispanics with both Blacks and Asians (4 percent less for each), Asians with Hispanics (6 percent less), and Blacks with Hispanics (1 percent less). However, Asians are 14 percent less likely to identify with Blacks than they are with Whites. These findings underscore a challenge for America’s racial minorities: in many cases, they do not identify with each other. Each racial and ethnic minority is not more likely to select people of other minority racial backgrounds. Much more significantly, they are less likely to select people of minority religious backgrounds.

AFFINITIES

While it is political differences that are largely driving partisan Americans apart, race and religion remain the demographic spaces that generate the greatest senses of commonality. Christians are 64 percent more likely to select a profile when it features a Christian background, and White people are 52 percent more likely to select a profile when it features a White background. Republicans are 57 percent more likely to identify with Christians and 43 percent more likely to identify with White Americans, as noted above. As a contrast to Republicans, Democrats are only 15 percent more likely to select a profile when it features a White racial background, and 36 percent more likely to select a Christian background. Importantly, America’s racial minorities do not exhibit such strong in-group racial affinities. Black Americans are 22 percent more likely to identify with other Blacks, Asian Americans are 17 percent more likely to identify with other Asians, and Hispanic Americans are 14 percent more likely to identify with other Hispanics. Indeed, for America’s racial minorities, religion drives a comparatively greater share of their decisions about with whom they identify (as noted above). While this suggests that America’s racial minorities feature alternative sources of solidarity and identity, it also points to the fact that these racial categories are ultimately broad categories of people with great ethnic, religious and cultural diversity. Notably, these polling data were collected in October and November 2020, near the 2020 U.S. general election — a period of heightened partisan polarization, which continued into 2021 after the Capitol insurrection and inauguration of President Biden. It is possible that the salience of political identity may change during the Biden administration. Ipsos continues to monitor developments in public attitudes.

Ipsos conducted an experiment to understand the attributes of individuals with whom Americans (and core subsets of Americans) feel they have the most in common. The experiment was conducted among 3,500 U.S. adults aged 18 and older who were interviewed online in English in an Ipsos survey conducted October 27– November 6, 2020. Each survey respondent participated in 10 comparison exercises. Each exercise presented respondents with the profiles of two different people identified as “Person A” and “Person B.” The randomly chosen profile of each person was defined along six demographic dimensions. When presented with each pair of separate, randomly chosen profiles, respondents were asked, “Do you think you have more in common with Person A or Person B?” For convenience, we refer to this perception as commonality. This type of experiment performed for this study is called a discrete choice conjoint experiment. A discrete choice conjoint experiment asks participants to make a series of choices between randomly assigned profiles of hypothetical entities (such as political candidates, products, policies, service offerings, and so forth). For this experiment, the profiles survey respondents were asked to compare were composite individuals assembled using attributes drawn from each of the following categories of demographic attributes:

• Level of education (has a high school diploma OR has a university degree);
• Nativity status (is American-born OR is an immigrant);
• Party identification (is a Republican, OR is a Democrat, OR is an Independent);
• Race/ethnicity (is White, OR is Black, OR is Latino/a, OR is Asian);
• Religion (is Christian, OR is Jewish, OR is Muslim, OR is an atheist); and
• Urbanicity (lives in an urban area, OR lives in the suburbs, OR lives in a rural area).

Each person could be shown any of 432 combinations of characteristics — all possible combinations except for Black-Jewish, Asian-Jewish, Latino-Muslim, and Muslim-Republican. Respondents were not given the option of indicating a “tie” between paired profiles.
The sample was randomly drawn from Ipsos’ online panel, partner online panel sources, and “river” sampling and does not rely on a population frame in the traditional sense. Rather, Ipsos used fixed targets specifically designed for this study in drawing the sample. After the sample was obtained, respondent characteristics were statistically calibrated to be representative of the U.S. population using standard survey adjustment procedures such as raking-ratio adjustments. The source of these population targets was the U.S. Census 2018 American Community Survey data. The sample drawn for this study reflects fixed sample targets on demographics. Post-hoc weights were generated based on the population distribution on gender, age, race/ethnicity, region, and education.
The data from this experiment were analyzed using methods for conjoint experiments that allow findings to be understood in two ways:

The first is via importance scores. Importance scores are expressed as percentages and represent how much each broad category of attributes contributes to respondent-perceived commonality. Importance scores thus provide a sense of which broad categories of attributes tend to be more (or less) responsible for shaping perceived commonality, and can be used to answer questions such as: “Overall, is race or ethnicity a stronger net driver of perceived commonality for Americans than religion?,” and “When Asian Americans think of commonality, is race overall more important than political affiliation or urbanicity?” The importance score of each attribute category (e.g., (race/ethnicity, religion, party ID, etc.) reflects its share of explained variance. Hence, the sum of the importance scores for all six attribute categories adds up to 100% within each group of respondents. Importance scores do not indicate the relative contribution of individual attribute levels within a category, however. For example, while importance scores might indicate that religion is the single largest factor driving respondent-perceived commonality, they do not provide information on the impact of specific religious or faith affiliations on perceived commonality.

• The second way conjoint findings can be expressed is with a metric called Average Marginal Component Effects (AMCEs). AMCEs are measures of individual profile attribute impacts on perceived commonality and indicate the change in probability that respondents will select a profile if a given feature of the profile is altered. AMCEs, which Ipsos refers to as “social affinity scores” help answer questions such as: “How much more (or less) likely are members of one subgroup of the population (e.g., Democrats or Whites) to respond that they feel they have something in common with another person, if that other person’s:
• Level of education was changed from high school graduate* to college graduate; or
• Nativity status was changed from U.S.-born* to immigrant; or
• Party identification was changed from Democrat* to Republican or Independent; or
• Race/ethnicity was changed from White* to Black, or Asian, or Hispanic; or
• Religion was changed from Christian* to Muslim, or Jewish, or atheist; or
• Urbanicity (i.e., type of community) was changed from urban* to suburban or rural.

* Reference level of profile attribute used in analytic models.
Note: in this experiment, respondents classified as Republicans and Democrats do not include self-described Independents who may “lean” Republican or Democrat. They are defined based on the following question: “Do you consider yourself a Republican, a Democrat, an independent or none of these?”