Is the United States becoming more culturally divided across racial, gender, income, religious, geographic and political lines? New research from SMU and UCLA finds rising divisions since the early 2000s. But this is not happening across all identity cleavages, and the divisions have not reached historically unprecedented levels.

Is America in the midst of a growing cultural divide along lines of race, gender, income, geography, etc.? Recent years have witnessed increasing concerns about rising income inequality and the urban-rural divide. Social movements such as Black Lives Matter and MeToo have highlighted the racial and gender divides. And certain regions of the country no longer seem to recognize themselves as part of a shared culture. Symptoms of this cultural divergence include a fraying social fabric, affective polarization, and the breakdown of bipartisan politics. Yet, at the same time, we live in an era of unprecedented globalization and exchange, enabling the rapid diffusion of cultural values through society. Rather than driving us apart, these forces should push us toward greater cultural convergence.

Which dynamic is stronger? Are different groups in American society inexorably drifting apart? Or is there more agreement than we might think? To answer these questions, we use the General Social Survey (GSS) to quantify the evolution of cultural heterogeneity between 1972 and 2016. Every two or three years the GSS asks individuals across the United States a series of questions on a variety of social attitudes, including religious beliefs, gender and marriage, civil liberties, and politics. The GSS thus provides a good barometer of Americans’ changing culture, as measured by how their views, norms and attitudes have evolved from the early 1970s to the present day.

In our paper, “The Cultural Divide”, we start by analyzing the degree of overall cultural heterogeneity in the United States, measured by the probability that two random individuals answer a random question from the GSS differently. To fix ideas, take one question, say, support for marijuana legalization. Cultural heterogeneity on that specific issue would be maximized if half of the population were in favor and the other half were against, and it would be minimized if there were complete consensus, either in favor or against. When averaging this measure across all the different questions of the GSS, we get a measure of society’s overall cultural heterogeneity.

Looking at the evolution of cultural heterogeneity over time, we see an overall U-shaped pattern, declining between the early 1970s and the early 1990s, and rising since then (Figure 1). Overall cultural heterogeneity is quite high—the US is a very pluralistic country—but today’s level is by no means unprecedented. Of course, the US has always prided itself on being culturally diverse. In that sense, the rising cultural diversity in recent decades may be a strength rather than a symptom of dysfunction.

Figure 1: Cultural Heterogeneity, 1972–2016. Source: Desmet and Wacziarg (2018). This graph represents overall cultural heterogeneity in the US, as measured by the probability that two random respondents answer a random question of the GSS differently. The dots denote the actual numbers, the curve represents a nonlinear approximation, and the shaded area the 95 percent confidence interval.

But what about divisions across groups? Is the US becoming increasingly fractured along the urban-rural divide, across race and gender lines, or along income classes and education levels? To quantify these divides, we borrow the fixation index measure, or FST, from population genetics. In our context, FST captures the share of a society’s overall cultural heterogeneity that is due to between-group differences. To illustrate this concept, go back to the question of marijuana legalization and focus on the gender divide. If the shares in favor of legalization are identical for men and women, then knowing someone’s gender reveals no information about his or her attitude towards marijuana legalization. In that case, the share of cultural heterogeneity that is due to between-group differences, or FST, would be 0 percent. There would be no gender divide on this question. Now consider the other extreme: men and women have opposite views, with all men being on one side of the debate and all women on the other side. If so, knowing someone’s gender perfectly predicts her or his attitude towards marijuana legalization. In that case, FST would be 100 percent. The gender divide would be maximized.

Different geographic regions of the US are not drifting apart, we see a falling urban-rural divide, and individuals of different age groups and genders are becoming less divided. Some of these findings might come as a surprise in light of what many commentators perceive to be a generalized fraying of the social and cultural fabric in the United States.

When analyzing the GSS, the share of overall cultural heterogeneity due to between-group heterogeneity is very small, with values of FST on the order of 0.5-3 percent (Figure 2). For example, the gender divide accounts for 0.6 percent of overall cultural heterogeneity for the time period 1972–2016. Essentially, both men as a group and women as a group are culturally very heterogeneous, but the part of that heterogeneity that is associated with gender differences is a mere 0.6 percent. This implies that knowing someone’s gender, on average, says very little about her or his attitudes, values, or norms.

The same is true for other cleavages: average FST is 1.1 percent across racial groups, 2.4 percent across education classes, 1.4 percent across self-identified political party affiliation, and 1.1 percent across the urban-rural divide. A word of caution is in order here: small divides do not mean they are unimportant. In previous work, we found that countries with higher cultural divides along ethnic lines suffer from more dysfunctional political economies (more civil conflicts, lower provision of public goods, lower incomes)—in spite of the small overall magnitude of the ethnic divide in most places.

Figure 2: Cultural divisions across different groups, 1972–2016. Source: Desmet and Wacziarg (2018). This graph represents FST across different identity cleavages (average, age, education, ethnicity, family income, gender, political party self-identification, race, region, religion, city size, and work status), as measured by the share of overall cultural heterogeneity that is due to between-group differences. The dots denote the actual level of FST, the curve represents a nonlinear approximation, and the shaded area the 95 percent confidence interval.

While small, cultural divides may be increasing over time. On average we find growing divides since the early 2000s, though this only comes from some cleavages (Figure 2). The religious divide saw an earlier increase, starting in mid-1990s. The divisions across income classes, education levels, and ethnic and racial groups started trending upward in the early 2000s. The growing cultural divide is nowhere more pronounced than along political party lines: after gradually increasing in the late 1990s, it rose sharply in the most recent decade. In contrast, we find stable or falling heterogeneity across some other divides. Different geographic regions of the US are not drifting apart, we see a falling urban-rural divide, and individuals of different age groups and genders are becoming less divided. Some of these findings might come as a surprise in light of what many commentators perceive to be a generalized fraying of the social and cultural fabric in the United States.

How can we interpret these results? Cultural evolution is a complex phenomenon that responds to at least three distinct forces: the intergenerational transmission of values; the pressure to conform to the majority; and the tendency to follow new trends. How these forces affect the cultural divide between groups depends crucially on whether social influence occurs mostly within or across identity cleavages. For example, if people who self-identify with a particular political party increasingly shun contact with individuals of the other party, this would lead to an increasing cultural divide along party lines. This widening gap would be further facilitated by the possibility of people sorting into the political party that best fits their own values. It would also be enhanced by new communication technologies, such as social media, the Internet and cable news TV. Those allow the emergence of so-called echo chambers, where people increasingly communicate with their own group, hence driving different groups apart.

Our attention to the most visible dividing lines and to the most divisive questions might mask a growing consensus along less visible dividing lines and questions.

These forces do not operate in the same way for all identity cleavages. For example, it is unlikely that people of one gender can create effective echo chambers to avoid contact with the other gender. Men and women, after all, can’t really avoid each other in day-to-day circumstances. Moreover, people generally cannot sort into the gender that best fits their values. It is therefore perhaps not surprising that we haven’t observed an increasing gender divide. Hence, while new technologies may help the creation of echo chambers and facilitate greater sorting on values, this does not affect all identity cleavages in the same way. In fact, this is precisely what we find in the data: identity cleavages for which the potential for both sorting and echo chambers is greatest are precisely the ones where we observe growing divides.

This leaves us with an important question: if the cultural divide is small and if it has not increased dramatically across all identity cleavages, why do so many feel the United States is becoming increasingly fractured? One possibility is our outsized focus on the political divide. Other divisions are either rising less fast or even declining. Our attention to the most visible dividing lines and to the most divisive questions might mask a growing consensus along less visible dividing lines and questions. Another possibility is that our memory is short. Cultural divides may be on an upward trend in the last two decades, but the divisions are not unprecedented. Remember the civil rights movement and the Vietnam War…

Klaus Desmet is Professor of Economics at Southern Methodist University.

Romain Wacziarg is Professor of Economics and Hans Hufschmid Chair in Management at UCLA Anderson School of Management.

Disclaimer: The ProMarket blog is dedicated to discussing how competition tends to be subverted by special interests. The posts represent the opinions of their writers, not necessarily those of the University of Chicago, the Booth School of Business, or its faculty. For more information, please visit ProMarket Blog Policy.