The U.S. faces increasing levels of political polarization. Bipartisan disagreements over moral, cultural, and socioeconomic issues have turned into social conflict, political gridlock, and personal animosity. How did we get here? The dominant narrative has argued that identity cleavages have caused increasing splits in values. Yet Klaus Desmet — Research Fellow at CEPR, Research Associate at NBER, and the Altshuler Professor of Cities, Regions, and Globalization at Southern Methodist University — painted a different picture in a CDDRL seminar series talk.

Leveraging data from seven waves of the World Values Survey (WVS), Desmet develops a new methodology to study the evolution of these social divisions. For Desmet, the problem with an identity-based measure of polarization is that these demographic traits do not necessarily align with people’s values. If we care about splits in values, we need to create social partitions based on values, not identity traits.

Desmet asks how individuals would optimally form groups based on “homophily in values.” If people associate with others who hold similar views, they will sort themselves into groups based on these preferences. Individuals leave groups with fewer shared values and join groups with more shared values. Eventually, this self-sorting process reaches what Desmet calls a “Global Values Identification Equilibrium (VIE),” where within-group value heterogeneity is lowest and between-group value heterogeneity highest. Importantly, Desmet can then compare the split in values between these underlying clusters (latent polarization) to the split in values between Democrats and Republicans (partisan polarization).

What does the model find? While many have argued that America has witnessed a shift from disagreements on redistribution to disagreements on culture, Desmet’s findings indicate otherwise. Since at least the early 1980s, the latent values-based clusters have been divided mostly along moral and religious values, and the level of disagreement has been remarkably stable. There is no evidence of latent polarization increasing over time, and the underlying conditions for the culture wars have been present for a long time.

Partisan polarization, on the other hand, is a more recent development. Using the same model, Desmet shows that in 1981, the average value positions of Democrats and Republicans were almost indistinguishable, nowhere close to aligned with the endogenous clusters. By 2017, however, the average positions of Democrats and Republicans have diverged, aligning with the positions of the values-based clusters. These findings suggest that there has been rising partisan polarization in spite of stable latent polarization.

How might we explain this sequence of events? Desmet suggests that increasing partisan polarization may be a consequence of politicians discovering which values are particularly salient for political mobilization. Instead of politicians engendering value splits in society, partisanship has become more representative of people’s values. The American public has long had the conditions to be divided — they just needed parties to catch on.