Democracy in America | Divided we stand

America’s cultural divide isn’t growing

So what explains increasing partisanship?

By C.K. | WASHINGTON, DC

THE idea of two Americas is a trope of political commentary: a population divided in mutual incomprehension by income, race, religion or region—flyover country versus coastal elite. The idea that cultural fissures are growing is used to explain increasing political rancour and the rise of Donald Trump. But those explanations may need tempering. Two papers on cultural distance, published by the National Bureau of Economic Research in June, suggest the idea of cavernous and expanding cultural fissures is over-wrought.

The papers both use data from the General Social Survey, a long-running poll of Americans’ attitudes towards issues including free speech, same-sex relations and crime. They examine how closely respondents’ characteristics including where they live, what they earn, their education level and religion, are associated to particular attitudes and suggest that, at the level of individual attitudes, the relationship is weak. Marianne Bertrand and Emir Kamenica, the authors of “Coming Apart? Cultural Distances in the United States over Time”, calculate that, in 2016, voting in the last election was associated with a 65% chance respondents are in the richest quarter of American households rather than the poorest quartile, implying a 35% chance they are in the poorest quartile. This is the strongest relationship between an individual General Social Survey question and income that they found. More generally, Klaus Desmet and Romain Wacziarg, authors of “The Cultural Divide”, report that most variation in attitudes and norms occurs within groups not across groups. That men as a whole have different attitudes to women as a whole accounts for about 0.6% of total variation in norms and attitudes across the country, while between-group differences in levels of education accounts for 2.4% of total variation in answers to General Social Survey questions.

The authors of “Coming Apart” broadly echo that result looking at the link between respondent characteristics and media consumption, consumer behaviour and time use. Regarding television-viewing habits, for example, watching American football’s Super Bowl was the best predictor of income, but only with an accuracy seven percentage points higher than a 50-50 chance. Unsurprisingly, purchasing expensive consumer items did slightly better as an indicator of wealth: owning an iPhone in 2016 was associated with a 69% chance respondents were in the highest rather than the lowest income quartile.

The association between personal characteristics and attitudes, consumption and time does strengthen when answers are aggregated—so, for example, using all of the data they have on consumption choices together allows the authors to predict which households are high or low income and what the race of the respondent is with around 85% accuracy. Utilising all data on consumer behaviour allows them to predict gender with almost perfect accuracy; race and gender are also strongly associated with aggregated responses to General Social Survey questions.

But looking over time, while there have been considerable changes in both attitudes and consumption choices overall, there are few signs of a trend in the relationship between individual characteristics and those attitudes (whether aggregated or not). Mr Desmet and Mr Wacziarg report, for example, that in 1991 only 1% of respondents said that courts were dealing too harshly with criminals; in 2016 the proportion was 25%. Support for keeping marijuana illegal fell from 83% in 1990 to 39% in 2016. But Ms Bertrand and Mr Kamenica suggest there is no significant sign of an increasing gap in consumption, time use or most attitudes over the past two decades between rich and poor, more or less educated, urban and rural, male and female, white and non-white (with the partial exception of consumption patterns). And Mr Desmet and Mr Wacziarg repeat that finding for norms and attitudes, suggesting that attitude changes have usually diffused across different groups at broadly similar speeds.

Take answers to the question “are homosexual relations always wrong?” In 1990, 83% of respondents living in locations with less than 10,000 people answered yes compared with 78% in cities with a population above 1m. By 2016 45% of people answered yes compared with 35% of those in cities.

There are exceptions: attitudes towards law enforcement have diverged somewhat between income groups while attitudes towards civil liberties and government spending have converged. The relationship between attitudes and the characteristics of age, rural location, and region appears to be declining overall while it is strengthening for religion. And the greatest change in cultural cleavage reported in both papers regards political ideology. It has become increasingly straightforward to infer party support from social attitudes over time. Mr Desmet and Mr Wacziarg report the relationship between party identification and attitudes doubles in strength in the period after 2000.

This finding fits with widespread analysis that supporters of the two major political parties in America are becoming increasingly ideologically distinct. But both new papers provide evidence that growing partisanship isn’t driven by broader cultural trends. That suggests it is more probably the result of changes within the political system. Perhaps politicians have no one to blame for growing ideological rancour but themselves.

Discover more

The fifth Democratic primary debate showed that a cull is overdue

Thinning out the field of Democrats could focus minds on the way to Iowa’s caucuses

The election for Kentucky’s governor will be a referendum on Donald Trump

Matt Bevin, the unpopular incumbent, hopes to survive a formidable challenge by aligning himself with the president


A state court blocks North Carolina’s Republican-friendly map

The gerrymandering fix could help Democrats keep the House in 2020