Appearance of knowledge should not be mistaken for actual knowledge

From Coming Apart? Cultural Distances in the United States over Time by Marianne Bertrand and Emir Kamenica.

From the Abstract.

We analyze temporal trends in cultural distance between groups in the US defined by income, education, gender, race, and political ideology. We measure cultural distance between two groups as the ability to infer an individual's group based on his or her (i) media consumption, (ii) consumer behavior, (iii) time use, or (iv) social attitudes. Gender difference in time use decreased between 1965 and 1995 and has remained constant since. Differences in social attitudes by political ideology and income have increased over the last four decades. Whites and non-whites have converged somewhat on attitudes but have diverged in consumer behavior. For all other demographic divisions and cultural dimensions, cultural distance has been broadly constant over time.

I want to believe the researchers. I am convinced that much of the media chatter of discord, divergence, polarization is 1) simple marketing to sell papers/eyeballs and/or 2) a reflection of the divergence of opinion between the chattering classes and the population, and/or 3) a reflection of intra-chattering class dysfunction rather than a reflection of the nation at large.

I encounter little in day-to-day life which mirrors the paranoia and hysteria reflected in the headlines of the increasingly marginalized chattering class.

All that said, Bertrand and Kamenica's research is behind a pay wall so I cannot comment on their paper directly. I do not know their sample size, randomization, controls, or methodology. While it sounds meaningful and intriguing, there is no substance that can be argued from the abstract.

Tyler Cowen has a run down of some of the key take-aways from his perspective.

1. From to 1995, the time use behavior of women and men converged a good deal, but not since then.

2. Differences in social attitudes by political ideology and income have increased since the 1970s. The rich and the poor have diverged the most in terms of their attitudes toward law enforcement.

3. Whites and non-whites “have converged somewhat on social attitudes but have diverged in consumer behavior.”

4. “Nevertheless, our headline result is that for all other demographic divisions and cultural dimensions, cultural distance has been broadly constant over time.” For instance, the media consumption gap between rich and poor has not been growing.

5. “The brand most predictive of top income in 1992 is Grey Poupon Dijon mustard. By 2004, the brand most indicative of the rich is Land O’Lakes butter, followed by Kikkoman soy sauce. By the end of the sample, ownership of Apple products (iPhone and iPad) tops the list. Knowing whether someone owns an iPad in 2016 allows us to guess correctly whether the person is in the top or bottom income quartile 69 percent of the time. Across all years in our data, no individual brand is as predictive of being high-income as owning an Apple iPhone in 2016.”

6. Voting and “trusting people” are among the “social attitudes” that best predict being rich.

7. Education is matched about as tightly to social attitudes now as it was in 1976.

8. “By 2016, watching Love It or List It and Property Brothers, both HGTV shows, were the most indicative of being educated.” [TC: yikes!]

9. Since the 1990s, there has been no divergence in the TV shows watched by liberals and conservatives. Note that in 2001, the three TV shows that best predicted ideology were The Academy Awards, Will and Grace, and Friends, all liberal. Nowadays it’s Fox shows, all conservative.

10. Liberals are more likely to drink alcohol, conservatives are more likely to go fishing.

11. Maybe this is the most important result: since 1976 there has not been much divergence between liberal and conservative attitudes toward civil liberties or law enforcement. The divergence on government spending is noticeable but not enormous (see p.39). the divergence on “Marriage, Sex, Abortion” is quite large. In another words, the true polarization is happening across gender issues, as I’ve argued numerous times in the past.

12. Here are related important results on the cultural divide. When will MSM articles catch up to the data?

I remain somewhat skeptical that there is a dramatic and energetic increase in polarization in the population at large. Within sub-populations? Intra, certainly, Inter, perhaps marginally. But averaged overall? Still not clear to me that there is.

And what a lot of this misses is nuance; two aspects in particular.

The first is definitional. Is anyone for increased police brutality? To a first order of approximation - no one. Fine. But was the Ferguson police stop and later shooting of Michael Brown an instance of police brutality (Black Lives Matter view) or an example of the tragic dangers police routinely encounter (majority of the population view). We may aspire to a common goal (zero police brutality) but mean very different things and interpret events very differently. In a survey like this, the nature and framing of the questions may lack sufficient specificity to reveal whether there is real difference in goals.

The second loss of nuance in studies such as this is in treating with averages rather than distributions, and in particular the appearance of the tails. This entails an Overton Window issue.

Is it acceptable to interpret the world in explicitly racial terms, i.e. is it acceptable to be a racist?

I suspect that were you to ask the population at large, the answer would be a reflexive no, but that there would be a normal distribution of views, no matter which groups you asked.

Click to enlarge.

The great bulk of the population (the blues and reds which overlap as green) basically agree on the proposition that it is inappropriate to view people based on their color. Even among those reds and blues who disagree (32% of the population), the great bulk are within a single standard deviation of one another - not in agreement but can comprehend the other's position.

The tails are where the action is, those who are more than two standard deviations from the norm. About 2.5% of the population are blue supremacists and 2.5% are red supremacists. Over decades, people's position may be rock-steady, the distribution doesn't change. There is no change in the large majority who consider judging people based on color as unacceptable. There is no polarization in reality, just the same old distribution of views.

What might change is the Overton Window. The absolute numbers and relative distribution might not change but the prevalence and acceptability might.

To take this from the abstract to the concrete, let's take race in America. Who are our racists? KKK - certainly. National Socialists, yes. White Supremacists, of course. OK. Now how many of them are there? That's where it get's really hard to measure. The Overton Window has closed on white racism. The number of those who electively self-identify as KKK, Nazi, or white supremacist are vanishingly small. Does that mean that there are hardly any white racists? Of course not. The definition is inadequate as is the measurement mechanism. I am happy to believe that there are more than the number who identify as KKK, etc. and I am also happy to believe that there might not be many more. I just don't know, and nor does anyone else. We have a nuance issue.

Likewise, how many Black Lives Matter, Nation of Islam, Ta-Nehisi Coates types are there out there who see everything through racially prejudiced eyes? If you listen to the media, there are a lot more. But are there really more black racists than white racists? The Overton Window is open for them but closed for white racists. There is political advantage and commercial advantage to these far left positions. Maybe the numbers at either end of the racism scale are the same but we hear from one end and not the other. We don't know.

So when Bertrand and Kamenica say that there is no polarization, is that true? Maybe. But I don't know that that is true. More likely, we still don't know how to ask the right questions. We lack definitional nuance, framing discipline, and fail to account for Overton Window skews which make things seem bigger or smaller than they actually might be.

We don't know.

UPDATE: Another report on this paper: No, America is not more divided than ever before by Howard R. Gold.