Gregory Clark on Economic Mobility

Tuesday, February 4th, 2014

Gregory Clark explains his research on economic mobility:

What gave you the idea to look at surname data?

Initially I was interested just in extending conventional social mobility estimates into the distant past. Estimating social mobility is very data intensive. You need to link individual parents and children. There are thus no such estimates for any society before 1850.

Tracking surname status was a convenient shortcut. In most societies, all the people with a surname such as Goodhart descended from the earlier set of Goodharts. We do not know the individual linkages, but we can ask what is happening to their status as a group across generations.

And what did you find in collecting this surname data?

I found that you get radically slower estimated mobility rates for all societies when you switch to surnames. The conventional estimates of status correlation across generations are 0.2–0.6. With surname groupings it is always 0.7–0.8.

The effect is dramatic in some countries. Modern Sweden has some of the most rapid social mobility rates estimated in the world. Yet surnames in modern Sweden show status persistence exactly in this 0.7-0.8 range. This result was completely unexpected. Understanding why that is the case is a key puzzle the book tackles.

You don’t just look at income either. You look at educational attainment, what occupation they’re in, etc.

The book mainly concentrates on measures such as education, occupational status, wealth and longevity as indicators of status. Another surprising puzzle that emerged is that with surnames, the persistence of status was the same for all these measures.

We might expect wealth to persist in a different way, since it can be transmitted across generations in a different way than education. You do not need any talent to inherit wealth. This is another regularity the book tries to account for.

Social mobility seems impervious to government intervention:

It is clear that families are very powerful determinants of children’s outcomes. But what do parents transmit to their children? Is it mainly some type of culture? Or is it mainly genetics?

The data does not exist to provide any conclusive answer to this question. But even if this is cultural transmission, it looks in all respects just the same as biological inheritance. The book performs a series of tests to see if biological transmission can be ruled out as the important link, and the empirical patterns never rule this out.

For example, if biological transmission is the most important, then elite groups will never be the product of the adoption of particular cultural traits. Instead they will always represent a selection from the upper end of abilities of a parent population. Modern Jews will not be elite because of the social and religious mores of Judaism, but because they are a selection based on ability from a larger parent Jewish population.

For all such elite groups we observe, they do indeed turn out to be a selection from a larger population. Egyptian Copts are such a social elite, for example, but they represent the descendants of the Copts rich enough at the time of the Arabian conquest to be able to afford the head tax levied on all who did not convert to Islam.

A recent book, The Triple Package [by Amy Chua and Jed Rubenfeld], argues the extreme opposite of biology in explaining social status, with the claim that successful cultural groups in the U.S. have three key features leading to success, one being impulse control.

But what is remarkable is how disparate the culturally successful groups they identify are — Jews, Chinese, Indians, Mormons, Iranians, Lebanese, Nigerians, Cubans. And it is demonstrable that most of the successful groups identified here were elites selected from the parent populations as a combined result of politics at home and immigration policy in the U.S.


Accounts that emphasize cultural transmission all have a hard time explaining why successful groups, and successful families in general, all experience regression to the mean. There is nothing to stop a cultural trait being inherited unchanged. We see the preservation of such cultural forms as religious rituals unchanged over many generations.

Only biological inheritance has an inbuilt mechanism to explain observed regression to the mean. It also has predictions about when this regression to the mean will not be observed (complete endogamy). It further implies that the rate of regression to the mean will be the same at the top of the status distribution as at the bottom.

Leave a Reply