Andrew Leigh discusses the remarkable persistence of power and privilege:
In the case of earnings, economists’ best estimate of intergenerational elasticity went from 0.2 when they used a single year of earnings (as did the studies Gary Becker was relying on) to 0.4 when they used a few years of earnings (Gary Solon’s approach). Over the next decade, US researchers threw better and better data at the problem, and each time they found less and less mobility. Using more than a decade of earnings data, Bhashkar Mazumder estimated in 2005 that the intergenerational earnings elasticity for the United States was 0.6. That would put it higher than the father–son height elasticity. Among American sons, fathers had a larger impact on their earnings than on their stature.
Using similar techniques, researchers began estimating father–son earnings elasticities for other countries. As one survey showed, Scandinavian nations tended to be extremely mobile, with elasticities below 0.2. In Latin America, there was much less class-jumping, with elasticities over 0.5. Compared with other nations, the United States is extremely immobile, a fact that Barack Obama has thankfully switched from denying (“In no other country on earth is my story even possible”) to decrying (“It is harder today for a child born here in America to improve her station in life than it is for children in most of our wealthy allies”).
In 2006, while I was working as an economist at the Australian National University, I produced the first (and so far, only) estimates of the father–son earnings elasticity in Australia, putting the intergenerational elasticity at around 0.25. This means that a 10 per cent increase in a father’s earnings translates to a 2.5 per cent increase in his son’s earnings. My estimate implied that we are more socially mobile than the United States but not as mobile as Scandinavia. Looking back through the twentieth century, I found no evidence that we had become markedly more or less mobile.
So what does the surname approach add to our understanding of mobility? Simply put, there are two reasons for using surnames. The first is that we only have good data on earnings (from surveys or administrative records) for the relatively recent past. If we want to understand mobility in centuries gone by, surnames may be the best torch for seeing into an otherwise dark statistical corner.
The second, and more important, reason for using surnames is that they may help to take out some of the transitory fluctuations. Recall how we got more precise estimates of the intergenerational earnings elasticity when we used data that smoothed out the fluctuations in an individual’s earnings over a career? Call it the “odd year” problem. Now let’s think about a different problem: a family where the social status dips down for one generation, before reverting to the long-run average. You might call this the “black sheep” problem. By looking at surnames, we are able to look not just at single father–son pairs, but also at patterns for entire lineages.
So once we take out the odd years and black sheep, how easy is it to jump between classes? Several assumptions need to be made in order to estimate an intergenerational elasticity from surnames. But if we accept Gregory Clark’s methodology, his results imply a very static society. For Britain, the United States, India, Japan, Korea, China, Taiwan, Chile and even Sweden, he concludes that the intergenerational elasticity is between 0.7 and 0.9. This would mean that social status is at least as hereditable as height. It suggests that while the ruling class and the underclass are not permanent, they are extremely long-lasting. Erasing privilege takes not two or three generations, but ten to fifteen generations. If you cherish the notion of a society where anyone can make it, these results are disturbing.
(Hat tip to Alex Tabarrok.)