The Son Also Rises: Surnames and the History of Social Mobility (The Princeton Economic History of the Western World) Hardcover – February 23, 2014

[major edit: I spent a couple days working through the math, and checking it with my own simulations, and have convinced myself that my earlier mathematical reservations were completely wrong. I've changed the review to reflect that]

The "Son Also Rises" was a fascinating read that seems likely to provoke controversy, but also to advance evidence-based discussions of equality and social mobility. Clark makes two major (somewhat separable) arguments in "Rises". First, that social mobility is much lower, and consistent across societies than anyone would have predicted. Second, that this low-mobility is biologically (in fact genetically) based. The first argument is better supported than the second. Clark's strong genetic conclusions seem rely on unassailable modelling (I tried) but some shakier genetic conclusions. They can't be dismissed entirely, however. Clark's evidence and reasoning is strong enough that the burden of proof is squarely on those who disagree with him. The implications the modern reader is left to draw are unsettling.

Clark's conclusions about the facts of mobility are astonishing. Typically, studies of mobility showed that intergenerational correlations (parent-offspring, typically father-son) in wealth are on the order of 0.4. This suggests ancestor-descendant correlations in wealth should be unobservable after about 4 generations. Across many cultures and times, and many different measures of status, Clark notes that identifiable elite or low-status groups regress to the mean at a rate between 0.75-0.85. This means that in fact differences in status persist for more than 10 generations.

Technically, Clark here models status as a single order Markov process, with three major components: time, [measurement] error, an underlying [latent] "social inertia" (my name) term. By this he emphasises we can model inheritance of social status from one's parents in exactly the same way we do height or eye color based on genetics. He notes that if we do so, we don't need to invoke any more complicated processes to explain the observed data (such as the status of extended family).

It turns out he's completely right about the models. I checked. If you model the inheritance process without the underlying latent term, you fail to match the data he's presented. If you model the process in the same way you would model additive genetic inheritance you get exactly the right answer. (I did this assuming a heritability of 0.4, parental-midpoint genotypes for the kids, renormalised mean and SD every generation, and a modelled range of assortative mating based on phenotype. I took beta and b vales from a number of the examples presented in the book.)

But here is where we begin to need to exercise caution. As a colleague is fond of quoting, "All models are wrong, but some are useful." We shouldn't let the simplicity of the model force us into a hasty overinterpretation of the underlying mechanisms. Clark jumps to a much less-cautious genetic interpretation of his results than almost any behavioural geneticist would (or at least should). Inheritance can be both genetic and epigenetic. Epigenetic is just a term that describes inheritance by any means but DNA (this isn't a magical thing: think language or religion). For instance, some primates and hyaenas inherit rank from their mothers. Fetal nutrition, maternal stress, early-life stress, and even languages and dialects, have effects on status and all have effects that are known to be transmitted across generations. Famously, maternal grooming in rats has profound (non-genetic) transgenerational effects on a range of personality measurements. It is extremely difficult to separate epigenetic and genetic effects when studying heritability.

Clark claims that because he can model inheritance of status as a first order Markov process, it actually is a first order Markov process based on transmitted characteristics inherent in the parents. Therefore, he claims, status is a deterministic product of a genetic "social competence" (his term). This is a strong claim. To his credit he discusses possible objections (such as inheritance of social networks). He also tries to quantify the non-genetic component of status in the best way possible, by examining adoption studies. Two studies, one on Korean adoptees in America, and another on adopted vs biological offspring in Sweden, seem to show a genetic heritabilty of income or education (here proxies for status) many times higher than conferred familial status.

The magnitude of these results is certainly far too high, as any number of factors (such as differences in the way parents and society treat adopted and biological children---see Hannah Williams) will bias these numbers. But at the very least we can find no reason to reject Clark's model, and I was persuaded that there is likely to be a higher effect of genetics on status metrics than I would ever have previously expected. Clearly more, and better, studies need to be conducted in this area.

At this point, any reasonable modern reader will be squirming. Raised under the spectre of the effects of early eugenics, racial determinism, and Manifest Destiny, we are rightly disturbed by attempts to reify social differences with biology. I'm reminded of the unproductive furor around "Sociobiology" and "The Bell Curve" (and Gould's error-filled attempt to rebut "The Bell Curve"). Clark spends much time demonstrating that there are no simplistic racial superiority claims to be taken from his data. His biologizing of hereditary class is inescapable, however. He tries to sugarcoat these interpretations with bland liberal prescriptions and platitudes, but they still rankle.

There have been notable failures in trying to increase social mobility (like Head Start in the US). But other recent studies have shown that good urban planning (access to public transport, and jobs, and good schools) can dramatically increase social mobility. Even if there is a genetic component to social status, Clark has almost certainly exaggerated it. Genetics certainly doesn't preclude other measures to increase social mobility. Then too, as Clark notes, inequality and mobility are different things, and we shouldn't confuse them.

In the end, "The Son Also Rises" was a thought provoking book, and one I'll read carefully again. I'd recommend it, as long as the reader doesn't accept any of the major conclusions without consideration.

Seven other reviews have preceded mine so I won't attempt to till plowed ground. Interestingly enough Clark begins his book by absolving the graduate students and paid research assistant who helped him in his research. Why do that? Here is the summary from book's end, "Most likely. . .the majority of status is actually genetically determined. You hit the jackpot in the great genetic casino or you go bust." In other words, the book is politically incorrect.

Some folks are familiar with Thomas Mann's Buddenbrooks, the novels that follow a family from shirtsleeves to shirtsleeves in three generations. Clark pretty much destroys this idea by following families with unique names in several distinct cultures, including communist China, Chile, England, Japan, and the United States. His null hypothesis is that over time families will regress to the mean in terms of status as measured by occupation and wealth. While his research shows a slight movement to the mean as generations come and go, for the most part those families who are on top stay there and those families not achieving stay at the bottom.

Clark comes up with some interesting tidbits. In Chile, for example, the leftist Allende government increased spending on education. The Pinochet government cut spending after taking power. Did the increase in education spending by the leftists have any effect on social mobility? Nope. None. What about the Cultural Revolution under Mao? Same story, the family names of those on top with the Nationalists stayed on top with the communists.

Now I don't believe Clark's hypothesis is all that new but his in depth research certainly has great merit. A book published in the early 1960's named The Geography of Intellect mentioned that families with the name Clark tended to have higher IQ's than the general population. The Clarks were medieval clerks. That should please the author and he notes the correlation on page 89..

Now for the negative. Clark has a conundrum. By going against the Standard Social Science Model (everybody is equal and all social traits are learned) Clark has to provide undeniable evidence or face ridicule. He does this with formulas and graphs. Unfortunately the book becomes something not for the general reader. Regression is easy enough to understand but when Clark gets in "first order Markov" I'm lost. And this may be my own ignorance and the fact that the last statistics class I sat in was over fifty years ago. Now Clark provides evidence for his hypothesis but in doing so leaves some folks (like me) in the dust.

Four stars for sure and a good book with an interesting conclusion. The difficult read in places costs it a five star rating.