Gregory Clark, The son also rises. Surnames and the history of social mobility (Princeton: Princeton University Press, 2014. Pp. xii 364. 132 figs. 42 tabs. ISBN 9780691162546 Hbk. 19.95)

The main contribution of this book is the ‘law of social mobility’: underlying social status in families regresses only slowly toward the mean, with a persistence rate of 0.75, and this persistence rate is constant across all societies. The persistence rate can be interpreted as the intergenerational correlation of underlying social status. Underlying social status, or social competence, is never observed. It is related to observed measures of status, but with some error. Two aspects of the law of social mobility are controversial. First, the intergenerational correlation of social competence is much higher than the intergenerational correlation of observed measures of status (such as income or occupational status). Second, the correlation does not change over time or vary between countries, whereas correlations of observed status show some change over time and considerable differences between countries. Not included in the law of social mobility, but claimed by Clark, is that the transfer of social competence between generations is largely genetic. The law of social mobility is derived using a novel method. It starts with identifying surnames among an elite, preferably some centuries ago (for example, Swedish noble families in 1626). Names are selected that are rare nowadays.Then it is calculated to what extent people with these surnames are overrepresented in later elites (for example, in high status occupations) compared to people with common surnames. For example, people born in 1910 with the rare surnames of the old Swedish nobility were eight times as likely to be registered as a physician among those born in 1910 and 2.5 times as likely among those born in 1980. Under certain assumptions an intergenerational correlation of status is derived from two or more of such observations of relative representation. In this particular case the implied intergenerational correlation is 0.74. Clark presents an amazing amount of empirical analysis to derive and test the law of social mobility. He and his colleagues analysed data from Sweden, the US, medieval and modern England, India, China, Taiwan, Japan, South Korea, and Chile. In all cases old data on the surnames of an elite had to be found as well as data on the surnames of members of newer elites. Besides, data are required on how many people wear these names in the total population. The researchers showed an extraordinary creativity in finding and analysing these data. Another strong point of the book is that several chapters are devoted to empirical investigation of implications of the theory; for example, whether revolutions really do not change the intergenerational correlation, whether endogamy strengthens the intergenerational correlation, and whether the correlation is similar for families with few and many children. Did the book convince me that the law of social mobility is true? Yes and no. It is convincing that there is more transfer (of resources or status) between generations than a simple correlation of occupational status or income between father and son suggests. People with the same surname as elite ancestors are more likely to have elite status than expected based on the observed correlations of status between generations. That implies that these (great)grandchildren are likely to be successful, even if their father had relatively low observed status. This is not a new observation. Sibling analyses show that the ‘total’ family influence is about twice as large as the observed influence. Also, multi-generation studies sometimes show direct effects of grandparents and great-grandparents on the status of their (great)grandchildren. Is this correlation of unobserved social competence 0.75? Is it constant over time? And is it the same in all countries? Here I am less convinced. It remains unclear to what extent the correlation is driven by the fact that only a very small elite (1 to 5% of the population) is investigated. Would it be possible to estimate correlations for less elite parts of the population and what would be the result? It is not so surprising that the intergenerational bs_bs_banner