The bell keeps tolling

Gregory Clark is an economic historian at Stanford who has been doing fascinating work using family genealogies to explore the persistence of advantage, SES turnover stability, causes of inequality etc. His first two books were A Farewell to Alms, which explored the link between morbidity, cultural norms, socioeconomic outcomes and fact that the earliest instance of the industrial revolution manifested in England, an otherwise unpropitious location.

What he found was that, unlike most other countries, England was characterized by persistent downward mobility in which high mortality rates among the lower classes combined with high rates of upper class fertility led to downward mobility among the upper classes bringing with them upperclass social norms which were conducive to economic productivity.

Clark's research was based on a novel form of labor intensive historical research - tracking names and family trees through the records reaching deep into history. I have viewed the approach as ambitious and clever but also had reservations based on record incompleteness, name mismatch with biological inheritance, etc. New methods have to go through a period of stress testing to earn respect for their validity.

It has been twelve years now since Clark's first publication. While the criticisms in some quarters have been shrill, and it appears others share my concern about the methodology, I have not yet, in those twelve years, seen a solid evidentiary criticism of the methodology. If there is a flaw to it, no one has yet been able to identify that flaw.

A Farewell to Alms was well received as an innovative approach to research but its conclusions alarmed social justice theory historians.

The Son Also Rises was even more bold and controversial. Wikipedia has a reasonable summary of the findings and inferences.

The book follows relatively successful and unsuccessful extended families through the centuries in England, the United States, Sweden, India, China, Taiwan, Japan, Korea, and Chile. Clark uses an innovative technique of following families by seeing whether or not rare surnames kept turning up in university enrollment records, registers of physicians, lists of members of parliament, and other similar contemporary historical registers. Clark finds that the persistence of high or low social status is greater than would be expected from the generally accepted correlations of income between parents and children, conflicting with virtually all measures of social mobility previously developed by other researchers, which Clark claims are flawed. According to Clark, social mobility proceeds at a similar rate in all of the societies and in all the periods of history studied – with the exceptions of social groups with higher endogamy (tendency to marry within the same group), who experience higher social persistence and therefore even lower social mobility.

The book attempts to explain the difference between Clark's estimates of social mobility rates and estimates by other researchers by noting that the effects measured by other researchers are based on only a few generations, and argues that Clark's posited hidden variable of inherited "underlying social competence" is swamped by chance variations in status from generation to generation - variations which Clark says are smoothed out in his longer-term study. This can be analogised to looking at a graph to understand the trend in the market price of a stock – a graph of a stock price over a one-day period may show large "zigzag" price swings and no apparent order, whereas a longer-term stock price graph, particularly if smoothed, may reveal a long-term trend for the price of the stock to increase or decrease.

From his finding that ethnically homogeneous societies, such as Japan and Korea, had similar rates of social mobility to ethnically diverse societies, such as the United States, Clark infers that racism may not be a significant factor affecting social mobility. From his finding that families who had many children were able to pass down their high social status just as well as families who had few children, Clark infers that simple inheritance of wealth cannot explain the persistence of high social status. From the referenced studies on the lack of correlation between the intelligence and adult family income of adopted children and their adoptive parents, Clark infers that family environment cannot explain the transmittal of social status from one generation to the next.

Clark controversially hypothesises that the main reason for the unexpectedly high persistence of social status in families (or, put another way, the unexpectedly low degree of social mobility he finds) is that high-status people are more likely to have genes that are beneficial to them achieving high status, and are therefore more likely to pass such genes on to their children.

There is way more than enough meat on the Clark table to focus on a single dish. His findings have impact on history interpretation, economics, public policy, etc.

Beyond my concern about the validity and reliability of the approach (in turn based on concerns about the integrity of the data records as you go further back) the residual concern is the emerging specter of genetic determinism. It is not a necessary conclusion from his findings. In fact, he is very insistent that within generation variance and contextual variability swamp the mere influence of genes and that the persistent influence of genes only manifests over long stretches of time. It is a weather versus climate issue; weather today says nothing about climate over time.

But long cycle genetic determinism is just about as alarming as short cycle genetic determinism. Not so much for the existence of the fact (if it does exist) but for the interpretations and conclusions with low sensitivity to nuance might draw from it. It is not dissimilar to the AGW hysteria when utter conviction in narrow interpretations of long cycle climate change lead to dramatic, consequential and destructive beliefs about short cycle economic activities.

In the midst of all the implications of his research, Clark diligently continues to mine historical family records. Russell Warne has a summary of some preliminary findings from Clark, presented at a conference.

Next #ISIR2019 talk is from Gregory Clark comparing family networks and genetics in determining social outcomesn England from 1750- 2019. #ISIR2019 #psychology

— Russell T. Warne (@Russwarne) July 12, 2019

Some of the findings of Clark via Warne, if you don't want to read in twitter format:

Economic outcomes among relatives are correlated. The big question is whether that is causal (that relatives help out one another) or whether shared genes create these correlations.

The correlation is stronger the more closely related people are. (Sir Francis Galton showed this in 1869.)

There is a way to untangle this through a family database. Different relatives will share an equal level of relationship but not an equal level of social interaction. If correlations differ when genetic correlations are equal, then this is a social effect.

Approach

Dead vs. living relatives. Dead people can't influence their relatives

Grandfathers' & uncles' influences. These groups are equally related but social relationship not the same.

Geographically vs. close relatives.

Clark is finding so far:

Dead grandfathers and uncles are as influential as living grandfathers (even if the person died before the child was born).

Living grandmothers might have some social influence.

No difference between influence from uncles and from grandfathers (who are equally related to a child).

No difference in influence from great-grandfathers and cousins, even though almost all g-gfathers are dead.

No difference between the influence of relatives who lived nearby and relatives who lived further away.

From this Clark concludes (tentatively)

All of these results show that social transmission of SES, educational outcomes, etc., is mostly (or entirely) genetic.

Non-parents give a lot of information about a child's SES. But it's not causal. The info is a proxy for genetic effects. "The amount of information relatives supply is proportionate to the amount of genetic information they share with the child."

"It's hard to construct a social transmission model that mimic the results we see here. Genetic effects explain the SES correlation among family members much better. I tried to create a model that was non-genetic, and I couldn't."

Warne - The conventional folk wisdom that parental SES is an important causal influence is probably false. And that's not just true in a modern society, but it was true in 18th and 19th century England!

Very intriguing. Quite possibly true.

And it should not be rejected because it has a whiff of genetic determinism.

Moral decisions we make today stand in the context of that moral system and do not depend on long cycle causal relations. Long cycle genetic determinism may be real but it has no baring on how we treat one another today. In part because personal and contextual circumstance swamps short duration outcomes anyway.