Eugenic thinking is not dead if you know where to look

There has recently been a lot of dicussion about the award of a prestigious fellowship at Cambridge University to Dr Noah Carl. I am not at all familiar with Dr Carl's work so not in a position to make any comment at all on whether it give comfort and succour to racists or eugenicists at the expense of scientific accuracy. Agreement broke out, at least on my Twitter time line, that critics have every opportunity to test rigorously his work as he uses openly available data. Which is an excellent argument for open data, but that is another story (see my earlier blog "In God we trust, the rest must bring data").

One direction in which the Twitter exchange led, however, resulted in me feelinig that a lot of sociologits have been leading somewhat sheltered live in terms of their familiarity with the influence of eugenic ideas in neighbouring disciplines. I work with epidemiologists. And the first thing one notices turning back to sociological literature is that no sociologist has used the Registrar General's Social Class schema (RG) for, oh, 30 years at least. Long ago, Gordon Marshall wrote of the "eugenic assumptions" that underlie this classification and I have also written a couple of blogs about this ("What is wrong with "SES""). But in social epidemiology, after a move in a different drection up to about 10 years ago , the use of measures and thinking that is based on the same ideas as the RG classification has come roaring back.

All this has coincided with the rise and rise of genetic epidemiology. Let me give a couple of quotes from recent papers. Here is one from a paper in a high ranked journal called "Intelligence in youth and all-cause-mortality: systematic review with meta-analysis" published in 2011:

"Twin studies to determine the extent to which intelligence shares genetic and environmental causes with health, education, and social class, in predicting mortality, will also help to inform this issue. With evidence of associations between cognitive performance and education showing substantial heritability, it is possible that these variables may share some genetic effects in predicting death."

Perhaps the recent paper that is being most discussed is called “Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals” published in Nature Genetics this year. Here is a quote from its Conclusions

For  research  in  social  science  and  epidemiology,  the polygenic scores that we construct—which explain 11–13% and 7–10% of the variance in educational attainment and cognitive performance, respectively—will prove useful across at least three types of applications. First,  by  examining  associations  between  the  scores  and  high-quality  measures  of  endophenotypes,  researchers  may  be  able  to  disentangle the mechanisms by which genetic factors affect educational attainment and cognitive phenotypes.”

The authors of this study were eager to avoid misinterpretation of their findings. On the website of the Social Science Genetic Consortium https://www.thessgac.org/ (stongly recommended) are a large number of "FAQs" that had been raised about the study, with associated sensible comments for example:

"it is important to keep in mind that the score fails to predict the vast majority (89%) of variation in years of education across individuals. Many of those with low polygenic scores go on to achieve high levels of education, and a large proportion of those with high polygenic scores do not complete college.

​

Thus, an important message of this paper and our earlier papers is that DNA does not “determine” an individual’s level of education, for multiple reasons: First, it is estimated that, at least in the environments in which we have been measuring it, the additive effects of common genetic variants will only ever predict about 20% of the variance in educational attainment across individuals. Second, today’s polygenic score is only able to predict a little more than half of that 20% (11 percentage points). 

Thus it seemed that “social genetics” (a term I hate) are too fascinating for many researchers to avoid, is being carried out by responsible people. And indeed, the GWAS studies have revealed important differences to the twin studies that have been carried on for many years .

But I guess it was never going to stop there. And in the last week or so a new paper has appeared: Genetic consequences of social stratification in Great Britain by Abdel Abdellaoui and colleagues, available at http://dx.doi.org/10.1101/457515 . In their words:

We show that the geographic clustering of genome-wide trait-associated alleles is related to recent geographic movement of people and that the resulting regional genetic patterns are associated with regional socioeconomic and cultural outcomes...... The strongest clustering was observed for Educational Attainment (EA). Among the rest of the geographically clustered traits are body dimensions, personality dimensions, and physical and mental health traits. ... Our results show that people with a genetic predisposition for higher cognitive abilities are leaving these (deprived) regions, likely attracted by better educational or occupational opportunities in other regions. In fact, the people who were born in coal mining areas and migrated to better neighbourhoods have higher average EA polygenic scores than people born outside of these regions. The regional clustering of cognitive abilities that follows may further affect the economic development of neighbourhoods.

The authors do go on to speculate that social policies might need to be devised to lower the tendency of geographical mobility to increase differences in deprivation between areas. If those with higher genetic scores for educational attainment did not have to travel to more prosperous areas in order to get better jobs and increase their income, the argument goes, this might slow down the tendency of health inequalities between areas to widen over time.

A clearly written cautionary commentary has also appeared in Nature at https://www.nature.com/articles/d41586-019-03171-6

But it is hard to escape the impression of an element of dog-whistling in this literature. At the very least, sociologists need to be on their toes in respect of it.

An interesting papear has recenty appeared in the Journal of Epidemology and Community Health by Murray et al, entitled  Inequalities in time from stopping paid work to death: findings from the ONS Longitudinal Study, 2001 to 2011

open access version at  https://discovery.ucl.ac.uk/id/eprint/10082463/1/Murray%2020190821_JECH_R2_manuscript_CLEAN%20copy.pdf

At first sight it might appear puzzling that the bottom line of the paper is that people who had worked in less advantaged, lower skilled & lower status jobs (defined by the Registrar General's class schema, which is conceptually based on skills and status) actually lived longer after leaving the labour force than those whose jobs were of higher skill and status. To cite the paper itself:

"LS members who had worked in lower social classes lived a greater number of years after they stopped work, with more time per decreasing social class."

How to make sense of this? Actually, a similar issue was also a big puzzle for the authors of the first book that decribed the contribution of the LS to the study of health inequality. To the amazement of almost everyone involved, the earliest analysis showed no social class gradient in health, or even a hint of a reverse gradient. Did this mean that all the previous research on social class differences in health including those in the Black Report, had been biased by either artefact or selection?

And what does this have to do with Murray et al.'s paper?

To cut a long story short, it turned out that high status, high skilled jobs also benefit from less arduous working conditions. Just think about the difference between a building worker and a university lecturer. Lets say both of them start to get angina pectoris. The lecturer, first of all, will probably not be affected by narowing of her coronary arteries as soon because she does not need to do much heavy lifting, digging, or other activities that equire increaed cardiac output. And when she does start to notice, it is not too hard to keep her job if she also needs to do less of these kinds of things (moving furniture, lifting large piles of books or files...).  By the time she even notices that her cardiovascular function is declining, let alone has to give up her job, her disease is far more advanced. Contrast to the building worker whose daily work activities will cause pain earlier in the natural history of the disease, and incapacitate her (or him) more quickly. Let's say the disease process is exactly the same in them both. The building worker will not die sooner than the lecturer but will have had to give up work earlier.

In other words, it is not that the people in the less skilled, lower status jobs died later, but that they had to give up work earlier in the disease process.

This phenomenon eventually resulted in a convention when analyzing social class differences in health in longitudinal follow up such as that in the LS, to allow several years to pass to "allow selection to wear off" before comparing the death rates in the different classes. After the 1st 5 years, the sicker people in the more advantaged social positions died and the social class inequalities in mortality risk re-appeared. All subsequent papers using the LS have used this adjustment, but often explain it in a rather abbreviated way (let me know if anyone wants references). With this adjustment, the class differences in mortality look about the same as in the previous, cross sectional ("unlinked") studies from 1951 onwards.

These comments are no criticism of the JECH paper, whose objectives were to show that extending the pension age is less fair to people in more advantaged social classes. But it does show the dangers in assuming that the nominal definitions of social position need to be taken with care. The social forces at work here are not derived from status or skills but from working conditions, which, while collinear to each other, do not measure the same thing.