Saturday, December 14, 2019

Self-refuting scholarship

[NB: this piece has been updated to incorporate links to further relevant scholarship.]

One of the signs of the increasing intellectual conformity of the academy, particularly the social science and (even more) the humanities, is the rise of scholarship that treats voting the "wrong" way as a pathology, to be explained pathologically.

In the US, analysis of voting Republican is increasingly treated in this way. Even more since Donald Trump won the Republican nomination and then the Electoral College and so the US Presidency. Brexit has received similar treatment.

A recent paper, Growing sense of social status threat and concomitant deaths of despair among whites, by Arjumand Siddiqi, Odmaa Sod-Erdene, Darrick Hamilton, Tressie McMillan Cottom, William Darity Jr. is a case in point.

The paper starts off well, carefully examining the data and assessing various explanations for the rise in mortality (and so falling life expectancy) about "white" Americans. Then it gets to its own explanation and the intellectual quality rapidly goes downhill.

The abstract sets out their findings and conclusion:
Rising white mortality is not restricted to the lowest education bracket and is occurring deeper into the educational distribution. Neither short-term nor long-term economic factors can themselves account for rising white mortality, because parallel trends (and more adverse levels) of these factors were being experienced by blacks, whose mortality rates are not rising. Instead, perceptions – misperceptions – of whites that their social status is being threatened by their declining economic circumstances seems best able to reconcile the observed population health patterns.
Conclusion: Rising white mortality in the United States is not explained by traditional social and economic population health indicators, but instead by a perceived decline in relative group status on the part of whites – despite no actual loss in relative group position.
To put the finding at its most blunt: they are dying because they're stupid. Or, to put it in a little more elevated way: they are dying because they have a false consciousness of their social position. 

This is a very large claim. So, what have the authors done to test the proposition about (lack of) falling social status? Nothing substantive. Which is remarkable, because a recent study of why people voted for Donald Trump finds no difficulty in identifying a range of status threats (pdf) that motivated voters. While another study founds that falling subjective social status has considerable explanatory power (pdf) for populism rising vote share in Europe.

Let's us put the question in another way: is there any other group, particularly any other racially defined group, that the authors would explain seriously adverse outcomes by arguing they had brought it on themselves because they are too cognitively incompetent to accurately assess their social position?

If they answer is no, as surely it is, then what does that say about the social status of "white" Americans? After all, such things as having various academic fuss over "it's OK to be white" posters are also, in their own way, signs of loss of status.

These authors have the gall to sneer at the reasons people are dying when the way they have "conducted" their own research shrieks sneering condescension towards the people they are studying because those people lack the social status to protect them from said sneering condescension.

This is self-refuting scholarship.  But self-refuting in a revealing way. It is scholarship unable to see itself critically. 

It is revealing that the paper's Fig. 4, which is adopted from a WHO document (pdf), puts in 'race' where the WHO document has 'ethnicity (racism)'. American race talk is typically a clumsy and unfortunate way of talking about ethnicity (i.e. ancestry and culture). For example, disaggregating PISA (Programme for International Student Assessment) results for the US by "race" is informative, because it disaggregates by ancestral commitment to formal education (with exactly the ranking results you would then expect). 

The only thing race talk is really good for is stigmatisation, as it either strips away a whole lot of relevant factors or reduces them to skin colour. 

Homo sapiens are a group-living, pair bonding species. We are typically good at social cues, especially social cues about status. We evolved to be, because that is how you acquired a mate and successfully reproduced. 

To claim that an entire group of people (who you have defined racially) are incompetent at such a basic human skill is a very big claim. It runs against evolutionary biology, anthropology, psychology and sociology. Such a claim requires a lot of supporting evidence. 

The first evidence offered is that "whites" still do better by various health and other indicators, such as median household income, than other groups whose mortality has not worsened. In fact, has improved. Yes, but that is perfectly compatible with a loss of status.

The authors agree that perceived status threat is the most powerful explanation for the mortality results. They then cite various studies of "white" Americans having inaccurate view about the situation of African-Americans ("blacks"). As they say:
These findings suggest whites perceive that blacks are economically catching up to them, even though this is not the case.
But status is not merely a matter of income. The authors note that "whites" show declining happiness and their sense of class status has generally fallen, particularly among the less educated. This is revealing, since class is a rather different metric than race. The authors state:
In other words, absolute declines in economic status of whites may produce a hyper-vigilance of sorts. This explanation would suggest that the short-term and long-term economic circumstance hypotheses forwarded by Case and Deaton (2017) are in some ways integral to the explanation of white status threat, rather than true competing explanations. However, we do not have sufficient evidence to test the veracity of this claim.
But on they march:
Put differently, given our modeling strategy, if county-level changes in the share of Republican voters is associated with changes in county-level white mortality, then it is highly likely that this association is indicative of a link between rising white perceptions of racial threat and rising white mortality, rather than traditional economic and social population health indicators.
So, if we treat voting Republican as a racially-based pathology, our modelling will demonstrate our case, despite the fact we have made no serious attempt to determine:
(1) How important "racial" identity is to the target group.
(2) What components of status matter to the target group.
(3) How good humans tend to be at rating social status, particularly their own.
(4) What it takes to misperceive status.

Scholarship often uses racial resentment as measured by answers to a particular set of survey questions as a measure of racism. Yet those questions do not correlate with any tendency to discriminate against African-Americans. Moreover, conservatives do not give different answers on such questions between ethnoracial groups while liberals do, but only more sympathetically to African-Americans. Finally, warmness to your own group does not correlate with negative feelings to other groups. This area is somewhat riddled with scholarship that does not show what it purports to show.

Donald Trump won the Electoral College because a whole group of counties that had twice voted for Barack Obama voted for Donald Trump. Obama was much more electorally successful than Trump, winning in 2008 (53% to 46%) and 2012 (51% to 47%) by absolute majorities of the popular vote while Trump famously lost the popular vote (46% to 48%). 

Given the nature of the Party coalitions, Obama did less well among "white" votes (43% in 2008, 39% in 2012). Trump got 54% of the "white" vote in 2016, but since 1976 among Republican nominees, only Gerald Ford (1976), Bush Snr (1992) and Bob Dole (1996), got a lower share of the "white" vote than Trump. So, if rallying "white" voters was Trump's thing, he did a comparatively bad job of it. Every other Republican elected President since 1980 did a better job.

But the author's analysis is concentrating on particular counties. There they find that:
Of all covariates, change in share of Republican voters (r = 0.24) and college degree attainment (r = −0.24) were the most strongly correlated with change in white mortality, suggesting that counties that became more Republican and that did not experience much change in college attainment also had increased rates of white deaths.
Folk in those counties changed votes because they perceived their situation deteriorating. At no stage have the authors demonstrated that white/black relations or standing is crucial to the target population. Yet the authors find that:
Rather, we hypothesize that the anxiety of whites is coming from a perception – a misperception – that their dominant status in society is being threatened, which is manifesting in multiple forms of psychological and physiological stress. While stratification economics suggests that this misperception may actually be quite functional for preserving relative group status, it may have health consequences. Indeed, the empirical test we provide of our hypothesis suggests this to be the case.
Yet, there is lots of public comment on how "whites" will become a minority in the US in the next few decades. The US as majority-minority country. Does that not constitute a threat to "white" Americans "dominant status in society"?

Let's put it another way: if you already think you are facing declining economic circumstances and various levels of cultural dislocation, with limited ability to influence events, how does no longer being part of the majority group sound? Not enticing. A loss of status. A basis for rational concern. Yet, migration is not mentioned as an issue once, except to bring up Trump's infamous comments about migration. Comments which, we might note, were not directed to African-Americans. 

The authors continue:
To be sure, this is a startling finding. The social status threat mechanism clearly has emerged as a way to explain the election of a presidential candidate who espoused highly racist views (Green, 2017), but we are now suggesting that this mechanism also explains the highly unusual phenomenon of worsening white mortality – and worsening white health more generally. Moreover, we are suggesting that the perception of racial threat among whites is occurring in the absence of substantive evidence of a decline in their relative social status, since both whites and blacks are experiencing parallel economic declines (Badger, 2017).
It is a startling finding that is derived only by carefully not enquiring about things that might get in the way and not noticing that Trump's infamous comments were not directed about African-Americans but about Mexican migrants. Nor are the areas of the study notable for their African-American population.

What they are generally notable for is rapid ethnic change, particularly increased Hispanic population. In Europe, experiencing, or being adjacent to, a locality experiencing rapid ethnic change is a strong predictor of votes for national populists. Indeed, the rate of change is a much more powerful predictor than the level of ethnic diversity.



Besides, think how many progressive narratives would be upset if the study concluded that the sense of loss of status was rational. The study reaches the progressively acceptable conclusion by progressively acceptable scholarly tropes.

This is bubble scholarship: we are going to characterise things racially, we are going reduce complexity to (literally) "black" and "white", we are going to treat voting Republican pathologically and we are going to say "white" people have rising mortality, and so declining life expectancy, because they are cognitively incompetent. A finding we would, almost certainly, not dare to make about any other racially-defined group. And then we have the gall to say that any sense the racially-defined group have of a loss of status is a misperception as we demonstrate in our own analysis how little status they have.

Self-refuting scholarship. A prime example of why more cognitive diversity is so needed in contemporary academe.

[Cross-posted at Skepticlawyer and at Medium]