UK data is unhelpful when trying to draw inferences to US employers/job market

I saw this going viral:

It’s not so much the study is wrong, but that commentators are way overreading into the implications or applicability to American tech jobs. From the study (hat tip to gwern.net, who uploaded it):

The analysis is based on Understanding Society, an ongoing longitudinal survey of more than 40,000 UK households that started in 2009
(Buck & McFall, 2011). Interviews are carried out face-to-face and a
complex stratified sampling design is employed to ensure representativity for the population of the United Kingdom (Benzeval et al., 2020).

The first problem is the study is based off of UK data. The UK is way different compared to the US in that the cognitive filtering is not nearly as strong, nor are salaries nearly as high compared to the US, especially at the tails, in which IQ becomes way more important for getting past the screening. In the UK, where jobs do not pay as well, then it follows there will be fewer applicants, and hence fewer IQ filtering mechanisms at play.

American employers are notorious for being extremely choosy. There is way more pre-employment screening compared to elsewhere in the world, such as background checks, credentialism, automated resume filtering, personality inventories (e.g. Wonderlic, which is also an IQ test), drug tests, long interviews, etc. These are all direct or indirect IQ filters. Post-Covid, such screening has only gotten worse. There are tons of stories, like on Reddit, of smart, competent, qualified Americans being unable to find work, or having to send out hundreds of applications to get an interview, let alone a job.

He’s right. White-collar employers want employees who are overqualified, for jobs that are not cognitively commensurate with those qualifications:

This is rational in the sense that bad employees are a bigger drain compared to the upside of good employees. IQ matters a lot in terms of saving time and money on training and fewer mistakes in the workplace. Not to mention, firing bad employees incurs legal risk and possible reputational harm on social media or job-ranking sites when those people complain.

In the past, overqualified employees were seen as ‘bad’ because they would be inclined to leave too soon for better opportunities after employers invested time and training, but nowadays employers do not have to worry about this because the labor market still heavily favors employers in terms of supply vs. demand, and the intensive interview/screening process ensures that prospective employees are already qualified, without as much training necessary, such as the whiteboard interview process.

Moreover, personality and other non-cognitive attributes matter less in the US in the era of automated resume screening and credentialism. To get to the interview stage, where EQ/personality matters, requires surviving the filtering.

The second problem is the study omits a large category of jobs that have a more selective hiring process, as someone notes in the comments:

The study ranks certain professions by cognitive ability:

If the classification of ‘tech job’ is expanded to include air traffic controller, solicitor, or ‘town planning officer’, then I guess the study is true. I dunno about you, but those occupations do not come to mind when I think of a tech job. Nowhere is coder mentioned, except maybe ‘IT business analysts’, which can mean many things. I am sure none of these jobs involve whiteboard interviews or ‘LeetCode’.

This is not the study’s fault, but rather how it’s misinterpreted as being representative of tech jobs in the US–when these are neither tech jobs, nor in the US. For someone who is a UK citizen with an IQ of at least 105 who aspires to be an air traffic controller, then perhaps the study is applicable, but it’s unsupported to generalize this to American tech jobs or hiring in general.

Similar studies compile data from the 70s-90s that is misinterpreted as evidence of IQ not being important today, as if half-century-old data is representative of America now. This is not the fault the studies, but rather those who draw this incorrect inference. I discuss this my November 2023 post Old studies about income/wealth and job success vs. IQ are unconvincing:

Overall, it’s not so much that the studies are wrong, but I don’t think the results mean what people hope they mean. In the context of debate online, such as Scott’s viral 2017 article Against Individual IQ Worries, people are not worried about only being in the middle like teachers–they aspire to be among the outliers at the top, like coders and other high-status, high-paying professions. In this case, such studies are not really applicable: the evidence suggests IQ does matter a lot, and more than it did 30 years ago at least for the highest-end of jobs.

A lot has changed, like the above the trends, so such studies are unhelpful in understanding America today. I think the best approach is to look at the qualifications of those who are hired for the type of job one seeks to attain, for specific locale, instead of putting too much faith is studies.