H-1Bs
Published on 27 December 2024
Recently there’s been a furore about H1b visas and I thought I’d write a post on it. I accumulated most of my labour economics knowledge after the 2008 crash when I was an unemployable grad, so it’s always been a focus of mine as to why employer and employee perceptions diverge so much (spoilers: it’s the employers who are delusional).
As usual a bit of history is required to understand the present. HR peaked around the mid-century, where large corporations hired bushy-tailed high-school and college graduates en-masse for entry-level jobs. Then it was the job of HR to identify the best-performing workers and promote them, while planning ahead to ensure that no workforce shortages occurred in the future (over a horizon of a decade or more). HR were able to make their own psychometric tests (they had post-grad qualifications in industrial-organisational psychology), as well as identify personality and social qualities which might suit certain workers for certain roles.
As the reader is no doubt aware, this system no longer exists, although one can identify vestigial elements still in place in the military. There are a variety of reasons for its disappearance, I’ll give the proximal ones but it’s worth noting that institutional quality and sophistication has generally been going backwards in the West since the 70s, so there are probably broader trends at play.
For starters there was the breaking up of the corporations themselves. The trend towards large vertically-integrated conglomerates reversed in the 70s, with a new-found emphasis on core competencies and the outsourcing of non-core functions. Partly this included an outsourcing of HR itself, which has worked out very poorly as incentives for recruiters etc. aren’t well-aligned with businesses that make use of their services. But even for corporations which retained their HR functions, the value of an ‘internal’ labour market diminishes as role diversity vanishes. I.e. if the semi-skilled and highly-skilled areas in a corporation go their separate ways, then the strategy of hiring lots of young people for the semi-skilled jobs and then promoting the best into higher-skilled work no longer can occur, the semi-skilled firm has no-where to be promoted to, and the high-skilled firm has no entry-level jobs.
This phenomenon was contemporaneous with the rise of professional managers who were trained at business schools. This meant that the best workers were much less likely to get promoted into management as those jobs were reserved for those specifically trained in that occupation. The mail-room to CEO success-story is a boomer-era story that simply can’t happen today. As meritocratic promotions decline, it becomes pointless even identifying talent beyond the hiring stage and so it justifies further cuts to HR (if not in headcount, then at least in function). What really was happening was the demise of the ‘company man’ and the disappearance of the internal labour market.
This trend was further accelerated in the 90s with the rise of internet hiring, which great reduced hiring frictions. Why bother planning your workforce requirements going forwards when you can just ask for exactly what you want online? Early on this worked especially well as your competitors ended up training your workers for you; there was a free-rider effect. Now that firms have disbanding their training programs and refuse to hire inexperienced workers, there’s a permanent shortfall of experienced high-skill workers which makes the experience of online hiring much worse than it was in the past.
To fully appreciate the problems this has caused requires a brief digression into theory. The simplest model was provided by Mincer, which relates log wages with years of schooling and experience (including a quadratic experience term). This canonical model has problems, although economists often claim the returns to education are very high, it’s clearly not independent of experience. In my country, Australia, new graduates earn close to minimum wage, thereby proving that the coefficient associated with education alone should be close to 0. An improvement to the model is to add in an interaction term between experience and education to help explain the steeper earnings trajectory for early-career professionals relative to unskilled workers.
Economists can’t deny that mathematically this interaction term improves the fit. But their interpretation of it is utterly bizarre, they often believe that education magically embeds a latent form of human capital which might lie dormant for decades only to be revealed, in earnings, after decades of experience in the workforce. Economists are obsessed with the idea that education alone is what creates human capital. The far more common-sensical interpretation is that higher education lets people snare an entry-level high-skilled job, and the skills acquired on-the-job is what causes wages to rise rapidly thereafter.
An obvious advantage of the second explanation is that explains the huge variance among individuals. It’s not uncommon for two graduates with comparable human capital variables (iq, major, grades, experience, etc.) to diverge dramatically after graduation if one of them manages to launch their career while the other gets stuck in a dead-end job (say bartending). Economists do all sorts of mental-gymnastics to pretend that there was always something seriously wrong with the grad that became a bartender, because they don’t believe in on-the-job skill acquisition and so can’t attribute the poor earnings growths for that individual due to them being unlucky to miss out on a proper job post-graduation.
So to recap, the reason an Australian grad might study at university for 4 years, get into a highly competitive grad program which pays only a bit above minimum wage, and then see their salary 3x in 4 years thereafter is because virtually no human capital is acquired in higher education, it’s all acquired on the job. Note I’m not expressly endorsing a signalling-model of education, it might be the case that university provides a tiny bit of human capital, just enough to let grads into grad-positions, which then pays massive dividends down the line. One reason the debate around signalling vs human capital is confused is because the signalling-advocates are right, there’s a super-abundance of evidence that people don’t learn much in higher ed, but the evidence in favour of signalling is also rather weak. My explanation reconciles both of these facts.
As an aside, there’s another plausible (but wrong) explanation for why grads see such insane earnings growth early on – because employers learn what human capital each employee has. Empirically this doesn’t explain what we see, because learning actually happens very quickly (you know if you’ve hired a dud within a few months) and because such learning doesn’t translate across firms very well, but early-career professionals have no problems switching firms. The lack of information flow is because it depends on the employee’s boss acting as a reference, but if the employee is incompetent the boss might giving a glowing reference just to be rid of him or her. The perverse incentive here means that references don’t provide reliable information about how good a candidate might be.
The second important piece of theory to understand is the distinction between general human capital, i.e. transferrable skills, and firm-specific human capital (e.g. the processes and other domain knowledge specific to that firm). The reason this is so important is that while both improve the productivity of the worker, in the case of transferrable skills this additional productivity is offset by the increased bargaining power that worker now has. So if you teach an employee Python, sure he might get more done, but you’ve also improved his CV and he will be offered more money by competitors who want to make use of his python skills so you will have to pay him more for having trained him. In other words, when a worker learns firm-specific things, it benefits the firm, when he learns general skills, all gains accrue to the worker.
This in itself isn’t enough to cause a major market failure. In jobs where the workers acquire transferrable skills, the skill acquisition itself can be a form of compensation, and this is seen empirically (there’s a dead-end job premium – i.e. the additional amount you need to pay workers who aren’t improving their skill base). However there are limits to this, for one young people might not have sufficient financial resources to get paid in the form of skill acquisition. Secondly, many countries essentially outlaw this, including my own (which has a high minimum wage and unpaid internships are effectively illegal). Thirdly, information asymmetries can prevent the labour market compensating workers in this way, e.g. if unpaid interns are promised useful training, but then exploited and told to get coffee, then eventually you won’t have an unpaid internship market as potential interns won’t know which offers are genuine.
Firms have a few strategies for getting around the problem that in a highly-liberalised labour market, training only benefits workers. One is to entrap young workers with golden-handcuff compensation schemes, such as shares which take years to vest. Another is to convert transferrable skills into firm-specific skills, e.g. by having a proprietary software stack. Or you can use various anti-competitive strategies like aggressive non-compete clauses. While these approaches are widely criticised, I think they’re solving a massive market failure which no-one has identified, i.e. by locking in workers they’re creating the pre-conditions where they can hire inexperienced workers and train them up in a mutually-beneficial way.
Having established the preliminaries let’s consider some stylized facts about the labour market. Studies show formal training programs have virtually disappeared in a generation, which makes perfect sense given the demise of the company man. While I don’t think most skill acquisition tended to occur in these programs, they’re probably a reasonable proxy for how much firms are willing to invest in their workers in general (including simply hiring inexperienced workers, which necessitates some sort of informal training even if it’s just a senior looking over their shoulder). So human capital development in general has likely, similarly, fallen of a cliff.
Contra some big twitter accounts, the explanatory power of cognitive ability in explaining wages has plummeted judging by NLSY and similar datasets. There’re two facets to this, firstly directly observing job performance and promoting the best workers is extremely meritocratic. In contrast the resume + interview approach to hiring is only weakly meritocratic, so the demise of internal labour markets has been a big loss for meritocracy. Secondly, because workers aren’t able to directly display their competence to get promoted, they’re increasingly reliant on credentials. Credentials have always been a mediocre proxy for merit, and they’ve gradually gotten worse as higher education standards have fallen, the SAT has fallen into disuse etc.
The graduate premium itself is plateauing a bit in the US and falling in many other countries (who’ve ramped up higher ed attainment rates much faster). This requires some careful interpretation. On the one hand underlying wage dispersion is inflating all wage premia when measured in dollars (hence the confusion over the weakening IQ – earnings relationship). This dispersion is because the returns to skills are increasing rapidly, but it’s being offset by ever-more grads being unable to enter into their chosen careers. So for those grads who do make it, the returns are bigger than ever, but fewer and fewer are. This reflects the dearth of entry-level positions which I suggested earlier was a likely outcome with the new approach to hiring. Observe that economists are nonplussed by the inability of the labour market to absorb the new graduates they advised get churned out in the UK, Australia, Canada, etc. Their simplified ‘all human capital comes from schooling’ idiocy precludes them from understanding what’s happening.
The bottleneck at the level of entry-level jobs also produces some easy predictions which it turns out are accurate. For instance there’s often a curious pattern where occupations which are supposedly in skill-shortage do not have good employment outcomes for graduates. E.g. CS majors have some of the worst employment outcomes in the UK, despite being supposedly a massive skill shortage. But if you imagine a labour market where the total supply of skilled workers has nothing to do with education, but instead the number of entry-level jobs, then you’d expect skill-shortages to be correlated with unemployable graduates.
Another interesting sidenote is that employers’ complaints about skill shortages do not seem to coincide with objectively tight labour markets. US employers were complaining about hiring difficulties more in 2012/2013 when unemployment was extremely elevated than they did prior to the GFC (2008) or prior to Covid. The way to understand this is that perceived hiring difficulty is based off expectations created in the recent past. So e.g. GFC -> massive glut of good workers looking for work -> employers decide they should only hire perfect candidates on low wages because market supports this hiring strategy -> market improves a bit (2010 – 2013) -> employers start claiming there’s 0 suitable candidates despite a very slack labour market. The moral of the story is to never trust employers about hiring.
The disappearance of entry-level jobs is accompanied by, although qualitatively different from, the decline in ‘middle-skill’ jobs. This is traditionally explained-away as an effect of computers and automation, although economists originally thought this would hit the low-skilled sector, not the middle-skilled sector. Regardless of the role of computers, I’ve already offered an alternative/additional explanation, that middle-skilled jobs were once intentionally used by HR as a training-ground for high-skill workers. Now that dual-purpose no longer exists and so there’s less supply of those types of jobs.
Another very important fact, somehow unknown outside labour economics (the media report the exact opposite to the truth), is the collapse of job mobility. You might think this contradicts my position, but not at all. Firstly, it can partly just be explained by an older labour force. Old people tend to switch jobs less. But more important, the new ‘just-in-time’ hiring approach means that you only ever hire someone who’s already done the job before. In this sort of market, where no-one can change careers or apply for a higher skilled job (both also are being gated by credentials) there’s simply less reason to change jobs. The highly varied careers the silent generation had is simply impossible today.
Another reason for the big decline in job mobility is that the job recruitment process is appalling, as reflected in time-to-hire which is ever-rising. This is undoubtedly because, in some sense, all candidates really are unsuitable, insofar as employers never want inexperienced workers but this means there’s no jobs for them so they spam all job ads. And there’s no experienced workers to apply for these ads because how would they have acquired said experience? HR’s job, instead of workforce planning, is nowadays to wade through giant piles of job ad spam (also reflected in statistics, applications per ad is generally rising - indeed this wasn’t even considered something worth measuring decades ago when you could get hired on the spot). This has been partly automated, using risible techniques like key-word searches, further degrading the integrity of the selection process. Another tactic for reducing the number of applications you have is to grossly exaggerate the skill and experience requirements for every job, nothing like demanding 10 years exp in a technology that has only existed for 2 to get the number of applications down. The result is that all honest candidates are filtered out, and then the employer later wonders why everyone they hire is awful.
On a tangential note, employers can’t even be trusted when they describe what qualities they want in a worker. When asked by surveyors, they tend to list qualities lacking in their current employees, i.e. precisely those qualities you provably don’t need to get hired. So they’ll say they want people skills, critical thinking etc. but if you ask employers why did you reject this application, it ends up being like 80% insufficient experience and 20% insufficient hard-skills / education. No-one is ever getting rejected due to lack of critical thinking skills (hardly surprising given it’s not assessed in virtually all cases). Young people would be well-advised to completely ignore employers whenever they talk to about the labour market, as everything they say is diametrically opposed to observable reality.
So this is the context that needs to be understood before even entering into a discussion on skilled migration. There are permanent skill shortages, which are really experience shortages, and these are because it’s genuinely not in the employers’ interests to hire inexperienced workers. It is not possible to educate yourself out of this problem, and no country has ever done so. Producing more STEM graduates won’t mollify employers who would never consider hiring graduates in the first place. Ramping up higher ed attainment rates has done nothing for GDP in those countries, proving human capital is not generated in HE.
H1Bs tend to substitute for inexperienced natives, being inferior to experienced ones. We have heaps of evidence for this. It’s literally a lottery, you can just directly observe that lottery winners hire fewer natives than lottery losers, the perfect natural experiment. Secondly, natives will rationally avoid areas where they have to compete more large numbers of high-skilled migrants. We have evidence of this too. While there’s some truth that westerners are nowadays too lazy and stupid to study hard mathy subjects, this cultural trend is being reinforced by migration policy which actively punishes anyone who does apply themselves.
In effect, training (which includes simply hiring inexperienced people) has been outsourced to India, where labour is much cheaper. Sometimes multinational firms do this knowing full-well if the worker excels they can get try to get them across to their American office. This is an equilibrium strategy in a standard neoclassical setting with free movement of labour, while existing engineers in the US might be sufficiently superior to pay them the extra salary, would-be engineers are not (after selecting on IQ using ability tests). Also observe that young workers in Western countries are increasingly saddled with student debt for their degrees where they learnt nothing, to the point that an American graduate can easily have higher interest payments than the salary of a junior SWE in India.
This creates a feedback loop. As Indian firms constantly lose their top talent, they become expert in taking young workers and upskilling them. Meanwhile firms in Western countries get rewarded for never training anyone, their unsustainable expectations for applicants never need to be diminished, they can use the resulting skill shortages to lobby for ever-greater migration while natives learn to simply avoid those industries. Skilled migration thus begets the need for ever-greater skilled migration, something easily observed in Australia.
It’s possible the US needs to do this to maintain its competitiveness in software development, when combined with its extremely good capital markets etc. although I’m skeptical. But what about the rest of the Anglosphere? The British economic model is apparently to build a services-economy by taking workers from third-world countries and paying them more in Britain. But no-one can articulate how there’s any comparative advantage there. Two competing narratives have been muddled together here, the idiotic one from economists about human capital = education where the UK would be leveraging Oxbridge, and the one from the business lobby that whitey is stupid and lazy and all workers in future must be imported. The combination of the two is utterly non-sensical though, like you’d have to suffer from really profound learning disabilities to entertain this as a viable economic strategy.
Finally, while I’m taking the position that skilled migration is much worse than economists claim, it’s not really the biggest issue here. Western countries have obliterated their ability to produce skilled workers, and the demand for skilled migration is just a response to that. Our dopey growth economists have largely failed to explain what happened in the 70s to cause the developed world to move towards a permanently lower growth trajectory. But I explained it in the first few paragraphs, it’s just human capital if you were to measure it correctly. We moved from a system which allowed workers to reach their potential through selection and upskilling, to one where there’s permanent gluts of underemployed graduates competing to get their resumes through ATS keyword searches. Everyone is chronically underemployed skill-wise. The dearth of skilled vacancies for young people means that employers who do offer them can quite literally work their young people to death with 100-hour weeks (happens occasionally in law, finance etc. in London and NY). Does this sound like a job market where natives don’t want to learn new skills?
Meanwhile managers, often MBAs who don’t know anything about high-skilled work because they’ve been parachuted in under the new system, screech that only workers from a country which has barely mastered improved sanitation have 20 years experience in Next.js so need to be imported in great numbers. Anyone who pushes for this needs to explain why Indians are so much better than natives. My thesis is that it’s a structural feature of the labour market and globalisation that upskilling happens in countries with cheaper labour. The alternative explanations are seldom forth-coming (short of white people being racially incompetent).
I’m hoping my explanation gets some legs as boomers retire, as they were the last generation to profit from the old system. My suspicion is that there’ll essentially be irreplaceable, every generation thereafter having been forced to languish in unskilled jobs due to a lack of upward mobility within the workforce. In any case the conventional model, that human capital is derived from education, is certainly hard to sustain as nations like the UK fail to see any sort of productivity boost from their huge education boom.