Preston Cooper recently published an excellent analysis of the return on investment (ROI) on a very large number of American university degrees. The results should be taken with quite a large pinch of salt as the report doesn’t have a credible way of establishing whether what degree someone took really caused a psychology graduate to have a lower salary than a nursing graduate. The study finds the correlation between degree type and income given various variables - race, parental income, local labour market conditions ect - but doesn’t make use of a natural experiment to find whether correlation is really causal. This is quite a large problem because we should expect lots of variables that are very hard to observe to have a large effect on people’s wages and to be correlated with an individual's degree choice. The sorts of people who do business degrees are much more likely to be the sorts of people who make making copious amounts of money an important goal in their life compared to someone who takes a degree in African-American studies. This is quite relevant from an individual choice perspective; it could be that what really matters is how hard someone tries to make money rather than the specific subject they major in.
Despite this weakness, there are some quite interesting inferences that can be drawn from the analysis.
Practical skills learnt matter a lot
There’s no magic STEM way of thinking that makes people a lot of money
Some subjects get returns mostly because they teach skills rather than they signal ability
Practical skills matter more than a STEM mindset
This table shows the degrees that generate the highest return on investment - how much more a graduate makes net of costs and earnings forgone while at university - over the lifetime of the graduate.
The standard narrative, at least amongst very online people, is that the route to financial success goes via a STEM degree and indeed the above table contains many of the degrees that the same people who advise English majors working at Starbucks to “learn to code” would expect to see. However, what’s interesting here is that Transportation and Architecture, Economics and Business beat out computer science and physical sciences (physics and chemistry) in terms of proportion of degree programs with a negative ROI and Economics beats the best physical science degrees in terms of top earnings by a large margin.
One theory of why STEM degrees are so good is the natural sciences by their empirical rigour and mathematics and computer science by the rigour demanded by proof and programming are better at teaching people to be high quality thinkers than the more wishy-washy methodologies employed by the humanities, arts and social sciences.
Below is the table showing the degrees that generate a low return on investment.
The results from this study suggest that a particular STEM way of thinking isn’t the secret code. That theory would suggest that mathematics and physics would generate at least as high returns as their slightly less rigorous cousins in engineering and computer science. It probably also suggests that biology and psychology which share experimental ground with physics and chemistry would generate similarly high earnings to their STEM brethren. However, what this data shows is substantial heterogeneity across the STEM subjects. Biology is one of the worst performing subjects and psychology very nearly the worst. If it was, rather than the rigorous application of the experimental method, mathematics that really separated STEM from the rest we would expect mathematics to be the highest performing STEM subject. Instead computer science and engineering beat maths and physics by a substantial margin when looking at the top performing programs. The success of economics may fit the mathematics narrative, but business really doesn’t and certainly doesn’t fit the scientific rigour theory.
Instead what I think this data suggests is that the practical skills that degrees confer matter a great deal. Nursing, agriculture, transportation and architecture stand out by the degree to which they prepare their graduates for specific fields. This is also the way in which engineering differentiates itself from the natural sciences and mathematics. It doesn’t seem like surprising success of agriculture and surprising success of engineering relative to other mathematical subjects can be explained by those subjects being a stronger signal of ability and conscientiousness than the others. In fact these more practical subjects seem anti-correlated to ability - the highest GRE scores belong to applicants for physics, mathematics and economics in the quant section and philosophy (by some distance) and English in the verbal section. Engineering (with the exception of chemical and materials) score lower than math, physics and econ and computer science substantially lower. Business (although not finance) is very substantially lower. One should be careful of selection of effects here. In the more academic subjects of mathematics and physics one should expect the best students to be applying for Phds whereas this seems less likely to be true for more vocational subjects like engineering and computer science and business. Even accounting for this, it doesn’t seem like the advantages of computer science, engineering and business can be explained by their having stronger students.
The weakness of the life sciences in increasing earnings, the advantage of business over physics and chemistry, engineering beating out math and the surprising success of agriculture degrees points towards there being no magic STEM sauce. Instead people who are engineers are able to solve engineering problems and this is a skill that is currently under-provided in the economy, as are people who are good at making business decisions and providing healthcare at a lower level of expertise than a doctor.
This is probably skewed by my own personality and experience of the world, but the explanantion that comes to mind to me for why those practical skills are underprovided is:
doing practical things can be boring, and by definition is less intellectually stimulating than more abstract fields. Hence for the same amount of money the people who are good enough at maths say to be great engineers or physicists would far prefer to be physicists. Thus this makes sense as an equilibrium where you are paid a premium for doing boring less intellectual jobs. This clearly doesn't cash out for the most boring unintellectual jobs (manual labour, retail etc) presumably because there is an oversupply of people willing and able to do them but unwilling or unable to do other things.
So if somehow we end up in a society where almost everyone is very clever and educated but we haven't fully automated menial tasks, I expect cleaners and so forth to be among the highest paid people in the economy to compensate them for doing boring work when they could otherwise have a fun job as a mathematician.