This chart from the Washington Post tells you that a degree in the humanities is, on average, “worth” less than a degree in engineering, something you already knew was true.
What you perhaps didn’t already know — and for which this article and chart were produced to remind you — was that a degree is a thing which should be understood through its convertibility into future earnings. After all, when you ask the question “What is a major worth” — as the chart’s headline does — you are really asserting that a college degree should be reduced to its instrumental value in getting money. To say that a major is essentially “worth” something is to claim that its essence is what it is worth. That’s why the headline omits the question mark: they aren’t asking, they are telling.
This is not to say that there’s no useful data in that chart, or in the report it comes from. But as I learned in college, a chart is a machine for narrativizing data, which is to say, for rendering smooth and simple the complex granularity of empirical reality. And the narrative this data is being processed to tell is misleading in precisely that assertion of apparent simplicity. For example, as the author of an article entitled “On path to riches, no sign of fluffy majors” puts it:
“The report is based on previously unreported census data that definitively links college majors to career earnings.”
But this is a complete red herring. The phrase “definitively links college majors to career earnings” cannot be meaningful, since no one has ever wondered whether majors are linked to earnings. Everyone understands that what you do in college has some effect on what you earn after college, and everyone also understands that starting salaries for software engineers are likely to be higher, on average, than whatever it is that people do with their humanities degree. The former is vocational training — in that it is built and designed to get you a particular kind of job as a software engineer — whereas the latter is not. Quelle surprise that getting a degree which is not designed to prepare you for a job turns out to be less good at getting you a job. The only interesting question is how dramatic that difference is, the extent to which you are ruining your life by picking a major based on something other than cold calculation of how to make the most money over the course of your life. Or, to put it another way, how justified you would be in choosing a life doing a kind of work you hate in order to make more money.
In this sense, while the article’s thudding subtext pounds it into your head that there is a quick and simple convertibility between major and earnings (and the title puts it about as stupidly as could be managed), it only takes a moment of looking at the full report to notice two things. First, the variance within individual majors is significantly greater than the variance between different majors. This means that a brilliant or lucky English major is going to make more money, on average, than a dull or unlucky engineer, roughly the same ball-park difference, in fact, as between the average humanities and engineering majors in the first place. Which is to say, the data could also be spun to demonstrate that the secret to future earnings is to emulate whatever it is that the people in the upper percentiles of your major-cohort are doing.
Which brings us to the second observation I would make: the statistical variance within individual majors is actually so huge that the predictive power of the data is negligible, contrary to the grand claims which are made for it by the researchers (and which are then dutifully repeated by the Washington Post). If you are an English major, this chart tells me that I have a fifty percent chance of being right when I guess that your future earnings will be between $37k and $71k. I also have fifty percent chance of being wrong in guessing that it is in that range, since precisely as many people are outside that range as are inside it. A quarter of the sample will make less than $37k a year, while a quarter will make more than $71k. Which means, then, that I can only predict with some moderate level of certainty that your future earnings will probably place you somewhere between lower-middle class and upper-middle class, and will often be wrong in making that prediction. By contrast, if you major in biomedical engineering (50% between $50k and $100k), exactly the same thing is true: I can only predict with some moderate level of certainty that your future earnings will probably place you somewhere between lower-middle class and upper middle-class. And I will often be wrong in making that prediction.
Again, as I learned in college, articles and charts like this one are where knowledge goes to die. When researchers collect a bunch of data, by a strange coincidence, they always just happen to discover that the data they just collected is really important and has clear consequences for how we see the world. And then, when they issue a report and a newspaper dutifully writes down the things the researchers said about that data (because, hey! free content!), we not only get an article declaring an even more simplified version of a truth that is, by its very nature, incredibly complex, but we are encouraged to think that by swallowing that simple truth — “what is a major worth” — we have understood something about the world. Both the researchers and the reporters have a vested interest in convincing us that this is true. And so, to ask the question “Can your choice of major predict your future?” will not only constantly get re-formulated as “Read my account of how your choice of major can predict your future!” but as the question mark disappears, the content of the sentence changes: instead of indicating that future income is the product of a vast multiplicity of different variables (as it manifestly is), we get the claim that it is mostly the product of one.
This is obviously wrong, of course, but because this study only tested one variable, the only way they could claim their study to be particularly newsworthy was by arguing that future earnings can be predicted by looking only at that single variable. And, again, this is something that even the most casual experience of reality would show you not to be true, and which even their own data demonstrate to be a deep oversimplification. But don‘t take my word for it; make a short list of people you actually know in actual life, and compare how well what they studied in college explains what sort of job they do now. For me, almost everyone I can think of has the job they do for reasons that are only vaguely, at best, related to what they did in college, if at all. And for the vast majority, it is obviously the case that the best predictor of their future income is something else entirely (a skill they developed along the way that happened to become a career, a job they acquired because they knew someone that then turned into a career, a complete fluke, etc). The one thing this study can’t say, it seems to me, is the one thing it should say: correlation exists but is much lessthan it would need to be for us to use it in making meaningful decisions about how to live our lives (or in making policy).
When the difference between the upper and lower halves of each major-group is as large as it is, the important economic question that will face new graduates will be figuring out what separates the successful ones from the less successful ones. And while the entire thrust of this discourse is to convince us that the choices made by college students is what determines their future — thus, both internalizing success and failure and encouraging government and universities to continue de-emphasizing everything but STEM subjects — the stark reality is that the best “choice” a college graduate could possibly make would be to have graduated in 2005 rather than 2011, when the unemployment levels was around half what it is now. If they were really smart, they would have graduated in the early 80′s, before crushing student debt was inescapable. Beyond that, they should have chosen to be white, male, to be born to money, and to be genetically healthy; statistically, by far, these are the best path to riches.
Cross-posted from Aaron Bady’s blog, zunguzungu.