I was quite impressed with the comments on the post, since most of you passed with flying colors and even touched on issues that were more subtle than what I was going for, such as the way unemployment is calculated and whatnot. The fundamental point that I was making, of course, was regarding the correlation versus causation issue. In that sense, there is nothing wrong with the graph itself, since it just shows data about the world. The problem is with the caption on the graph that interprets the data as the benefits of education, since this implies that you can take a randomly selected person, give him some education and therefore bump him into a different one of these buckets. I particularly liked reader Warren’s J’s comment, which states “Even more pronounced is the correlation between employment and the education of your highest-educated parent. So, the conclusion is that you increase employment by sending parents to school, right?”
This discussion highlights the classic problem that labor economists face in trying to estimate the returns to education. On the up side, labor economists have a lot of data at their disposal- they can calculate wages versus education and can control for things like age, tenure at job, gender, race, etc. On the down side, they can only control for those things that they can see, and ability and motivation are, for the most part, not on that list of things. Why is this a problem? Let’s say that our regression tells us that a year of schooling is associated with a $5000 increase in yearly wages. Does this mean that I can increase my income $5000 a year by signing up for some night classes? I don’t know. What it does mean is that those people who have CHOSEN to have an extra year of schooling make $5000 more. Put this way, it’s pretty clear that at least some of the $5000 should be attributed to whatever qualities made the person choose the schooling in the first place. If it’s the case that smarter and/or more motivated people choose more schooling, then the $5000 is an overestimate of the return to school itself. On the other hand, if people seek out extra education to make up for what they are lacking in terms of ability or motivation, then the $5000 is an underestimate of the true return to schooling.
In an ideal world, economists would randomize people into different groups, each of which gets a different amount of schooling, and then compare the resulting wages and unemployment levels. Unfortunately, the government hasn’t gotten on board with that plan yet, so economists have to be more creative. For example, economists use the fact that there is an imperfect correlation between age and number of years in school to estimate the effect of mandatory schooling on financial outcomes. This works because most mandatory schooling laws are written such that a student must go to school until he reaches a particular age, often 16 years old. However, depending on when the student was born and/or allowed to enter school, this requirement could translate to completing 9th grade for some students and completing 10th grade for others. If you believe the assumption that birth month is not correlated with innate ability or motivation, this approach provides a limited but valid causal estimate of the return to a (particular) year of education.
When estimating these sorts of effects, it’s important to remember that even a causal interpretation merely answers the question “what does an extra year of school do for this person under the assumption that no one else decides to up and get more school also?” If a large number of people increase their level of schooling, the supply of workers for jobs that require higher education increases, which pushes down the wages that these workers get. (The baby boomers would probably be more than happy to shed some more light on this concept for you.) To some degree, the mix of available jobs might change to meet this supply, but it’s certainly not the case that if everyone went to college we’d end up all doing jobs that required or could financially reward a college education. (If nothing else, this would make your morning cup of coffee pretty expensive.) In fact, you’d likely see a ratcheting effect illustrated a few years back by one of my freshmen who earnestly explained to me that she was discussing with her parents how “grad school is the new college.” In a lot of ways, she’s not wrong.
I initially put this topic out there as a quiz because I feel like I’m beating a dead horse with this whole correlation vs. causation thing- I’ve tried stick figures, TV dinners, SAT scores, baseball, and boobs, and yet not everyone seems to grasp (or be willing to acknowledge) the concept. This wouldn’t really be a big deal if the distinction between correlation and causation were just a matter of nerdy semantics, but instead the issue has clear choice and policy implications. For example, there would be a lot of pissed off people if everyone saw this graph and ran out to get their University of Phoenix degrees, only to find that they were still unemployed and $50,000 in debt afterwards. Before you say “that’s absurd, clearly people aren’t that dumb,” take a look at the following clip at around the 6:52 mark:
|The Daily Show With Jon Stewart||Mon – Thurs 11p / 10c|
(Click here if you can’t see the embedded clip.)
“You want to create jobs as rapidly as China? The Chinese pay zero capital gains tax. If we had zero capital gains tax in the United States, we’d be building factories, founding companies, creating jobs, we’d be dramatically better off.”
This actually made me look up Newt Gingrich’s educational background to confirm that he had not gone to law school, since I can’t imagine him scoring very well on the LSAT with this sort of logic. The most frustrating thing is that I don’t actually think that Newt or others are actually unintelligent or incompetent, so I feel to a large degree that they should know better than to go around trying to convince people that if something leads to a particular result in China that it’s going to lead to a similar result in the U.S. I’m not sure I could think of a worse natural experiment if I tried. Maybe Newt was suffering from the same lack of recollection as Homer Simpson:
On a side note, it seems like Jon Stewart is on Paul Krugman’s side of the octagon. I am also pleased that my readers seem to be more with it than the average politician. (What’s the line about damning with faint praise again? 🙂 )