Community college alumnae earn $6k more per year than high school grads | Hacker News Reader

Community college alumnae earn $6k more per year than high school grads

eurekalert.org

17 pointstoni18 comments

As a former PhD student in economics I can tell you that this kind of research is often not very well done. Having not read the paper I will reserve final judgement, but based on the journal it's published in this probably doesn't hold much weight. Here's why:

In order to say anything causal (as the author of this paper would like to) in a field where you can't run experiments and only have observational data you need to have a source of identification. It's not enough to compare means. Yes this has everything to do with unobserved-variable bias.

In labor economics the huge unobserved variable is ability. People who are naturally more "able" will likely earn more than less able people and also will likely obtain more education. If that is true then it will look like education is much more important in determining earnings/wages than it actually is.

So there are all these ways of trying to eliminate that unobserved-variable bias. The popular one in the 90s and early 00s (and still popular today despite a lot of criticism) is the instrumental variables technique. Here's the idea:

We need to introduce a source of randomness into our otherwise non-random data. If we can find a variable that is correlated with the causal variable we are interested in, but (arguably) uncorrelated with the unobserved variable, then we can use that new variable as an "instrument." I won't get into the technical details of how to run the regressions with actual data, but this method basically uses an external source of randomness to attempt to say something causal.

The classic paper on this is Josh Angrist's and Alan Kreuger's paper estimating returns to education. They used Vietnam draft lottery numbers as their instrument, since your draft lottery number should be completely uncorrelated with your natural "ability," but definitely would correlate with how much education you got (those having to serve couldn't go to school). More on that paper is here: http://econpapers.repec.org/paper/priindrel/670.htm.

So basically I doubt that this community college paper does much of the required analysis to actually "identify" the causal relationship it is claiming to have found.

UPDATE: Having read the methodology (page 7) portion of the paper (link to the pdf at the bottom of this page: http://userpages.umbc.edu/~marcotte/pubs.htm) I can confirm that my doubts were justified (the author also makes some highly questionable assumptions in order to estimate individuals' work experience as well, but I won't get into that). The author does acknowledge that unobserved variables will confound the results, and attempts to eliminate the bias by controlling for underlying ability with standardized high school math and reading exam scores. This is a common method, but it's really not good enough. We are still dealing with observational data, and we still haven't been presented a source of identification. That is, even if two people scored exactly the same on their high school math and reading exams, but one went to community college and the other didn't, you still haven't accounted for whatever underlying differences existed to lead each to make a different choice.

Wonderful commentary, thank you!

Can you describe some other techniques used to eliminate unobserved variable bias, or do you know of a resource that describes these?

I remember in one of the other studies on the impact of "elite" education versus state-provided university education, they used people who were accepted to an elite university, but chose not to attend. Is that using "acceptance" as an instrumental variable?

I've been out of the econ phd world for a little while now (see blog post here: http://hellosorld.com/personal/2009/07/16/why-i-quit-my-phd-...) so it's hard to remember what all the resources were, but i'll ask around for some good material and post it here. Natural random experiments are another good source of identification (i.e. elimination of unobserved variable bias).

For the elite vs state university question, choosing to attend isn't an instrument since it's still an observed decision. If there was some source of randomness that was uncorrelated with all your other variables besides attending vs not, then that would be an instrumental variable. Say the government decided to take a pool of 10,000 elite college acceptees, gave them each a number from one to 10,000 and then said only the even numbers could actually attend. That would be your instrument. Trouble is that those kinds of natural random experiments don't occur often. That's why economists can get tenure simply by finding a good instrument for some study.

Ugh! Repeat after me: correlation does not imply causation, correlation does not imply causation, correlation does not imply causation.

People who work straight out of HS are very different from those who decided to save up money and go to college, even before they go to college.

Very true. It's entirely possible that people who save up their own money in order to make an investment in themselves through education are simply more shrewd by enough that even if they went another route, they'd make $6k more than the median.

I know we interviewed a CC grads (more out of kindness/curiosity than anything else) and while the guy wasn't what we were looking for, he definitely put a lot more effort into it than most people his age likely would have.

There is a high likelihood of causation though, simply because many jobs (especially in the healthcare industry) require an associate's degree. For many people earning those degree help them go from minimum wage jobs to 30-40K per year.

Yes, the confusion of correlation and causation that you find in every article on post-college earnings certainly causes the benefit of college to be overstated, but that's not even the real problem. Even with an ideal randomized experiment that shows the exact size of the causal effect on individuals, we would know little about the causal effect on a macro level.

Imagine if all doctors had to put $10,000 into a fire pit or else be prohibited by law from practicing a certain lucrative form of medicine. If you ran a controlled randomized experiment, you could easily establish a causal (yes, causal) relationship between throwing money in the fire pit and increased future earnings. This would not make throwing money in the pit any less stupid.

College is a resource sink that no rational individual acting in their own self-interest (with a few rare exceptions) can refuse. It's rational for individuals to go to college to get better jobs, and it's rational for employers to discriminate against non-grads to avoid bad employees, but at the end of the day we're still throwing money into an ever-growing fire pit of crushed beer bottles, cannabis seeds, and Playstation controllers.

You kind of contradict yourself. I'm sensing that you feel college education is purely a signaling mechanism, and I don't disagree completely. You correctly acknowledge that the behavior of both the signal's sender and receiver (student and employer respectively) is rational, but then lament that we are throwing money into a fire-pit as though we could all agree to just stop using this signal and save money. The reason we have the signal in the first place is because high-quality individuals want a way to differentiate themselves, and they have rationally decided that such differentiation is worth the cost.

To the extent that it's only a signaling mechanism, though, it doesn't make sense to subsidize it. Encouraging everyone to go to college with the state paying half their tuition doesn't confer any benefit on the public. It just appears to because it's inflating a common metric of intelligence, thus providing impetus for further subsidy, and so on. It's as if the government noticed that people with large diamonds are happier, healthier, and smarter than people without (which they probably are), so they started subsidizing diamond mines for the good of the public.

It might be advantageous for any given peacock to have a bigger, flashier, and more cumbersome tail than the next peacock, but the arms race makes it ever more certain that the peacocks will go extinct if a new predator shows up. Money that's spent on escalating signals is money that could be spent on other things, and once the public cost becomes high enough that people get disgusted and stop respecting the signal (e.g., large SUVs) its value plummets.

I don't think college is at the breaking point just yet, and it might still be rational to get that generic bachelor's. But if tuition continues to rise at its current rate, which greatly outpaces inflation, there will come a day when the bubble bursts.

The finding that an incomplete degree still has an effect on earning is the most surprising to me. I thought college had mostly devolved into a signal about how committed you were to your career, but clearly at least some courses are imparting actual skills. Maybe the effect is greater in community colleges because they tend towards degrees in more practical subjects? (I don't know anything about community college, I'm not from the US)

I attended a community college (mostly during high school) and also have taught at one.

In my experience, the community college offers two general styles of education programs: one is more or less equivalent to the first two years of a 4-year college, and the other is primarily focused on skill development.

Many students who plan on completing a 4-year degree start at the community college for a year or two, because they can complete a lot of the general education requirements for the 4-year degree cheaper and/or easier.

Other students attend a community college basically to learn a trade. While the college might have a 2-year program in computer science, they may also have a 2-year program in "computer programming", offering classes like "Introduction to IBM AS/400", "COBOL Programming", "Java Web Services", etc. For some IT jobs, this sort of training is really more useful (at least in the short term) than a study in computer science.

Other "trade" programs that I saw included culinary arts, industrial craftsmanship (welding, usage of AutoCad, etc.), dental assistantship... these are basically terminal associates degrees, and the college tells you up front to not expect these classes to transfer toward any other university program.

You also forget the adult education population that are there for career changes because theirs was cut short, further education for promotions etc. and the foreign student population attending to learn English. The community college system in my area if packed with ESL students from Asia and Africa looking to either return to their home country with a new skill, or move into better life in the U.S. using the CC system as a way to open the doors into the 4-year state schools they wouldn't have been eligible from coming fresh off the boat.

IMHO, community college systems are one of the backbones of the American Economy. Cheap, customizable, readily available, flexible and practical education that nearly anybody can benefit from at any age in almost any subject.

Taking individual courses can be just as much of a signal as completing an entire degree. There's nothing here that implies that "some courses are imparting actual skills." What's implied is that the people who actually take courses at community college are likely to earn more than the people who don't. There could be any number of underlying causal factors though, many of which fit with an interpretation of college courses as pure signal.

I barely made it out of high school. I learned almost nothing academic after the 8th grade. That was back in the days when you passed if you showed up enough. I went to community college and paid for it myself just to learn basic college level english, math, public speaking, finance, economics. I have to say it was worth every dollar. I was never interested in a degree. I didn't have time and money to waste on Women's Studies and Interpersonal Communications and host of other garbage that was required. Today I earn far more than 45% more than the average high school only grad. Not all, but part of my earning ability, I credit to the community college. For the price, you can't go wrong on the core skill stuff. I've never worked for an employer who seemed to care about my education, they all seem to care only about experience and performance. I've never even been asked where I went to college during an interview.

I agree with you completely. I'm not trying to say that community college has no value. I'm just saying that the causal relationship isn't clear. Certainly you shouldn't discount your education's affect on your earning power. The problem is that there's no counterfactual world in which we can observe a version of you who didn't go to community college, so we really don't quite know exactly what it was worth.

Lots of CC programs are certificate or licensing programs designed to get people into a trade. Anything from electrician to nursing assistant.

Those jobs pay far better than most unskilled labor jobs HS only grads are eligible for.

Take 1M high school graduate and compare their salary with 1M community college graduate and you probably are right. Fast forward 20 years later, and I am sure the difference will be extremely significant.

Mandatory XKCD comic... http://xkcd.com/552/

Reply on news.ycombinator.com