Hacker News Reader

Comment by cm2012 | Hacker News Reader

cm2012

I can't find the citation now but the biggest argument against college I've seen is that once you control for family income, life success is way more correlated to SAT scores/the college's you apply for than the ones you actually attend. This would imply that college success stats merely indicate that they are selective in who they choose to enroll.

EDIT: Its not the one I'm looking for, but I found a good one. http://conference.iza.org/conference_files/CoNoCoSk2011/gens... shows that if you control for IQ, the average person makes $100k more over the course of their whole working life by getting a bachelor's.

- This doesn't even account for the four years of work you lose by going to college instead of working.

- Its based on data from people who mostly graduated before the year 2000, and by all accounts the value of college is becoming less, not more. It's also based on a calculation when college was much cheaper.

View on HN

I don't think your first dash is correct. It looks like they just integrate the curve in figure 4, multiplied by a 5% discount. This includes both tuition cost and missed work during college. Also, it's not really fair to represent this as a difference in lifetime income because of the discount rate: the college attendee makes significantly more than $100k more over their life, but the gains accrue later in life while the cost accrues earlier, so the discount rate makes the benefit smaller.

Noted, everyone ignore the first dash. Thanks!

Those studies say something more like "UW Madison/U Washington/UIUC really do give you as good an education as the ivies" and quite a bit less like "college doesn't matter/is just a signal".

I have read that study, but I'm actually digging for a separate study now. I remember it split SAT score, educational attainment, and average salary into separate variables. For people with high SAT scores, average salary 10 years post graduation was basically the same regardless of level of education.

Ivy League matters if you want to be in finance or white shoe consulting.

Everything else, you just work hard at your first job and nobody gives a shit. I went to a big SUNY school and was a mediocre student. We all made out fine.

Agreed on the Ivy League part. I work with a guy that went to Princeton. I went to a public University in Florida.

Sure he learned a lot more and had to do a thesis for his undergrad, but we have the same job now.

He has no more leverage than me when it comes to that.

And when that guy’s roommate strikes it big, who gets the call to join the rocketship?

Steve Ballmer was Bill Gates’ roommate. You are getting free networking lottery tickets with high achievers at an Ivy.

IQ correllates somewhat with income and even more with school performance, so by “controlling for IQ”, the authors are intentionally flattening and hiding some of the income variation. They’re effectively discounting for the probability of going to college in the first place. That will, of course, hide differences in income due to going to college.

The actual data shown in that reference seems to tell a different story, it shows people with college degrees earning 20-50k per year, on average, after taxes. That will sum to more than $100k over a lifetime.

They account for that plus a lot of other factors in the 100k number. Basically the conclusion is, for people born from 1920 to 1980, college was net value add accounting for everything else, but not a huge one.

Yes, I know they account for that. Why does that make sense to account for IQ? That is explicitly discounting for college!

The 100k number is a fictional number, it’s trying to answer the question of assuming the same person didn’t go to college, is it possible to calculate the true value of college.

We could argue whether they actually calculated that number correctly, but what is absolutely true -- my point -- is that the actual people who had actual degrees actually earned a lot more than 100k extra over their lives, on average. The differences in wealth, illustrated by the data in that paper (see figure 1), are 2x-3x in annual income and accumulated savings much higher than 100k.

Yeah, but those people could have earned that much without going to college. If you just take the correlation between college and income, all it tells you is that colleges skim the top percent of society, which we knew already. The question we're trying to answer is that if you have the choice to go to college or not, should you?

I believe I understand the question & goal, and maybe this is the right way to go about it, but I’m not sure. It somewhat strikes me as the wrong question -- people don’t choose college equally at all income & IQ levels, and some of the factors that (statistically) lead people into to choosing to go are already accumulating before they’re even born. As a result, the hypothetical solution that hides the factors that lead to that choice in the first place in the data might not be fairly or adequately answering the question of whether someone should invest in college, given the choice, right?

Under this model, you could have someone with a higher IQ and a degree earning more money than a non-degree holder, but have a ‘negative’ return. Does that accurately reflect the situation, should the advice be to skip college, even though the absolute financial returns might be positive?

And what does this answer tell someone who has a low IQ or a low socioeconomic status? How should they try to evaluate this choice? Does discounting high IQ and high family education give an accurate picture of the benefits of post-secondary education to poorer families?

Controlling for IQ here is difficult because the higher your IQ is, the more you would be attracted to learning for learning’s sake in the first place.

Exactly!

The 100k is a net present value calculation, which includes a 5% per year discount, which is quite steep. 1/1.0515<0.5, 0.9530<0.25. The undiscounted difference in lifetime earnings looks, just eyeballing the figure, to be around 300-500k.

Yes, precisely. It seems like a good idea to keep an eyeball on the absolute, unadjusted numbers, even if controlling for IQ and other factor is a valid way to look at it, since the experience of being someone in the disadvantaged bracket is not a minuscule difference. Flattening the perceived income difference by controlling for IQ hides the absolute magnitude of income differences. Plus there’s always the chance that the model is wrong, and/or not accounting for everything.

They have a really good discussion of this in the middle of the study, admitting it erodes predictive power.

They point out that much of your lifetime earnings are predicted by your parents' level of education.

Reply on news.ycombinator.com