[djoldman]

This piece is almost entirely about the cost of college. There is little about the benefits.

As far as I can tell, all the statements about benefits are:

> According to repeated analyses by economists at the Federal Reserve Bank of New York, a four-year degree generates an annual return of 14 percent over a 40-year career.

> ...a weakening job market for new grads (true, but better than the market for non-grads)...

The first quote links to a much better analysis:

https://libertystreeteconomics.newyorkfed.org/2019/06/despit...

As someone who works with statistics, this is the oldest trick in the book: talk about the average when it tells the story you want to tell as opposed to the distribution. Given the massive variability in total cost to attend, the average loses meaning.

[ttul]

And what about causality? People who go to college are likely the smartest, wealthiest, and best supported people to begin with. How do you find a matched control group of people who didn’t go to college but who could have? What were their outcomes in life?

I think college is overrated. By all means, go to college if you’d like to learn at the highest level academically. But also consider not going to college and achieving great things without burning 4-5 years of opportunity cost.

[dahart]

> what about causality?

A bunch of researchers have asked that question and tried to answer it the best they can, by using statistical methods to discount for things like high family income, race, family history of college attendance.

From what I’ve seen (and I’ll try to dig up links in a minute if I can) papers often conclude that the causality is definitely not 100% responsible for the income difference between grads and non-grads, but they also generally find the number is much higher than 0% too. It’s somewhere in the middle.

It’s not really a stretch to suggest that something happens during college, right? People do learn things when they spend their time learning, and we have plenty of data to show that people who get an elementary and secondary education are better off than people who get no education… or take it a step further and look at basic literacy. Why is it hard to believe there’s value in college when it’s not hard to believe that literacy and pre-college education has value?

> consider not going to college and achieving great things without burning 4-5 years of opportunity cost

While that can be true for a few naturally motivated and smart (or lucky) people, the aggregate stats don’t really support that view. Most people end up better off going to school, and opening the door to the many higher paying jobs that require a degree.

* edit: here’s an example: https://www.stlouisfed.org/-/media/project/frbstl/stlouisfed...

See “Section III: “The True (Causal) Return on a College Education: Evidence from the 2016 SCF”

The conclusion: “In order to examine the effect of these variables in accounting for some of the relationship between education, income and wealth, we utilized multiple regression. All variables, including age variables, own education, parents’ education, and financial acumen, were regressed onto income and wealth. This model was compared to the simple model of only lifecycle and own education. Results are in Tables 6 and 7. (See Tables B1-B6 for black, Hispanic and other race results).

“Clearly, parents’ education and financial acumen were important variables previously omitted in estimations of the college and post-graduate premiums (see Table 8). Together, these variables reduced the income premium by 32 percent for white terminal bachelor’s degree holders and by 29 percent for white postgraduate degree holders. The reductions of the wealth premium were even starker, with this premium being reduced by over half for graduates and postgraduates (54.4 percent and 60.4 percent, respectively).”

[than3]

Well none of those studies take into account US-style weed-out classes where there may be a 12% pass rate for a bottleneck transfer class, and the class is structured to fail students.

Any value-based research that neglects these is just junk science.

[dahart]

That’s not true. The study I cited does account for your pet-peeve example, because they’re treating university as a black box and examining the outcomes. This captures all the weed-out classes along with everything else that happens in college.

If the weed-out classes were having a statistical effect, then we’d see it in the outcomes. So either we are seeing it, and the success rate of a college degree would be even higher without the weed-out classes, or your contrived example in reality has little to no effect on the outcomes. Either way, your example might disenfranchise a few people who aren’t prepared or capable of handling college at all, but on the whole most people fare well with the education. Maybe the weeder classes don’t matter because people try again, or switch majors, or continue without transferring. Or maybe the number of people hitting such a weeder class is low, or maybe the 12% pass rate is very exaggerated and not at all representative of the average pass rate?

[than3]

> That's not true. This captures all the weed-out classes.

No it doesn't. You are wrong. There are a number of well understood ways where you would not see deviations like this in the statistical outcomes. Flawed, skewed, or under-representing metrics being just one, as I mentioned in another post. Garbage in Garbage out.

If you compare the success rate of US college against the EU, there are big differences but this is apples to oranges comparison with arguably the same outcomes, but there is no reasonable way to discount or map certain differences in implementation. EU has generally a higher pass rate for less investment.

I fared well with education in every class except the structured to fail classes. What does it say when you have a student who passes Calculus 3 and other math classes above with a high 90, and can't pass Mechanics that solely tests Math in 3 question, two test format with causality properties between the questions on each test. Where the accompanying required lab scores for the same person are top 90s, even scored top of all pooled lab classes for an egg drop design contest, but can't pass the test. The only egg drop among all classes at the college to survive 4 stories without a crack with paper, water, and a baggy (by design at my insistence). Far from a contrived example when you have the transcripts to back it up.

While not necessarily a good example because probabilities have poor results in relation to likelihood in reality, it still demonstrates the margin for error, the likelihood of a passing score assuming the material was taught is roughly .3^6 power for 2 exams. You need to get a 70 in the class to pass, so you could only get 1 question wrong for the entire class between both 2 tests, which means you could only get the final question wrong on either of those two tests if the test had causality properties. That doesn't seem so bad, but in reality each one of those questions would be 10-20 steps each to get the correct answer for one question. So that probability shrinks further to .3^(6*120). You have perfection, or you don't pass.

Honestly it just sounds like you are trying to make excuses and minimize my experience, like most people do because you want to believe that something is true (even when it isn't). In my 'experience', which spans 15+ years in education, alongside many others; that is not the case. It was a lie, a pipe-dream sold with lies that had nothing to do with actual merit.

Its a drain on financial resources with moving goal posts with no determinable way to pass for those classes that are structured to fail, and those courses are bottleneck courses needed for GE and transfer.

In learning system's theory, you learn about the specific requirements needed for determinism. There's only 3 or 4 fundamental ways you can structure exam questions where you can legitimately have 1 correct answer. Most of those test questions in those structured to fail classes, fail those specific requirements needed for a determinable answer. Worse they optimize for the least work, and max fail rate (for repeat customers)

If a question on an exam doesn't have a determinable answer... it is testing how well you guess which is to say its not testing anything at all. It is using an unsound test to keep your money and require you to take the course again. That's fraud.

Many of those exam's questions are simply cleverly disguised guesses; with 3 or more different possible correct answers.The point of testing is with the instruction from class, there is only one right answer.

Anything else is fraud; but they are state-funded institutions, and you have no due process because there is no obligation to act on legitimate complaints and those institutions are largely shielded legally. It suffers all the classic bureaucratic failures.

[dahart]

I don’t know what you’re talking about, but you’ve said next to nothing that is verifiable and haven’t included any links to references or stats in your rant, so with nothing verifiable, you’ve left me with no choice but to mostly ignore the comment. FWIW, using words like ‘lies’ and ‘fraud’ and ‘junk science’ are all red flags to me and make it sound like you have more of a bone to pick than an objective viewpoint. I have no idea what your experience is, it’s not possible for me to minimize it intentionally, and I’m sincerely sorry you feel that way as it’s not my intention, but it should be as obvious to you as it is to me that your experience could have deviated from the norm, maybe you were unlucky. In any case I’m left completely unconvinced by your argument that there’s anything missing from the Fed’s statistics relating to weeder classes that affects the broad outcomes of students. Even part of your explanation, under-represented metrics, is exactly what I suggested as well. Maybe the weeder classes aren’t experienced by most people, and therefore aren’t detrimental to the overall outcomes, in average. Or- and your wall of text failed to consider this possibility- that weeder classes are affecting outcomes negatively, and the outcomes would be even better without such classes. It’s possible that you’re right and the statistics are also right.

[dahart]

> this is the oldest trick in the book: talk about the average when it tells the story you want to tell as opposed to the distribution.

That’s a little bit of FUD though; use of an average does not imply it’s being misused. All statistics can be misused when not cross-checked with other statistics, but just because it can be does not imply it is. And to be fair, citing emotional anecdotes and failing to look at the average or any other stats seems to be abused in reporting a lot more often than trying to be tricky with numbers.

The St. Louis Fed published a paper a few years ago [1] showing the distributions of earnings of college degree holders compared to non-degree-holders (called the “income premium”), and it backs up the story of the average here. The average graduate in the U.S. earns approximately double what people that stop at a high school education earn. I was very surprised by that stat, I had no idea the income premium was that big. Even though this Fed paper is arguing that the relative savings (“wealth premium”) is going down for college grads, it’s also pointing out that the income premium is not really going down, and on top of that they’re a little quiet about the absolute number of dollars people have in savings. When grads and non-grads have the same savings rate but one has twice the income, then that one also has twice the money at the end of the day, and the Fed paper argues that’s actually a break-even. (This is another old statistics trick…)

[1] https://research.stlouisfed.org/publications/review/2019/10/...

[theGnuMe]

Ok so talk about the median that should be sufficient. I think average is appropriate here because the outliers are rare. Not many people rack up 200k in debt with nothing to show for it. Apart from extremely pricey art institute degrees (whose absolute numbers are low) I don't see an issue.

Median debt is $20k. Avg debt is $37k. The college wage premium is about 30k. A new car is $45k.

[lapcat]

> outliers are rare. Not many people rack up 200k in debt with nothing to show for it.

Citation needed. Numerically, what does "rare" and "Not many" mean?

[AwaAwa]

If you need to speak about averages, doesn't that imply a net loss? Unless the average is overwhelmingly made up of folk with college degrees, which I don't believe holds water.