I asked Google Gemini to take top 100 small lib arts colleges from US News and plot avg net cost vs. acceptance rates. It first balked saying it’s too much work (lazy bones!), so I told it to leave out schools w/ acceptance <10% or cost <$15k (just one, Berea). Check it out:
Is it just me or is there a U-shaped curve/normal distribution? The values aren’t totally accurate, I’m sure… like, Reed is too cheap. There must be others. But the pattern is interesting. Thoughts? Comments? Obvious errors?
My first question is– what do you think this data shows?
And my follow-ups will be based on that. I think one flaw in the methodology is that no individual student cares about the average net price. The student cares about what his or her actual cost will be. The fact that a Pell grant recipient gets more money than you do from a school with generous need based is– just a fact. That student has greater need then you do. So the ticket price is lower. This is not news.
What is the source of the net price data? If it’s coming from IPEDs, note that the net price includes only students who receive federal and/or institutional financial aid.
So, full pay families are not included…meaning average net price for the vast majority of these schools is higher, maybe much higher. We know at some/many of these private schools more than half the students are full pay.
At a guess, two different effects are being overlaid here.
Understanding we are looking specifically at students getting aid, we are looking at different average prices for students with aid, and in a relatively narrow range overall. Like whether you are looking at, say, $27500 in net price or $32500 in net price, in both cases the student isn’t likely paying close to full price.
OK, and then my hypothesis for the right side of this graph is that you may be looking at colleges that feel the need to compete just a bit more on price, and that is (very roughly) correlating with higher acceptance rates, since higher acceptance rates tend to correlate with: (a) lower yields, which is a measure of how much competition a college is facing for the students it wants to enroll; and (b) also lower application volumes, which is a sort of general popularity measure. So, a little less popular in terms of applications, and facing tougher competition for the students you want, you might try to compete more aggressively on price.
Then on the left side, I think you may just be looking at colleges with more of a budget for need aid, which: (a) may more directly correlate with higher applications and higher yields and so lower acceptance rates, because of course students like more generous aid and lower prices; and (b) may indirectly correlate with other valued uses of ample institutional resources which then also correlate with higher applications/yields and lower acceptance rates.
I think your theory is fine at 20,000 feet but somewhat ignores the numerator/denominator problem of “selectivity”. And if there’s anything that all of us on CC have learned over the years, it’s that for many kids, applying to schools that OTHER kids in their high schools are applying to (and presumably have deemed “worthwhile”) is very, very important.
So this is counter intuitive of course. Avoiding “bunching” is an extremely powerful tool for boarding school and high end prep school college counselors, but the typical kid at a large HS just shows the counselor the list, gets the nod, goes off without too much dialogue.
So at the end of the day, I think average net price and selectivity are less linked than most people think, even accepting for a moment that some colleges need to work harder (and discount more) to fill the seats. Numerator, meet denominator. Selectivity as a percentage is driven by both, only one of which is stable enough to be predictive.
I remember the year Denison was ‘“discovered” in my part of the country, which is funny because it was a year after Lehigh was “discovered” and only a few years after Franklin and Marshall (hiding in plain sight in Amish country, I guess) was dicovered.
A U curve is not the same as normal distribution. The first is plotted on a scattergram, while the latter is plotted on a histogram. This is a scattergram, so it cannot show a normal curve.
I don’t disagree, but I also don’t think this data “disagrees” either. This scatter plot is VERY noisy. So even if we did fit some sort of curve to it, I am sure the statistical analysis would say we are still missing a lot of the factors that actually go into determining both acceptance rate and average net price.
As is often the case with social science, identifying factors that may play some role in determining something is not the hard part. Identifying ALL those factors, and then understanding how they are interacting, is the much harder part.
And honestly, if you are choosing colleges, all this modeling of institutional practices is of little value anyway. What you ultimately need to know is what price they will charge you. Why they are doing that is at best a very distant consideration.
Re obvious errors. If you simply asked Gemini about USNWR, be aware that Spelman College, Bard College, Colorado College and Reed College do not provide USNWR with data, so I am guessing that’s going to skew your graph.
Can you please clarify what your goal is with this post? Are you trying to see if a more highly ranked college is worth the cost? Or if any “top 100” LAC is worth its pricetag?
I will also tag @Data10 who might have some insight on your gemini result. Perhaps he can tweak your result to give you some more meaningful overview if you provide more context.
First off. Pull some schools at random and check the information given. Otherwise, none of this matters but fun trial.
I am currently vibe coding a student app for applying to college (more on that at another time). What I am finding out is the source and accuracy of the information given. I am like over 90% done but now need to “talk “to the software to isolate more where it is getting its information. I took 21 colleges somebody I am helping applying to college as a start. Most of the information is correct. I wanted just to use the colleges websites for accuracy but for some coding reasons some can’t do that. Now I am going to ask to use the CDS and compare and update to this year. Before it was only giving me information from 2024 … I am learning but my point is. Make sure the data is actually correct.
As an aside. My daughter lives a few blocks from Reed College. We are going for a week to visit her (Getting her Masters in SLP on Wednesday). The walking Caynon there is lovely and will definitely walk through that.
… I am learning but my point is. Make sure the data is actually correct.