Skip to main content

Are graduation rates an input or an output? (redux)

This is a refresh of a popular post I've done a few times, asking the important question in the title.

People tend to think of graduation rates as an output of the institution, and of course, in some sense, they are; they are certainly measured that way.  But what if I told you that a college's six-year graduation rate (and, to a lesser extent, its four-year graduation rate) can be easily predicted by a single variable that we know before a student ever sets foot on campus?

I bet you'd want to see the proof.

The variable is the mean SAT (or ACT) score of the incoming class of first-year students.  AHA! some of you are saying.  How can you be test-optional then?

Relax.  The SAT and ACT are nice proxies for all the things that help determine who is going to graduate from college in America.  In one nice little number, it encapsulates several of those factors: Academic preparation (remember, the scores are good proxies for high school GPA, the best predictor of student performance in college), parental wealth, parental educational attainment, and race or ethnicity.

The tests don't tell us much beyond the high school GPA, and they don't help predict how well an individual student will do in college, but when you get a large pool of students together and average the results and all they encapsulate, they work quite well. It's the exact same reason your stock portfolio will avoid risk if you have 30 stocks in it.  (This is not financial advice, of course, and you should only invest in stocks after getting advice from a professional.)

Here is the data presented in five views:

The first two show SAT and then ACT means of every institution arrayed on the x-axis, and the six-year graduation rate on the y-axis.  Hover over the regression lines to get the r-squared values and ask your friendly social scientist if that's telling.  But don't stop there.  Use the filters to see if you can get it better.  Look at percent admitted, or percent with financial aid; you might suspect that the highly rejective colleges who can choose whom they admit, or the ones with fewer students on aid, might cluster together.  Do they?  Yes, they do.  Or use the other filters if you wish, in any combination. 

The third view is for the skeptics who continue to think the SAT and ACT are different tests.

The fourth view shows each institution's four- and six-year rates plotted against each other.  And the fifth view shows the percentage of students who graduate who managed to do so in four years.  Some programs (engineering, or accounting, for instance) will often be five-year programs by design, so lower rates at institutions with these types of programs will usually show up lower on this criterion.

A note: I used 2019 mean SAT and ACT scores because that is the last "normal" year prior to COVID.  While the 2020 entering class was admitted with test scores at most institutions, these numbers show enrolling students, and we know who enrolled was greatly affected by the pandemic.

If you use this in your for-profit work, you can help defray my costs of software and hosting by buying me a coffee (or beer) by clicking here.  If you're an interested parent or you work for a not-for-profit, just skip right to the visualization, below.

And as always, let me know if you see anything interesting or important (but don't ask me about Northeastern's four-year graduation rate...I just use what IPEDS publishes.)


Comments

Popular posts from this blog

Changes in AP Scores, 2022 to 2024

Used to be, with a little work, you could download very detailed data on AP results from the College Board website: For every state, and for every course, you could see performance by ethnicity.  And, if you wanted to dig really deep, you could break out details by private and public schools, and by grade level.  I used to publish the data every couple of years. Those days are gone.  The transparency The College Board touts as a value seems to have its limits, and I understand this to some extent: Racists loved to twist the data using single-factor analysis, and that's not good for a company who is trying to make business inroads with under-represented communities as they cloak their pursuit of revenue as an altruistic push toward access. They still publish data, but as I wrote about in my last post , it's far less detailed; what's more, what is easily accessible is fairly sterile, and what's more detailed seems to be structured in a way that suggests the company doesn&

The Highly Rejective Colleges

If you're not following Akil Bello on Twitter, you should be.  His timeline is filled with great insights about standardized testing, and he takes great effort to point out racism (both subtle and not-so-subtle) in higher education, all while throwing in references to the Knicks and his daughter Enid, making the experience interesting, compelling, and sometimes, fun. Recently, he created the term " highly rejective colleges " as a more apt description for what are otherwise called "highly selective colleges."  As I've said before, a college that admits 15% of applicants really has a rejections office, not an admissions office.  The term appears to have taken off on Twitter, and I hope it will stick. So I took a look at the highly rejectives (really, that's all I'm going to call them from now on) and found some interesting patterns in the data. Take a look:  The 1,132 four-year, private colleges and universities with admissions data in IPEDS are incl

Freshman Migration, 1986 to 2020

(Note: I discovered that in IPEDS, Penn State Main Campus now reports with "The Pennsylvania State University" as one system.  So when you'd look at things over time, Penn State would have data until 2018, and then The Penn....etc would show up in 2020.  I found out Penn State main campus still reports its own data on the website, so I went there, and edited the IPEDS data by hand.  So if you noticed that error, it should be corrected now, but I'm not sure what I'll do in years going forward.) Freshman migration to and from the states is always a favorite visualization of mine, both because I find it a compelling and interesting topic, and because I had a few breakthroughs with calculated variables the first time I tried to do it. If you're a loyal reader, you know what this shows: The number of freshman and their movement between the states.  And if you're a loyal viewer and you use this for your work in your business, please consider supporting the costs