Monday, October 19, 2009

Why Do Private Schools Have Higher Graduation Rates?

By Richard Vedder

It is well known that the graduation rates of students from private colleges and universities is significantly higher than is the case at public institutions. Of course, a huge part of the reason is that the private schools, on average, get better students --higher SAT scores, higher rank in their high school class, etc. Yet even when these factors are controlled for, usually researchers find that graduation rates are still higher at private institutions?

Why? My own hunch is that the differences between "public" and "private" universities with respect to the way they operate and do business are relatively small. I have not found a huge difference between the private schools with which I have been associated (as a student and parent, Northwestern, as a professor, Claremont McKenna College and Washington University in St. Louis) and the public ones (University of Illinois, University of Colorado, Ohio University), except in one vital respect --personal attention to student needs.

On average, in schools which are more tuition-intensive in terms of revenue sources, there is a sense that the student is indeed a paying customer who needs to be treated with respect, kindness, and even affection. I sense that, on average, advising students is taken a bit more seriously at private schools, especially liberal arts colleges. Many students get lost between the cracks, and some personal attention from a relatively senior faculty member can often get them back on track. I am working more closely than usual with a student who has failed to earn his degree, helping him through the hoops of completing his academic work in a satisfactory fashion. But it takes a bit of time and effort on my part, and many professors don't want to make the effort --publishing articles is much more rewarding in a pecuniary and career sense. When I visit a good liberal arts college (Kenyon College comes immediately to mind), I feel that the faculty have a greater sense that their duties include attentiveness to student needs, come of which extend beyond the narrow classroom experience.

Jonathan Leirer and I are playing around with statistical models trying to explain variations in graduation rates. We do find, other things equal, that student performance varies positively with tuition levels, consistent with the discussion above. Clearly, as Bowen, Chingos and McPherson argue at length in their new book, socioeconomic status matters also in any discussion of graduation rates --the more Pell Grant oriented a school is, the lower the graduation rate (especially the four year rate). But I have seen some results that hint that high amounts of research activity, other things equal, may lower graduation rates, as research activity crowds out the advising of students.

The incentive systems are all out of whack. Publication in obscure journals on topics of tangential interest to the human race that few even bother to read can reap significant rewards --worth thousands of dollars per article. Time spent helping a struggling undergraduate yields nothing in the way of rewards-- or even less, denial of tenure for example, if student concern leads to publication rates below the accepted minimum.

Here would be an interesting proposition, the kind Jeff Sandefer might enact at his Acton School of Business. Set up a large bonus pool that goes to faculty members in each department if, and only if, the department raises the three year moving average of the graduation rate over time upward, subject to the constraint that the department's accumulative grade point average cannot move up at all (indeed, a second bonus should be awarded if higher graduation rates are achieved with falling average grades). This, plus coupling salary increases more to teaching performance and less to research, could lead to improved student outcomes. Will it happen? Not without some fundamental change in the way the academic game is played in the U.S.

5 comments:

capeman said...

So the Doc has a hunch about why Washington U., Claremont, Kenyon etc. have higher graduation rates.

I have a hunch, too. Give me students with SAT scores like those places and I'll see higher graduation rates too.

Oh, SAT scores don't account for everything. (How much is the difference?)

I have another hunch. Look at the expenditure per student at those places, I'm not talking research, football, etc. but real expenditure per student.

You'll find it's 3-4 times what it is at the public university where I work.

That will buy a lot of handholding and personal attention to get Johnny through, at Washington U. there are about 4 times as many faculty per student as where I toil.

Plus, the students at those expensive schools probably don't have to worry about their finances so much. Not at a place like Washington U. with its $4 billion endowment.

Really, the Doc is outdoing himself with his shoddy "analysis" and recklessness.

This is pathetic.

Jonathan L. said...

@Capeman: While we are still in the preliminary stage of the analysis, I'll share with you some results. Using a sample of 600 schools (the same 600 used for the Forbes/CCAP rankings) we ran a regression model looking at four year graduation rates as a function of the acceptance rate, the admissions yield, the percentage of students receiving pell grants, the bottom 25th percentile SAT score, in state tuition (in thousands), undergrad enrollment (in thousands). The results:
coef.est coef.se
(Intercept) 18.56 8.56
undergrad.enroll.0607.000 -0.39 0.08
in.tuition.000 0.68 0.07
accept.rate -0.16 0.03
admit.yield -0.02 0.04
pell.perc -0.43 0.05
sat.25th 0.04 0.01
---
n = 600, k = 7
residual sd = 11.07, R-Squared = 0.75

through in a public dummy and you get this:
coef.est coef.se
(Intercept) 27.61 8.42
public.dummy -14.70 2.34
undergrad.enroll.0607.000 -0.19 0.08
in.tuition.000 0.15 0.11
accept.rate -0.18 0.03
admit.yield -0.04 0.03
pell.perc -0.46 0.05
sat.25th 0.05 0.01
---
n = 600, k = 8
residual sd = 10.72, R-Squared = 0.77

run them separately on subsets of only public and only private

lm(formula = grad.form.all, subset = public)
coef.est coef.se
(Intercept) -24.74 14.37
undergrad.enroll.0607.000 -0.04 0.09
in.tuition.000 1.64 0.35
accept.rate -0.28 0.05
admit.yield 0.02 0.05
pell.perc -0.31 0.07
sat.25th 0.07 0.01
---
n = 217, k = 7
residual sd = 10.54, R-Squared = 0.67

lm(formula = grad.form.all, subset = private)
coef.est coef.se
(Intercept) 39.88 9.96
undergrad.enroll.0607.000 -0.80 0.15
in.tuition.000 0.24 0.12
accept.rate -0.17 0.04
admit.yield 0.02 0.04
pell.perc -0.50 0.06
sat.25th 0.03 0.01
---
n = 383, k = 7
residual sd = 9.90, R-Squared = 0.65

The graduation rate variable is in whole numbers (that is, 10% is 10). So based on the regression using all schools and a public dummy, the sat.25th coefficient of .05 says that increasing the SAT score of the bottom 25th percentile by 200 points is associated with an increase in graduation rates by 10 percent.

Granted, we are limited by the availability of data, and these results are preliminary, but they are based on a sample of 600 (rather than one), which lends some credibility.

To try to address your spending question, lets throw in instructional spending per full time equivalent student (in thousands)

coef.est coef.se
(Intercept) 29.59 8.59
public.dummy -14.80 2.34
undergrad.enroll.0607.000 -0.19 0.08
in.tuition.000 0.14 0.11
accept.rate -0.18 0.03
admit.yield -0.05 0.03
pell.perc -0.46 0.05
sat.25th 0.05 0.01
instruction.per.fte.000 0.03 0.06
---
n = 597, k = 9
residual sd = 10.70, R-Squared = 0.77

The spending variable doesn't seem to work. Still, there are a lot of avenues to explore and I, for one, appreciate that someone is there to keep us on our toes.

capeman said...

I'm glad you're trying to run serious regression analysis, but when you compare Washington U. and Kenyon and Claremont to low-funded state U., you make yoursevles sound ridiculous, with an ax to grind.

But since you are at least somewhat willing to listen, I'm graciously willing to help you out.

Another thing to consider: the graduation rates only consider the specific institution, as far as I am aware; this is a major weakness of all graduation rate comparisons.

If I drop out of Harvarad and finish at Princeton, I'm counted as a graduation "failure".

Not too common, I'm sure, but much more in state systems -- I suspect, not having much data.

e.g. I start at Iowa, finish at Iowa State. All within the state system. Sounds like a success to me. But it's counted as a "failure" at Iowa.

Unless you can come up with means to take this kind of stuff into account, your (and others') studies of this won't mean too much.

Jonathan L. said...

@capeman: You have a valid point with the transfer problem. I am working on finding a way to address it (it appears the delta cost project has made some progress which I may duplicate). However, according to a study released in 2002 by the NCES, Descriptive Summary of 1995-96 Beginning Postsecondary Students: Six Years Later , after six years, 23% of students who begin at a 4-yr institution intending to earn a BA transferred to another institution, but only 30% of those eventually earn a BA, which would add 7.4% to the graduation rate. (table 10 pg 23). Note that the percentages are rounded, so even though 30% of 23 is 6.9, the actually addition to average 6-yr graduation rates is 7.4%.

While 7.4% won't explain everything, it is a significant, and I suspect that while the average change would be 7.4%, there is much variability from school to school.

We will, of course, keep you updated on the progress of our dubious and insidious research agenda.

capeman said...

Glad you recognize the problem.

Here's another thing: you say you looked at

"instructional spending per full time equivalent student"

This is important, but not the whole story in greater financial resources of many private schools (such as the ones you mentioned, Washington U etc).

There's also "academic support", "student services", "scholarships" etc. which can contribute heavily to students persisting to graduation. The first two include things like counseling for basket cases, hand-holding, extracurricular activities to keep students interested, including extras like dorm programs, etc. etc.

Another thing: the fields of study matter a lot, too. Look at sci/tech schools like Caltech, Harvey Mudd, others. They tend to have low graduation rates, compared to other schools with comparable SAT's, resources, etc. It's not clear this represents "failure". Maybe it's just damn hard to get a degree in electrical engineering or physics or biochem.

One last thing: the grade inflation has been worse at private schools. Could this be related to their need to suck up to the students, long enough to get them to a degree, to keep the money rolling in?

I've seen this at second-tier expensive private schools. When you're charging $40K for something that is basically available for $7K or so at a good state school, it's pretty hard to bite the hand ...

In other words, it's not clear that a higher graduation rate is better. It all depends.