Studies of Financial Aid’s Effect on Tuition are Rorschach Tests
A December 2001 report by the National Center for Education Statistics found no
associations between financial aid and tuition prices… dispelling the myth that increases in federal and state aid result in increases in college prices…
With all due respect to Mark, who is and should be the go-to guy for all things financial aid related, this is nonsense.
Big picture: the story that emerges from the literature is that the evidence is mixed. The Bennett Hypothesis is neither confirmed nor denied (see the links here for examples)
As I wrote before:
the original Bennett Hypothesis does not hold, but those that repeatedly declare this often ignore all the fragments of evidence that there is some truth to the idea… The appropriate response when faced with mixed evidence is not to declare the idea dead and buried, but to explore more deeply both the idea and alternatives to try and figure out why the evidence is mixed.
To be fair to Mark, many of the studies on this issue came out after his paper. Moreover, I am only highlighting his statement to illustrate how widespread this view is. Which brings us to the smaller picture – no one should be drawing sweeping conclusions from the 2001 ED study, for 3 reasons.
First, the authors themselves were aware of the caveats to /problems with their study, and were repeatedly very cautious about not over generalizing.
In the 1998 mandate, Congress asked NCES to explore how institutional financial aid affects price increases, and the extent to which federal financial aid is involved in this relationship… Although the analysis presented in this chapter cannot fully answer the questions posed by Congress (emphasis added), it contributes to the knowledge framing the debate…
Second, they had one year of financial aid data. They essentially looked at whether that one year’s worth of financial aid data was correlated with changes in tuition (did schools that increased tuition more have larger financial aid). This is a bit like trying to write a book review when all you have is the dust jacket. Again, the authors were aware of this, and warned readers:
“the change in financial aid, if available, may be a better variable to include than financial aid in a given year”
Third, they used a backward elimination method to select variables for the regression. This approach basically involves throwing the kitchen sink at the problem and seeing what sticks. Such data mining does have its uses, but the results are generally more of a guide of what to look into rather than a method for settling an issue once and for all. One of the reasons for that is that you can get some odd results – for instance the authors find that:
None of the variables remained… for private not-for-profit research/doctoral institutions…
If we take these results at face value, tuition at private not-for-profit research/doctoral institutions is apparently randomly determined.
Again the authors were aware of this:
these regression models allow a preliminary exploration (emphasis added) of the relationship between changes in tuition for undergraduate students and student financial aid…
All of this jumps out at readers, which leaves me baffled about why so many folks keep thinking that this study is a nail in the coffin of the Bennett Hypothesis.
All of which brings me to my hypothesis of the day. Most people made up their minds when it comes to the Bennett Hypothesis years ago, and the mixed evidence that we keep producing won’t convince anyone to convert from one side to the other. These studies are Rorschach tests where people see what they want to see. And given the mixed evidence we generally find, everyone will continue to be able to find something to latch onto to justify their preexisting position. In retrospect, I should have been clued into this phenomenon when many, including the NASFAA, tried to critique my study without reading and/or understanding it.
By the way, my study offers a refined Bennett Hypothesis 2.0, which could provide an explanation for why we keep finding mixed evidence.
The claim in this report (that financial aid often leads to higher spending per student), while similar to the Bennett Hypothesis, is different from it in two important respects. First, the focus here is on the interaction between aid and spending, rather than aid and tuition. The next step, higher spending leading to higher tuition, is a perfectly logical result, but there could be other complicating factors, such as a cap on tuition rates (or their growth) imposed by state legislatures, which prevent this from occurring.
The second way in which this report differs from the Bennett Hypothesis is that it is explicit about when the effect occurs (and the types of aid likely to suffer from it). Specifically, aid will fuel increases in spending when it is given to students whose ability and willingness to pay is in excess of current costs at the school. Because costs and ability to pay vary by school, this implies that a much more nuanced view is warranted. The same aid program can have different effects based on the characteristics of the school and the students attending…
Thus, lumping all federal aid together when analyzing its impact, or even all aid of a given type, is unlikely to yield accurate results.