We thank Andrew Gillen for his thoughtful comments on our recent paper. We have two reactions that highlight where we concur and where we don’t. First, we agree that structural modeling is crucial in helping analysts and policymakers alike understand the complex ways that federal policy affects tuition in the diverse American higher education system. But we disagree with his contention that our model’s foundational assumptions make it unsuitable for analysis of the Bennett Hypothesis. Secondly, Gillen argues that the “revenue theory” is a better alternative to our enrollment management approach. In our view, the revenue theory suffers from serious methodological flaws. We will describe a few of them in this short reply.

Gillen claims that our enrollment management model is flawed because “to explain the empirical evidence, [our] model needs to violate one of its foundational assumptions.” This is not correct. Our foundational assumption is that “the size of the institution is set.” This is a standard assumption in microeconomics when we explain short-run behavior of firms. The assumption about size implies a given physical plant, a certain faculty, and a certain amount of expenditures to deliver programming. The institution uses historical information from past applicant classes and information from the current applicant pool to attract the highest-quality class that meets its goals for number of students and tuition revenue per student. Discounting is the key tool. Some schools use it primarily to meet revenue needs while more elite programs use it to sculpt the highest-quality incoming class.

If there is a change in federal financial aid policy, we assume that the institution does not immediately change its size. We are still doing short-run microeconomics. The change in policy means that the institution’s revenue options have changed. Our model points out the many ways an institution may react to its new opportunities. Indeed, one possibility is that it will earn more revenue from its students. When we point out this possibility, we are not violating our foundational assumptions. The institution’s initial assessment of the revenue it needs from its students was conditioned on its size, its programming, the quality of the students it hoped to attract, and the amount of tuition discounting required to hit its initial objectives.

The change in federal financial aid policy affects the amount of tuition discounting that is required to hit these initial objectives, so it is perfectly sensible for the institution to adjust the amount of revenue it extracts from its students. As we suggested, the institution might use this golden opportunity to increase the quality of its programming by spending more per student. But it also might choose to pass the increased subsidy on to poorer students as a lower net price.

It is a fair criticism of our approach that we do not fully model what each category of college or university will do. In our defense, we were writing down a descriptive account of an abstract school’s options, not a mathematical model for professional economists. It’s also fair to note that we do not fully describe the process by which the institution sets the quality of its programing, its long-run size, or its aspirations for student talent, all of which determine its initial revenue needs. These are tasks for another day and a more technical approach.

With regard to for-profit schools, Gillen thinks it bizarre that they might see increased profits from more generous federal policy. He believes that competition between schools would eliminate the rents they could collect. If higher education is highly competitive and the number of seats available to students expands elastically with demand, then the Bennett Hypothesis makes no sense for any type of school. We would note that in the large nonprofit sector there is good evidence that the number of seats has indeed grown rapidly as demand has surged. But there is no obvious reason to presume the industry is perfectly competitive. Many for-profits likely have local market power that could allow them to push up price in the event that the federal subsidy grows. The empirical work on the Bennett Hypothesis that uses some structural model of university behavior (at nonprofits and for-profits alike) usually presumes that schools can exercise some form of market power.

Local market power may be eroded in a fully online world, and fully online training is growing fastest in the for-profit sector. But 60% of students at for-profit institutions currently experience a fully face-to-face education. And many fully online programs deliberately keep classes small to ensure some degree of personalization and brand differentiation. Lastly, the “90–10” rule, requiring that 10% of school revenues come from sources other than federal funds, is not likely the driver of the Bennett Effect at for-profits, since most for-profits are distant enough from the 90% threshold that they would not be forced to increase tuition just to remain in compliance with the federal rules.

Gillen’s suggestion that Howard Bowen’s revenue theory of cost would provide a better model is not at all appealing to us. In fact, Gillen’s own discussion of the revenue theory illustrates how it is internally inconsistent. One of Bowen’s Laws is “Each institution raises all the money it can.” Yet Gillen says, “But their goal is not revenue maximization.” In other words, to use this theory, one has to violate one of its laws.

More generally the revenue theory of costs is based on a tautology. All nonprofit institutions spend as much revenue as they earn. We tested and critiqued the revenue theory in earlier work (“Explaining Increases in Higher Education Costs,” Journal of Higher Education, Vol. 79 [May/​June 2008]). In brief, the revenue theory is a higher education specific theory. It claims that the driving factor shaping the long-term evolution of higher education costs is higher education revenue. Yet many other industries display a very similar evolution of cost over time. If higher education revenues are the force driving higher education costs, then it must be a coincidence that costs in many other industries have behaved so similarly. In our view, features the industries share in common likely drive the similar behavior of cost. In our last book on college cost, Why Does College Cost So Much?(Oxford University Press, 2011), we show that the industries whose price growth looks most similar to higher education tend to be personal services, not goods, and most personal services have experienced very slow labor productivity growth.

The tendency for service prices to grow more rapidly than goods prices is well known. It can be traced back to the work of David Ricardo in the 19th century, and in higher education it’s called Baumol’s cost “disease.” (See William Baumol and Sue Anne Blackman, “How to Think About Rising College Cost,” Planning for Higher Education, Vol. 23 [Summer 1995]). Any industry that experiences slower productivity growth than the national average is likely to experience higher than average price growth because that industry still must pay national wage rates for its labor, but without the cost-reducing benefit of high productivity growth.

Second, in many personal services the workforce is highly educated compared to the national average. This is quite true in higher education. When the premium on educated workers is growing, all industries with a highly educated labor force face extra cost pressures. The wage premium on college-educated labor has grown steadily since the early 1980s. Lastly, many personal service industries do not just adopt new technology when it lowers cost. They are driven to adopt new techniques because their service must evolve to meet new needs, even if this raises cost. In higher education, much like in medicine, an evolving standard of care drives the adoption of new and often expensive technologies. These three broad economic forces that have reshaped the global economy offer a story for why higher education costs have risen more rapidly than the overall inflation rate.