CHAPTER 1

Part 1

Costs and Productivityin Higher Education

AS MY WIFE keeps reminding me, I have a Don Quixote–liketendency to fl ail away at windmills—to take on topics suchas race in America and affirmative action; the insidious problemswith college sports at all levels, including Division IIIand the Ivy League (which cause me to cringe whenever theNCAA refers to its legions of "student-athletes"); and, yes,the un forgiving economics of labor-intensive industries, suchas the performing arts and higher education. But, my DNAis what it is, and so I am now adding to this list the potentialimplications of online learning for college costs.

Context matters, and I will begin by outlining as succinctlyas I can aspects of the economics of higher education that arerelevant to my topic:

• trends in costs, the "cost disease," and how to think aboutchanges in productivity;

• other forces, some deeply ingrained in the fabric of highereducation, that also push up costs; and

• growing worries about affordability, especially in the publicsector, where reductions in public support have beencoupled with significant increases in tuition.

Then, in the second part of the book, I will discuss what Ithink—or, better said, what I suspect—about the potentialimpact of the variety of approaches to online learning that areeverywhere present, including, of course, at Stanford Universityand at Stanford spin-offs such as Coursera and Udacity. Isthere, as President Hennessy has suggested, a tsunami of somestill ill-defined kind coming? Is it realistic to imagine thatonline learning is a "fix" (at least in part) for the cost disease?Throughout, I will maintain a system-wide perspective, since itwill not do to think about these large questions solely from theperspective of individual institutions.

Cost Trends, the "Cost Disease," andProductivity in Higher Education

It is fitting that I gave these Tanner Lectures in close proximityto Clark Kerr's neighborhood, since it was President Kerr, inhis capacity as chairman of the Carnegie Commission on theFuture of Higher Education, who commissioned a study ofmine in the mid-1960s that became The Economics of theMajor Private Universities. In that study I documented theseemingly inexorable tendency for institutional cost per student(which is, of course, different from tuition charges) to risefaster than costs in general over the long term. Kerr christenedthis finding Bowen's Law, although he was, he said, "originallyskeptical about it."

What is important today is not the exact numbers containedin that study (which were based largely on a detailed examinationof the experiences of the University of Chicago, PrincetonUniversity, and Vanderbilt University between 1905 and 1966)but the underlying pattern, which has been found to hold forpublic as well as private universities, and for colleges too. Ireproduce here, as something of a historical relic, a figure frommy 1960s Carnegie study (figure 1). The figure shows that,excepting war periods and the Great Depression, which requireseparate analysis, cost per student rose appreciably faster thanan economy-wide index of costs in general. The consistency ofthis pattern suggested to me then, as it does today, that we areobserving the effects of relationships that are deeply embeddedin the economic order.

Running through all the factors at play (and there are many,as I will indicate shortly) is a key proposition that my teacherand lifelong friend, William J. Baumol, and I first articulated inour study of the performing arts, which also dates from themid-1960s. The proposition is known to this day in the literatureas the "cost disease." The basic idea is simple: in labor-intensiveindustries such as the performing arts and education,there is less opportunity than in other sectors to increase productivityby, for example, substituting capital for labor. Yetmarkets dictate that, over time, wages for comparably qualifiedindividuals have to increase at roughly the same rate in allindustries. As a result, unit labor costs must be expected to risefaster in the performing arts and education than in the economyoverall.

Robert Frank of Cornell University provided this succinctexplanation of the cost disease as recently as March 2012:"While productivity gains have made it possible to assemblecars with only a tiny fraction of the labor that was oncerequired, it still takes four musicians nine minutes to performBeethoven's String Quartet No. 4 in C minor, just as it did inthe 19th century." In short, productivity gains are unlikely tooffset wage increases to anything like the same extent in thearts or education as in manufacturing; hence, differential ratesof increase in costs are to be expected—a finding Baumol and Ireported for major orchestras at about the same time that myCarnegie study of higher education was under way.

About a decade after the Carnegie study, I reported a similarpattern in my 1976 President's Report at Princeton: "Whileprices in general have risen about 50% [over the previous10 years alone], the most widely used price index for highereducation has risen about 70%." And in 2012, three and a halfdecades later, Sandy Baum, Charles Kurose, and Michael S.McPherson reported basically the same pattern. In their paper"An Overview of Higher Education," presented at Prince tonUniversity, they cite a careful study using data from the DeltaCost Project that shows that "educational expenditures per FTEstudent increased at an average annual rate of about 1%beyond inflation at all types of public institutions from 2002 to2008." There is no need to burden this argument with moredata about trends in institutional costs, which are notoriouslyhard to interpret, in part because they often involve aggregationsof various kinds. It is easy to get mired in the underbrush,and we do well to remember the admonition of the architectRobert Venturi: "Don't let de-tails wag the dog."

There is, however, a final big point to note about trends—namely, the reversal that has occurred in the last decade or soin the respective positions of private and public institutions.When I wrote my 1976 report, from the perspective of thepresident of a private university, there was widespread concernabout the widening gap in charges between the privates andthe publics (with the privates becoming ever more expensiverelative to the publics). In those years, the privates were hitespecially hard by the stagflation of the time, with its dampeningeffect on stock market values that, in turn, affected bothreturns on endowments and private giving. Today, it is the publicsthat have suffered more than most of the privates (and certainlymore than the most selective privates), largely as a resultof sharp cutbacks in state appropriations.

During a discussion session the day after I originally madethese remarks at Stanford, President Hennessy contrastedtrends in tuition and student aid in the public sector with therecent experience at Stanford. He observed that while Stanford's"sticker price" has continued to increase, as it hasthroughout almost all of higher education, Stanford has hadthe financial wherewithal to increase its outlays on student aidby even more than the increases in its tuition and has chosen tospend some part of its resources in this highly commendableway. Only a small number of other wealthy private institutionshave been able to do the same thing, and the fortunate circumstancesof these relatively well-off "outlier" institutions shouldnot be allowed to obscure the general pattern pertinent to thepublic colleges and universities that educate three-quarters ofthis country's college students—or, for that matter, the trendspertinent to the large number of private colleges and universitiesthat have also been compelled to raise tuition faster thanthey have been able to raise student aid. I will return in duecourse to the broad subject of increasing stratification in highereducation and its implications.

I am aware that thus far I have been using an importantword—productivity—without defining it. Put simply, productivityis the ratio of outputs to the inputs used to produce them. Butthis formulation conceals at least as much as it reveals, since it ismaddeningly difficult in the field of education to measure bothoutputs and inputs—even within a single institution, never mindacross institutions serving different missions. If only we producedstandardized widgets or harvested blueberries!

As one illustration of how treacherous this terrain is, theNational Academy of Sciences released, in 2012, a massivereport of over two hundred pages devoted to the measurementof productivity in higher education. A major virtue of thereport, which in turn cites a voluminous literature, is that itdebunks the idea that productivity in higher education is unidimensional.It warns against a multiplicity of dangers thatlurk behind the use and misuse of (inevitably) simplified measures.The report insists that "quality should always be a corepart of productivity conversations, even when it cannot befully captured by the metrics." It also emphasizes the complicationsstemming from joint production of outputs such asteaching and research, and the need to recognize a complexmix of inputs, including capital and student time.

In thinking about the implications of these myriad complicationsfor the ways in which technology might impact the costdisease, I have been helped greatly by the authors of an articlein the New England Journal of Medicine (NEJM), who havecaptured quite skillfully factors that explain what is known asthe IT productivity paradox—the apparent tendency, noted byRobert Solow of MIT in 1987, for computerization to fail toimprove standard measures of productivity. Solow notedfamously, "You can see the computer age everywhere but in theproductivity statistics," an observation said to have launchedmore than two decades of research into the sources of theparadox.

The authors of the NEJM article argue that explanationsfor the IT productivity paradox fall into various categories.Under the heading of "mismeasurement," they note that"important dimensions of service output such as accessibilityand convenience—factors that are greatly improved by IT—aredifficult to quantify and are rarely captured by productivitymetrics." For example, ATMs increased consumer conveniencein banking, but this increase in convenience, and all thetime saved by customers, was not captured by traditional measuresof productivity.

The authors go on to point out: "In terms of 'mismanagement,'the introduction of new technologies usually forcesreexamination of the assumptions that underpin less productiveprocesses." They give a telling example concerning the introductionof electricity in manufacturing: early on, "factoriessimply swapped large electronic motors for waterwheels andsteam engines but retained inefficient belt-and-pulley systemsto transmit power from the central power source. Realproductivity gains came only after manufacturers realized thatmany small motors distributed throughout a factory couldgenerate power where and when it was needed."

This discussion in the NEJM, aimed at implications for thehealth industry, resonates with the uses of IT in education. It iseasy to think of examples, including the tendency in the earlydays of online teaching simply to mimic typical classroomteaching methods, often by videotaping lectures, rather thanre-engineering the teaching process as a whole.

From the standpoint of our interest in the cost disease, itis critical to keep in mind that the productivity ratio has botha numerator and a denominator. Productivity improvementscan be either output-enhancing (raising the numerator)or input-conserving (lowering the denominator). It seemsevident that information technology has been extremelyconsequential in higher education over the last twenty-fiveyears, but principally in output-enhancing ways that do notshow up in the usual measures of either productivity or costper student. It is important to distinguish between at leasttwo broad types of educational "output": research findingsand student learning outcomes. We should also recognize thatthere is a consumption component in the output numerator.The veritable revolution in information technology has had anespecially large impact on research output. Data managementsystems and powerful number-crunching capacities havepermitted research that would have been simply impossibleotherwise. Work in particle physics and studies of the humangenome are but two examples from the physical and lifesciences. To cite a much more mundane example from thesocial sciences, the work that Derek Bok and I did on theeffects of race-sensitive admissions would have been impossiblewithout the construction of the large College and Beyonddatabase. More generally, advances in communications, andthe development of networks and systems for managing textand exchanging perspectives with colleagues at a distance,have revolutionized the way papers are prepared and revised—again and again! Yet these innovations do not show up at allin the usual measures of output.

Technology has also led to dramatic improvements in thescholarly infrastructure. If I may again cite activities that I knowwell, the creation of JSTOR (a highly searchable electronicdatabase of scholarly literature) has changed fundamentallythe way scholars use the back issues of journals and has hadprofound effects on libraries. Similarly, ARTstor (a digitalrepository of high-quality images) now permits art historians tostudy, for example, images of a Bodhisattva on the wall of acave in Dunhuang, an oasis town on the Silk Road, alongsideimages of the same Bodhisattva on a silk painting at theGuimet Museum in Paris. It is worth emphasizing that thesebenefits generally do not accrue to the institutions that madethe investments necessary to realize them. For example, theextra ordinary time savings for scholars made possible by bothJSTOR and ARTstor do not prompt the institutions that employthe scholars to harvest these savings by, for example, increasingteaching loads (unimaginable!).

Although faculty and students have certainly benefited inmany ways from easy Internet access, relatively little hashappened with respect to classroom teaching—until quiterecently. In the second part of this book, I will suggest that weare only at the beginning of the kind of re-engineering that couldin time transform important parts—but only parts—of how weteach and how students learn. Most fundamentally, I will arguethat we need to improve productivity in two ways: (1) throughdetermined efforts to reduce costs—that is, we need to focusmore energy on lowering the denominator of the productivityratio; and (2) through new ways of increasing the student-learningcomponent of the numerator of the ratio, principally byraising completion rates and lowering time-to-degree.

Factors Other Than the Cost DiseasePushing Up Educational Costs

As important as I believe the cost disease to have been (and tobe) in putting upward pressure on instructional costs, I certainlydo not think that it is the sole villain. Let me now mention everso briefly three other forces behind the rise in costs. I recognize,of course, that this list is by no means comprehensive.

Inefficiencies

I am not one of those who looks with disdain at how poorlymanaged colleges and universities are often alleged to be. (Ihave seen too much of other organizations in all sectors of theeconomy, including the for-profit sector.) It is at least mildlyannoying when, with no attempt to provide evidence, a businesspublication such as Forbes blithely asserts on its cover (ofNovember 19, 2012) that "no field operates more inefficientlythan education." I wonder if that is really true. Perhaps thetime has come—if it is not past—when we should cease makingsuch sweeping pronouncements (recognizing that collegesand universities in general have had much greater stayingpower than many for-profit businesses, which have been seento fall by the wayside in surprisingly large numbers).

(Continues...)