Where Are the Savings?

Love MOOCs or hate them, there’s something to disappoint everyone in the Udacity/Georgia Tech services contract amendments that Inside Higher Ed’s Ry Rivard obtained through a public records request.

Although MOOCs have monopolized definitions of higher education reform for nearly two years now, some academic managers have wondered whether they shouldn’t extend online instruction on the base of their existing online programs, rather than partnering with an MOOC platform like Udacity.  A consortium of Midwestern research universities recently took a major step in this direction in suggesting that their members might evolve their own "coordinated platform for the development and delivery of online or blended courses” for the whole consortium’s use.

In contrast, the strongest argument to skip internal development and hire MOOC companies has been the companies' claim to bring revolutionary cost savings to colleges and their students with their revolutionary technology. Unfortunately for all concerned, there is no sign in the Udacity spreadsheets of massive online cost-cutting services. Nor can the savings that do appear be traced directly to the Udacity platform.

The contract, between Udacity and the Georgia Tech Research Corporation (GTRC), aims to create a MOOC master’s degree in computer science — described as "the first professional online master of science degree in computer science (OMS CS) that can be earned completely through the ‘massive online’ format." The hook is the low low price -- $6,630, according to Rivard, or one-seventh of the $40,000-plus price of a face-to-face computer science M.S. at the same institution.

But when we look for massive cost savings in the Georgia Tech-Udacity spreadsheets, what do we find?  In Year 1 (Exhibit H), things don’t look so cheap. The two entities together will spend about $3.1 million running a program for an estimated 200 students in the first semester. This comes to around $15,700 per year per enrolled student. The figure is close to what the University of California Office of the President says it spends educating all UC students averaged together (Display 6). In Year 1, Udacity-Georgia Tech costs look like those of a good, conventional public university program.

But wouldn’t Semester 1 naturally be burdened by start-up costs and a steep learning curve? Yes, were Georgia Tech designing the platform from scratch.  But Udacity is supposed to have already solved higher education’s "cost disease" with its technology. We’ll note that Year 1 is not plug-and-play. Years 2 and 3, when student volume increases, feature courses that are in the can, and low marginal costs of instruction could kick in.

Here’s where things get disappointing.

First of all, the budgets don’t fit with the enrollment plan. Each student takes "6 credit hours (2 courses) per semester" in a 12-course master's. The enrollment forecast "assumes 200 pending full standing (degree-seeking) students begin the program each semester and all 200 in semester 1 become full-standing students .... in semester 2." So in semester 6, the enrollment forecast has semester 1 students in their final term, and five more semesters of 200 new students apiece, for a total of 1,200 students. That is a lot of growth in the existing program that now admits 150 new students a year (page 2). But on the spreadsheet, year 3 revenues have increased nearly 14-fold, to over $19 million. This income requires over 2,800 student FTE in Year 3 (based on Year 1 per-student revenues), which is more than double the enrollment forecast found in the footnotes.

But in fact, the partnership is not collecting $6,630 per year but per degree, and the degree is estimated to take three years. So the typical student in any given year is paying $2,210 in tuition. At that annual price, $19 million of Year 3 revenue requires 8,700 paying student FTE. This figure is larger than the total number of computer science master’s degrees granted in 2009-10 in the United States (Table 4).  Even after noting an untapped global market, this Year 3 number is not credible for degree program enrollment.

At least two conflicting enrollment scenarios are supported by different parts of these documents: (1) growth to 1,200 in Semester 6, and (2) growth to 8,700 in Year 3. (1) fits with reasonable estimates of the time it takes to design and produce new classes, hire and train many "course assistants" (CAs), solve infrastructure issues, and so on, while abiding by the Georgia Tech faculty’s desire to enforce the institution’s high academic standards.

Scenario (2), however, fits with MOOC hype about the "digital revolution" cracking open closed universities to massive global markets by teaching at "effectively zero dollars marginal cost per additional student," in the words of Coursera's co-founder, Daphne Koller. With $14.4 million in expenses in Year 3, a OMS CS with an implausible enrollment of 8,700 students would spend about $1,655 per year per student, or under $5,000 per degree. The cost is ultra-low for a master’s degree, or for any other kind.

How would the OMS CS get to this cost?  The simple answer is that it wouldn’t. The faculty working group expected a 30:1 student:TA ratio. The contract says 1 contact hour per student credit hour, or 104,400 contact hours per year for 8,700 students, each of whom gets a total of 12 hours of personal attention in that time, or 36 hours over the 3-year program. At the same time, the first-year co-sponsor, AT&T, expects 100 percent online instruction, which is cheap only when stripped of personal attention (whether online, via Skype, or in person).  There is a big difference between 290 CAs with the 30:1 ratio, the 87 who at 40 hours a week could deliver 12 hours per student per year, and zero for the online model that eliminates personal attention.  (The budget for student support suggests something close to 87.) With this slippage, Udacity can appeal to the corporate belief that the future of teaching is no teachers, while hiring quasi-teachers to suppress that belief’s results.

Let’s turn from price to cost. Regardless of real enrollment growth, the spreadsheets undermine two key assumptions about commercial MOOCs. The first is that MOOCs offer an automation of teaching that will allow the elimination and/or the cheapening of most teaching staff. This is how Taylorism worked in assembly-line industrialization, and how robotics has worked in manufacturing.

But in the budget, the category of "student support" grows in lockstep with revenue (up 13.8x and 13.9x respectively). One simple reason is that the Georgia Tech faculty wants the OMS CS to have "world-class quality," and that means "blended" or "hybrid" courses. Fully online courses are cheaper, but they generate the highest attrition rates in the history of higher education. Quality MOOCs are always blended MOOCs, and blended MOOCs have lots of CAs (coming mostly from Udacity), which means they aren’t actually MOOCs in the sense of the imagined near-zero personnel costs that has set the business and policy worlds on fire.

This is an important admission that a MOOC is "as good or better" than hands-on instruction only through much hands-on instruction. (Coursera’s recent announcement of 10 public university partnerships also focuses on blended services.) This blocks online’s alleged cost revolution — although online support for instruction obviously helps with costs.

Next, there’s an unpleasant surprise in the equally relentless growth of "Operations, Materials, & Supplies." This is where Udacity’s proprietary technology was supposed to rescue teaching budgets from the medieval methods that currently bloat them. In fact, looking at the budget, its platform does nothing to cut operating costs. The cost of examinations is particularly large. The big savings, ironically, come by squeezing innovation — payments to course creators flatten out — and by leveraging overhead. But there’s nothing novel in these practices. It’s easy to reduce expenses by giving the same lecture over and over, which is what existing online courses are designed to do. The same goes for running more volume on the same equipment, which is another time-honored university tactic.

The Udacity-GTRC contract raises the question of what exactly Udacity brings to the table. First, it brings its platform. Yet its platform is not transformative — not visibly better, faster, or cheaper than what Georgia Tech’s computer science department has already created or could create with new resources for this purpose.

Second, there are some net revenues. With a profit of $1,665 per degree, the program would earn $14.5 million on the untenable 8,700 graduates, or about $4.8 million per year. Sixty percent of this total (or $2.9 million) would go to GTRC. This amounts to a bit over 1 percent of GTRC’s annual research revenues, even with this high number of enrollments. On top of this, GTRC would get 20 percent of gross revenues and 20 percent of gross profits for non-OMS students that use Georgia Tech courses through Udacity. But these are MOOC students who will in general not pay anything. Georgia Tech probably has better margins on its existing extension programs, and could also support its institutional needs with new, smaller programs that it runs on its own.

What Udacity does bring to the table is platform branding. The company has positioned itself as a first mover and dominant player in what it describes as a new global market. Its founder is high-status, famous, and influential. Sebastian Thrun is associated with Google’s driverless car and with Stanford’s artificial intelligence program. He co-signed a deal to provide three entry-level courses to San Jose State University in the presence of the governor of California.

A similar story of brand dominance can be told about Coursera and its co-founder Daphne Koller, whose access to decision makers extends to the World Economic Forum conference at Davos. The three main MOOC companies have had the clout to sign deals directly with a given institution’s senior managers, over the heads of the university faculty. Since Internet and communications technologies seem always to lead to oligarchy (Google/Bing/Yahoo) or duopoly (Apple OS /Microsoft Windows), Udacity can pitch its platform as one of the very few ways for universities to stay in a global online game.

In exchange for presenting itself as an oligarch in waiting, Udacity extracts quite a bit from Georgia Tech. Udacity gets the intellectual content for a master’s program of 20 courses at an upfront cost of $400,000. It borrows Georgia Tech’s reputation as its own, at a huge discount (no training of graduate students, no support for labs, no decades of accumulated know-how through which Georgia Tech earned its reputation).  It acquires these courses for a proprietary platform: Georgia Tech cannot offer these OMS CS courses, created by its own faculty, to a competing distributor.

Udacity expects Georgia Tech faculty members to maintain and update course material, and can use their latest version. While requiring that Georgia Tech not compete with it, it can take Georgia Tech-created courses and offer them to tens or hundreds of thousands of non-registered students — and sell a program certificate for those courses. These courses will differ from Georgia Tech’s in being "minimally staffed to rely on course assistants only for student assessment," but will use Georgia Tech’s content to compete with Georgia Tech’s and all other masters’ programs.  With these courses, Udacity enters the master's certification business, selling a complete degree program without a degree’s intellectual ecology, physical infrastructure, interpersonal venues, and sunk costs.

Udacity’s business model requires that it become a dominant platform. With a series of Georgia Tech-style deals for entire degree programs, it could leverage university content — with sustained free-riding -- to appear to the public as a global university.

For two years, claims about the cheapness of the MOOC format overcame widespread doubts about their educational and social effects. During this time, the main MOOC companies did not release specific financial projections.  Now we finally have two spreadsheets, and their claims to cheapness are not confirmed.

If these costs are typical, it will be more efficient for universities to partner directly with other universities to develop online instruction for underserved students -- and avoid taking on yet another middleman to do one of their two basic jobs.