«The earnings premium associated with additional education in the United States has risen markedly since 1980, suggesting that the supply of educated ...»
Can Online Learning Bend the Higher Education Cost Curve?
David J. Deming, Harvard University and NBER *
Claudia Goldin, Harvard University and NBER
Lawrence F. Katz, Harvard University and NBER
Noam Yuchtman, University of California - Berkeley and NBER
The earnings premium associated with additional education in the United States has risen
markedly since 1980, suggesting that the supply of educated labor has not kept pace with
demand (e.g., Goldin and Katz 2008). Yet the status of U.S. higher education funding is a key impediment to growing the supply of skills. Inflation-adjusted state appropriations for higher education have been stagnant since 1990 and have declined by 16 percent since 2007 (Baum and Ma 2014). Moreover, since education is a relatively labor-intensive industry, costs per student have risen faster than inflation and are likely to do so in the future (Baum, Kurose and McPherson 2013; Baumol and Bowen 1966). These financial pressures have lent urgency to the search for innovations that can “bend the cost curve” in higher education (Bowen et al. 2014).
Online learning technologies are regarded by most observers as the best hope for cost-saving innovations in higher education, with the primary channel being reduced labor costs through larger class size and less face-to-face interaction (Bowen 2012). Much research on the use of online technology in the classroom has focused on its potential for “disruptive innovation” in higher education, from superstar professors teaching elaborate courses developed by teams, to Massively Open Online Courses (MOOCs) offering top notch instruction to students worldwide (Christensen and Eyring 2011; Cowen and Tabarrok 2014).
In this paper we study whether online education can “bend the cost curve” in traditional higher education. We focus on degree-seeking undergraduates attending open access and lessCorresponding author Deming: Harvard Graduate School of Education, Gutman 411, Appian Way, Cambridge, MA 02139 (email: email@example.com). We gratefully acknowledge support from the Institute of Education Sciences, U.S.
Department of Education, through Grant R305C110011 to Teachers College, Columbia University. The opinions expressed are those of the authors and do not represent views of the Institute or the U.S. Department of Education.
selective postsecondary institutions, a group that accounted for about 59 percent of all U.S.
postsecondary enrollments in 2013 based on data from the Integrated Postsecondary Education Data System (IPEDS) of the U.S. Department of Education. This tabulation does not include MOOCs and other non-degree programs that are difficult to track in the IPEDS. Few degreeseeking students attend selective institutions through online programs.Cost pressures from public funding cuts are most important for non-selective postsecondary institutions offering standard college coursework in a fee-for-service model (Hoxby 2014).
Using the IPEDS, we first document the extent of online education among U.S. degreeseeking undergraduates and how it varies by school sector and selectivity. We then ask whether schools with higher shares of students enrolled in online programs charge lower prices, controlling for detailed institutional characteristics and for geographic market and institution fixed effects. Finally, we compare the trend in tuition prices for online education to tuition prices in more traditional, non-selective postsecondary institutions. Overall, we find that institutions with more online students charge lower prices. However, the impact of online technology on the quality of education remains uncertain. Thus, one needs to be cautious before concluding that lower costs and prices in online programs will raise the productivity of U.S. higher education.
I. Background on Online Education in the United States The growth of distance education was potentially limited by the 50 percent rule of the Higher Education Act (HEA) of 1992, under which schools offering more than 50 percent of their courses through distance education were not eligible to distribute Title IV student financial aid.
Distance education from the early twentieth century took the form of correspondence courses administered by mail. Online higher education emerged in the 1990s with the Internet. The 1998 HEA created the Distance Education Demonstration Program (DEDP), granting waivers from the 50 percent rule to selected institutions including large for-profits such as the University of Phoenix. In 2006 the U.S. Department of Education discontinued the 50 percent rule thereby allowing “online only” postsecondary institutions to flourish (Deming, Goldin and Katz 2012 provide more details). The share of bachelor’s degrees awarded by “online only” institutions grew from 0.5 percent in 2000 to over 6 percent in 2012 (Deming et al. 2014).
Table 1 presents descriptive statistics on online enrollment by sector and selectivity. 1 In 2013, 11.1 percent of all U.S. undergraduate, degree-seeking students were enrolled in programs in which all instructional content was delivered online. More than one in four undergraduates in 2013 took at least one course online. Online education is concentrated among less-selective institutions. Only 1.3 percent of students in selective institutions were enrolled in fully online programs. Large for-profit “chains” such as the University of Phoenix, DeVry, and Kaplan enrolled more than half of all their students in programs operating exclusively online. For-profit “chains” accounted for about 32 percent of enrollment in fully online programs.
Combining online and traditional classroom instruction is most common in the non-selective public sector, where nearly 20 percent of students took at least one (but not all) courses online in
2013. This category most likely represents a variety of approaches to online education, from hybrid designs (i.e., reduced lecture time or “flipped” classrooms) to students enrolling in some fully online courses and some traditional courses.
Non-selective two- and four-year public institutions enrolled only 8.5 percent of students in fully online programs. Moreover, most of these students resided in the same state as the institution despite being enrolled online (column 5). The pattern suggests some reliance on the physical campus infrastructure (online students might come to campus for the exam or to receive IPEDS has collected online enrollment data at the campus level since 2012.
help), although it might also reflect differential pricing for in-state students. 2 Out-of-state enrollment is far more common in the private not-for-profit and for-profit sectors (column 6).
The market for online education is rapidly becoming a national one, but it is not yet global.
Only about 1 percent of all enrollment (just 1,800 students) in fully online programs came from outside the United States in 2013, but this situation may be changing rapidly. At least three large flagship public universities (Penn State, Colorado State and Arizona State) recently opened “global” campuses that operate exclusively online and are available to students worldwide.
The IPEDS, regrettably, does not contain school-level information on the demographic characteristics of students who are enrolled in online education. However, we can learn about aggregate demographic characteristics of students in online programs using the 2011/2012 National Postsecondary Student Aid Study (NPSAS), a nationally representative cross-section of institutions and students. The NPSAS data show that online students are older, have lower levels of parental education, are more likely to be single parents themselves, and are more likely to be working full-time while enrolled in school than are other college students. 3 II. Online Education and Prices To study the impact of online education on prices, we regress tuition and fees, from the IPEDS, on the share of students enrolled in “all” and “some” distance education for U.S.
undergraduate institutions in 2012 and 2013, controlling for a variety of campus-level covariates such as sector, urbanicity, and measures of selectivity. The results are presented in Table 2.
Column 1 reports estimates for public institutions, and column 4 reports results for for-profit and The IPEDS survey specifies that “requirements for coming to campus for orientation, testing or academic support services do not exclude a course from being classified as distance education.” All calculations were performed on 2011/12 NPSAS data, using the PowerStats application (http://nces.ed.gov/datalab/powerstats/default.aspx).
private not-for-profit institutions combined. 4 Columns 2 and 5 add geographic market fixed effects, which may help adjust for differences in costs and student demand across locations.
A major concern in interpreting the results in Table 2 is that differences in price are correlated both with the share of students enrolled online and with unobserved differences in quality. This issue is apparent in the descriptive statistics in Table 1, where selective institutions are far less likely to have students enrolled in online coursework. To address the concern, we first control directly for indicators of selectivity in all the results reported in Table 2. In addition, our preferred specification (columns 3 and 6) controls for institution-by-year fixed effects. 5 The estimates in columns (3) and (6) exploit price and online enrollment differences across institutions with multiple campuses. For example, Penn State World Campus (where 98 percent of students are fully online) charged $13,466 in 2013, compared with $14,525 at a regional campus such as Penn State, Harrisburg and $17,332 at the flagship Penn State campus in State College. Kaplan University, Davenport IA (with more than 99 percent of students taking all courses online) charged $14,162 in tuition in 2013, compared with $15,659 at the Cedar Falls IA branch (having 35 percent of students fully online).
We find modest evidence of lower prices for schools with higher shares of students enrolled online. The results are most robust for the public sector, where a 10 percent (about 1 standard deviation) increase in the share of students taking all courses online is associated with a decline in prices of about 1.4 percent (column 3).
Few not-for-profits have multiple campus branches and we pool them with for-profits to increase power and conserve space. The results are highly similar when we estimate separate results by sector.
Our geographic unit is the Core Based Statistical Area (CBSA), but results are similar using Combined Statistical Area (CSA) or county fixed effects. IPEDS collects data (e.g., tuition and fees) at the campus level. We group institutions by the first six digits of the eight-digit Office of Postsecondary Education (OPE) ID, with the last 2 digits unique to each campus. Although our main outcome is published tuition and fees for full-time students, our results are robust to using the per-credit-hour price (when available).
Columns 4 and 5 show a statistically significant impact of online enrollment on price in the not-for-profit and for-profit sectors of around 1.5 percent for a 10 percent increase in the online share. But there is no detectable impact of the online enrollment share on price for private institutions when controlling for institution-by-year fixed effects (column 6). The IPEDS data indicate that many large for-profits “chains” charge a single price across multiple campus branches. As seen in Table 1, more than half of students in for-profit chains are enrolled exclusively online, and only 13 percent of those students reside in the same state as the campus’s physical location. For-profit chains appear to operate in a national market for online students, and national competition may limit their ability to charge different prices across campuses.
Although online enrollment data is only available from IPEDS since 2012, it is possible to look at trends in prices for campuses that were “online only” prior to 2012. We define “online” campuses as those where more than 50 percent of students were taking all their courses online in
2012. 6 Figure 1 examines the time trend in tuition at online institutions. We compare prices at online schools to schools where more than 50 percent of students take courses in-person. The 50 percent rule ended in 2006, opening up the online education market to additional competitors.
From 2006 to 2013, the enrollment-weighted median price of a full-time undergraduate online education declined by 34 percent. Over the same period, the price of a traditional education at a large for-profit or not-for-profit school dropped by about 8 percent, and tuition at all nonselective four-year public institutions increased by 9.2 percent. 7
The time trend in Figure 1 is very similar when we define campuses as online if no more than 33 percent of the school’s students are from one U.S. state, as in Deming, Goldin and Katz (2012).
The three public flagships (Penn State, Arizona State and Colorado State) that opened “global” online campuses charged prices ($13,735, $8,184 and $8,800 respectively, all reported in 2014 dollars) that hew closely to the 2013 average for for-profit and not-for-profit online schools ($11,923).
III. Conclusion We find some evidence that colleges are charging lower prices for online coursework, suggesting that advances in online learning technology might be able to “bend the cost curve” in higher education. There are two main caveats to our results.
First, it is possible that the quality of education suffers when more content is delivered online. An initial randomized trial of a college statistics course found no difference in student achievement in online versus in-person course sections (Bowen et al. 2014), but two recent studies have found negative impacts of switching from in-person to online instruction on course final grades in an introductory economics class (Alpert, Couch and Harmon 2014; Joyce et al.