The Labor Market Returns to Private Two-Year Colleges

Transcription

1 The Labor Market Returns to Private Two-Year Colleges Stephanie Riegg Cellini George Washington University Latika Chaudhary Scripps College November 2009 FIRST DRAFT: PLEASE DO NOT CITE! Abstract Private two-year colleges are a growing component of the two-year college market in the United States, yet we know little about these institutions and their graduates. This paper offers a first look at labor market outcomes, such as wages, hours worked and employment status, of students who attend private two-year colleges as compared to public two-year colleges. Combining data from the 1997 panel of the National Longitudinal Survey of Youth (NLSY) with an individual fixed effects estimation strategy, we find that students completing associate s degrees and vocational certificates in private institutions earn roughly eight percent more than their public sector counterparts, or about four percent per year of education. In contrast to findings of previous studies, however, the wage returns appear to be driven by degree completion, rather than attendance. Our results suggest that today s private two-year colleges provide their students with measureable earnings gains that may justify their large price tag relative to public community colleges. * We thank Janet Currie, Robert Pedace, and Wes Hartmann for helpful comments and Faith Fried and Phil Gross for excellent research assistance. We are grateful for financial support from the Ford Foundation (Grant Number ). The views expressed in this paper are the sole the responsibility of the authors and do not necessarily represent the official views of the Ford Foundation.

2 I. Introduction In the United States two-year colleges enroll more than 6.6 million students every year (U.S. Department of Education 2007), and while much is known about the public community colleges that serve these students, their private sector counterparts largely remain a mystery. At the intersection of postsecondary education and job training, private two-year colleges (also known as proprietary schools, vocational institutes, technical colleges, occupational colleges, or for-profit colleges), offer short-term certificates and associate s degrees in fields ranging from computer programming to hairdressing. These institutions hold both promises and perils for the low- and middle-wage workers that they serve. Understanding proprietary schools and their students is essential for the design of effective policies involving the regulation and licensing of for-profits, eligibility and formulation rules for federal and state financial aid programs, optimal investments in public education systems, strategies for fostering economic growth, and individual incentives for human capital development. The most important question about private two-year colleges for students, policymakers, and taxpayers is undoubtedly the question of college quality. Economists have long focused on the labor market returns to education or earnings gains to assess college quality and the value of education. To date, the literature has found that a year of education in a public two-year community college generates roughly the same earnings gains as a year in a four-year college (Kane and Rouse 1995), yet no one has asked whether private two-year colleges generate similar returns. Our study fills this gap in the literature, offering a first look at the labor market outcomes of students who attend private two-year colleges and evaluating whether these students outcomes differ from those of public two-year college students. To address this question, we 1

3 must first overcome a crucial endogeneity problem: students in private institutions may differ on both observable and unobservable dimensions from those in public institutions. If these differences are correlated with a student s choice of institution and her labor market success, a simple OLS estimate of the impact of private two-year colleges on employment and earnings will be biased. We implement individual-level fixed effects to mitigate this problem. Unlike students in four-year colleges, two-year college students often work before, during, and after they attend college, allowing us to compare an individual student s wages after attendance to her wages before. In so doing, the individual fixed effects can control for all time-invariant student characteristics that may bias cross-sectional estimates of returns. Using the 1997 panel of the National Longitudinal Survey of Youth (NLSY), we find that students completing associate s degrees and vocational certificates in private institutions earn an 8 percent higher return on average than their public sector counterparts, or about four percent per year of education. Compared to individuals who do not complete any postsecondary education, the earnings gains for private two-year college graduates are estimated to be about 7 percent higher per year of education, well within the range of values found in studies of high school and four-year college attendance. In contrast to findings of previous studies of public community colleges, however, earnings gains appear to be driven by degree completion, rather than attendance. Further, we find that while wages are higher for private two-year college students than for public, we find no significant effect of private college attendance on hours worked. We do, however, find a small negative effect of private two-year colleges on employment. Conditional on finding a job, graduates of two-year private colleges earn higher wages. The pattern reverses for the older 1979 NLSY cohort, as we observe a positive effect of private 2

4 college attendance on hours and employment, but no significant difference in wage gains for private sector students. The results suggest that, despite their sometimes dubious reputation, private two year colleges may offer students superior instruction or career services relative to public community colleges. On the other hand, we caution that our analysis focuses primarily on young adults and we explicitly exclude students attending four year colleges. If public community colleges generate additional value by encouraging and enabling students to transfer to universities, our results may overstate differences between sectors. II. Background Over the past half-century, an expansive body of literature has developed to measure the returns to schooling. Reviews of the literature by Card (1999, 2001) and Ashenfelter, Harmon, and Oosterbeek (1999) report that one additional year of education causes a wage gain in the range of 6 to 9 percent, depending on the method, sample, and time frame of the analysis. The vast majority of the research in this area has focused on high school and four-year college-going, with relatively few studies emphasizing differences in various types and levels of schooling (Ashenfelter, Harmon, and Oosterbeek 1999). A handful of studies examine two-year colleges, and those that do focus exclusively on public community colleges. Reviewing the literature on community college returns, Kane and Rouse (1999) find that a year at a community college generates returns between 4 and 6 percent just marginally below those for a four-year college and on average, students attending community colleges earn 9-13 percent more than their counterparts who do not attend any type of postsecondary education. 3

5 Central to the literature on returns is a debate over the accuracy of various methods to identify the causal effect of education on earnings. Students who enroll in college are likely to be very different on both observable and unobservable dimensions from those who do not. If these differences are correlated with subsequent earnings, OLS estimates of the returns to schooling will be biased. Unobserved innate ability is a central concern in the literature, as students with higher ability are likely to have both higher wages and higher levels of schooling. In the high school and four-year college literature, a myriad of studies using instrumental variables and sibling comparisons have attempted to address this endogeneity, 1 but few studies in the community college literature have implemented similar identification strategies. Of the six studies reviewed by Kane and Rouse (1999), all but one (Jacobson, LaLonde, and Sullivan 1997) rely on cross-sectional variation, comparing students who attended college with those who did not, generally controlling for ability with imperfect proxies, such as IQ scores. Jacobson, LaLonde, and Sullivan (1997, 2005) offer the only studies of two-year college returns that adequately controls for unobservables. The authors use an individual- or personspecific fixed effects approach, comparing the wages of displaced workers before and after they attend a public community college, thereby controlling for time-invariant individual characteristics. They find returns of 9 percent for men and 13 percent for women, with much higher returns to quantitative and technically-oriented vocational coursework than lessquantitative coursework in the humanities, social sciences, and basic skills. In this study, we follow Jacobson, LaLonde, and Sullivan, and earlier work by Angrist and Newey (1991) in implementing a similar individual fixed-effects approach. We apply this approach to study returns to private two-year colleges colleges that have been left out of 1 These methods may create additional biases, see Card (2001). 4

6 previous analyses using as a comparison group, students who receive their AA degrees from a public community college or ever attend a public community college. Private two-year colleges Research on private two-year colleges is scarce, primarily due to a lack of data. Most studies of these colleges rely on a non-random sample of schools and students based on the U.S. Department of Education s Integrated Postsecondary Education Data System (IPEDS) and most are purely descriptive in nature (Apling 1993; Bailey, Badway, and Gumport 2001; Rosenbaum, Deil-Amen, and Person 2006; Turner 2006). 2 Administrative licensing data of two-year colleges in California has added to our knowledge of these institutions in recent years and allowed for causal studies of institutional behavior (Cellini 2005, 2009), but in spite of these advances surprisingly little is known about private two-year colleges and their students. What we have learned from these studies and data sources is that there are more than 3,800 private two-year colleges in California alone, but most of these colleges are quite small. Average enrollment is just 350 students (Cellini 2009), a figure that pales in comparison to community colleges that average 6,000 students each nationwide (NCES 2007). Private two-year colleges are primarily organized as for-profit institutions: only 8 percent of California private two-year colleges are reported to receive non-profit or religious exemptions (Cellini 2009). Turner (2006) reports that for-profit colleges award a disproportionate share of all two-year associates degrees and less-than-two-year certificates, particularly in pre-professional and vocational fields in which skills are easy to verify (e.g., by certification examinations) and physical plant requirements are modest. Private two-year colleges tend to offer more programs in fast-growing vocational fields, 2 The IPEDS severely undercounts the number of private two-year colleges in the U.S. For many years the survey relied on snowball sampling and did not require the participation of two-year private colleges. In recent years, greater efforts have been made to track down institutions receiving federal financial aid, but many colleges remain unaccounted for in the data (Cellini 2005). 5

7 such as information technology and real estate (Cellini 2009), and award a much larger share of associates degrees in business and administrative services, engineering-related technologies, mechanics, and production trades than their public sector counterparts (Turner 2006). Community colleges offer significantly more programs and award a much greater share of associates degrees in the liberal arts than private colleges (Cellini 2009, Turner 2006), in part due to their role as transfer institutions. In most states, community colleges have clear articulation agreements with public four-year colleges, allowing students to transfer coursework easily, pursuing their first two years of their BA degree at a community college before transferring to a university. Nonetheless, only a small portion of community college students appear to take advantage of the opportunity: in California, the transfer rate is just four percent (CPEC 2005). Despite these slight differences in focus, both public and private two-year colleges offer associate s degrees and certificates in a wide range of overlapping vocational and academic fields ranging from health and medicine to construction and transportation. For the most part, two-year college students are likely to find their needs met in both types of institutions. Moreover, public and private two-year colleges have been shown to compete for students, suggesting that they operate in the same product market (Cellini 2009). The biggest difference between private and public two-year colleges is undoubtedly their price: required tuition and fees for public community colleges average just $2,017, while private two-year colleges charge more than $10,000 more, averaging $12,620 in (NCES 2007). Private two-year college students undoubtedly receive substantial federal, state, and private financial aid awards to bring costs down, but, as in the four-year college market, the price difference remains substantial (Cellini 2009b). 6

8 III. Conceptual Framework If the market for two-year college education is perfectly competitive, the public-private price differential should reflect differences in the quality of education. With many of the same programs available to them in the public sector, informed consumers should only attend private two-year colleges if the present value of future earnings gains from a private college is greater than their higher cost. There are two primary mechanisms through which two-year colleges can influence wages and employment outcomes. The first is through the quality of instruction. Because private colleges are not restricted by government bureaucracy, they may be able to attract more knowledgeable or dedicated faculty than community colleges. They may also be able to add and modify courses more quickly to meet industry demands, build new classrooms or laboratories, or make other changes to enhance instruction. If instructional quality is indeed superior at private colleges, we would expect private two-year graduates to be employed full-time and earn higher wages. Private colleges may also impact earnings through career development services. In case studies comparing seven of the best private two-year colleges to community colleges in Illinois, Rosenbaum, Deil-Amen, and Person (2006) find that the private colleges they study provide students with extensive job placement services as well as frequent and mandatory advising. The colleges also engage in substantial outreach to build long-term relationships with local employers and teach soft skills to students in preparation for the workplace. These services are almost nonexistent in the public sector and may have an important impact on labor market outcomes. 7

9 Competition in the private sector provides the incentive to allocate resources to their most efficient use, whether those are student services or instructional resources, yet many observers have concerns over the quality of private two-year, and particularly for-profit institutions. The more complicated the skill, and the more difficult it is to verify skill acquisition, the less likely employers are to trust small mom-and-pop and online private colleges. Allegations of fraud and abuse in the sector have further contributed to a sense of mistrust recent years. Colleges have been accused of, among other things, misusing federal financial aid money, closing down midsemester, and making misrepresentations about time-to-degree and the transferability of coursework (e.g., Arenson 2005, 2006, Hechinger 2005, Hefftor 2007, Moore 1995). In the presence of widespread improprieties, or simply pervasive rumors of fraud, employers may be hesitant to hire students coming out of these colleges. In that case, we would expect earnings gains from private two-year colleges to be low, and significantly lower than their more trusted and regulated public sector counterparts. Two-year private graduates would also be less likely to be employed or employed full-time, as compared to public community college graduates. IV. Estimation We estimate log wages and other employment outcomes (i.e., log of hours worked per week, and indicators for whether the individual was employed and whether she was employed fulltime),yy iiii, for individual i in year t as a function of two-year college completion, private college attendance, and individual characteristics as specified in equation (1): yy iiii = ββ 0 + ββ 1 (AAAA iiii ) + ββ 2 (PPPPPPPPPPPPPP iiii AAAA iiii ) + ββ 3 XX iiii + εε (1). The variable AAAA iiii controls for the timing of each student s degree completion. It equals 1 in the year the individual receives their degree or certificate, and remains 1 in all-subsequent years. 8

10 We then interact this variable with PPPPPPPPPPPPPP iiii, a dummy for whether the individual received the degree or certificate from a private two-year college. The result is that the variable of interest (PPPPPPPPPPPPPP iiii AAAA iiii ) equals 1 for private two-year college students in each year after degree completion and 0 otherwise. Finally, we control for individual characteristics XX iiii, including sex, race, age, and age squared. 3 OLS estimates of equation (1), comparing individuals with different levels or types of schooling, are plagued by well-known endogeneity problems. Cross-sectional estimates will be biased if omitted variables, such as innate ability, are correlated both with wages and the choice of schooling. In the case of public and private two-year colleges, innate ability is only one of the omitted variables we might be concerned about. A student s motivation, location, social network, or knowledge of the local labor market may cause additional biases in cross-sectional estimates. Previous studies of the returns to education have used a range of methods to mitigate the endogeneity problem, including proxy variables, instrumental variables, and sibling or twin comparisons, yet few of these studies can adequately control for all unobservable characteristics of individuals that might be correlated with schooling and wages. In this study we implement a stronger identification strategy, employing fixed effects for each individual in the sample. 4 Unlike high school and four-year college students, two-year college students typically work either before or during their coursework, making it possible to observe wages before and after 3 We run robustness checks using experience (defined as age schooling 6) rather than age. The results are essentially unchanged. 4 Our strategy is similar to that of Jacobson, LaLonde, and Sullivan (2005) who use individual fixed effects to examine the returns to different types of coursework in a public community college. It is worth noting, however, that in their preferred specification, Jacobson, LaLonde, and Sullivan include individual-specific time trends. We are not able to implement this approach due to our small sample size. 9

11 attendance. 5 We can therefore estimate: y iiii = ββ 0 + ββ 1 (AAAA iiii ) + ββ 2 (PPPPPPPPPPPPPP iiii AAAA iiii ) + ββ 3 XX iiii + dd ii + εε where dd ii is a vector of dummy variables for each individual. This strategy has the advantage of controlling for all time-invariant unobservable characteristics of the individual that might bias cross-sectional estimates, including innate ability, motivation, and other correlated idiosyncrasies. The model is identified off of changes in individual earnings before and after college attendance. In our final specification, we add calendar year fixed effects to the model, denoting the vector of calendar year dummies, dd tt : y iiii = ββ 0 + ββ 1 (AAAA iiii ) + ββ 2 (PPPPPPPPPPPPPP iiii AAAA iiii ) + ββ 3 XX iiii + dd ii + dd tt + εε This approach further controls for time trends common to all individuals in the sample. While this specification goes a long way in mitigating endogeneity from omitted variables, it is possible that time-varying individual-level unobservables remain. In particular, it is possible that student migration, information, or social networks may change over time. If these are correlated with college choice and wages, our estimates may still be biased. However, given that we control for age, individual fixed effects and year fixed effects, any bias resulting from time-varying individual-specific unobservables is likely to be small. V. Data To implement our analysis, we draw on the National Longitudinal Surveys of Youth (NLSY), two major nationally representative panel surveys that follow the same individuals annually. We focus on the 1997 panel for our main regressions and then present additional 5 Angrist and Newey (1991) point out that in the 1979 panel of the NLSY, almost 20 percent of workers increased their schooling over the five years they study. 10

12 results using the 1979 panel. Each panel created a baseline survey of youth to track throughout secondary school, college and beyond. Both panels contain in-depth questions on educational attainment, earnings, and other related topics, but only the 1997 panel asks specifically whether a student s college was public or private. The NLSY 1997 is based on a representative sample of 8,984 youths who were years old when they were first surveyed in Similar to the NSLY 1979, the survey was designed to document the transition from school to work and into adulthood tracking each participant s educational experiences and behavior in the labor market. The youths are interviewed each year and currently data are available through We thus have a group of individuals ranging from age 22 to 28 by To identify the effects of private two-year colleges on wages, we restrict our analysis to three specific samples of individuals. Given our focus on two-year colleges, in all three samples we drop individuals that went on to receive a BA or higher degree and anyone who completed 16 or more years of schooling even if they do not have a BA. This ensures we have wages for individuals both before and after they complete their two-year program. By restricting our focus to two-year college students, we are limiting the analysis to a pool of individuals more similar to each other relative to four-year college students. Sample A focuses on youth that have completed an associate s (AA) degree. Sample B includes all of the associate degree holders as well as individuals that have completed a vocational degree or certificate program. Sample C is our largest group and includes individuals in sample B as well as anyone who ever enrolled in an associate s degree, vocational degree, or certificate program. We construct sample C to explore whether there are any effects of simply 11

13 attending, rather than completing, a degree program in a private two-year college. If the effects of private colleges in sample C are driven by individuals with completed degrees, adding those who only enrolled, but did not complete programs, will attenuate our estimates of the impact of private two-year colleges. Because of our three samples, the definition of the variable AAAA iiii in equation (1) varies slightly for each. In sample A, AAAA iiii represents the effect of completing an AA degree. In sample B, it represents the effect of completing an AA degree, vocational degree, or certificate. For sample C, AAAA iiii represents enrollment in an associate s degree or vocational program. We use a strict definition of private college to ensure that our measure captures the type of institution (public or private) where an individual received her degree or certificate. The NLSY 1997 has several variables that report the first, second, third, etc. type of college attended and the type of college attended since the last interview by each individual. Using this detailed information, we create a dummy for private college enrollment if the respondent ever reported attending a private college in our sample period. Since an individual could attend both public and private colleges, however, we only code as private those individuals who do not report attending a public college in the year the degree/certificate was received or any year prior. Our main dependent variable is the log of average hourly rate of pay across the first five reported jobs weighted by the weekly hours worked at each job. In the NLSY 1997, individuals report their hourly wage for up to nine jobs but the number of people reporting wages for more than 5 jobs is less than 1 percent over our 10 year period. Hence, we focus on wages for the first five jobs and report results on alternate measures of hourly wages as robustness checks. While wages are a good measure of returns, we also construct three other labor market outcomes, namely log of mean hours worked per week across the first five jobs, a dummy for full time 12

14 employment equal to one if an individual reports working more than 35 hours per week, and a dummy for any employment if an individual reports working non zero hours per week. Our hours and employment variables are not conditional on paid employment and also count hours worked by a small group of individuals reporting positive hours but no wages. These are probably cases of unpaid employment such as internships. Table 1 presents summary statistics for the main variables used in our analysis by sample. Hourly wages are highest in our sample of associate s degree holders at $9.83 between 1997 and Almost 50 percent of the individuals across our three samples are employed full time and work 28 to 30 hours per week. Almost 22 percent of the youths in samples 1 and 2 received their AA degree or vocational certificate from a private college. This proportion however drops to 10 percent when we extend the sample to individuals that ever attended a two-year college or certificate program. Our sample C is thus picking up many people that attended a public community college or two-year private college but never received an AA or vocational certificate. Although we focus primarily on the NLSY 1997, we also present estimates from the NLSY This survey is based on a representative sample of 12,686 young men and women who were years old when they were first surveyed in The respondents were interviewed annually through 1994 and have been interviewed on a biennial basis since, with the latest round in Unlike the 1997 survey, in 1979 participants were never asked whether the college they attended was public or private. To get around this problem, we create two different definitions of private college attendance based on a variable identifying the type of school of a student s vocational/technical training. 13

15 Under the first definition, we identify students attending private colleges if they attended a vocational/technical institute. While some states may indeed have public vocational/technical colleges, we contend that particularly with the wording of institute the majority are likely to be private. The second definition is slightly broader, including students identifying their type of school as a vocational/technical institute, business college, nurses program, barber/beauty school, flight school, or correspondence. We omit students who identify the type of school as company training or apprenticeships. All other students who have completed either associates degrees or vocational certificates who do not identify one of the aforementioned schools are considered public community college students. We acknowledge that our definitions of private college attendance for the 1979 cohort are imperfect proxies and may introduce substantial measurement error: the misclassification should attenuate our estimates of the returns to private colleges. Despite these difficulties, we present results for a sample of students completing associate s degrees or vocational certificates in the 1979 NLSY. 6 When compared to the results using the 1997 cohort, the 1979 cohort can provide some limited insight into the differences in returns over time and at later stages in life. VI. Results Table 2 presents our first set of findings on log hourly wages for each of the three samples described above. Specification 1 is the standard OLS, which exploits cross sectional variation across individuals and finds large effects of both an AA degree and receiving it from two-year private colleges on wages. In the OLS regressions, we also include age, age squared, dummies for sex and race, but still there are several unobservable factors such as ability that could affect 6 Alternative samples based on attendance yielded similar results. 14

16 private college attendance and wages. 7 The OLS coefficients suggest that completing an AA or vocational degree from a public college generates a return of 8.3 percent (coefficient on AA degree/certificate in specification 1), but completing the degree from a two-year private college generates an additional return of 8 percent relative to a public community college. Either attending or completing a degree at a private two-year college, however confers smaller benefits on the order of 3.8 percent (specification 7 for sample C) suggesting that the biggest effect comes from completing the degree or certificate versus merely attending a private college. Given the numerous endogeneity problems inherent in cross-sectional estimates, we present our individual fixed effects results for sample A (AA degree holders) in specifications 2 and 3. After controlling for individual fixed effects, our findings on AA degree completion are no longer significant but the effect of receiving the degree from a private college is still significant at the 5 percent level. Moreover, it is robust to the inclusion of age, age squared and year fixed effects (specification 3). Receiving an AA degree or vocational certificate from a private college generates a 7.5 percent return on wages relative to public community colleges. Returns to twoyear private colleges are slightly higher for sample B (AA and vocational degree/certificate holders) averaging 8.9 percent. Because we find no statistically significant effects of two-year private colleges in our sample C, we believe there are no significant gains to wages from simply enrolling in, but not completing, a degree or certificate in a private two-year college (specification 8 and 9). In Table 3, we subject our findings on hourly wages to several robustness checks. For each of our three samples, the first regressions replace age with experience and experience squared. As seen, the findings on hourly wages are robust to using experience rather than age. 7 Race is missing for some individuals so the number of observations is different between the OLS and individual FE regressions. 15

17 The second and third regressions experiment with alternate measures of hourly wages. For each sample, the second regression uses average hourly wages across all reported jobs weighted by the weekly hours worked as opposed to relying on just the first five jobs (specifications 2, 5 and 8). Specifications 3, 6 and 9 focus on just the first reported job for each of the three samples. The findings on average wages across all jobs are very similar to those for first 5 jobs. In comparison, the findings on wages for first job are smaller in magnitude and statistically insignificant. Since more than 50 percent of the people in our sample report working more than one job in a year, we are more comfortable with measures that account for multiple wages and hours worked. Similar to Table 2, we again find no statistically significant results of attending two-year colleges or attending two-year private colleges on our alternate measures of wages in sample C. We thus find evidence for substantial returns to two-year private colleges on the order of 7 to 8 percent relative to public community college attendance and the findings are robust to various samples, the use of experience rather than age, and multiple measures of hourly wages. In Table 4, we explore the effects of private colleges on other labor market outcomes such as hours worked, full time employment, and any employment. Panel 1 focuses on sample A, panel 2 on sample B and panel 3 on sample C respectively. We find no difference in hours worked for private versus public two-year college students even in our crude OLS specifications. While people receiving degrees from two-year private colleges receive higher wages, they are no more likely to work longer hours relative to people at community colleges. For this group of individuals, the income and substitution effect from higher wages thus appear to cancel each other. We do, however, find some evidence in Sample B, that private college students are less likely than their public college counterparts to be employed full-time. Evidence on any 16

18 employment also reveals consistently negative effects in the fixed effects estimates (specifications 8 and 9), although these are only significant at the ten percent level. To place our findings within the context of the returns literature on community colleges, Table 5 extends each sample to also include individuals who stopped their schooling after completing high school or a GED. Controlling for individual fixed effects, year fixed effects, age and age squared, community college graduates earn about a 6 percent return on their investment (specification 3), or 3 percent per year of education, relative to those who do not continue their education beyond high school an estimate well within the range found in previous literature (e.g., Kane and Rouse 1999). Unlike previous literature, however, the returns appear to be driven by degree completion, rather than simply enrollment, as the effects are small and insignificant when we include those who do not complete degrees in sample C (specifications 7-9). Adding the coefficients on Private*AA and AA in Table 5, suggests a return of roughly 14 percent, or 7 percent per year of education, for private college students relative to high school graduates, and even higher if we include vocational degree or certificate holders. Our results thus far reveal much higher returns to private two-year colleges than public community colleges among students that complete their degree or certificate. On average private colleges appear to provide students with the skills they need to thrive in their careers. These colleges may adapt more quickly to changing workforce demands, offer more technical or practical coursework, or provide better student services and career counseling than community colleges. Further, the fact that returns increased in Sample B as we added vocational degree and certificate holders to our original sample of AA degree holders, suggests not only that these credentials are more valuable to employers, but also that private colleges may excel in providing 17

19 them. On the other hand, vocational degree and certificate students pursuing their education in private colleges are less likely than others to find full-time employment. While the reason for this pattern is not clear, it may be that the types of vocational programs provided by private colleges differ from public. Private colleges tend to dominate the market in fields that lend themselves more to part-time work, consulting, contracting, and self-employment, such as information technology, construction, and real estate (Cellini 2009), making full-time employment less likely, but driving wages higher. Finally, two important caveats are in order. First, our analysis is limited to students who worked immediately following their two-year degree. We exclude BA holders, and in so doing, we have eliminated students in the upper tail of the ability distribution. These students may be particularly attracted to public community colleges because coursework is more likely to transfer to a four-year institution. Future research will look further at these students. Second, due to the nature of the NLSY97, our results may only hold for young people and those who attended a two-year college in the late 1990s or early 2000s. To address this second concern, we compare differences in outcomes over time and across the lifecycle using the 1979 NLSY. The results are reported in Table 6. In contrast to the 1997 cohort, students receiving associate s degrees or vocational certificates in private colleges in the 1979 cohort see no additional gain in hourly wages relative to those completing programs in the public sector: point estimates are insignificant for both definitions of private college and very close to zero in all specifications. Interestingly, we do detect positive effects of private college attendance on other margins. Students attending private colleges appear to work about 6 percent more hours per week and are between 6 and 11 percent more likely to be employed and working full-time than students in the public sector. 18

20 The differences between the results for the 1997 and 1979 cohorts may reflect differences in wage growth across the lifecycle. Our 1979 panel includes much older adults (age in 2006) who may have attended college at later ages. As such, wage trends before and after attendance should show fewer differences due to the concavity of the earnings function over the lifecycle, as wages rise at a decreasing rate (Mincer 1974). Differences in labor market conditions may also explain differences in the results by cohort. The students we study in the 1997 cohort finished their education in the early 2000s a time of economic growth and low unemployment, but growing income inequality. In this context, we might expect to see greater differences in average hourly wages than in employment. In contrast, students in the 1979 cohort were subject to more varied labor market conditions, as they may have completed their education at any time in the 1980s, 90s, or early 2000s, allowing for greater differences at the margin of employment and unemployment. Changes in the two-year college market may also explain the differences in returns between the cohorts. In the 1980s and 90s, the expansion of the Pell Grant gave rise to a new breed of corporate for-profit colleges (Turner 2006). The national chains that are so common today may provide more workplace relevant coursework or have better reputations among employers than the mom-and-pop colleges of past decades. VII. Concluding Remarks This study takes a first step in assessing the quality of private two-year college education, comparing the earnings gains of students attending these institutions to students attending public community colleges. Using individual fixed-effects methods and data from the 1997 NLSY, we find that students completing associate s degrees, vocational degrees, and certificates in private 19

21 institutions see roughly 8 percent greater returns than their public sector counterparts. Compared to individuals who do not complete any postsecondary education, returns for private two-year college graduates are estimates to be 14 percent higher or roughly 7 percent higher per year of education. Results from the 1979 cohort of the NLSY revealed no impact on wages, but positive effects on hours worked and employment. Taken together the results suggest that today s private two-year colleges the vast majority of which are for-profit institutions provide their students with measureable earnings gains that presumably justify their average $10,000 difference in price. However, the mechanisms driving these gains are unclear. It may be that private two-year colleges offer superior instruction than community colleges. Facing few regulations, these colleges may be able to hire more talented instructors, build better computer labs, modify courses or programs more quickly, or simply allocate instructional resources more efficiently than their public sector counterparts. Alternatively, the difference may come from their greater emphasis on career services and links with employers, as suggested in recent case studies (Rosenbaum, Deil-Amen, and Person 2006). Much more research and data are needed to fully understand the reasons behind the differences and their implications for students, and we caution that our analysis does not include students who later transfer to four-year colleges the vast majority of whom likely attend community colleges. What this study has shown, however, is that despite their sometimes questionable reputations, for at least some students, private two-year colleges may be worth their expensive sticker price. 20

22 References Apling, Richard N Proprietary Schools and Their Students. Journal of Higher Education. 64(4): Arenson, Karen W College Accused of Cheating in Aid Process. New York Times. December 6, Arenson, Karen W Investigators Say Flaws at School are Deeper. New York Times. July 24, Ashenfelter, Orley, Colm Harmon, and Hessel Oosterbeek A Review of Estimates of the Schooling/Earnings Relationship, with Tests for Publication Bias Labour Economics, 6(4): Bailey, Thomas, Norena Badway, and Patricia J. Gumport For-Profit Higher Education and Community Colleges. National Center for Postsecondary Improvement, Stanford, CA. Card, David The Causal Effect of Education on Earnings. Chapter 30 of the Handbook of Labor Economics, Vol. 3A, eds. Orley Ashenfelter and David Card. California Postsecondary Education Commission (CPEC) website At-a-Glance. Retrieved July 17, 2005, from: < Hechinger, John Battle Over Academic Standards Weighs on For-Profit Colleges. Wall Street Journal. A1. September 30. Hefftor, Emily Crown College Employees Suspected of Fraud, Seattle Times. August 18. Jacobson, Louis, Robert LaLonde and Daniel G. Sullivan Estimating the Returns to Community College Schooling for Displaced Workers. Journal of Econometrics. 125: Jacobson, Louis S., Robert J. LaLonde and Daniel G. Sullivan The Returns from Community College Schooling for Displaced Workers. Federal Reserve Bank of Chicago, WP , June. Kane, Thomas J. and Cecilia Elena Rouse The Labor Market Returns to Two- and Four- Year College. American Economic Review. 85(3): Kane, Thomas J. and Cecilia Elena Rouse The Community College: Educating Students at the Margin between College and Work. The Journal of Economic Perspectives. 13(1): Mincer, Jacob Schooling, Experience, and Earnings. New York: Columbia University Press for the National Bureau of Economic Research. 21

23 Moore, Richard W The Illusion of Convergence: Federal Student Aid Policy in Community Colleges and Proprietary Schools. New Directions for Community Colleges. 91: National Center for Education Statistics (NCES) Digest of Education Statistics. Retrieved from: < Turner, Sarah E For-Profit Colleges in the Context of the Market for Higher Education, Chapter 3 of Breneman, David W., Brian Pusser, and Sarah E. Turner, eds. Earnings from Learning: The Rise of For-Profit Universities. Albany: State University of New York Press. Rosenbaum, James E., Regina Deil-Amen, and Ann E. Person After Admission: from College Access to College Success. Russell Sage Foundation: New York. 22

24 Table 1: Summary Statistics for NLSY 1997 Variables Mean Std. Dev N Sample A: Individuals with Associate Degree Avg Wage $9.83 $ Hrs Worked/Week Full Time Employment 47.12% 49.92% 6040 Any Employment 70.70% 45.52% 6040 Age Male 48.51% 49.98% 6040 Degree/Cert. from Private College 22.52% 41.77% 6040 Sample B: Individuals with Associate Degree or Vocational Degree/Certificate Avg Wage $9.77 $ Hrs Worked/Week Full Time Employment 47.64% 49.95% 6450 Any Employment 70.79% 45.48% 6450 Age Male 48.37% 49.98% 6450 Degree/Cert. from Private College 22.33% 41.65% 6450 Sample C: Individuals with Associate Degree, Vocational Degree/Certificate or Ever attended an Associate Degree Program, Vocational Degree/Certificate Program Avg Wage $9.24 $ Hrs Worked/Week Full Time Employment 47.37% 49.93% Any Employment 66.24% 47.29% Age Male 51.41% 49.98% Earned Degree/Cert. from Private College 10.88% 31.14% Notes: Avg wage is for first 5 jobs weighted by hours worked per week. Hours worked is the mean hours worked/week for all reported jobs. Full time employment is a dummy if individual worked more than 35 hours/week at all jobs. Any employment is a dummy if individual reported positive hours worked per week and average wage across all jobs was greater than zero. In all three samples, we dropped individuals with a BA or higher degree and those reporting sixteen or more years of schooling.

25 Table 2: Returns to Two-Year Private Colleges on Log Hourly Wages, NLSY 1997 Sample A Sample B Sample C (1) (2) (3) (4) (5) (6) (7) (8) (9) AA Private*AA *** ** *** [0.0294] [0.0261] [0.0277] [0.0261] [0.0237] [0.0249] [ ] [0.0092] [0.0092] ** ** ** ** ** ** *** [0.0364] [0.0361] [0.0361] [0.0362] [0.0352] [0.0352] [0.0117] [0.0184] [0.0183] Age * 0.174*** 0.131*** * 0.168*** 0.125*** 0.117*** 0.211*** 0.177*** [0.0474] [0.0343] [0.0413] [0.0452] [0.0328] [0.0400] [0.0170] [0.0129] [0.0152] Age ^ ** ** *** *** *** [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] Constant * *** *** [0.452] [0.330] [0.399] [0.431] [0.317] [0.387] [0.164] [0.127] [0.148] Sex & Race Yes No No Yes No No Yes No No Individual FE No Yes Yes No Yes Yes No Yes Yes Year FE No No Yes No No Yes No No Yes No. Obs No. Individuals Robust standard errors in brackets. *** p<0.01, ** p<0.05, * p<0. Notes: See table 1 and text for details on variables and samples.

26 Table 3: Robustness Checks on Log Hourly Wages, NLSY 1997 Sample A Sample B Sample C Avg Wage Avg Wage Avg Wage Avg Wage Avg Wage Avg Wage Avg Wage (all jobs) (1st job) Avg Wage (all jobs) (1st job) Avg Wage (all jobs) (1st job) (1) (2) (3) (4) (5) (6) (7) (8) (9) AA Private*AA * * [0.0277] [0.0277] [0.0324] [0.0251] [0.0249] [0.0293] [ ] [ ] [0.0105] ** ** ** ** [0.0361] [0.0360] [0.0422] [0.0352] [0.0352] [0.0415] [0.0183] [0.0183] [0.0208] Constant 1.456*** *** *** *** *** [0.0390] [0.399] [0.467] [0.0368] [0.387] [0.455] [0.0135] [0.148] [0.169] Exp & Exp ^2 Yes No No Yes No No Yes No No Age & Age ^2 No Yes Yes No Yes Yes No Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes No. Obs No. Individual Robust standard errors in brackets. *** p<0.01, ** p<0.05, * p<0.1 Notes: See table 1 and text for details on variables and samples.

27 Table 4: Effects of Two-Year Private Colleges on Hours and Employment, NLSY 1997 Log Hours Worked/Week Full Time Employment Any Employment (1) (2) (3) (4) (5) (6) (7) (8) (9) Sample A: Individuals with Associate Degree AA * *** 0.128*** 0.150*** *** ** *** [0.0240] [0.0262] [0.0277] [0.0232] [0.0214] [0.0228] [0.0177] [0.0189] [0.0200] Private*AA * * * * [0.0284] [0.0359] [0.0357] [0.0272] [0.0298] [0.0297] [0.0217] [0.0262] [0.0261] N Sample B: Individuals with Associate Degree or Vocational Degree/Certificate AA *** 0.126*** 0.146*** *** *** *** [0.0210] [0.0236] [0.0247] [0.0209] [0.0196] [0.0206] [0.0166] [0.0173] [0.0181] Private*AA ** ** * * [0.0277] [0.0349] [0.0347] [0.0267] [0.0292] [0.0292] [0.0213] [0.0257] [0.0256] N Sample C: Individuals with Associate Degree, Vocational Degree/Certificate, or Ever Attended AA *** *** *** *** 0.110*** 0.117*** *** *** *** [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] Private*AA *** *** * * [ ] [0.0184] [0.0182] [ ] [0.0142] [0.0141] [ ] [0.0134] [0.0133] N Sex & Race Yes No No Yes No No Yes No No Age & Age ^2 Yes Yes Yes Yes Yes Yes Yes Yes Yes Individual FE No Yes Yes No Yes Yes No Yes Yes Year FE No No Yes No No Yes No No Yes Robust standard errors in brackets. *** p<0.01, ** p<0.05, * p<0.1 Notes: See table 1 and text for details on variables and samples.

28 Table 5: Comparing Returns at Two Year Private Colleges to Public CC and HS Diploma/GED, NLSY 1997 Log Hourly Wages Sample A Sample B Sample C (1) (2) (3) (4) (5) (6) (7) (8) (9) AA *** *** *** *** ** *** *** * [0.0201] [0.0176] [0.0176] [0.0195] [0.0173] [0.0173] [ ] [ ] [ ] Private*AA ** ** ** ** *** *** *** [0.0363] [0.0352] [0.0351] [0.0360] [0.0347] [0.0345] [0.0117] [0.0188] [0.0188] Constant 0.290** *** *** 0.293** *** *** 0.401*** *** *** [0.146] [0.105] [0.126] [0.146] [0.105] [0.126] [0.149] [0.110] [0.129] Sex & Race Yes No No Yes No No Yes No No Individual FE No Yes Yes No Yes Yes No Yes Yes Year FE No No Yes No No Yes No No Yes No. Obs No. Individuals Robust standard errors in brackets. *** p<0.01, ** p<0.05, * p<0.1 All specifications include controls for age and age squared. Notes: Sample A, B and C refer to Sample A, B and C from earlier tables plus individuals with High School Diploma or GED