![Sample our Education journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5932/TOC-Subject-Sample-2021-Education.jpg"})
Abstract
I estimate the impacts of higher 2-year and 4-year tuition on the outcomes of community college attendance. Higher 2-year tuition is associated with higher dropout rates in both academic and occupational programs and lower rates of terminal degree receipt in occupational programs. Dropout increases are especially large in late semesters and are stronger among men than women. Higher 4-year tuition reduces dropout rates in the early semesters of both academic and occupational programs. Much of the effects of tuition on academic-program enrollment disappear when including campus fixed effects.
Acknowledgements
The author thanks Ronald G. Ehrenberg, Robert M. Hutchens, anonymous referees, and seminar participants at Minnesota State-Mankato and the University of Montana for comments and suggestions regarding earlier versions of this paper. He also thanks George Jakubson and especially Jeffrey Groen and Paul Cichello for advice concerning estimation strategies, and John Porter, Gary Blose, and Pinky Chandra for data availability.
Notes
1. Students are classified by the SDF as either ‘fulltime’ or ‘part-time’. Another variable records whether a student is a first-time, transfer, continuing, or returning student. This paper discusses only students who record themselves as both fulltime and first-time in a fall semester. Nutting (2008) finds that enrollment of first-time fulltime equivalent students – equal to fulltime enrollment plus one-third of part-time enrollment – responds the same to tuition changes as first-time fulltime enrollment.
2. There are 30 campuses in SUNY classified as community colleges. One, the Fashion Institute of Technology (FIT) in Manhattan, is for institutional funding reasons labeled a community college despite offering baccalaureate and advanced degrees. Students who ever attended FIT are removed from the analysis. I omit two other campuses for which I lack full data over the sample period.
3. Students who transfer to a SUNY 4-year college with either an AA or AS degree are almost always awarded first-semester junior status.
4. First-time fulltime students who do not choose a specific program of study are removed from the study. Nutting (2008) finds that enrollment of these students rises when tuition rises, suggesting that IPEDS tuition data do not accurately represent the direct cost for these students. Students who enroll in certificate programs, which follow different graduation timelines than associate's degree programs, are also removed from the analysis.
5. DesJardins, Ahlburg, and McCall (Citation2006) suffer from the same data limitation. Wellman (Citation2002) reports that approximately 35% of students who transfer from New York state 2-year institutions (SUNY and CUNY combined) move to out-of-state schools or to in-state private schools. The City University of New York (CUNY) is a completely separate system from SUNY. Security precautions taken by SUNY during the creation of this data set disallow the matching of its observations to nonSUNY observations via the National Student Clearinghouse.
6. The data does not include information regarding student financial aid. Community college students are much less likely to receive financial aid than 4-year college students (Romano and Millard Citation2006).
7. Regressing 1990–1999 real tuition at SUNY community colleges on year fixed effects dummies results in an adjusted R2 of 0.664. By contrast, regressing 4-year tuition on a set of year dummies results in an R2 of 0.954.
8. Betts and MacFarland (Citation1995) and Kane (Citation1995) find that higher unemployment levels cause increases in community college enrollment. Nutting (2008) finds these increases in enrollment to be especially large in occupational programs. Retail weekly wage corresponds strongly with the wage of workers with less than a college education (Gould, Weinberg, and Mustard Citation2002).
9. New York State is generally divided into ‘cities’ and ‘towns’ (in some states, towns are called ‘townships’). Villages are population centers located in towns. The mailing address for each campus is used to determine whether it is located in a city or a village; those that are located in neither are assigned the population of the surrounding town. Two campuses are located in ‘Census Designated Places’, and are assigned populations accordingly.
10. Nationwide, Hispanic presence in community college is relatively high, because of large Hispanic presences in states, such as California and Texas, with large community college systems (Cameron and Heckman Citation2001). The share of Hispanics in SUNY is less than the share in CUNY (Leinbach and Bailey Citation2006), see also footnote 5.
11. Cameron and Heckman (Citation2001) examine schooling choices up to the point of college enrollment. Groen et al. (Citation2008) analyze attrition and graduation from PhD programs in the humanities.
12. The data set is restricted to students who enroll fulltime in their first semester. Enrollment in subsequent semesters is permitted to be part-time.
13. This allows autocorrelation in the decisions of students who have left community college.
14. A small number of students who receive associate's degrees before the end of their fourth semester are removed from the sample. For students who remain enrolled after having earned an associate's degree, choices j = 1 and j = 2 are merged, as are choices j = 3 and j = 6.
15. One exception is students who have earned a degree yet remain enrolled in community college. They are followed regardless of their program of enrollment. Thus, choices j = 5 and j = 7 are merged for students who have already earned their associate's degree. So have choices k = 2 and k = 3.
16. Multinomial logit estimations allow for choices among unordered outcomes. Since community colleges often cater to students with different educational ambitions (Kane and Rouse Citation1999; Leigh and Gill Citation2004), and since it is difficult to discern whether some outcomes are better than others (e.g. transferring to another program versus continuing in this program), it is preferable to other hazard rate models.
17. A quadratic time trend is preferable to year fixed effects because baseline 4-year tuition is fixed between SUNY campuses in any one year. The variation in 4-year tuition when including year fixed effects would thus be determined by the relatively small changes between 4-year campuses in fees, which would not be very informative.
18. Spring semesters use January county-level unemployment rates, and fall semesters use July county-level unemployment rates.
19. Very few students enrolled in any specific semester are enrolled at a 2-year campus differing from their first-semester campus.
20. There are only six ultimate outcomes, even though there were seven choices in each multinomial logit estimation, because transferring to a 4-year college with an associate's degree (l = 1) can be done either immediately after a student earns an associate's degree or in later semesters.
21. Six-semester probabilities are used because student choices are observed through t = 8 and it is not possible, for example, to determine whether a student who leaves community college in when t = 7 remains nonenrolled for three consecutive semesters.
22. Increasing a tuition value by 20% includes increasing the value of next year's tuition by 20% in spring-semester estimations.
23. Multinomial logit tuition coefficients for the restricted model when omitting campus fixed effects are available from the author upon request.
24. This is consistent with the finding of Nutting (2008) that new enrollment in academic programs is significantly more responsive to 2-year tuition than enrollment in occupational programs.
25. An alternative possibility is that financial aid is less generous in late semesters, and therefore that the simulation's stated tuition increases are much lower than the real tuition increases experienced by late-semester students. Singell (Citation2004) finds that financial aid significantly affects retention at a 4-year college.
26. Over 90% of the variation in 2-year tuition is explained in the simulations that include campus fixed effects.
27. When including campus fixed effects, large standard errors on the transfer-without-degree and transfer-programs hazards cause random draws from underlying parameters to frequently yield exceptionally small probabilities of continuing beyond a third semester. Thus, the probabilities of ultimate outcomes which require four or more semesters at community college (i.e. transferring with a degree or earning a terminal degree) are tiny – well under 0.001 – in a large share of the 750 random draws. This yields a tiny standard error in the probability of these two ultimate outcomes. Bootstrapping, an alternative technique of computing simulation standard errors, is infeasible due to extreme time constraints.