Redistributive fee increases, net attendance costs, and the distribution of students at the public university

Redistributive fee increases, net attendance costs, and the distribution of students at the public university

Economics of Education Review 20 (2001) 551–562 www.elsevier.com/locate/econedurev Redistributive fee increases, net attendance costs, and the distri...

84KB Sizes 0 Downloads 5 Views

Economics of Education Review 20 (2001) 551–562 www.elsevier.com/locate/econedurev

Redistributive fee increases, net attendance costs, and the distribution of students at the public university Michael J. Hilmer

1,*

Department of Economics, Brigham Young University, 180 Faculty Office Building, Provo, UT 84602-2363, USA Received 6 December 1998; accepted 31 March 2000

Abstract This paper examines the effect of policies that increase tuition at public four-year colleges while returning a substantial portion of the revenue to economically disadvantaged students in the form of increased financial aid awards. Such “redistributive fee increases” are demonstrated to have potentially important effects on the distribution of students choosing to attend in-state public four-year colleges. Specifically, if in-state public four-year net attendance costs increase by themselves high-test students are far more likely than low-test students to choose alternative paths while if the net attendance costs of alternative paths also increase Black and Hispanic students are most likely to choose alternative paths.  2001 Elsevier Science Ltd. All rights reserved. JEL classification: JEL I20 Keywords: State federal aid; Costs; Economic impact

1. Introduction Public attention has recently focused on the escalating costs of higher education. Since 1980, average undergraduate tuition and fees at public four-year colleges have increased by 371%, at private four-year colleges have increased by 357%, and at public two-year colleges have increased by 328% (Digest of Educational Statistics, Office of Educational Research and Improvement, 1997, 1985). This disturbing trend has been blamed on the combination of increasing educational costs and, within the public sector, decreasing support for higher education. The Commission on National Investment in Higher Education (1997) reports that since 1976 public support per student has just kept up with inflation while real educational costs per student have increased by 40%. Such a combination places public universities in a pre-

* Tel.: +1-801-378-2037. E-mail address: [email protected] (M.J. Hilmer). 1 Visiting Assistant Professor at Brigham Young University.

carious situation. Public university funds derive from two main sources: government contributions and student tuition receipts.2 Subsequently, as the cost of educating students at a given quality level increases, states can respond by either increasing public support or by increasing student tuition charges at their public fouryear institutions. In recent years, the primary response has been to increase tuition, which has lead to the skyrocketing attendance costs at public universities. There is a policy dilemma associated with increasing public four-year tuition, however. Higher education remains one of the main avenues to economic mobility. A consistent finding in the economics of higher education literature is that college graduates enjoy a large earnings premium relative to their peers who never

2 According to the Digest of Educational Statistics (Office of Educational Research and Improvement, 1997, 1985) during the 1994–95 academic year, public four-year institutions in the US received roughly 18% of their total revenue from tuition and fees and roughly 36% from state governments.

0272-7757/01/$ - see front matter  2001 Elsevier Science Ltd. All rights reserved. PII: S 0 2 7 2 - 7 7 5 7 ( 0 0 ) 0 0 0 3 6 - 4

552

M.J. Hilmer / Economics of Education Review 20 (2001) 551–562

attend college. For instance, among 25–34 year-old males, the difference between the median 1994 earnings of Bachelor’s and high school degree holders was roughly 86% (National Center for Education Statistics, 1996). Moreover, recent studies by Brewer, Eide and Ehrenberg (1999), Eide and Grogger (1995), Bound and Johnson (1992), Katz and Murphy (1992), and others find that the positive return to higher education has grown rapidly over the past two decades, suggesting that college attendance may be even more important for the current generation of students. Hence, state governments recognize that it is important to maintain access for all groups, especially the economically disadvantaged who rely on higher education to help bridge the gap between the economic “haves” and the “have-nots.” Unfortunately, the need to raise tuition to help cover the escalating costs of educating public four-year college students is likely at odds with the goal of maintaining access for all students. Upon graduation from high school, potential students have thousands of options to consider. They can attend a two-year college, they can attend a public or private four-year college in their home state, they can migrate to attend one of those institutions in a different state, or they can bypass college attendance altogether. Presumably, students make their college attendance choices by comparing the expected utility of each potential option and choosing the one that provides the highest. The problem facing state governments is that increasing tuition at public four-year colleges increases the net attendance cost of that path, thereby decreasing its expected utility relative to the expected utilities of the alternative attendance paths. Hence, it is possible that increases in public four-year tuition will encourage many students to choose attendance options other than public four-year attendance. Especially problematic from a policy standpoint would be if economically disadvantaged students were more likely than their financially-able peers to switch to non- or two-year attendance. Nonetheless, several studies tend to confirm this fear as they suggest that low-income students are more responsive to changes in net attendance costs than high-income students (McPherson & Schapiro, 1991; Leslie & Brinkman, 1987; Manski & Wise, 1983). This fact is troubling as a student’s financial capacity is likely correlated not only with income but also with important diversity factors such as gender, ethnicity, and academic ability. Consequently, a realistic concern for policymakers is that increases in public four-year tuition may have important effects on the distribution of students choosing to matriculate at those institutions. In response to this concern, states have enacted what may be termed “redistributive” fee increases. A redistributive fee increase is one in which tuition is increased for all public four-year students with the caveat that a significant fraction of the increased revenue is returned to less financially able students in the form of increased

financial aid awards. In theory, under such a policy, students who could initially afford attendance would see an increase in net attendance costs while those who were initially financially constrained would see no change in net attendance costs.3 Proponents hope that by keeping net attendance costs fixed for financially constrained students, such policies will achieve the goal of increasing tuition revenues without pricing economically disadvantaged students out of the economic engine that is college attendance. To determine whether redistributive fee increases will successfully achieve this goal, it is necessary to generate some idea of the effect that such policies would be expected to have on the composition of the student body at the public university. The current study does so by estimating the effect that increases in net attendance costs have on student attendance decisions and examining how increases in net attendance costs at in-state public four-year colleges might affect the decisions of students who do not receive financial aid. The results suggest that students respond to increases in net attendance costs in expected ways. Namely, when the net cost of a particular attendance path increases students are significantly less likely to choose that path. More importantly for this study, simulations based on those results suggest that increasing the net attendance cost of in-state public four-year attendance for financial aid nonrecipients may significantly alter the distribution of students at the public four-year college. Specifically, it appears that if in-state public four-year costs increase by themselves high-test students are nearly three times more likely than low-test students to choose an alternative attendance path. On the other hand, if the net costs of the alternative paths are also increased Black and Hispanic students are most likely to choose alternative attendance paths. Hence, the results suggest that redistributive fee increases are likely to have potentially troubling effects on the diversity and/or overall ability level of the student body at a state’s public four-year colleges.

2. Econometric specification There is some debate about the correct way to model the college attendance decision. Two prominent methods are the multinomial logit (Ordovensky, 1995; Savoca, 1991; Weiler, 1989) and the ordered probit (Hilmer, 1998; Broomhall & Johnson, 1994). Both models recognize that the college attendance decision is a discrete choice problem (Maddala, 1983) in that rather than

3 It is even possible that under such a policy, financial aid awards for some students might actually increase by more than the increase in tuition, thereby leading to a reduction in net attendance costs for some students.

M.J. Hilmer / Economics of Education Review 20 (2001) 551–562

observing the probability that a student chooses each possible attendance path the researcher only observes the student’s chosen path. The fundamental difference between the models is that the ordered probit assumes there is a natural ordering of the attendance path choices while the multinomial logit does not.4 This fact makes the ordered probit inappropriate for the current application. To have a natural ordering, we would have to be able to say that private four-year attendance was preferred to public four-year attendance and public fouryear attendance was preferred to two-year attendance and so on (to say nothing of the in-state/out-of-state ordering). Because evidence such as Ganderton (1992) suggests that students often opt for lower quality public universities over higher quality private universities due to the in-kind nature of the public higher education subsidy, it is not clear that a natural ordering exists among the alternative attendance paths. Consequently, the estimation method used here is the multinomial logit. Estimation of the multinomial logit model is straightforward (Greene, 1997, pp. 915–917). For this study, we define the student’s attendance decision as being between non/vocational college attendance,5 community college attendance (both in-state and out-of-state),6 in-state public four-year attendance, in-state private four-year attendance, out-of-state public four-year attendance, and outof-state private four-year attendance. The non/vocational college attendance group is the base group for the estimation. This study makes use of a unique data set constructed by merging individual-specific data drawn from the High School and Beyond (HSB) survey with state-specific information on American colleges drawn from published college handbooks. Data from the sophomore and senior cohorts from the HSB (National Center for Education Statistics, 1987) were combined to produce a data set containing information on 16,551 students.7 Within that

4 For an excellent discussion and application of the ordered probit model see Hausman, Lo, and MacKinlay (1992). 5 The model certainly could be estimated with vocational school attendance defined as a separate attendance path. As with Weiler (1989), however, doing so does not significantly alter the results. 6 In-state and out-of-state community college attendees are lumped together because nearly all students who attend community colleges do so in their home state. 7 Combining the base samples results in a sample of 25,820 students. Students were dropped for not participating in all subsequent surveys and for failing to provide adequate information to calculate variables of interest. The fact more than a third of the sample were omitted due to inadequate data leads to the possibility of respondent bias. The distribution of explanatory variables does not appear to differ significantly between the remaining subsample and the full sample, however, and thus we do not consider respondent bias to be a problem.

553

data set, 6984 students did not attend an academic postsecondary institution, 3951 students attended a two-year college, 3568 attended a public four-year college, and 2048 attended a private four-year college. As expected, students attending private four-year colleges appear to be far more mobile than their peers attending public fouryear colleges. Namely, among private four-year attendees, roughly 40% migrated outside their home state, while among public four-year college attendees only 12% chose to do so. Table 1 provides descriptions of the individual characteristics used in this analysis. Individual-level data are chosen to represent individual and family background characteristics affecting a student’s proclivity towards college attendance. State-level data are chosen to represent the relative net costs of the six attendance options available within a student’s relevant choice set. Calculation of the net attendance cost measures is a bit tricky and merits some discussion. Our first concern is the definition of a student’s relevant choice set. The standard approach relies on the fact that roughly 80% of students attend college in their home state (Digest of Educational Statistics, Office of Educational Research and Improvement, 1997, 1985) and defines a student’s choice as consisting solely of the institutions in his or her home state (Brewer et al., 1999; Hilmer, 1997 etc.).8 This is not sufficient for the current study, however, as unlike those previous studies we want to consider the student’s decision whether to migrate to a different state or remain in his or her home state. Clearly, this decision will depend not only on the net attendance costs in the student’s home state but also the net attendance costs in other states to which the student might choose to migrate. While students are surely free to attend college in any of the 50 states, it is likely that students in different states are more likely to choose colleges in particular states than in others. Therefore, we calculate the out-ofstate choice set for students in each given state as a

8 A well-known property of the HSB data set is that it does not include identifiers of the student’s home state. Most previous studies have imputed home states following the backtracking technique described in Ganderton (1992). According to that technique, the state in which most students from a high school attend college is likely the home state of the schools students. Using that procedure, it possible to unambiguously identify the home states of roughly 90% of all students. This study makes use of an alternative method for identifying home states. The HSB file contains the 1980 unemployment rate in each student’s home state. For 1980, there was enough variation in unemployment rates across census regions that it is possible to unambiguously identify each state. As a consequence, our data set contains a fuller sample as we do not have to exclude any students for missing state data.

554

M.J. Hilmer / Economics of Education Review 20 (2001) 551–562

Table 1 Description of variables used in analysis Individual characteristics Male, female, black, Hispanic, other Binary dummy variables indicating the student’s gender and ethnicity. race Parent college, college prep, receive aid Binary dummy variables indicating whether the student had at least one parent who from parents graduated from college, classified his or her high school program as academic/college preparatory, and reported that his or her parents contributed toward college expenses. Test scores Continuous variable representing the student’s average score on mathematics and reading test administered during the senior year in high school. Family income Categorical variable indicating the income level of the student’s family during high school. In 1980 US$ the categories are (1) less than $7000; (2) $7000–$11,999; (3) $12,000–$15,999; (4) $16,000–$19,999; (5) $20,000–$24,999; (6) $25,000–$37,999; (7) $38,000 or more. HS grades, family size Categorical variables indicating the student’s self-reported high school grades (four-point scale) and the number of members in the student’s family during the student’s senior year in high school.

weighted average of the states in which migrating students from that given state choose to attend college.9 Once we have defined a student’s choice set, our second concern is the calculation of the net costs of attending each of the six different types of institutions we define. The first part of this calculation is the direct cost of attendance. Because we want to examine how the costs of different alternatives affect the student’s choice, we need to look at the direct cost of each potential attendance path rather than the specific direct cost of the attendance path actually chosen. We therefore calculate direct attendance costs as the mean tuition for two-year and public and private four-year colleges in the student’s home state and for the weighted out-of-state tuition for public and private four-year institutions attended by other high school graduates from the student’s home state. For in-state measures we use tuition only while for out-of-state measures we use both tuition and room and board, to reflect the fact that students who migrate must move away from home while in-state attendees do not necessarily have to. The second part of the net cost calculation is the expected financial aid for each potential attendance path and is somewhat more complicated. College attendees in the HSB only self-report the amount of 9

As an example of this calculation, consider the state of Arizona. Eight students from Arizona chose to attend a public four-year college in a different state. The breakdown of states for those students is as follows (public four-year college fees in parentheses): two students in Arkansas (US$2184), one in California ($2491), one in Colorado ($2954), one in Iowa ($2518), and three in New Mexico ($2280). Thus, the measure of out-of-state public four-year fees for students in Arizona is calculated to be (2∗2184⫹1∗2491⫹1∗2954⫹1∗2518⫹ 3∗2280)/8=2396.38. The calculation is similar for the remaining out-of-state tuition measures.

financial aid they actually receive from the institution they initially choose to attend. As such, we lack direct information on the amount of financial aid that the student would have received at each of the other potential attendance options. We solve this problem by calculating the desired predicted financial aid measures at the nonchosen attendance paths in the same manner as Brewer et al. (1999). To do so, we regress actual aid received on individual characteristics (gender, ethnicity, high school GPA and athletic status, test scores, parental income, and family size) and a set of state-specific dummy variables (to reflect potential cross-state differences in financial aid practices) for the subsets of students choosing each attendance option.10 The resulting coefficient estimates are then used to construct predicted financial aid for each of the six attendance options for each student in the sample. Predicted financial aid is then subtracted from average in-state and out-of-state tuition to determine a student’s predicted net attendance cost for each attendance option. Finally to capture potential cross-state differences in access to different attendance options, we include the average number of “slots” available at each type of institution.11

10 Because financial aid is zero for many students, the data are left-censored and we use a maximum likelihood tobit (Greene, 1997) to obtain these predictions. Results are available upon request. 11 The definition of “slots” merits some discussion. We want this measure to capture the access to different types of institutions. One possible measure then is total attendance at the different institution types in a given state. This is not quite accurate, however, as to have access students need to be admitted but that does not necessarily mean they must attend a particular institutions. Hence, for four-year colleges we define “slots” as being the total number of students admitted to public

M.J. Hilmer / Economics of Education Review 20 (2001) 551–562

555

Table 2 Summary statistics for individual characteristicsa Individual characteristics

Do not attend

CC attendees

Public four-year In-state

Male Female Black Hispanic Other race White College prep Test score HS grades Parent college Family income Family size Receive aid from parents Number of observations a

0.5118 0.4882 0.1233 0.1481 0.0175 0.7111 0.1945 0.4592 (0.0764) 2.3583 (0.7191) 0.0663 3.8252 (1.7588) 3.6017 (1.8051) 0.0711 6984

0.4554 0.5446 0.0902 0.1072 0.0288 0.7738 0.4334 0.5065 (0.0810) 2.7273 (0.6240) 0.1486 4.3805 (1.7410) 3.3345 (1.6311) 0.6435 3951

0.4740 0.5260 0.1014 0.0644 0.0228 0.8114 0.6920 0.5561 (0.0779) 3.1283 (0.6186) 0.2454 4.6428 (1.7625) 3.3609 (1.5606) 0.7596 3133

Private four-year Out-of-state 0.4919 0.5081 0.1299 0.0362 0.0313 0.8025 0.7514 0.5660 (0.0773) 3.0249 (0.6540) 0.3133 4.9914 (1.7529) 3.2910 (1.4587) 0.7791 435

In-state 0.4574 0.5426 0.0857 0.0619 0.0170 0.8354 0.7829 0.5638 (0.0786) 3.1622 (0.6116) 0.2304 4.6320 (1.7618) 3.3445 (1.5313) 0.7899 1228

Out-of-state 0.4836 0.5164 0.0569 0.0449 0.0154 0.8827 0.8464 0.5764 (0.0781) 3.1491 (0.6294) 0.4099 5.0577 (1.7331) 3.2969 (1.4890) 0.8190 820

Data are weighted. Standard deviations in parentheses.

Table 2 presents summary individual characteristics for students in the sample. The distribution of students across attendance paths in this sample is similar to the distribution of students nationwide. Within the sample, roughly 74% of all postsecondary students chose to remain in their home-state, while nationwide in 198212 slightly more than 80% chose to do so. Likewise, roughly 40% of all students who attend either a two- or a four-year college chose to start at a two-year college, while among the national population in 1982 roughly 35% chose to do so (Digest of Educational Statistics, Office of Educational Research and Improvement, 1997, 1985). Comparing gender and ethnicity across attendance paths suggests that there are marked differences among students choosing the different attendance paths. Nonattendees are most likely to be male and Hispanic and least likely to be white. Among the five groups of college attendees, those starting at two-year colleges are most likely to be female and Hispanic and least likely to be white. Out-of-state public four-year attendees are less and private four-year colleges in each state (source: Barron’s Profiles of American Colleges, Barron’s College Division, 1984). 12 Students from the senior cohort made their initial attendance decision in 1980 while students from the sophomore cohort made their initial attendance decision in 1982. Therefore, 1982 is a representative year for when these students made their decisions.

likely than their in-state counterparts to be female, Hispanic, and white and more likely to be black and other race. Conversely, private four-year attendees who choose to migrate out-of-state are less likely to be black and other race and more likely to be white. Comparing public to private four-year attendees, public four-year attendees (both in-state and out-of-state) are more likely to be male, black, and other race and less likely to be white. Hence, as students must first be admitted in order to attend a particular institution, it appears that gender and ethnicity might be prominent factors in the college admission process (Brewer, Eide & Goldhaber, 1999; Bowen & Bok, 1998). Factors affecting college attendance are not surprising. Family background, measured academic ability, and high school performance are all significantly higher for the five subsets of college attendees than for non-attendees, while average family size (an expected deterrent to attendance) is lower. A more interesting comparison is between in-state and out-of-state four-year college attendees. In general, within a particular institution type outof-state students are more likely than in-state students to have followed a college preparatory high school program and possess higher average test scores, high school grades, parental education, and family income and lower average family sizes. Note, however, that in-state private students tend to possess lower average values than outof-state public students for those same variables, except for average family size, which is higher. This, as might be expected, suggests that students need to possess

556

M.J. Hilmer / Economics of Education Review 20 (2001) 551–562

higher subjective probabilities of graduation in order to make the choice to incur the higher cost of attending college in a different state, regardless whether the institution is public or private.13 Table 3 reports summary statistics for the net attendance cost measures. While it is difficult to draw any firm conclusions about a student’s college choice from these averages, a few noticeable cross-group differences do appear. Most prominently, college attendees appear to choose attendance options that provide the lowest relative net attendance costs. Specifically, among all six subsets of students, two-year college attendees have the lowest average net two-year attendance cost, in-state public four-year attendees have the lowest average in-state public four-year net attendance cost, and out-of-state public four-year attendees have lower average out-of-state net public four-year attendance costs than all but one group. The story appears to be different for both subsets of private four-year students, however, as the net attendance costs of the attendance paths actually chosen by those students tend to be among the highest. This confusion is likely due to the fact that among private institutions cost and quality are highly related. As these subsets of students have the highest financial capacities (see Table 2), they are likely choosing their colleges based on quality without even considering the relative net attendance costs. Hence, the higher averages for these groups might simply reflect the fact that these students are choosing higher quality colleges. Finally, access to different types of institutions appears to affect student attendance decisions in predictable ways. Two-year attendees come from states with the highest average number of two-year slots while public and private four-year attendees come from states with the highest relative proportion of those types of institutions.

3. Results Table 4 presents the results of estimating the college attendance equation by maximum likelihood multinomial logit. The entries have been converted to marginal effects and should be interpreted as the effect that changes in the independent variables have on the probability of choosing one attendance path relative to choosing the base path of non/vocational attendance, holding all else 13 Upon graduation from high school, potential students have imperfect information about their probability of one-day graduating from a university. According to the “option value” theory of education (Manski, 1989; Comay, Melnick & Pollatschek, 1973) the best way to gain additional information about one’s graduation potential is to attend either a university or a community college in order to get additional observations of educational aptitude and/or motivation. Such information can then be used to make more informed persistence decisions.

constant. For ease of exposition, we only present the estimated marginal effects for the predicted net attendance cost variables, which are the focus of this study. Results for the remaining variables are consistent with standard findings (Hilmer, 1998; Behrman, Kletzer, McPherson & Schapiro, 1992; Ganderton, 1992 etc.) and suggest that high school performance, measured academic ability, parental education, and family income are positively correlated with college attendance and that students possessing higher values for those characteristics are increasingly likely to choose private four-year colleges over public four-year colleges and two-year colleges.14 The first row in Table 4 suggests that one of the most important determinants of college attendance is whether a student’s parents contribute towards the cost of attendance. As might be expected, if parents help defray attendance costs students are significantly more likely to pursue higher education. Results for the net attendance costs measures are generally consistent with the central tenants of consumer theory. Own-price effects follow the Law of Demand for most of the attendance paths. Namely, students from states with higher net two-year, in-state net public four-year, and out-of-state net public four-year attendance costs are less likely to choose each of those attendance paths. The lone exception to the expected findings is that students from states with higher net attendance costs at out-of-state private colleges are more likely to attend a private four-year institution in a different state. The most likely explanation for this anomalous result is twofold: (1) elite private institutions are concentrated in a minority of states so that most attendees will be from out-of-state and (2) the students who choose to attend such institutions are more likely to come from privileged backgrounds so that they can afford to choose to attend despite the higher net attendance costs.15 The estimated effects of out-of-state net attendance costs are less obvious. As expected, students from states with higher out-of-state public four-year costs are less likely to attend out-of-state public fouryear colleges. At the same time, however, such students are also significantly less likely to attend two-year or instate public and private four-year colleges. Finally, higher out-of-state private four-year net attendance costs do not have statistically significant effects on any of the attendance paths. Again, this is not surprising given the above evidence that out-of-state private four-year attendees are likely unaffected by differences in net attendance costs. The bottom two panels of Table 4 suggest that access

14

Full results available upon request. In an attempt to account for this possibility, we included measures representing the average quality (as defined by median SAT of entering freshmen) of institutions in different states. However, doing so did not alter this finding. 15

M.J. Hilmer / Economics of Education Review 20 (2001) 551–562

557

Table 3 Summary statistics for state cost measuresa Do not attend

CC attendees

Public four-year In-state

Home state cost measures: Two-year college: Tuition Predicted aid Net cost Public four-year college: Tuition Predicted aid Net cost Private four-year college: Tuition Predicted aid Net cost

Out-of-state

In-state

Out-of-state

674.62 (291.02) 295.85 (222.82) 378.76 (328.29)

577.80 (329.81) 296.07 (215.07) 281.74 (326.51)

661.89 (298.54) 357.89 (233.98) 303.99 (325.07)

701.95 (270.12) 329.33 (219.14) 372.63 (296.84)

771.63 (297.69) 383.51 (223.69) 388.11 (329.32)

728.55 (261.32) 378.41 (222.33) 350.15 (275.58)

812.33 (324.06) 499.86 (388.37) 312.47 (507.21)

720.45 (340.87) 518.81 (374.18) 201.64 (494.88)

768.45 (302.15) 579.83 (391.07) 188.62 (501.83)

845.60 (283.00) 544.58 (368.06) 301.01 (462.16)

913.72 (329.14) 607.99 (386.73) 305.72 (512.78)

880.65 (263.89) 518.16 (350.94) 362.49 (436.99)

3699.74 (742.87) 1143.32 (812.95) 2556.42 (1006.11)

3883.60 (766.46) 1323.30 (812.30) 2560.30 (937.68)

3697.90 (740.08) 1478.92 (840.97) 2218.98 (1033.11)

3797.40 (712.43) 1367.52 (822.15) 2429.88 (1005.73)

3988.01 (645.42) 1542.65 (790.13) 2445.36 (945.67)

4118.68 (721.66) 1470.14 (786.76) 2648.53 (1022.31)

2746.64 (185.97) 598.98 (735.38) 2147.65 (793.28)

2741.57 (203.53) 829.22 (847.05) 1912.35 (909.98)

2773.98 (233.90) 869.00 (840.31) 1904.98 (928.69)

2790.70 (205.75) 925.24 (873.76) 1865.46 (956.49)

2898.70 (255.88) 752.53 (732.82) 2146.17 (832.68)

5729.92 (436.48) 1148.59 (966.23) 4581.33 (1086.66)

5603.41 (566.45) 1427.71 (1031.62) 4175.70 (1207.00)

5721.96 (568.82) 1304.66 (974.87) 4417.30 (1114.79)

5812.25 (483.47) 1576.25 (1007.32) 4236.00 (1132.16)

5974.50 (501.48) 1361.86 (972.68) 4612.64 1082.65

41.69 (23.49) 30.39 (27.71) 320.79 (362.83)

36.77 (23.19) 25.80 (28.10) 190.30 (263.63)

35.95 (24.30) 28.88 (29.84) 183.98 (237.13)

46.93 (25.51) 44.54 (35.09) 202.36 (216.34)

38.36 (27.40) 34.39 (33.84) 172.78 (219.39)

29.84 (7.49) 23.05 (8.50) 189.90 (122.94) 3951

32.23 (9.61) 21.78 (8.73) 198.79 (130.73) 3133

30.54 (7.85) 25.72 (11.89) 205.70 (132.01) 435

30.71 (8.41) 25.79 (8.49) 194.35 (104.21) 1228

31.47 (10.29) 29.81 (12.35) 203.27 (167.36) 820

Out-of-state cost measures Public four-year college: Tuition 2754.63 (205.80) Predicted aid 502.56 (722.82) Net cost 2252.08 (792.96) Private four-year college: Tuition 5671.74 (543.28) Predicted aid 1023.01 (1006.63) Net cost 4648.73 (1158.11) Home state slots Public four-year/1000 37.28 (22.51) Private four-year/1000 28.72 (29.31) Two-year/10,000 186.14 (235.62) Out-of-state slots Public four-year/1000 31.00 (8.56) Private four-year/1000 23.18 (9.27) Two-year/10,000 212.82 (138.46) Number of observations 6984 a

Private four-year

Data are weighted. Standard deviations in parentheses.

558

M.J. Hilmer / Economics of Education Review 20 (2001) 551–562

Table 4 Marginal effects for college attendance equationa Two-year students

Public four-year In-state

Received aid from parents In-state cost measures Net two-year Net public four-year Net private four-year

0.3022** (49.02)

0.0207** (23.80)

⫺0.0137** (⫺4.82) ⫺0.0001 (⫺0.92) 0.0024** (3.78)

0.0021 (⫺1.27) ⫺0.0066** (⫺3.41) 0.0003** (2.18)

Out-of-state cost measures Net public four-year ⫺0.0019** (⫺3.51) Net private four-year 0.0014 (1.52) In-state slots Public four-year/1000 ⫺0.0007 (⫺1.00) Private four-year/1000 0.0003 (0.84) Two-year/10,000 0.0003** (5.14) Out-of-state slots Public four-year/1000 ⫺0.0013 (0.16) Private four-year/1000 0.0013 (0.35) Log-likelihood Number of observations

Private four-year Out-of-state 0.0575** (33.18)

In-state

Out-of-state

0.0300** (29.07)

0.0742** (9.56)

0.0010 (0.76) 0.0008 (1.34) ⫺0.0004* (⫺1.90)

0.0005 (⫺0.97) ⫺0.0004 (⫺1.03) 0.0007** (3.35)

⫺0.0026** (⫺5.42) 0.0021** (3.70) 0.0002** (2.57)

⫺0.0005** (⫺2.63) 0.0006 (1.28)

⫺0.0001* (⫺1.67) ⫺0.0003 (⫺1.36)

⫺0.0013** (⫺5.88) ⫺0.0006 (⫺1.56)

0.0003 (0.82) ⫺0.0003 (⫺1.28)

0.0014** (2.57) ⫺0.0009** (⫺2.03) ⫺0.0002** (⫺4.78)

⫺0.0003** (⫺3.44) 0.0000 (0.74) 0.0000* (1.79)

⫺0.0004** (⫺2.02) 0.0008** (5.59) 0.0000 (⫺1.56)

⫺0.0002** (⫺2.46) 0.0002** (2.36) 0.0000 (0.13)

0.0022** (4.59) ⫺0.0044** (⫺6.64)

⫺0.0002 (⫺0.71) 0.0005** (4.07) ⫺16,148.68 16,551

0.0003** (2.05) 0.0005 (1.26)

0.0004** (4.22) 0.0011** (8.91)

a

Marginal effects are the derivatives of the probability function evaluated at the sample means for continuous variables, and the difference between 0 and 1 for dummy variables. Estimation also includes dummy variables that are equal to one if values for a variable are missing (in which case those variables are set to zero). Z-scores in parentheses. Data are weighted. **,*Significant at the 5 and 10% levels.

to different types of institutions also have significant effects on student attendance decisions. The higher the access to in-state two-year colleges the more likely the student is to attend a two-year college, while the higher the access to in-state public and private four-year colleges the more likely the in-state student is to choose that type of institution and the less likely to choose the opposite type of institution. Likewise, the higher the access to in-state public and private four-year colleges the less likely the student is to attend those institutions in different states. The results in Table 4 suggest that students actively respond to changes in net attendance costs. However, with redistributive fee increases, only students who initially do not receive financial aid observe increases in net attendance costs while students who initially do receive financial aid do not observe any change in net attendance costs. Hence, the results discussed above do

not clearly demonstrate the expected effects of redistributive fee increases, as they do not distinguish between the two groups. Obviously then, to examine the effects of redistributive fee increases, we need to separate financial aid recipients from non-recipients. As a first step in this process, Table 5 presents differences in summary individual characteristics across the six attendance paths for financial aid recipients and non-recipients. This differentiation reveals interesting differences between the two groups. Among financial aid recipients, nonattendees, two-year attendees, and in-state public and private four-year attendees are more likely to be female and Hispanic than non-recipients while out-of-state public and private four-year attendees are less likely to be female and Hispanic. This is not surprising given the evidence that Hispanic students disproportionately benefit from access to two-year colleges (Ganderton & Santos, 1995). At the same time, financial aid recipients in all

M.J. Hilmer / Economics of Education Review 20 (2001) 551–562

559

Table 5 Summary comparison of financial aid recipients and non-recipientsa Do not attend

Two-year students

Public four-year

In-state Received financial aid Male Female Black Hispanic Other race White College prep Test scores HS grades Parent college Family income Family size Number of observations Received no aid Male Female Black Hispanic Other race White College prep Test scores HS grades Parent college Family income Family size Number of observations a

Private four-year

Out-of-state

In-state

Out-of-state

0.3137 0.6863 0.1638 0.1570 0.0112 0.6681 0.2839 0.4901 (0.0876) 2.6257 (0.7487) 0.0685 3.4364 (1.6154) 3.5172 (1.7096) 219

0.4146 0.5854 0.1430 0.1171 0.0277 0.7122 0.4431 0.5133 (0.0847) 2.8830 (0.6324) 0.1096 3.6134 (1.7086) 3.4892 (1.7077) 1134

0.4289 0.5711 0.1523 0.0717 0.0228 0.7531 0.7091 0.5603 (0.0814) 3.2794 (0.6118) 0.2003 3.9915 (1.7829) 3.4925 (1.6332) 1510

0.5359 0.4641 0.1719 0.0503 0.0198 0.7580 0.7719 0.5713 (0.0812) 3.2337 (0.5746) 0.3043 4.4275 (1.7321) 3.4167 (1.5098) 205

0.4502 0.5498 0.0950 0.0664 0.0204 0.8182 0.8165 0.5736 (0.0769) 3.3096 (0.5955) 0.1741 4.3599 (1.7017) 3.4372 (1.5676) 790

0.4919 0.5081 0.0881 0.0420 0.0132 0.8566 0.8665 0.5844 (0.0792) 3.2945 (0.6297) 0.3692 4.5379 (1.7275) 3.3858 (1.5713) 424

0.5166 0.4834 0.1223 0.1479 0.0177 0.7122 0.1923 0.4584 0.0760 2.3518 (0.7171) 0.0662 3.8357 (1.7615) 3.6038 (1.8074) 6765

0.4697 0.5303 0.0717 0.1037 0.0292 0.7954 0.4300 0.5041 0.0795 2.6727 (0.6118) 0.1617 4.6694 (1.6640) 3.2801 (1.6001) 2817

0.5088 0.4912 0.0621 0.0587 0.0228 0.8564 0.6787 0.5529 0.0749 3.0116 (0.5983) 0.2791 5.1849 (1.5495) 3.2593 (1.4947) 1623

0.4604 0.5396 0.0999 0.0262 0.0396 0.8343 0.7367 0.5622 0.0743 2.8756 (0.6674) 0.3195 5.4393 (1.6401) 3.2011 (1.4171) 230

0.4692 0.5308 0.0704 0.0545 0.0115 0.8636 0.7277 0.5478 0.0787 2.9203 (0.5590) 0.3221 5.1166 (1.7654) 3.1922 (1.4588) 438

0.4764 0.5236 0.0299 0.0474 0.0174 0.9053 0.8289 0.5695 0.0766 3.0229 (0.6020) 0.4444 5.6704 (1.5299) 3.2197 (1.4109) 396

Data are weighted. Standard deviations in parentheses.

five attendance groups are less likely to be white and have, as expected, lower average family incomes and parental education and higher average family sizes than non-recipients. Comparing public four-year attendees to public two-year attendees suggests marked cross-gender and ethnicity differences in the likelihood of financial aid receipt at the different types of institutions. Namely, at public four-year institutions (both in- and out-of-state) financial aid recipients are more likely to be female, black, and Hispanic and less likely to be other race and white than at private four-year colleges. This difference is especially large for black students, for whom the per-

centage of public four-year attendees receiving financial aid is nearly twice as large as the percentage of private four-year attendees.16 Hence, the differences in summary statistics across the two groups suggest that changes in net attendance costs for one group and not the other

16 For this reason the average predicted net attendance cost for blacks at public four-year colleges is negative while for all other groups the predicted net attendance costs at all six types of institutions are positive.

560

M.J. Hilmer / Economics of Education Review 20 (2001) 551–562

might significantly affect the composition of students at public four-year colleges. To further explore this possibility, we use the multinomial logit results to calculate predicted outcomes for the group of students who will be affected by the proposed policy. As the redistributive fee policies we are considering will be enacted by in-state public four-year colleges, the students who will be affected are in-state public four-year attendees who do not receive financial aid. Therefore, to examine the potential effects that such policies have on the distribution of students at public four-year institutions, we need to consider how the policy affects the attendance decisions of students who initially choose to attend in-state public four-year colleges. We do this by: (1) calculating predicted attendance probabilities for a hypothetical student possessing characteristics equal to the average characteristics of a public four-year non-recipient, (2) increasing net attendance costs in accordance with the redistributive fee increase policy and recalculating the attendance probabilities, and (3) examining the difference between the predicted attendance probabilities. This exercise replicates the desired experiment of taking a random hypothetical public four-year attendee who does not receive financial aid, increasing his or her in-state public fouryear net attendance costs, and observing how that affects his or her college attendance decision. Table 6 presents the results of this experiment. An important facet of the calculations is whether we consider the policy changes as occurring in a vacuum in which only the in-state public four-year net attendance costs are increased, or whether we consider the changes as occurring in a world in which the net attendance costs of the alternative attendance paths are also increasing. To put bounds on the possible expected outcomes of redistributive fee increases, the top and bottom panels of Table 6 present the results calculated under each of those scenarios. The signs of the results are generally similar across the two sets of results. The lone difference in signs is that if only in-state public four-year costs increase students are more likely to attend out-of-state public four-year colleges while if all attendance costs increase students are less likely to choose that path. The biggest difference between the two sets of results is in the magnitude of the predicted changes. Increasing only in-state public four-year costs results in larger predicted decreases in the probability of in-state public four-year attendance and increases in the probability of out-of-state private four-year attendance. Increasing all attendance costs results in larger increases in the probability of twoyear attendance and larger decreases in the probability of in-state four-year attendance. These differences are not at all surprising, however. As Ganderton (1992) demonstrates, students often choose to attend lower quality public universities rather than higher quality private universities due to the in-kind subsidy provided by the former.

Because increases in net attendance costs at in-state public four-year colleges represent decreases in the in-kind subsidy, we might expect students to respond by switching back to the higher quality private institutions which are likely out-of-state. Clearly, this effect should be larger when the in-state public four-year costs are the only ones increasing. Similarly, when all attendance costs increase we would expect students to be more likely to attend two-year colleges as those institutions represent low-cost alternatives to the first two-years of college attendance. Whichever view of the world we have, both sets of results suggest that redistributive fee increases might be expected to have significant effects on the composition of the student body at in-state public four-year colleges. As expected, the results demonstrate that financial aid non-recipients of all genders, ethnicities, and ability levels would be expected to respond to such policies by switching away from the in-state public four-year colleges. The primary beneficiaries of such policies appear likely to be two-year colleges and out-of-state private universities. A potentially troubling finding is that students of each type are also predicted more likely to respond by foregoing college attendance altogether. Focusing on the change to the probability of in-state public four-year attendance, the predicted distributional effects differ depending on which attendance costs are increased. If only the in-state public four-year net costs are increased, the biggest decreases are for male, white and high-test score students. In fact, high-test students are predicted to be nearly three times more likely than low-test students to defect from the public four-year college, suggesting that redistributive fee increases might substantially alter the overall ability level of students at the public university. This is cause for concern if there are positive cohort effects associated with a higher quality student body. If all attendance costs are increased, the results are markedly different. In that case, the biggest decreases in the probability of in-state public four-year attendance are for black, Hispanic, and other race students. Again this is a potential cause for concern to policymakers as it suggests that the proposed policies might have the undesirable effect of decreasing the diversity of the student body at a state’s public four-year colleges. Either way, the results suggest that state policymakers should consider the effects that redistributive fee increases might have on the composition of the public four-year college’s student body before enacting such policies.

4. Conclusion Due to the combination of increasing educational costs and decreasing public support, state policymakers have been forced to consider new methods of raising revenue

M.J. Hilmer / Economics of Education Review 20 (2001) 551–562

561

Table 6 Predicted attendance probabilities for average in-state public four-year attendees not receiving financial aid a Do not attend

Two-year Students

Public four-year

In-state Change due to increase in in-state public four-year: net cost only. Male 0.0066 0.0040 ⫺0.0613 Female 0.0064 0.0060 ⫺0.0588 Black 0.0129 0.0070 ⫺0.0568 Hispanic 0.0119 0.0083 ⫺0.0490 Other race 0.0034 0.0073 ⫺0.0596 White 0.0059 0.0042 ⫺0.0611 High test score 0.0025 0.0023 ⫺0.0726 Low test score 0.0201 ⫺0.0026 ⫺0.0281 Change due to increases in all in-state and out-of-state: net costs. Male 0.0084 0.0325 ⫺0.0153 Female 0.0089 0.0353 ⫺0.0126 Black 0.0056 0.0354 ⫺0.0236 Hispanic 0.0082 0.0343 ⫺0.0224 Other race 0.0008 0.0506 ⫺0.0309 White 0.0090 0.0331 ⫺0.0119 High test score 0.0040 0.0315 ⫺0.0106 Low test score 0.0144 0.0146 ⫺0.0146

Private four-year

Out-of-state

In-state

Out-of-state

0.0125 0.0118 0.0153 0.0055 0.0185 0.0123 0.0159 0.0033

⫺0.0031 ⫺0.0027 ⫺0.0018 ⫺0.0017 ⫺0.0021 ⫺0.0032 ⫺0.0042 ⫺0.0017

0.0413 0.0373 0.0235 0.0250 0.0326 0.0419 0.0561 0.0089

⫺0.0148 ⫺0.0142 ⫺0.0136 ⫺0.0068 ⫺0.0201 ⫺0.0151 ⫺0.0184 ⫺0.0052

⫺0.0296 ⫺0.0339 ⫺0.0172 ⫺0.0248 ⫺0.0242 ⫺0.0333 ⫺0.0361 ⫺0.0124

0.0189 0.0164 0.0135 0.0116 0.0239 0.0182 0.0295 0.0032

a Values calculated by inserting sample average values for each subset of students into estimated college attendance equation. High and low test scores refer to students with test scores more than one standard deviation above and below the mean. In-state net public four-year attendance costs increased by one standard deviation (US$501.82). Net attendance costs for alternative attendance paths increased by the same percentage (155%).

without pricing economically disadvantaged students out of higher education. One proposal is to increase tuition for all public four-year college attendees and return a portion of the revenue to financially constrained students in the form of offsetting financial aid increases. This study examines the potential effects that such a policy might have on student attendance decisions. The results suggest that increasing the net costs of in-state public four-year attendance for students who do not receive financial aid might significantly alter the composition of a state’s public four-year colleges with high-test, black, and Hispanic students potentially being the most likely to leave such institutions.

References Barron’s College Division (1984). Profiles of American colleges. Woodbury, CT: Barron’s Educational Series. Behrman, J. R., Kletzer, L. G., McPherson, M. S., & Schapiro, M. O. (1992). The college investment decision: direct and indirect effects of family background on choice of post-secondary enrollment and quantity. USC Economics — Arthur Anderson Working Paper Series No. 9201. Bound, J., & Johnson, G. (1992). Changes in the structure of wages in the 1980s: an evaluation of alternative explanations. American Economic Review, 82, 371–392.

Bowen, W. G., & Bok, D. (1998). The shape of the river: longterm consequences of considering race in college and university admissions. Princeton, NJ: Princeton University Press. Brewer, D. J., Eide, E. R., & Ehrenberg, R. G. (1999). Does it pay to attend an elite private college? Cross-cohort evidence on the effects of college type on earnings. Journal of Human Resources, 34, 104–123. Brewer, D. J., Eide, E. R., & Goldhaber, D. D. (1999). An examination of the role of student race and ethnicity in higher education since 1972. Paper Presented at the CHERI Higher Education Policy Research Conference, Cornell, NY. Broomhall, D. E., & Johnson, T. G. (1994). Economic factors that influence educational performance in rural schools. American Journal of Agricultural Economics, 76, 557–567. Comay, Y., Melnick, A., & Pollatschek, M. A. (1973). The option value of education and the optimal path for investment in human capital. International Economic Review, 14, 421–435. Commission on National Investment in Higher Education (1997). Breaking the social contract: the fiscal crisis in higher education. ERICFNO-ED414806. Eide, E. R., & Grogger, J. T. (1995). Changes in college skills and the rise in the college wage premium. Journal of Human Resources, 30, 280–310. Ganderton, P. T., & Santos, R. (1995). Hispanic college attendance and completion: evidence from the high school and

562

M.J. Hilmer / Economics of Education Review 20 (2001) 551–562

beyond surveys. Economics of Education Review, 14, 35– 46. Ganderton, P. T. (1992). The effect of subsidies in kind on the choice of a college. Journal of Public Economics, 48, 269–291. Greene, W. H. (1997). Econometric analysis. (3rd ed). New York: Macmillan. Hausman, J. A., Lo, A. W., & MacKinlay, A. C. (1992). An ordered probit analysis of transaction stock prices. Journal of Financial Economics, 31, 319–379. Hilmer, M. J. (1998). Post-secondary fees and the decision to attend a university or a community college. Journal of Public Economics, 67, 329–348. Hilmer, M. J. (1997). Does community college attendance provide a strategic path to a higher quality education? Economics of Education Review, 16, 59–68. Katz, L. F., & Murphy, K. M. (1992). Changes in relative wages, 1963–1987: supply and demand factors. Quarterly Journal of Economics, 107, 35–78. Leslie, L. L., & Brinkman, P. T. (1987). Student price response in higher education. Journal of Higher Education, 58, 181–204. Maddala, G. S. (1983). Limited dependent and qualitative variables in econometrics. Cambridge, UK: Cambridge University Press. Manski, C. F. (1989). Schooling as experimentation: a reappraisal of the post-secondary phenomenon. Economics of Education Review, 8, 305–312.

Manski, C. F., & Wise, D. A. (1983). College choice in America. Cambridge, MA: Harvard University Press. McPherson, M. S., & Schapiro, M. O. (1991). Does student aid affect college enrollment? New evidence on a persistent controversy. American Economic Review, 81, 309–318. National Center for Education Statistics (1996). Remedial education at higher education institutions. Washington, DC: US Department of Education. National Center for Education Statistics (1987). In High school and beyond 1980 senior cohort third follow-up (1986). Data file users manual, Vols. I and II. Washington, DC: US Department of Education. Office of Educational Research and Improvement (1997, 1985). Digest of educational statistics. Washington, DC: US Department of Education. Ordovensky, J. F. (1995). Effects of institutional attributes on enrollment choice: implications for postsecondary vocational education. Economics of Education Review, 14, 335–350. Savoca, E. (1991). The effect of changes in the composition of financial aid on college enrollments. Eastern Economic Journal, 17, 109–121. Weiler, W. C. (1989). A flexible approach to modelling enrollment choice behavior. Economics of Education Review, 8, 277–283.