Download - Losing HOPE
1
LOSING HOPEFinancial Aid and the Line Between College and Work
February 2013
Celeste K. Carruthers and Umut Ozek
University of Tennessee and AIR/CALDER
2
Essential Questions• Financial aid has been shown to increase college
enrollment, persistence, and graduation1 • Caveat: results are not universal2 • Why is financial aid important for students?• Why do students work so hard to avoid losing aid? 3
• Non-pecuniary loss aversion?• Financial need?
1 Dynarski 2000,2003; Deming & Dynarski 2009; Scott-Clayton 2011; Castleman & Long 20122 Hansen 1983; Cohodes & Goodman 2012; Sjoquist & Winters 20123 Cornwell et al. 2005; Scott-Clayton 2011
3
Contributions• To date, we know that students who lose Georgia's HOPE
accumulate fewer credits and receive fewer degrees (Henry et al. 2004) and tend to be concentrated in more technical majors (Dee & Jackson 1999).
• We provide the first causal evidence of students' college/work substitution patterns in the wake of losing financial aid.• Preview: less college, more work
• Beyond the scope of this paper: normative statements on optimal scholarship retention policies.
4
The Tennessee HOPE scholarship• Financed by the state lottery• First available to fall 2004 freshman (and some
sophomores)• Up to $4,000/year grant for four-year students, $2,000 for
two-year students• Low eligibility threshold and (comparatively) high renewal
threshold.• Renewed with continuous enrollment, GPA of 2.75 after
24 hours, 3.0 after 48, 72, and 96 hours.• About 1/2 of HOPE scholars lose their scholarship
midway through college.
5
Data• Administrative records for 2003-2006 entering cohorts in
Tennessee two-year and four-year institutions, followed through fall 2008.• Scholarship status, credits attempted per term, major(s), college
GPA, demographics, distance from permanent address.
• Merged with FAFSA and ACT records for income and ACT controls.
• Merged with quarterly earnings data for every quarter (enrolled or not), 2003-2008.
• Merged with full-time undergraduate tuition and fees (IPEDS).
6
Descriptive Statistics
All4-year students
2-year students
Credit load (attempted credits this term) 13.2 14.2 11.4(3.4) (2.7) (3.9)
Different major from last term 0.15 0.17 0.12Any earnings while enrolled (half-year) 0.73 0.70 0.78Earnings while enrolled (half-year, thousands, 2005$) 2.6 2.2 3.4
(3.2) (2.9) (3.4)Exiting college: working but not enrolled next term 0.12 0.06 0.22 Undeclared major 0.23 0.30 0.13Student-years 584,925 369,044 215,881Students 116,440 73,816 67,823
Standard deviations in parentheses below continuous variable means
7
Descriptive Statistics
All 4-year students 2-year studentsLost HOPE 0.19 0.21 0.14Scholarship value (in thousands) 1.39 1.95 0.43
(1.82) (1.98) (0.88)Cumulative GPA 2.20 2.38 1.90 (1.32) (1.26) (1.36)Missing GPA 0.05 0.05 0.06Tuition and fees (000s) 4.04 4.98 2.42 (1.33) (0.60) (0.20)Male 0.44 0.45 0.43Black 0.16 0.17 0.15White 0.78 0.77 0.80Family income $<$ 60,000 0.47 0.44 0.52Missing family income 0.08 0.05 0.14Student-years 584,925 369,044 215,881Students 116,440 73,816 67,823Standard deviations in parentheses below continuous variable means
8
Empirical Strategy
• 1(losthopeit) – equal to one in semesters after HOPE loss.
• η1(t-t0)*1(beforeit) + η2(t-t0)*1(afterit) - local linear function of time until/since HOPE loss
• βt*GPAit - college GPA, effect allowed to vary by t
• αi, αt - individual and semester sequence FE
• Zitγ - time-varying student and institution controls, linear time trend.
9
Empirical Strategy
• Hit - value of HOPE scholarship in 2005$.
10
Identifying Variation in HOPE• 2004-2006 cohorts
• Deviations from mean HOPE status or value, across and within students.• Within-student, across αt.
• Within-semester sequence, across αi
• Within-student variation is from the failure to renew HOPE.
• 2003 cohort• Deviations from mean HOPE status or value, across and within
students.• This cohort was eligible to gain HOPE in 2004, then held to
renewal standards thereafter.
11
Internal Validity Challenges• Students who lose HOPE are of lower ability.
• αi controls for this class of heterogeneity.
• Students about to lose HOPE are on Yit trajectories that confound treatment with other factors.• η1(t-t0)*1(beforeit) + η2(t-t0)*1(afterit) controls for pre-loss and post-
loss trends common to students who lose the scholarship.
• Typically, GPA starts low and climbs just before HOPE loss.
12
College GPA before and after HOPE loss
13
Credits attempted before and after HOPE loss (regression-adjusted)
14
Exit rates before and after HOPE loss (regression-adjusted)
15
Results – 2004-2006 cohorts(1) (2) (3) (4)
Outcome Credits this term Changed major (0,1)
Lost HOPE (0,1)
-0.996***(0.025)
-0.006**(0.003)
HOPE value (000s)
0.684***(0.007)
0.012***(0.001)
Four-year public (0,1)
1.573***(0.117)
0.408***(0.107)
0.109***(0.010)
0.088***(0.010)
Terms until loss (negative)
0.149***(0.012)
0.390***(0.012)
0.007***(0.001)
0.014***(0.001)
Terms since loss
0.067***(0.009)
0.167***(0.009)
-0.001(0.001)
0.002*(0.001)
Tuition and fees (000s)
0.734***(0.045)
0.772***(0.040)
-0.010**(0.004)
-0.009**(0.004)
Adjusted R2 0.13 0.18 0.11 0.11
Nit = 421,615 student-years. Cluster-robust standard errors in parentheses
16
The impact of HOPE loss versus an equivalent tuition increase on attempted credits
17
Results – 2004-2006 cohorts(1) (2) (3) (4)
Outcome Any earnings this term (0,1) Earnings (000s)
Lost HOPE (0,1)
-0.006*(0.003)
0.170***(0.019)
HOPE value (000s)
0.001*(0.001)
-0.027***(0.004)
Terms until loss (negative)
0.005***(0.002)
0.005***(0.001)
0.022**(0.009)
0.042***(0.008)
Terms since loss
0.002(0.001)
0.002(0.001)
0.112(0.073)
0.032**(0.012)
Adjusted R2 0.02 0.02 0.09 0.09
Nit = 421,615 student-years. Cluster-robust standard errors in parentheses
18
Results – 2004-2006 cohorts(1) (2) (3)
Outcome Undeclared major (0,1)
Work but no college next term, and no degree (0,1)
Lost HOPE (0,1)
0.054***(0.003)
0.070***(0.002)
HOPE value (000s)
-0.017***(4.5E-05)
Terms until loss (negative)
-0.021***(0.002)
-0.034***(0.001)
-0.029***(0.001
Terms since loss
-0.002*(0.001)
-0.002***(0.001)
0.001(0.001)
Adjusted R2 0.15 0.16 0.16
Nit = 421,615 student-years. Cluster-robust standard errors in parentheses
19
Results – 2003 cohort(1) (2) (3) (4)
Outcome Credits this term
Changed major (0,1)
Earnings (000s) Leaving college (0,1)
HOPE value (000s)
0.334***(0.012)
0.013***(0.001)
-0.046***(0.008)
-0.013***(0.001)
Terms until loss (negative)
-0.040***(0.012)
0.009***(0.001)
0.014(0.010)
-0.017***(0.001)
Terms since loss
0.123***(0.018)
-0.003*(0.002)
0.014(0.018)
-0.003***(0.001)
Adjusted R2 0.11 0.11 0.12 0.12
Nit = 163,310 student-years. Cluster-robust standard errors in parentheses
20
Inference• For students who stay in college after HOPE, the first
term without HOPE is associated with • 0.996 fewer credits (7.6% of the mean)• $170 additional earnings (6.4% of the mean)• 5.4 ppt higher likelihood of not declaring a major (23.2% of the
mean)
• Little to no impact on labor force participation.• Extensive margin impacts appear much stronger than
intensive margin impacts.• In the first term without HOPE, there is a 7.0 ppt higher likelihood
of exiting college for the workforce without a degree (58.8% of the mean exit rate)
21
Heterogeneous Impacts of Losing HOPE: 2004-2006 cohorts
(1) (2) (3)
Outcome Credits this term Earnings (000s) Leaving college (0,1)
All students -0.996***(0.025)
0.170***(0.019)
0.070***(0.002)
Family income < 60,000
-1.092***(0.038)
0.150***(0.027)
0.090***(0.004)
Family income at least 60,000
-0.949***(0.036)
0.199***(0.028)
0.054***(0.003)
Started in a four-year college
-0.667***(0.028)
0.184***(0.025)
0.050***(0.003)
Started in a two-year college
-1.561***(0.056)
0.116***(0.033)
0.114***(0.005)
Nit = 421,615 student-years. Cells report “lost HOPE” coefficient estimates for different subgroups. Cluster-robust standard errors in parentheses
22
Conclusions• Loss of financial aid decreases engagement with college
and shifts the college/work margin towards work.• Attempted hours decrease.• The likelihood of declaring a major decreases.• Earnings increase.• The propensity to leave college increases.
• Findings are pronounced for the lower-income half of students, and for students who start in two-year colleges.
• Implications: money matters, especially at the extensive margin of college enrollment.