View
213
Download
0
Tags:
Embed Size (px)
Citation preview
Wisconsin Scholars Longitudinal Study:An Introduction
SARA GOLDRICK-RAB & DOUGLAS N. HARRISCo-DirectorsUniversity of Wisconsin-Madison
WWW.FINAIDSTUDY.ORG
Thanks for the Extensive Support!
• Funders: Bill and Melinda Gates Foundation, William T. Grant Foundation, Spencer Foundation• Partners: University of Wisconsin System, Wisconsin
Technical College System• Board of directors of the Fund for Wisconsin Scholars• Our staff, past, present, and future• And especially our project manager, Alison Bowman
Policy Context
• The nation faces a college completion problem• Completion rates are especially poor among
students from low-income families• Governments spend $155B+ on need-based
financial aid each year (inc. grants & loans)• To what degree is aid exerting an independent
contribution to degree completion rates?• Is it a cost-effective contribution?• Through what mechanisms do effects arise?• For which students is aid most effective?
How Financial Aid Could Help Students
• The conditions required to get the aid could provide motivation• “I will continue to enroll in college in order to get this
money.”• “I will attend full-time rather than part-time in order to get
this money.”
• The income could be used in positive or negative ways• “This money helps me work a little less and study a little
more.”• “This money makes me less worried that I can get through
college.”• “With this money, I’ll stay in a dorm where I can hang out
with my friends.”
Prior Research
• Studies suggest modest positive impacts of financial aid on college attendance and persistence for the average student• The effects seem to vary depending on the conditions
associated with the program• They also depend on who the program is serving• Isolating a causal effect of financial aid is particularly
difficult
Grant vs. No Grant Comparison
Ways to Evaluate the Effects of a Grant Program
ddf 100 students sign up to receive a grant
Grant vs. No GrantGrant
100 students sign up
No Grant 1000s of students don’t sign up
Grant vs. No GrantGrant
College Completion Rate= 70%
No Grant College Completion Rate= 60%
Grant vs. No GrantGrant
College Completion Rate= 70%
No Grant College Completion Rate= 60%
Students receiving the grants have higher college completion rates… Can we attribute that difference
to the grants?
Scholarship vs. No ScholarshipScholarship
College Completion Rate= 70%
No Scholarship College Completion Rate= 60%
The observed differences in completion rates could be due to unobserved differences among
students—which led them to get the grants.
Grant vs. No Grant
• This method compares apples to oranges• The reason for not getting the grant could be
associated with the reason for not finishing college (“selection bias”)• If we know what the reason is, we can
“control” for it• If we don’t know what it is, or can’t observe it
in the data we have, we can’t control for it• The estimated impact of the grant mixes up
these factors
Fund for Wisconsin Scholars
• A new, generous grant program• Allocates grants with a method that facilitates
rigorous estimation of impacts• Willing to participate in an evaluation
• An uncommon opportunity to help large numbers of Wisconsin college students—while generating research knowledge needed to help students nationwide
Fund for Wisconsin Scholars
• Grants to students enrolled in UW System & the Wisconsin Technical College System• Began making awards in fall 2008• Eligible students graduated from Wisconsin
public high schools, enrolled in college within 3 years of graduation, for the first-time, and received a Pell Grant• Had to have at least $1 of unmet need after all non-
repayable sources of aid were accounted for• Had to enroll full-time (12 credits) by the date of
record
Fund for Wisconsin Scholars
• Grant amounts:• $1,800 per year for students at 2-year colleges• $3,500 per year for students at universities
• Terms of renewal:• Continued receipt for up to 5 years• Transferrable among publics• Must register for 12 credits by the start of each new
term• Maintain ‘satisfactory academic progress’ (~C
average)• Terms are very comparable to the federal Pell Grant
Fund for Wisconsin Scholars
• Identification of eligible students:• Fall of first year of college (after enrollment)• FFWS works with Wisconsin’s financial aid officers• Students don’t have to sign up for a “chance” to get the grants;
they have already done the hard work of completing the FAFSA• Selection process:• Students are chosen (2-year and 4-year separately)• Potential recipients are notified• Students respond and verify their eligibility
• Award process:• After verification, payments are disbursed via financial aid
offices each term• First payment arrives by 2nd semester, year 1
Random Assignment of FFWS Grants
All eligible students have an equal chance of receiving a grant
A coin toss helps ensure recipients and non-recipients are equivalent groups
FFWS Grant vs. No FFWS Grant
• This method compares apples to apples• The reason for not getting the grant could NOT be associated
with the reason for not finishing college• The estimated impact of the FFWS grant is clean– if we
observe differences in students’ outcomes they are due to the grant
How Students Get the FFWS Grant
• They are enrolled and already have financial aid• The FFWS grant must go to the financial aid
office & by law the student’s package must be adjusted if receiving the grant means their financial aid exceeds the institution’s cost of attendance• Usually this means reducing loans and/work-study
money; schools agreed not to reduce institutional aid
• So they get the grant as loan reduction, cash refund, or a combination, depending on the size and composition of their initial aid package
Wisconsin Scholars Longitudinal Study
• What are the average effects of the FFWS grant on college attainment?• For which students is the grant most effective?• Under what conditions is the grant most
effective? (e.g. loan vs. cash, type of college)• How does the FFWS grant affect students?• How does it change how they spend their time?• In what ways does it alter their relationships
with other people?• How does it affect how they think and feel?
Wisconsin Scholars Longitudinal Study
• Includes 3,000 students who were eligible for the FFWS grant in Fall 2008• 1,200 students randomly chosen to receive the
grant• 1,800 students not chosen (at random) to
receive the grant• The sample is split between universities and 2-
year colleges• We also plan to include students who will be
eligible for the FFWS grant in Fall 2012
Wisconsin Scholars Longitudinal Study
• We observe students’ outcomes with the following kinds of data:• National Student Clearinghouse (enrollment
anywhere)• College transcripts• Surveys (administered yearly, by mail)• Interviews (conducted in-person, every
semester, with a stratified sample of 36 students at 4 universities)
• We also plan to consider employment and earnings outcomes
Questions?
• About the study’s research questions?• Design?• Data?
Education Impacts of the Wisconsin Scholars Grant on University Students:2008-2011
SARA GOLDRICK-RAB, DOUGLAS N. HARRIS,JAMES BENSON, & ROBERT KELCHENUniversity of Wisconsin-Madison
WWW.FINAIDSTUDY.ORG
Sample
• 1,500 students who enrolled at one of Wisconsin’s 13 public universities in Fall 2008• All students in this analysis received a Pell Grant
in that term• We compare the outcomes of 600 students
randomly selected to be offered the FFWS grant and 900 eligible students who were not chosen
Data
• Enrollment: National Student Clearinghouse (2008-2011)• Obtained for all students • Captures enrollment anywhere in the U.S., if NSC can
find a “match.” Our “match” rate ~98%.
• Credits & GPA (2008-2010)• Financial aid packages (2008-2010)• Student surveys (2008, 2009)• In-depth interviews (50 students, every 6 months, 2008-
11)
Were Students Offered and Not Offered the Grant Similar Before the Grant Was Awarded?
Full Sample Non-Recipient
Mean
Treatment Difference
P-Value
% Female 57.3 56.7 1.0 .711
% Minority 24.6 24.0 1.1 .681
Average Age 18.2 18.2 0.0 .942
% First Gen 53.4 53.5 -0.2 .958
% Dependent 97.3 97.0 0.5 .572
EFC ($) 1,633 1,603 53 .669
% $0 EFC 30.6 31.9 -2.3 .362
Parent AGI ($) 29,963 29,403 1,014 .314
% 1st time in college 95.9 95.7 0.5 .681
Terms prior enroll 1.8 1.8 -0.1 .856
N 1500 900 600
Question 1: How did the FFWS grant affect students’ financial resources?
Impact on Aid Package: Year 1 (2008-2009)
CONTROL TREATMENT IMPACT ($3,500)
Total Aid $11,426 $1,665***
Pell (%) 99.8 - 0.1
State Grant (%) 99.0 - 3.7 ***
SEOG (%) 63.6 -9.8***
ACG (%) 80.4 0.3
Institutional Aid (%) 54.7 1.6
Work Study (%) 18.3 - 5.0 *
Sub. Staff (%) 77.9 - 11.1***
Unsub Staff (%) 38.9 - 3.8
Total loans ($) 3428.3 -909.5
93% of treatment group received FFWS this year.Aid amounts are unconditional on receipt—impacts reflect supplanting + student decisions
Impact on Aid Package: Year 2 (2009-2010)
CONTROL TREATMENT IMPACT ($3,500)
Total Aid $10,082 $814.3 **
Pell (%) 74.3 - 0.6
State Grant (%) 67.2 0.5
SEOG (%) 39.2 1.2
ACG (%) 24.1 1.5
Institutional Aid (%) 48.1 -5.4
Work Study (%) 16.3 -3.6
Sub. Staff (%) 66.7 -7.2*
Unsub Staff (%) 45.6 -9.8**
Total loans ($) 3581.2 -830.9
67% of treatment group received FFWS this year. Aid amounts are unconditional on receipt.
Summary
• About 1/3 of the amount of the FFWS grant was used to reduce students’ loans and work-study (as well as replace some state grant aid)• After two years, students offered the FFWS grant had
about $1,800 less debt than students not offered the grant
Question 2: How did the FFWS grant affect academic outcomes?
Average Impacts of the Grant on Enrollment:2008-2011
Control Mean Treatment Effect
Total # terms enrolled (f/s, %) 5.192 0.05 (.09)
Ever enrolled summer (%) 21.4 0.8 (2.6)
Ever enrolled winter (%) 5.8 3.9 (1.7) **
Ever transferred (%) 23.7 -0.4 (2.7)
Ever attended 2-year college (%)
14.4 -0.4 (2.2)
Completed associate’s degree (%)
3.1 0.4 (1.1)
These are average effects, comparing FFWS recipients vs. control group
Notes: *** p<.01, **p<.05, * p<.10
Average Impacts of the Grant on Credits:2008-2010
Control Mean Treatment Effect
Credit Accumulation
Average credits completed 46.9 0.9 (1.0)
Earned 1-29 credits (%) 18.2 0.4 (2.4)
Earned 30-47 credits (%) 16.0 2.6 (2.4)
Earned 48-59 credits (%) 42.3 - 8.2 (3.1) ***
Earned 60+ credits (%) 22.2 6.3 (2.8) **
Progress toward 4-year Degree
60+ credits, 2-2.5 GPA 2.2 0.0 (0.9)
60+ credits, 2.5-3.0 GPA 6.8 -0.5 (1.6)
60+ credits, 3.0-3.5 GPA 8.5 4.7 (2.0) **
60+ credits, 3.5+ GPA 4.3 2.4 (1.4) *
Notes: *** p<.01, **p<.05, * p<.10
Effects of the Grant Varied
• We compared the effects based on a student’s propensity to persist, using pre-FFWS factors:• Age, parental education, race/ethnicity, gender, number of
siblings• Family income, assets, expected family contribution• Years of high school math and science, GPA, ACT score,
hours worked in high school, receipt of the Academic Competitive Grant
• Help from parents completing FAFSA• # credits registered for 2nd week of freshman year• Receipt of the Supplemental Educational Opportunity Grant• College attended
Propensity to Persist
• Overlapping disadvantages means that no single factor describes the category but parental education and ACT scores are some of the defining characteristics• High (94%): Parents typically have at least a bachelor’s
degree, students have higher than average test scores• Middle (82%): Parents typically have a high school
education, students have moderate test scores• Bottom (55%): Parents typically have attended some
college but not necessarily a university, students have lower test scores
Impact of the Grant on Enrollment by Propensity to Persist
2008-2009 2009-2010 2010-2011
Top: Control 99 94 94
Top: FFWS 94 83 79
Middle: Control
96 85 82
Middle: FFWS
93 84 81
Bottom: Control
94 62 55
Bottom: FFWS
97 83 72
-15
-1
+ 17
Impact of Grant on Total Number of Enrolled Terms: 2008-2011
Bottom tercile Middle tercile Top tercile0
1
2
3
4
5
6
TreatmentControl
p-=.001
Propensity to Persist
Impact of Grant on Number of Attempted Credits: 2009-2010
Bottom tercile Middle tercile Top tercile0
5
10
15
20
25
30
TreatmentControl
p-=.034
Propensity to Persist
Note: NoImpacts on AttemptsIn ‘08-09
Impact of Grant on Number of Completed Credits: 2008-2010
Bottom tercile Middle tercile Top tercile0
10
20
30
40
50
60
TreatmentControl
p-=.331
Propensity to Persist
Impact of Grant on Percent Completing 1-29 Credits: 2008-2010
Bottom tercile Middle tercile Top tercile0102030405060708090
100
Treatment
Control
p-=.025
Propensity to Persist
Impact of Grant on Percent Completing 60+ Credits: 2008-2010
Bottom tercile Middle tercile Top tercile0102030405060708090
100
Treatment
Control
p-=.424
Propensity to Persist
Summary
• On average, the grant appeared to accelerate time-to-degree for some students. This is promising since less than 30% of Pell recipients in Wisconsin complete bachelor’s degrees in 4 years.
• The grant appeared most effective for the students who were the least likely to persist in college
Are FFWS Requirements Driving the Effects?• We think this is unlikely
• The requirements are not unusual• Students think of the FFWS grant as simply as part of their
aid package – it requires exactly what Pell does (12 credits, satisfactory academic progress)
• Many think the FFWS grant requires a B average– but they think the Wisconsin Higher Education Grant does too
• Awareness of the grant and its requirements is limited:• In Fall 2008 survey just 37% of those awarded the grant
reported getting it– a year later, just 49% of those awarded it seemed to know they had it
Decisions About Full-Time Enrollment
• Interviews do not indicate that students took specific grants into account when making decisions about the number of credits to take• But they did consider overall financial constraints and the
needs of their family members• More important for determining the # credits was their
perception of how well they were doing in school
• There was an overall lack of awareness of how number of credits relates to time-to-degree• Lots of shifting from “4-year track” (15+ credits) to “5-
year track” (12-14) credits, and sometimes to <12 too
Are FFWS Resources Driving the Effects?
• Resources could have many different types of effects
• For example, increased resources could allow students to substitute study time for work time
• The increased resources could also help students meet their other obligations, for example to family members, helping them stay enrolled
• The resources could prove necessary for continued enrollment, but insufficient to help them manage a higher credit load
Who Helped Students Make Decisions?
• Students who were unlikely to persist in college were first-generation students with close family ties:• Their families may have worked extra hard to help them
succeed in college, and they may have made decisions about college (including how to spend $) with their parents (Minikel-Lacocque & Goldrick-Rab, 2011)
• Students who were very likely to persist in college came from families where attending college was more normative, these were “emerging adults”:• Students may have been afforded more
independence about how to use the grant
Policy Implications
• Targeting of social programs can maximize their cost-effectiveness• The challenge lies in identifying the best ways to target• It can be difficult to find politically feasible approaches
Policy Implications
• Costs of the Pell Grant are substantial and growing (~$20-40B)• The Obama administration wants to keep the grant maximum
high while finding other places to cut• One proposal is to require students to take 15 credits per term
(instead of 12) to get the maximum Pell ($5,500)• Only 31% of Pell recipients today take 15+ credits per term—most
are juniors or seniors.• This change means the Pell would be cut by $1000 or more for
the 41% of Pell recipients currently taking 12-14 credits per term• Assuming some of these succeeded in registering for 15+ credits,
savings would be approximately $1-2 billion per year– more, if those at 12-14 credits drop down—or dropout
• Cost-effectiveness of the approach is unclear