the impact of postsecondary remediation using a regression discontinuity approach: addressing...
TRANSCRIPT
The Impact of Postsecondary Remediation Using a
Regression Discontinuity Approach:
Addressing Endogenous Sorting and
Noncompliance
Juan Carlos CalcagnoMathematica Policy Research, Inc.
IES Research ConferenceJune 11th, 2008
Juan Carlos Calcagno, June 11th, 2008 2 :: 12
>> Introduction: Why is this Important?
Little is known about the causal effects of remediation on student outcomes:» Lack of data
» Nonrandom selection into remediation
Controversial topic in postsecondary education:
» Remediation increases access but is costly to states and colleges
» Should we pay twice for academic preparation that should have
occurred in secondary school?
» Large proportion of entering students requiring remedial courses
» Remedial credits do not count toward degrees, but students pay
tuition for these courses
Juan Carlos Calcagno, June 11th, 2008 3 :: 12
Estimate the causal impact of remediation in Florida community colleges using a unique administrative dataset.
Use a quasi-experimental design (regression discontinuity) based on a remedial placement cutoff rule to solve the selection problem.
Discuss threats to internal validity and implications for practice: noncompliance and endogenous sorting
>> Contribution of this Study
Juan Carlos Calcagno, June 11th, 2008 4 :: 12
>> FL Administrative Dataset
All first-time, degree-seeking community college students
starting in the Fall of 1997 to 2000 (over 100,000 records)
Term-by-term transcript data through Spring of 2006
All test scores reported (CPT, SAT, ACT) plus other controls
Outcomes: passing Algebra/English 101, retention,
certificate, associate degree, transfer to state 4-year college,
credits earned (remedial and non-remedial)
Juan Carlos Calcagno, June 11th, 2008 5 :: 12
>> The Regression Discontinuity Design: Intuition
Beth and Becky are observationally similar
Compare the outcomes of Beth and BeckyCompare the outcomes of Beth and Becky
Beth and Becky both take the College Placement Test (CPT)
Beth scores just above the cut-off score
Becky scores just below the cut-off score
Crossover: Beth take remediation
No-show: Becky never enroll in remediation
Beth go to college-level courses Becky go to remediation
Endogenous Sorting: Beth
retests to place out of remediation
Juan Carlos Calcagno, June 11th, 2008 6 :: 12
>> Noncompliance
Probability of Enrollment in Remediation by CPT Score and Subject
-50 -40 -30 -20 -10 0 10 20 30
CPT Score Relative to Math Cutoff-50 -40 -30 -20 -10 0 10 20 30
CPT Score Relative to Reading Cutoff
-50 -40 -30 -20 -10 0 10 20 30
CPT Score Relative to Writing Cutoff
`
CPT Score Relative to Math Cutoff CPT Score Relative to Reading Cutoff
Juan Carlos Calcagno, June 11th, 2008 7 :: 12
-50 -40 -30 -20 -10 0 10 20 30
>> Endogenous Sorting around Cutoff
Density of Reading CPT for Institution E
0.0
1.0
2.0
3.0
4
-50 -40 -30 -20 -10 0 10 20 30
Juan Carlos Calcagno, June 11th, 2008 8 :: 12
>> The Econometric Model
D: assignment to remediation (below cutoff)
is the intention-to-treat estimator (ITT)
Z: CPT test score
Robustness tests: (1) add controls; (2) narrow sample; (3) endogenous sorting; (4) functional form
& standard errors
iiii )Z(fDy ε+δ+β+α= 1
1β
is the local average treatment effect (RD-IV)
T: enrollment in remedial education
2βiiii )Z(fT̂y ε+δ+β+α= 2
iiii v)Z(fDT +γ+δ+α= 1
Juan Carlos Calcagno, June 11th, 2008 9 :: 12
ITT RD-IV ITT RD-IV ITT RD-IV ITT RD-IV
(1) (2) (3) (4) (5) (6) (7) (8)
-0.014 -0.022 -0.018 -0.030 0.006 0.012 -0.001 -0.002
(0.012) (0.020) (0.011) (0.019) (0.029) (0.057) (0.031) (0.064)
0.020 0.035 0.014 0.026 0.004 0.008 0.007 0.015
(0.012) (0.021) (0.011) (0.019) (0.032) (0.062) (0.025) (0.051)
-0.004 -0.006 -0.003 -0.006 -0.004 -0.008 -0.007 -0.014
(0.002) (0.004) (0.002) (0.004) (0.005) (0.009) (0.006) (0.012)
-0.006 -0.010 -0.006 -0.011 -0.016 -0.032 -0.014 -0.027
(0.006) (0.011) (0.006) (0.011) (0.016) (0.031) (0.012) (0.025)
-0.001 -0.002 -0.003 -0.005 -0.022 -0.043 -0.033 -0.067
(0.006) (0.010) (0.006) (0.010) (0.017) (0.033) (0.018) (0.037)
3.590** 6.169** 3.290** 5.690** 3.797* 7.453* 3.515* 7.282*
(0.657) (1.099) (0.613) (1.023) (1.698) (3.425) (1.621) (3.252)
0.233 0.400 0.011 0.019 1.398 2.744 -0.118 -0.244
(0.649) (1.113) (0.596) (1.031) (1.836) (3.622) (1.759) (3.641)
Institutions 28 28 28 28 28 28 19 19
Observations 96,724 96,724 96,724 96,724 14,493 14,493 9,593 9,593
Without Controls With Controls
All students
Transfer to State University System (SUS)
Total Credits Earned
Total Non-Remedial Credits Earned
Fall-to-Fall Persistence
Earning a Certificate
Associate Degree Completion
No-Retesting & Narrow Band Sample
Completion of First College-Level Course
Narrow Band Sample
>> The Impact of Math Remediation
Juan Carlos Calcagno, June 11th, 2008 10 :: 12
ITT RD-IV ITT RD-IV ITT RD-IV ITT RD-IV
(1) (2) (3) (4) (5) (6) (7) (8)
-0.066** -0.095** -0.060** -0.086** -0.053** -0.090** -0.028* -0.036*
(0.008) (0.012) (0.008) (0.012) (0.009) (0.017) (0.013) (0.017)
-0.009 -0.012 -0.003 -0.003 -0.017 -0.029 -0.009 -0.013
(0.008) (0.011) (0.008) (0.011) (0.010) (0.017) (0.018) (0.028)
-0.002 -0.003 -0.002 -0.002 -0.004 -0.007 -0.003 -0.005
(0.002) (0.003) (0.002) (0.003) (0.004) (0.007) (0.005) (0.008)
-0.025** -0.037** -0.020** -0.029** -0.024** -0.040** -0.020 -0.031
(0.004) (0.006) (0.004) (0.006) (0.009) (0.014) (0.017) (0.026)
-0.016** -0.024** -0.009* -0.013* -0.015* -0.025* -0.004 -0.005
(0.004) (0.005) (0.004) (0.006) (0.007) (0.011) (0.016) (0.022)
1.527** 2.266** 2.048** 3.025** 0.854 1.437 1.858 2.889
(0.447) (0.647) (0.461) (0.653) (0.496) (0.818) (1.158) (1.740)
-1.751** -2.599** -1.190** -1.758** -1.182 -2.159 -0.935 -1.590
(0.467) (0.685) (0.431) (0.636) (0.684) (1.271) (1.252) (2.124)
Institutions 28 28 28 28 28 28 7 7
Observations 97,938 97,938 97,938 97,938 37,747 37,747 8,775 8,775
No-Retesting & Narrow Band Sample
Completion of First College-Level Course
Narrow Band SampleWithout Controls With Controls
All students
Transfer to State University System (SUS)
Total Credits Earned
Total Non-Remedial Credits Earned
Fall-to-Fall Persistence
Earning a Certificate
Associate Degree Completion
>> The Impact of Reading Remediation
Juan Carlos Calcagno, June 11th, 2008 11 :: 12
>> Summary of Findings
Math remediation increases persistence to 2nd year (2 to 3.8%)
for students on the margin of passing the cutoff.
Impacts on total credits earned are positive (7 and 3 points, M
& R), but not different from zero for college-level credits.
No statistical differences for all other outcomes
The likelihood of passing English 101 was slightly lower for
reading remedial students while no difference was found in
passing Algebra 101 for math remedial students.
Juan Carlos Calcagno, June 11th, 2008 12 :: 12
>> Policy Implications
Remediation might promote early persistence in college, but it
does not necessarily help students on the margin of passing the
cutoff to make progress toward a degree.
Costs of remediation should be given careful consideration in
light of the limited benefits.
State Departments of Education should explore noncompliance
and retesting practices and consider potential consequences.