Transcript Developmental Education in North Carolina Community Colleges

Developmental Education in North
Carolina Community Colleges
Charles T. Clotfelter
Helen F. Ladd*
Clara Muschkin
Jacob L. Vigdor
Sanford School of Public Policy
Duke University
*[email protected]
Introduction and Summary
• We use policy variation across NC community colleges
to examine effects of remediation on college success
and other outcomes.
• Findings: Remediation
– Reduces probability of college success
– Reduces the probability of passing a college level course in
the remediated field
– Has no adverse effects on short run college persistence
(Some differential effects by prior achievement quartile,
gender, and income )
Policy context and conceptual
foundation
• North Carolina Community College System
– 58 CCs (56 in this study)
– No uniform test or policy for determining college
readiness
– About 48 percent of our sample takes remedial math
and 37 percent remedial English
• Possible functions of remedial education
Developmental function
Discouragement of students
Diversion function
Prior research
Policy variation: Bettinger and Long (JHR 2009) – positive effects
Ohio colleges, both 4-year and CC
Restricted to students who took ACT
Positive effects of remediation
How relevant to CC students?
Regression discontinuity studies- small or mixed effects
Martorell and McFarlin (2011) Texas
Calagno and Long (2008) Florida
Scott-Clayton and Rodriguez (2012)
How generalizable ?
Experimental designs to look at specific programs- mixed effects
Barnett et al. (2012); Visher et al, 2012
North Carolina Data
Linked student level admin data from NCCCS
and NC public schools
 CC students can be linked to their 8th grade test
scores -- and various other variables.
Sample. All 8th graders in 1999 who
subsequently enrolled in a CC before 2006
(other than Wake and Mecklenburg)
Initial Model
Outcomeij = α + βCij + γXij + εij
(i is student, j is college attended)
Four outcomes : two related to probability of college
success, two related to short run persistence
C : enrollment in a math or reading developmental course,
different levels (only one type of course in each equation)
X : student variables, including quartile of 8th grade
achievement distribution
Parameter of interest is β
Problems with initial model
• Potential bias in estimate of β
– Downward because of left out variables
– Upward because of differential compliance
• Confounding effects at the college level
Solutions
Use an instrument for actual enrollment in a
developmental course
Add community college fixed effects
Construction of the instrument(s)
For each student:
• Use college-specific probability of enrollment that differs by
the student’s position in the 8th grade achievement
distribution (i.e. Q1, Q2, Q3 or Q4) – based on actual patterns
in each college.
• Use the college that is closest to the student’s high school
rather than the community college actually attended .
Example. The value of the instrument for a students whose high school is
closest to Wilson CC would differ depending on the student’s 8th grade
achievement quartile as follows:
Q1 = 0.91 , Q2 = 0.74, Q3 = 0.63, Q4 = 0.10
Evaluation of the instruments
First stage regression:
Cijk = ak + bZk + cXijk + eijk
(i is student, j is college attended, k is closest college)
Z is the relevant instrument.
Coefficients on instruments are typically in the range of 0.6
to 0.8, all with small standard errors, sigh large F statistics in
all cases.
=> The instruments are strong
Testing for developmental function
of remedial education
Effective requirement of remedial education
reduces probability of college success
Math
Any develop -0.198*
Course <= 70 -0.227***
Course <= 60 -0.179***
(Reduced form estimates)
English
Any develop –0.155**
Course <=85 -0.174***
Developmental function (cont.)
Effective requirement of remedial education
reduces probability of ever passing a college level
course in the remediated subject.
Math
Any develop -0.211**
Course <= 70 -0.266***
Course <= 60 -0.222***
(Reduced form estimates)
English
Any develop –0.225***
Course <=85 -0.235***
Testing discouragement
No adverse effects on short run persistence
(based on two measure of persistence).
=> Main function of remedial education is to divert
students from college level courses.
Effects by subgroup
Remediation has more negative effects on college
outcomes for :
•
Students who are in the lowest 2 quartiles of the 8th
grade achievement distribution in math; lowest
quartile in English
• Female students relative to male students in both
math and English (but females more likely to succeed
than males)
No differential effect of remediation by poverty status
(as measured by FRP lunch in school)of students (but
poor non-remediated students are less likely to succeed
than poor students)
Conclusions
• No support for the developmental function of CC
remediation. (Consistent with most studies other than
Bettinger and Long, 2009)
• No adverse effects on short run persistence
• Main effect is to divert students away from college
level courses.
May be a reasonable thing to do.
Policy implication: Assure better skills in high school
Cautionary note. Our results apply only to traditional age
students.