Teacher Credentials and Student Achievement in High School: A Cross Subject Analysis with Student Fixed Effects

1
Teacher Credentials and Student Achievement in
High School A Cross Subject Analysis with
Student Fixed Effects

• Charles T. Clotfelter
• Helen F. Ladd
• Jacob L. Vigdor
• Presentation at CALDER Conference, Oct. 4, 2007

2
Motivation

• Focus on teacher credentials.
• Lively policy debate about whether teacher
  credentials are predictive of student
  achievement.
• Credentials are potentially important policy
  levers.
• Focus on high schools
• Compare use of administrative data sets to
  examine teacher credentials in the elementary
  grades

3
North Carolina data

• End-of- course (EOC) tests in high schools.
• -- Preferable for this purpose to the
  comprehensive exit exams used in many states
• -- Based on standard course of study
• -- Scores count for 25 percent of a students
  grade in the course.
• gt Teachers have strong incentive to teach the
  material and students to learn it.
• Availability of data on a wide array of teacher
  credentials.

4
Approach

• We use test scores on five tests typically taken
  by 9th or 10th graders English I, algebra I,
  biology, geometry, and ELP (economic, legal and
  political systems)
• Four cohorts of students matched to their
  specific teachers
• Students in 10th grade in 1999/2000 2000/01
  2001/02 2002/03
• Concern about selection into courses

5
Models

• Preferred model.
• Student achievement in subject S
• f(teacher credentials and characteristics,
  classroom characteristics, student fixed effects)
• Student fixed effects VIP. They control for all
  subject-invariant characteristics of students,
  such as basic ability or motivation. Equivalent
  to expressing everything relative to the mean for
  that student (Some remaining technical concerns,
  but see paper)
• Alternative model.
• No student fixed effects but include student
  level time-invariant characteristics, such as
  race and gender.

6
Strategy

• Basic model and many variations on each
  credential to examine in detail the achievement
  effects of credentials.
• All test scores are normalized by subject and
  year to have mean of 0 and s.d. of 1. .
• Magnitude for purposes of comparison.
• Effects of being black with low SES -0.116
  s.d.
• (From alternative model)

7
Credentials I

• Teacher experience (base no experience)
• 2-3 years 0.050
• 6-12 years 0.061 not diff. from 0.050
• (With addition of teacher fixed effects,
  coefficients rise with experience)
• Teacher licensure (base regular license)
• Lateral entry -0.061
• other -0.046

8
Credentials II

• Masters degree (base no grad. degree)
• Received after 5 years of teaching 0.009
• Comparable results for NC elementary teachers
  - 0.010 in reading
• National Board Certification (base never
  certified)
• Pre-certification 0.022
• Cert. app. year 0.048
• Has certification 0.051 (sig. dif from
  0.022)
• Compare NC elementary results. No human capital
  effect.

9
Credentials III

• Teacher test scores (base within 1 sd of
  average)
• lt 1 sd below average -0.027
• gt 1 sd above average 0.010
• (By subject matters most for math courses)
• Teacher certification (base no certification)
• -- in subject 0.081
• -- in related subject 0.074
• -- in other subject 0.012 (not sig.)
  
• (Disaggregated. Biggest for math and biology)

10
Magnitudes

• Predicted achievement difference for teacher at
  the 90th percentile compared to a teacher at the
  10th percentile of the predicted achievement
  distribution
• 0.183 standard deviations
• Large effect
• Relative to class size reduction of 5 students
  0.013
• Relative to SES effects (from alternative model)
  0.116
• But, nonetheless, credentials explain only 1/5 to
  1/3 of the overall variation in teacher quality.

11
Distributional concerns

• Uneven distribution of teacher credentials
  across high poverty schools, defined by poverty
  quartile
• Q 1 (high) Q4 (low)
• Non reg. license 20. 5 13.3
• Lic.test score 0.057 0.117

12
Across classrooms algebra I
Novice teacher () Lateral entrant () Teacher test score lt1 sd
Blacks 20.24 7.23 8.27
Whites 16.75 4.02 3.80
13
Distributional effects

• Consider black vs. white differences in teacher
  credentials in algebra 1 multiplied by estimated
  achievement effects.
• Across a large number of credentials total
  adverse effect on black achievement lt0.02
• May seem tiny, but looms larger relative to the
  coefficient for a black student (controlling for
  other factors) of -0.057
• gt Maldistribution of teachers contributes to
  black-white achievement gap.

14
Conclusions

• Teacher credentials matter in high school
• Aggregate effects are quite large
• Some differences with findings at the elementary
  level. E.g. masters degree National Board
  Certification
• Uneven distribution of teacher credentials across
  students by race contributes to the black-white
  achievement gap.