Teacher Credentials and Student Achievement Presentation to Civil Rights Research Roundtable on Education March 10-11, 2011 Helen F. Ladd Duke University and CALDER Hladd@pps.duke.edu 1
Key Questions Do teacher credentials affect student achievement? Lively policy debate Which credentials matter? Early work. Only agreement is teacher experience in the early years. Are the estimated effects large or small? Room for debate. Compared to what? Why relevant for policy? My particular concern: Uneven distribution of teachers as defined by their credentials across schools and students defined by race or poverty. 2
Research by Charles Clotfelter, Helen Ladd and Jacob Vigdor based on North Carolina data Teacher-student matching the assessment of teacher effectiveness. Journal of Human Resources, 2006. How and why teacher credentials matter for student achievement. CALDER and NBER working paper. Teacher credentials and student achievement: Longitudinal analysis with student fixed effects. Economics of Education Review, 2007. Teacher credentials and student achievement in high school: A cross subject analysis with student fixed effects. CALDER and NBER working paper; Journal of Human Resources, 2010. 3
NC Data Grades 3-8. End-of-grade (EOG) tests; math and reading. High school. End-of course (EOC) tests in 10 subjects (we use 5 subjects designed for grades 9 and 10). Linked to NC standard course of study At high school level, tests count for 25 percent of student s grade in the course. => incentive for teachers to teach the material and for students to learn it. But we can identify correct teachers only in grades 3-5 and in high school => no estimates for middle schools. 4
Key challenge: systematic sorting of teachers and students Strategies to address this problem Elementary level Cross sectional (5 th grade) include school fixed effects Longitudinal (3 rd, 4 th and 5 th grades) include student fixed effects High school Multiple subjects. Include student fixed effects. Note. All test scores are normalized by grade and year with mean 0 and standard deviation =1. Coefficients are fractions of a standard deviation. 5
Teacher experience From full models, lower bound estimates for elementary Years of exp. Elementary (base = 0) Math Reading High school 1-2 0.057 0.032 0.050 3-5 0.072 0.046 0.061 6-12 0.079 0.053 0.061 13-20 0.082 0.062 0.059 21-27 0.092 0.067 0.062 >27 0.084 0.062 0.043 6
Licensure status Elem. math (lower bound) High school Regular (base) --- --- Lateral entry -0.033-0.061 Other -0.033-0.046 7
National Board Certification Elem. Math High school (lower bound) NBCT -2 0.024 Pre. 0.022 NBCT -1 0.018 --- NBCT current 0.018 0.048 NBCT has 0.022 0.051 8
Master s degree MA before teaching MA 1-5 years into teaching MA degree 5+ years into teaching Elem. Math High school (lower bound) -0.001 (not sig.) -0.006 (not sig.) 0.004 (not sig.) 0.009 (sig.) -0.010 (sig) 0.009 (sig.) 9
Teacher test scores Teacher test score Math score for alg. and geo. Elem. Math (Lower bound) 0.011 0.011 High school 0.031 Biology 0.013 Other science for bio. ELP English 0.009-0.013 unexp. -0.015 unexp 10
Teacher certification -- high school Alg. and geo. : Teacher certified in math 0.120 Alg and geo.: Teacher certified but not in math Biology Teacher certified but not in biology ELP teacher certified in a related subject ELP teacher certified in some other subject English certified in some other subject 0.053 0.066 0.082 0.100-0.145 unexp. 11
Magnitude of the effects Bundles of characteristics add up the effects Elementary teachers. Strong vs. weak set of credentials Math 0.150 to 0.206 Reading 0.081 to 0.120 High school teachers 90 th to 10 th percentile of the predicted achievement distribution 0.183 12
Are the effects large or small? Large relative to reductions in class size. Comparable in magnitude to estimated effects of SES disadvantage except for elementary reading. But still a lot of unexplained variation in teacher effectiveness. 13
Policy implications Many credentials matter but not master s degrees at elementary level. Effects are relatively large Important because disadvantaged and minority students are far more likely than more advantaged students to have teachers with weak credentials. 14
Teachers with less than 3 years of experience Percent by poverty quartile and school level (NC data 2004. Quartile 1 is high poverty; quartile 4 is low poverty) 25 20 15 Elementary Middle High 10 Quartile 1 Quartile 2 Quartile 3 Quartile 4 15
Teachers who are National Board Certified Percent by poverty quartile and school level (NC data 2204. Quartile 1 is high poverty; quartile 4 is low poverty ) 10 8 Elementary 6 Middle High 4 2 Quartile 1 Quartile 2 Quartile 3 Quartile 4 16
Algebra 1 Exposure of students by race to teachers with weak qualifications Probability of having teacher with given qualification. ( NC data; from Clotfelter, Ladd and Vigdor, 2010) Novice teacher Lateral entrant Teacher test score more than 1 s.d. below the mean Blacks 20.2 7.2 8.3 Hispanics 16.7 5.7 5.8 Whites 16.8 4.0 3.8 Black males 21.1 7.8 8.6 White females 16.4 3.9 4.0 17
Conclusion Many specific teacher credentials matter for student achievement Bundling of teacher characteristics can lead to quite large achievement effects Poor and minority students tend to have teachers with weaker qualifications than white or more affluent students. 18
Are the elementary school teacher effects large? Much larger than effects of class size Estimated effect of reducing class size by 5 students Math +0.010 / +0.025 Reading +0.010 / +0.020 More comparable to effects of parent education (model 2) Math Reading Parents are college grads (base) -- -- Parents are HS grads -0.112-0.114 Parent are HS drop outs -0.223-0.254 Compare teacher effects (model 5) -0.206-0.122 19
Standard value-added model Definition: A it = k + aa it-1 + b TQ it + c SC + error it where A = student achievement; and TQ = teacher qualifications SC = student characteristics Justification: Education is a cumulative process a = estimate of persistence of knowledge from one year to the next. a =1 => complete persistence (no decay) a = 0 => no persistence (100 percent decay) b = estimate of the effects of the qualifications of the student s teacher in year t on her achievement in year t. Alternative form. Gain = (A it -A it-1 ) = k + b TQ it + c SC it + error it 20
Baseline teacher Illustrative differences in achievement between subject and baseline teacher (Math and reading, models 4 and 5) Subject teacher (weak credentials) Math Reading 10 years of experience No experience -0.079/-0.094-0.053/-0.072 Competitive undergraduate college Licensure test score is average Non-competitive undergraduate college Licensure test is 1 SD below average -0.007/-0.010 * / * -0.011/-0.015-0.003/-0.004 Regular license Other license -0.034/-0.058-0.018/-0.024 Has advanced degree No advanced degree */ * +0.004/+0.008 National Board Certified Not Board Certified -0.020/-0.028-0.012/-0.012 Total difference -0.150 / -0.206-0.081/ -0.122 21