1. Human Capital Policies in Education: Further Research on Teachers and Principals 5 rd Annual CALDER Conference January 27 th, 2012

2. How teacher turnover harms student achievement Matthew Ronfeldt, Hamp Lankford, Susanna Loeb, and Jim Wyckoff January 27, 2012 CALDER Research Conference

3. Motivation Teacher turnover rates are high, particularly in schools serving low-income, non-white and low-achieving student populations – NYC about 18% of teachers leave their school each year; 30-40 percent not uncommon. Turnover may harm students – Transition confusion – Institutional memory loss – Hiring and mentoring costs But it may not – If poor teachers leave – Infusion of new ideas Surprisingly, no direct evidence on the effect

4. Research Questions What is the average effect of teacher turnover on student achievement? Are the effects different for different kinds of schools? What explains the relationship between teacher turnover and student achievement? Are the effects different for depending on who leaves?

5. What is the average effect of teacher turnover on student achievement? Typically measured at the school level – Attributes that simultaneously lead to turnover and influence achievement, e.g., principal Here we use grade level turnover – Turnover in one grade might not affect educational experiences for students in other grades – Easier to account for unmeasured factors affecting both turnover and achievement if we look at grade-by-year level turnover Allows to identification strategies – School by grade and year fixed effects – School by year and grade fixed effects

6. School by grade and year fixed effects A itgsy = β 0 + β 1 A itgs(y-1 ) + β 2 OtherA itgs(y-1) + β 3 Χ itgsy + β 4 C tgsy + β 5 S sy +ϕ y + ν gs + β 6 T gsy + ε itgsy – Compares a given grade within a school to that same grade in other years School by year and grade fixed effects: substitute ν sy for ν gs – Compares a given grade within a school to other grades in that same school in same year Consider principal turnover or other school-level shock causing turnover: – Negative effect of shock would probably be the prior year – Plus school by year effects should adjust for this (not school by grade) Consider teacher conflict in a given grade causing turnover – Negative effect of shock would probably be the prior year – School by grade effects should adjust for long-run within grade conflict Consider teacher attrition because know next year will be bad. – Don’t adjust for this, except if whole school will be bad

7. Data New York City Department of Education and the New York State Education Department. 625,000 observations of 4 th and 5 th grade students five academic years (2000-2002; 2004-2007) link student test scores in math and ELA to student, class, school, and teacher characteristics

8. measuring school-by-grade level turnover in each year “lagged attrition” -- defines turnover as the proportion of teachers in a given grade level in year t-1 who left the school by year t. “proportion new” -- defines turnover as the proportion of teachers in a given grade level who are new (movers or first year teachers) to the school in year t. Note that we are not including movement within the school to another grade as turnover (unless the teacher is replaced by a new teacher)

9. TABLE 3: Examining Measures of Turnover in Growth, Decline, and Constant Years Hypothetical: Grade 4 in School A Example Turnover Rate Using Lagged Attrition (# who left in 04-05) / (total # in 04-05) Turnover Rate Using Proportion New (# new in 05-06) / (total # in 05-06) Growth: Increase in Number of Teachers 2004-05: 6 teachers 2005-2006: 7 teachers (6 stayers, 1 mover) Turnover Rate = 0/6 Turnover Rate =0 Turnover Rate = 1/7 Turnover Rate = 0.14 Decline: Decrease in Number of Teachers 2004-05: 7 teachers 2005-2006: 6 teachers (6 stayers) Turnover Rate = 1/7 Turnover Rate = 0.14 Turnover Rate = 0/6 Turnover Rate = 0 Constant: Number of Teachers is Constant 2004-05: 6 teachers 2005-2006: 6 teachers (5 stayers, 1 mover) Turnover Rate = 1/6 Turnover Rate = 0.17 Turnover Rate = 1/6 Turnover Rate = 0.17

10. TEACHER-YEAR CHARACTERISTICSMean Experience6.03 (5.84) Proportion Stayers0.82 Proportion Movers0.04 Proportion First Years0.12 Proportion Unknown Status0.02 Observations (Teacher-Year)24917 GRADE-BY-SCHL-BY-YEAR CHARACTERISTICSMean Teachers5.95 (2.83) Turnover Rate (Lagged Attrition)0.20 (0.12) Zero Lagged Attrition0.37 (0.27) Total Lagged Attrition0.02 (0.07) Turnover Rate (Proportion New to School)0.21 (0.13) Zero New to School0.36 (0.26) Total New to School0.03 (0.09) Observations (School-Grade)1395

11. WHAT IS THE AVERAGE EFFECT OF TEACHER TURNOVER ON STUDENT ACHIEVEMENT?

12. Table 2: Effects of Teacher Turnover, Using School-By-Grade Fixed Effects TestTurnover MeasureModel 1Model 2 Math Proportion New To School-0.0709***-0.0640*** (0.010) Lagged Attrition-0.0613***-0.0590*** (0.010) ELA Proportion New To School-0.0613***-0.0511*** (0.010) Lagged Attrition-0.0294**-0.0296** (0.010) School-By-Grade Fixed Effectsxx Year Indicatorsxx Controls x

13. Table 4: Effects of Teacher Turnover, Using School-By-Year Fixed Effects TestTurnover MeasureModel 1Model 2 MathProportion New To School-0.0704***-0.0682*** (0.012) Lagged Attrition-0.0544***-0.0559*** (0.012) ELAProportion New To School-0.0723***-0.0713*** (0.013) Lagged Attrition-0.0359**-0.0387** (0.013) School-By-Year Fixed Effectsxx Grade Indicatorsxx Student, Class, School Controls x

14. Table 6: Estimates by Quartile of Turnover (Q4 0.03 worse than Q1) Turnover MeasureModel 1Model 2 Lagged Attrition Q20.0036-0.0128* Lagged Attrition Q3-0.0167**-0.0199*** Lagged Attrition Q4-0.0294***-0.0296*** Turnover MeasureModel 1Model 2 New to Grade-by-School Q2-0.0088-0.0055 New to Grade-by-School Q3-0.0192**-0.0145** New to Grade-by-School Q4-0.0309***-0.0336*** Student, Class, School Controlsxx School-by-Year Fixed Effectsx Grade Indicatorsx School-by-Grade Fixed Effects x Year Indicators x

15. HOW DOES THE AFFECT DIFFER ACROSS SCHOOLS?

16. Table 7b: Different Kinds of Schools (achievement and race), School-By-Year FEs TestTurnover MeasureHigh Ach.Low Ach.High BlackLow Black MathProportion New To School-0.0514**-0.0600***-0.0589***-0.0517** (0.019)(0.015)(0.016)(0.018) Lagged Attrition-0.0443*-0.0999***-0.0930***-0.0431* (0.018)(0.015) (0.018) ELAProportion New To School-0.0032-0.0735***-0.0557**-0.0188 (0.020)(0.016)(0.017)(0.019) Lagged Attrition-0.0317-0.1164***-0.1006***-0.0366~ (0.020)(0.016)(0.018)(0.019) School-By-Year Fixed Effectsxxxx Grade Indicatorsxxxx Student, Class, School Controlsxxxx

17. Table 8b: Different Kinds of Schools (age and size), Using School-By-Year Fixed Effects TestTurnover MeasureNewOldSmallBig MathProportion New To School-0.0368*-0.0659***-0.0547***-0.0539 (0.017)(0.019)(0.014)(0.038) Lagged Attrition-0.0407*-0.0992***-0.0425**-0.1377*** (0.016)(0.019)(0.014)(0.037) ELAProportion New To School-0.0386~-0.0404**-0.0241-0.0951* (0.023)(0.015) (0.038) Lagged Attrition-0.0568*-0.0743***-0.0503**-0.0896* (0.023)(0.015) (0.040) School-By-Year Fixed Effectsxxxx Grade Indicatorsxxxx Controlsxxxx

18. WHAT EXPLAINS THE RELATIONSHIP BETWEEN TEACHER TURNOVER AND STUDENT ACHIEVEMENT?

19. What explains the relationship between teacher turnover and student achievement? Two potential mechanisms 1.teachers who replaced those who left were either more or less effective 2.turnover itself may cause a broader disruption that impacts all students Approaches 1.Control for teachers’ average prior value-added to see whether the effect is driven by changes in the effectiveness of teachers leaving and entering grade level teams (much reduced sample) 2.Add teacher experience indictor variables to see whether any observed turnover effects are driven by changes in experience of the teachers; 3.Rerun analyses but only for students who had teachers who were in the same grade-by-school group in the prior year (i.e. students of “stayers”). If there were no disruption effects, we should not see an effect of turnover for students in this group.

20. TABLE 10b: Whether Prior VA Explains Effects of Teacher Turnover, School-by-Year Fixed Effects MathAll SchoolsLow PerformingHigh Performing Lagged Attrition-0.0406**-0.0196-0.0417*-0.0286-0.0392~-0.0094 (0.0152)(0.0130)(0.0197)(0.0180)(0.0232)(0.0190) Proportion New-0.0224-0.0120-0.0571**-0.0347~0.00160.0020 (0.0155)(0.0134)(0.0204)(0.0188)(0.0229)(0.0189) ELA Lagged Attrition-0.0150-0.0090-0.0431*-0.0502*0.01090.0329 (0.0161)(0.0149)(0.0207)(0.0196)(0.0243)(0.0221) Proportion New-0.0433**-0.0291~-0.0984***-0.0919***0.00010.0229 (0.0167)(0.0156)(0.0220)(0.0207)(0.0241)(0.0225) School-by-Year FEsxxxxxx Grade Indicatorsxxxxxx All Controlsxxxxxx Average Prior VA x x x

21. Table 9b : Whether Teacher Experience and Migration Explains of Turnover Effects, School-by-Year FEs TestTurnover MeasureModel 1Model 2Model 3 MathProportion New To School-0.0682***-0.0333**-0.0246* (0.012) Lagged Attrition-0.0559***-0.0439***-0.0405*** (0.012) ELAProportion New To School-0.0713***-0.0380**-0.0322* (0.013) Lagged Attrition-0.0387**-0.0265*-0.0204 (0.013) School-By-Year Fixed Effectsxxx Grade Indicatorsxxx Student, Class, School Controlsxxx Experience Indicators xx Mover Indicator x

22. TABLE 11a: Stayers, Movers, and Leavers in High and Low Achieving Schools (school-by-grade FEs) MATH ELA Hi AchLo Ach Hi AchLo Ach StayersLag Attrit-0.0023-0.0593***0.0030-0.0253~ (0.018)(0.015) (0.019)(0.014) Prop New0.0088-0.0338* 0.0095-0.0390** (0.017)(0.015) (0.019)(0.014) 1 st YrsLag Attrit0.00590.0030 0.02740.0383 (0.035)(0.028) (0.034)(0.030) Prop New0.02130.0228 0.0516-0.0068 (0.037)(0.028) (0.036)(0.028)

23. Summary Estimate overall negative effect of attrition Somewhat larger effect in lower performing schools Explained in part by the replacement of leaving teachers by teachers with less experience But even teachers who remain are negatively effected by attrition Caveat – measuring attrition at the grade level and thus do not capture any school-wide effects of turnover

25. It does matter who leaves MathELA School-gradeSchool-yearSchool-gradeSchool-year Math teachers ELA teachers Lagged attrition-0.063**-0.039**-0.031**-0.023* Percent of leavers with low VA0.043**0.013*0.043**0.016**