Labor Force Quality and How to Produce It

1
Labor Force Quality and How to Produce It

• Eric A. Hanushek
• Stanford University

2
Outline of Talk(s)

• Impact of cognitive skills on economic outcomes
• Individual earnings
• Earnings distribution
• Economic growth
• Government resource policy
• Measured inputs (class size, teacher education,
  ..)
• Credentials
• Identification and measurement of teacher quality
• Impacts of school choice

3
Human Capital Measures

• Quantity
• Mincer earnings functions
• Economic growth
• Schooling inputs
• Spending
• Real resources
• Cognitive test scores

4
Earnings and productivity

• Mincer structure (Murnane et al., Lazear,
  Mulligan)

5
(No Transcript)
6
(No Transcript)
7
(No Transcript)
8
Earnings and Test Inequality
9
Human Capital Quality and Growth

• Hanushek-Kimko (2000)
• Other
• Barro (2001)
• Wößmann (2002, 2003) Gundlach et al. (2002)
• Bosworth and Collins (2003)
• Jamison, Jamison, and Hanushek (2006)

10
Empirical Concerns

• Measurement of human capital
• Enrollment rates
• Attainment
• Quality
• Causation
• Income gt schooling
• Institutions gt schooling growth
• Heterogeneous impacts

11
Quality Measures

• International testing (since early 1960s)
• Voluntary testing, varying set of countries
• Difficulties
• Tests not aligned
• Varying set of countries
• Possibility of trends in quality

12
Baseline Estimates growth 1960-90 31 countries
Initial per capita income (1960) -0.61 (0.19) -0.47 (0.10) -0.48 (0.10)
Quantity of schooling 0.55 (0.21) 0.10 (0.13) 0.11 (0.12)
Labor force quality 0.13 (0.02) 0.13 (0.02)
Population growth -0.04 (0.21)
R2 0.33 0.73 0.73
13
Causality

• Growth causes schooling

14
(No Transcript)
15
International Production Functions
Adult schooling 1.54 (0.64) 2.04 (0.82) 1.62 (0.76)
Pupil-teacher 0.066 (0.16)
Expend/ Pupil -0.69 (0.19)
Expend/ GDP -165.9 (90.1)
16
Causality

• Omitted productivity factor

17
Mincer Earnings of Immigrants
All Home only U.S. and home U.S. only
S 0.09 (0.002) 0.08 (0.002) 0.10 (0.004) 0.13 (0.005)
QL 0.0019 (0.0004) 0.0021 (0.0005) -0.00018 (0.00006) 0.00004 (0.0016)
18
Causality East Asian Miracle
All Exclude HK, Korea, Sing., Taiwan Japan Indonesia, Malaysia, Thailand
Quantity of schooling 0.103 (0.12) 0.117 (0.12) 0.106 (0.12) 0.085 (0.11)
Labor Force quality 0.134 (0.02) 0.101 (0.02) 0.095 (0.02) 0.091 (0.02)
No. countries/R2 31 0.70 27 0.49 26 0.39 25 0.40
19
Extensions

• More countries (31 -gt 45)
• Better tests (Hanushek-Wößmann)
• Comparison of Bratsberg-Terrell (2002)
• Longer time period (1960-2000)
• Extension to mortality
• Evaluation of heterogeneity/avenues of growth

20
Growth GDP/pop, 1960-2000
YPC60 -0.36 -0.33 -0.35
ED60 0.46 0.25 0.07
EQTEST 0.015 0.008
TFR -0.36
OPEN 1.48
Tropical -0.32
N/R2 45/0.29 45/0.56 43/0.75
21
(No Transcript)
22
(No Transcript)
23
Continuing Research Agenda(Ludger Wößmann)

• Effects of institutions
• Tracking
• Determinants of changes
• Impacts of changes

24
(No Transcript)
25
Changing Performance

• Resources
• Overall effects
• Study quality
• Teacher quality
• Alternative approaches
• Output based analysis

26
Big Issues in School Policy Debates

• Relating analysis to policy interests
• Confidence in causation
• Generalizability

27
Analytical designs

• Random assignment experiments
• Natural experiments
• Data solutions
• Trade-offs
• Credibility
• Expense
• Questions that can be addressed

28
Estimated School Resource Effects
N sig sig i insignif
Teacher-pupil ratio 276 14 14 72
Teacher education 170 9 5 86
Teacher experience 206 29 5 66
Teacher salary 118 20 7 73
Expenditure per pupil 163 27 7 66
Facilities 91 9 5 86
Administration 75 12 5 83
Teacher test scores 41 37 10 53
29
Value-added Studies
N sig sig insignif
A. All estimates
Teacher-pupil ratio 78 12 8 80
Teacher education 40 0 10 90
Teacher experience 61 36 2 62
b. Estimates within a single state b. Estimates within a single state b. Estimates within a single state b. Estimates within a single state b. Estimates within a single state b. Estimates within a single state
Teacher-pupil ratio 23 4 13 83
Teacher education 33 0 9 91
Teacher experience 36 39 3 58
30
Teacher Policy Views Shaped by Distinct but
Related Research Lines

• Aggregate labor markets
• Determinants of teacher characteristics
• Impacts of teacher characteristics
• Outcome based perspectives

31
Aggregate Factors

• Expanding opportunities outside of teaching
• Rising female wages
• Much wider career opportunities
• Teacher salaries have not kept up
• One factor model of skill

32
Percent college educated workers below average
teacher earnings
33
Impact of Measurable Characteristics

• Pay parameters
• Teacher education
• Experience
• Salary
• Movements along v shifts in schedule
• Merit pay
• Compensating differentials

34
Teacher Policies and School Choice

• Eric Hanushek
• Stanford University

35
Basic model
36
UTD Texas Schools Project

• Stacked panels of students and teachers
• Annual testing reading and math
• Lone Star district
• Match students-teachers
• Grades 4-7 math
• 1996-2001

37
Estimation

• Estimate teacher or teacher by year fixed effects
• Control for student demographics or student fixed
  effects
• Consider both total variance and within school
  variance
• Use stability over time to estimate measurement
  error variance

38
Measurement Error and Calculation of Variance of
Teacher Quality

• Observe teachers in two years
• Correlation across years

39
Estimated Variance in Teacher Quality
Within district Within district Within district Within school and year Within school and year
unadjusted demographic controls Unadjusted demographic controls
Teacher-year variation Teacher-year variation 0.210 0.179 0.109 0.104
Adjacent year correlation Adjacent year correlation 0.500 0.419 0.458 0.442

Teacher quality variance / (s.d.) Teacher quality variance / (s.d.) 0.105 (0.32) 0.075 (0.27) 0.050 (0.22) 0.047 (0.22)
40
Measured Characteristics
No fixed effects Student fixed effects
Masters degree 0.015 0.004
Pass certificate exam 0.002 -0.002
Pass certificate exam 1st time 0.015 -0.039
plt.1
41
Returns to Experience(comparisons with
experience gt5 years)
student fixed effects student and teacher fixed effects

1st year -0.16 -0.16 -0.12
2nd year -0.03 -0.03 -0.00
3rd year 0.03 0.02 0.02
4th year 0.05 0.08 0.06
5th year 0.04 0.03 0.01
plt.05 plt.01
42
Teacher-Student Interactions
Teacher and student race Student fixed effects Student and teacher fixed effects
Tb -0.082 -0.015
Tb x Sb 0.047 0.102 0.105

plt0.05 plt0.01
43
Teacher Transitions by Experience
No move Change campus Change district Exit public schools
1 year 70.4 11.5 4.0 14.0
2-3 years 70.8 11.2 5.0 13.0
44
Teacher Quality by Transition Status
With student fixed effects With student fixed effects With student fixed effects With student fixed effects With student fixed effects
Within district Within district Within district Within school Reclassified women returnees

change campus -0.089 -0.089 -0.061 -0.061 -0.061 -0.054 -0.060

change district -0.011 -0.011 -0.031 -0.031 -0.031 -0.023 -0.028

exit public schools -0.044 -0.044 -0.089 -0.089 -0.089 -0.072 -0.095

plt0.1 plt.01

45
(No Transcript)
46
(No Transcript)
47
Extensions

• Quality gap for exits highest 2-3 years of
  experience
• Noisy evidence that mean quality higher for young
  district switchers
• Transition year fall off exit-stayer gap
  disappears with lagged quality

48
Estimated Effects of Salary and Student
Demographic Characteristics on the Quality of
Newly Arrived Teachers
Destination campus characteristics      
Destination campus characteristics standardized gains passed certification examination advanced degree
log salary 0.12 -0.15 0.22

Black 0.0000 -0.0009 -0.0024

LEP 0.0016 -0.0017 -0.0017
plt0.1 plt0.01
49
Conclusions

• Substantial variation in teacher quality
• Most within schools
• Most unexplained by observable characteristics
• Experience effect concentrated in 1st year
• Sizeable differences by race/ethnicity
• Little evidence that urban district loses best
  teachers
• School leavers not significantly better
• Younger district switchers slightly better
• Districts not obviously pursuing quality
• Teachers with advanced degrees
• Importance of retention process

50
(No Transcript)
51
(No Transcript)
52
Charter School Evaluation issues

• Most analysis of entry and participation
• Shortage of reliable information on performance
• Difficulty of selection issues
• Very political

53
Evaluation approaches

• Model selection process Heckman (1979)
• Instrument for attendance Neal(1997)
• Intake randomization Howell and Peterson (2002),
  Hoxby and Rockoff (2004)
• Matching

54
Difficulties with traditional approaches

• Hard to find factors affecting attendance but not
  achievement
• Results of random assignment experiments may not
  generalize
• Aggregate matches uncertain
• (Note differences with micro-matches)

55
Innovations in Our Work

• Use sector differences in school value-added
• Identify charter school effects from students who
  switch sectors
• Control for direct effect of school switches and
  any changes in family income
• Consider heterogeneity across schools
• Model consumer responsiveness to quality

56
UTD Texas Schools Microdata Panel

• Four cohorts followed 1996-2002
• Achievement in grades 4-7 (TAAS math and reading)
• Each cohort gt 200,000 students in over 3,000
  schools
• gt250 distinct charters of varying vintage

57
Texas charter schools

• Introduced in 1995
• Variety of legislative changes and limits with
  215 permitted in 2002
• Most charters very young

58
Charter enrollment
1997 2002
4th grade lt0.01 0.8
7th grade 0.07 0.9
59
Participation rates by race/ethnicity
1997 2002
Blacks 0.06 2.1
Hispanics 0.04 0.8
Whites 0.01 0.5
Low income 0.03 1.0
60
Charters by vintage (analytical)
Age 1997 1998 1999 2000 2001 2002 Total
startup 17 10 70 83 43 47 270




61
Charters by vintage (analytical)
Age 1997 1998 1999 2000 2001 2002 Total
startup 17 10 70 83 43 47 270
two 2 16 9 69 78 40 214



62
Charters by vintage (analytical)
Age 1997 1998 1999 2000 2001 2002 Total
startup 17 10 70 83 43 47 270
two 2 16 9 69 78 40 214
three 0 2 15 8 68 73 166
Four 0 1 2 15 8 66 92
Five 0 0 1 3 17 22 43
63
Annual exit rates
Move to other public school from Move to other public school from
Charter Regular
4th to 5th 24.3 13.0
5th to 6th 24.2 12.6
6th to 7th 21.7 10.9
64
Annual exit rates
Move to other public school from Move to other public school from Exit Texas public schools from Exit Texas public schools from
Charter Regular Charter Regular
4th to 5th 24.3 13.0 22.8 7.2
5th to 6th 24.2 12.6 19.3 6.6
6th to 7th 21.7 10.9 15.9 7.0
65
Empirical framework (value-added)

• Identify charter school from sector switches
• Control for confounding influences associated
  with sector changes

66
Concerns

• Ashenfelter dip
• Enter charter because of bad school experience
• Negative family experience
• Mobility has independent effect
• Psychometrics of TAAS
• Competitive responses
• Generalizability

67
Average Charter School Effect
Student fixed effect yes yes yes
Own mobility yes yes yes
Peer turnover yes yes
Peer achievement yes
Charter school -0.18 -0.28 -0.13 -0.17
68
Sensitivity

• Not sensitive to peer achievement variable
• Distribution of charters by year
• Ashenfelter dip

69
Charter school effect by school age
Age Student fixed effects, own and peer mobility, peer achievement
Startup -0.53 -0.33
Two -0.20 -0.25
Three -0.08 -0.08
Four -0.01 0.00
Five 0.13 0.06
70
Interrupted Panel Estimates
Charter age All Before-after entry only Omit before switch
One -0.43 -0.46 -0.46
Two -0.26 -0.30 -0.27
Three -0.03 -0.06 -0.00
Four -0.04 -0.16 -0.01
Five or more 0.05 0.08 -0.02
71
Measure of school quality

• Mean adjusted student gains
• use fixed state weights to aggregate
• Regression adjust mean gains
• Race/grade/disadvantaged/migrant shares plus year
  dummies
• Charter age

72
Quality distributions
73
Do parents make good decisions?

• Parents cannot see value added
• Considerable mobility/exiting
• Models
• Exitf(quality, age, year, race, grade)

74
Exits and school quality
(1) (2)
Quality 0.01 0.02
Quality x charter -0.14 -0.13
Own gain -0.01
Own gain x charter -0.01
75
Exits and school quality by income
(1) (2)
w/own gain
Quality 0.00 0.01
Quality x charter -0.19 -0.18
Quality x low income 0.01 0.01
Quality x charter x low income 0.09 0.08
76
Conclusions

• Charters have rough beginning
• After startup, do as well as regular publics
• Parents much more responsive to quality