the lifetime impacts of a kindergarten classroom n.
Download
Skip this Video
Download Presentation
The Lifetime Impacts of a Kindergarten Classroom

Loading in 2 Seconds...

play fullscreen
1 / 68

The Lifetime Impacts of a Kindergarten Classroom - PowerPoint PPT Presentation


  • 95 Views
  • Uploaded on

The Lifetime Impacts of a Kindergarten Classroom. Diane Whitmore Schanzenbach Northwestern University and National Bureau of Economic Research. Introduction. What are the long-term impacts of high quality early childhood education ?

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about 'The Lifetime Impacts of a Kindergarten Classroom' - stone-kirk


Download Now An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
the lifetime impacts of a kindergarten classroom

The Lifetime Impacts of a Kindergarten Classroom

Diane Whitmore Schanzenbach

Northwestern University

and National Bureau of Economic Research

introduction
Introduction
  • What are the long-term impacts of high quality early childhood education?
    • We link data from a widely studied education experiment to administrative records on earnings for the first time

 How does kindergarten classroom assignment affect you 25 years later?

research context
Research Context
  • Important policy question but very limited evidence to date
    • Why? Few datasets link information on early childhood test scores with data on adult outcomes
  • Perry Preschool
    • High quality preschool intervention in 1962
    • Long term positive impacts across a wide range of outcomes
    • 123 participants, all low income
slide4

Design of STAR Experiment

  • 11,571 Children
  • 79 Schools
  • 1 Cohort, Started School in 1985-86
  • Random Assignment
  • Small (13-17 Students)
  • Regular, Aide (22-25 Students)
  • Teachers Randomly Assigned
  • KG-3rd Grade
    • 45% of cohort entered after KG
    • ~25% of cohort in STAR all 4 years
  • 4th grade all returned to regular classes
  • Most children born in 1980 -> 30 years old today
slide5

Students in small classes scored higher on standardized tests, K-8

Black students

Switch from SAT to CTBS and return to regular-size classes

Free-lunch students

Full sample

Note: Effect of class size after controlling for student’s race, gender, free-lunch status and initial school-by-entry-wave fixed effects.

slide6

Students randomly assigned to small classes

more likely to take a college entrance exam

  • Percent Taking SAT or ACT

Notes: Figure shows percent of students who took the ACT or the SAT exam, by their initial

class-type assignment. Students initially assigned to regular classes with and without a teacher aide

are combined in the figure. Sample consists of 9,397 STAR students who were H.S. seniors in 1998.

Source: Krueger and Whitmore (1999).

other benefits of small classes
Other benefits of small classes
  • African American boys less likely to have juvenile criminal record
  • Girls less likely to give birth as teens
  • More likely to graduate from high school
  • Students today are ~30 years old
    • Opportunity to investigate a wider range of adult outcomes
united states tax data
United States Tax Data
  • Access to selected variables in anonymous U.S. tax records to conduct research on behavioral responses to economic policies
  • Dataset covers full U.S. population from 1996-2008
  • Approximately 90% of working age adults file tax returns
  • Third-partyreports yield data on many outcomes even for non-filers
    • Employer and wage earnings from W-2 forms
    • College attendance from 1098-T forms
  • 95% of STAR records were linked to tax data
    • Match rate unrelated to treatment
outline
Outline
  • Test scores and adult outcomes in the cross-section
  • Re-evaluate validity of STAR experimental design
  • Impacts of observable classroom characteristics
  • Impacts of unobservable classroom characteristics
  • Fade-out, re-emergence, and non-cognitive skills
  • Cost-benefit analysis
1 cross sectional correlations
1. Cross-sectional correlations
  • Begin by correlating KG test scores with adult outcomes
    • Useful to benchmark estimates from randomized interventions
  • Show both raw correlations and OLS regressions with controls
    • Quartic in parental household income interacted with marital status
    • Mother’s age at child’s birth
    • Parent 401K contributions, home ownership
    • Child gender, free-lunch status, race and age
  • Test score: percentile score on the KG Stanford Achievement Test (average of math & reading)
slide12

What is a kindergarten test?

  • Instructions:
    • You should be on the part of the page with the keys across the top. Look at the cups in the first row. How many cups are there? Fill in the circle under the number that tells how many cups there are.

Mathematics

4

5

6

slide13

What is a kindergarten test?

  • Instructions:
    • You should be on the part of the page with the strawberries across the top. Put your marker under the first row. In this box there are a sentence and three words. Read the sentence and then read the three words under the sentence. Decide which word makes the most sense in the sentence. Fill in the circle under the word that goes best in the sentence.

Reading

The dog likes to _________.

can

jump

big

slide14

Figure 1a: Wage Earnings in 2007 vs. KG Test Score

$25K

$20K

Mean Wage Earnings from Age 25-27

$15K

$10K

0

20

40

60

80

100

KG Test Score Percentile

slide16

Figure 1b: College Attendance Rates vs. KG Test Score

80%

60%

Attended College before Age 27

40%

20%

0%

0

20

40

60

80

100

KG Test Score Percentile

slide17

An Earnings-Based Index of College Quality

  • We construct an index of college quality using tax data
  • Tuition paid to any higher ed. institution automatically generates a 1098-T form linking student and institution
  • Find everyone age 20 enrolled in college in 1999
  • Calculate average wage earnings in 2007 (from W-2s) by college
  • For those who do not attend college, define college quality index as mean earnings for those not in college in 1999
slide19

College Mean Wage Earnings by US News Ranking

$80K

$70K

$60K

Mean Earnings at Age 28

$50K

$40K

0

25

50

75

100

125

US News Rank of Colleges

slide20

Figure 1c: College Quality vs. KG Test Score

$28K

$26K

$24K

Earnings-Based College Quality Index

$22K

$20K

$18K

0

20

40

60

80

100

KG Test Score Percentile

slide21

Figure 2a: Home Ownership vs. KG Test Score

40%

Owned a Home by Age 27

30%

20%

0

20

40

60

80

100

KG Test Score Percentile

slide22

Figure 2b: Retirement Savings vs. KG Test Score

45%

40%

35%

Made a 401(k) Contribution by Age 27

30%

25%

20%

0

20

40

60

80

100

KG Test Score Percentile

slide23

Figure 2c: Marriage by Age 27 vs. KG Test Score

55%

50%

45%

Married by Age 27

40%

35%

30%

25%

0

20

40

60

80

100

KG Test Score Percentile

slide24

Figure 2d: Cross-State Mobility vs. KG Test Score

35%

30%

Lived Outside TN before Age 27

25%

20%

0

20

40

60

80

100

KG Test Score Percentile

slide25

Figure 2e: Percent College Grads in ZIP code vs. KG Test Score

22%

20%

18%

Percent College Graduates in 2008 ZIP

16%

14%

0

20

40

60

80

100

KG Test Score Percentile

slide26

From now on, concentrate on 4 outcomes:

  • 1. College attendance
  • 2. College quality
  • 3. Mean earnings (ages 25-27)
  • 4. Summary index of other outcomes:
  • Index = 401K + Home Owner + Married + Moved (Leave TN) + Pct. College Grads. in Zip
slide27

Figure 2f: Summary Outcome Index vs. KG Test Score

.4

.2

Outcome Summary Index

0

-.2

-.4

0

20

40

60

80

100

KG Test Score Percentile

slide28

Part 2: Validity of the STAR Experimental Design

  • Experimental analysis rests on two assumptions:
  • Assumption #1: Successful Randomization
    • All pre-determined variables (e.g. parent characteristics) are balanced across classrooms
  • Assumption #2: No Differential Attrition
    • 95% match rate  little attrition here
    • No evidence of differences in match rates across classrooms
    • No evidence of differences in death rates across classrooms
slide29

Part 2: Validity of the STAR Experimental Design

  • Threat #1: Failure of Randomization
    • We test for balance across class types with an expanded set of parent/sibling characteristics in two ways:
    • 1. Class size (type) intervention
      • Students were assigned to small vs. large class sizes
      • Do characteristics vary across small vs. large class types?
    • 2. Class room intervention
      • Students were assigned to specific classroom within class size
      • Do characteristics vary across classrooms within schools?
slide31

Validity of the STAR Experimental Design

  • Threat #2: Selective Attrition
  • Much less attrition than in prior studies of STAR because we follow 95% of the sample
  • Test for selective attrition through two channels:
    • Does match rate vary across treatment groups?
    • Does death rate vary across treatment groups (Muennig et al. 2010)?
  • Find no difference in rate of follow-up observation
slide33

Part 3: Class Size Impacts

  • Replicate specifications in previous studies
  • Independent variable: dummy for small class assignment (ITT)
  • Focus on four outcomes:
  • 1. College attendance in 2000
  • 2. College quality index
  • 3. Mean earnings (ages 25-27)
  • 4. Standardized (SD = 1) summary index of other outcomes:
  • Index = 401K + Home Owner + Married + Moved (Leave TN) + Pct. College Grads. in Zip
slide34

Figure 2a: Effect of Class Size on College Attendance by Year

30%

25%

Percent Attending College

20%

15%

10%

2000

2002

2004

2006

Year

Large Class

Small Class

slide35

Figure 2b: College Earnings Quality by Class Size

Frequency

$20K

$30K

$40K

$50K

$60K

Earnings-Based Index of College Quality

Large Class

Small Class

slide36

Figure 2c: Effect of Class Size on Wage Earnings by Year

$18K

$16K

$14K

$12K

Wage Earnings

$10K

$8K

$6K

2000

2002

2004

2006

Year

Large Class

Small Class

slide37

Table 5: Impacts of Class Size on Adult Outcomes

Note: All specifications control for school-by-entry-wave effects, class size, and student and parent demographics, and standard errors are clustered on initial school.

slide38

Part 3: Teacher/Peer Effects

  • Students randomly assigned to classrooms that differ in teacher and peer quality
  • For example, a school has two small classes
  • One small class taught by experienced teacher, one by inexperienced teacher
  • By luck of draw, students assigned to one or other teacher
  • Can tease out impact of being assigned to more experienced teacher
    • After accounting for class size
  • Similarly, by luck of draw some classes have more/fewer
    • Girls
    • African American students
    • Students on free lunch
do teachers peers affect adult outcomes
Do Teachers & Peers Affect Adult Outcomes?

First test: does random assignment to a more experienced KG teacher improve adult outcomes?

  • Not necessarily causal effect of teacher experience per se
  • Experienced teachers not only have more experience but may also be different along other dimensions like dedication to teaching
slide40

Figure 3a: Causal Effect of Teacher Experience on Test Scores

56

54

52

KG Test Score Percentile

50

48

0

5

10

15

20

Kindergarten Teacher Experience (Years)

slide41

Figure 3b: Causal Effect of Teacher Experience on Earnings

$19K

$18K

Mean Wage Earnings, 2005-2007

$17K

$16K

0

5

10

15

20

Kindergarten Teacher Experience (Years)

slide42

Figure 3c: Effect of Teacher Experience on Earnings by Year

$20K

$1104

$18K

$16K

$14K

Wage Earnings

$12K

$10K

$8K

2000

2002

2004

2006

Year

Teacher Experience <=10 Years

Teacher Experience > 10 Years

slide43

Table 6: Observable Teacher vs. Peer Effects

Note: All specifications control for school fixed effects and class size, as well as student and parent demographics.

slide44

Part 4: Unobserved Class Effects

  • Many elements of teacher and peer quality (e.g. clarity of instruction, enthusiasm) not captured by observable measures
    • Well known problem in literature on teacher effects
  • Modern literature captures unobserved teacher characteristics using analysis of variance
    • In other words, does across-class difference in outcomes vary by more than normal statistical variation?
  • Other studies have estimated whether certain teachers usually have classes that show larger-than-average achievement gains
    • Large differences across individual teachers
slide45

Part 4: Unobserved Class Effects

  • We do not observe the same teacher multiple times
    • Only one cohort of students
    • Cannot isolate “teacher effects”
  • We can test for “classroom effects” on adult outcomes
    • Is there significant intra-class correlation in student’s outcomes?
    • Note that this “class effect” includes effect of teachers, peers, and any class-level shocks such as noise outside classroom (barking dog on test day)
slide46

A Statistical Model of Class Effects

  • Test scores and earnings for individual iin class c in school n:
    • zcn = class-level intervention (e.g. better teaching) that affects scores and earnings
    • zYcn = intervention that affects earnings but not scores
    • aicn = academic ability
    • nicn = earnings ability orthogonal to academic ability
    • b + g= impacts of interventions on earnings
    • b= covariance of class effects on scores and earnings
slide47

A Statistical Model of Class Effects

  • Test scores and earnings for individual iin class c in school n:
  • Thus far, we have estimated bdirectly by using observable z’s that affect test scores (e.g. teacher experience)
  • How can we estimate bwhen z is unobserved?
slide48

Strategy 1: Analysis of Variance

  • Test for class effects on earnings (β + γ > 0) using ANOVA
  • Do earnings vary across classes by more than what would be predicted by random variation in student abilities?
    • F-test for significance of class fixed effects
    • Random effects estimate of class-level SD for outcomes
slide49

Table 7: F-Tests for Kindergarten Class Effects

Note: All specifications control for school fixed effects and class size as well as student and parent demographics.

measuring class quality the intuition
Measuring class quality: the intuition
  • We see that there is variation across classrooms from the same schools in adult outcomes
    • Want to know: is it same classrooms that have higher test score gains that produce higher earnings?
  • Because of random assignment, class “pre-test” scores should on average be the same
  • Difference in end of year scores across classrooms is measure of “class quality”
    • Includes teacher effects, peer effects, common shocks
  • Each student’s class quality is the difference between classmates’ scores and schoolmates’ scores
    • How did classmates score, relative to the classmates you could have had?
    • Higher scores -> better class quality
slide51

Strategy 2: Covariance of Class Effects on Scores and Earnings

  • Two limitations of ANOVA:
    • Does not tell us whether class effects on scores are correlated with class effects on earnings (β > 0)
      • Do (unobserved) interventions that raise test scores also improve adult outcomes?
    • Inadequate power to implement ANOVA for grade 1-3 entrants (50% of sample)
  • Turn to a second strategy to measure covariance between class effects on scores and earnings (β)
    • Class effect on scores zcn is unobserved  proxy for it using peer test scores
slide52

Peer-Score Measure of Class Quality

  • Average end-of-year test scores in class relative to school is a (noisy) measure of class effect on scores:
  • Motivates regression of the form:
  • “Own-observation” bias: in finite sample, EbM > 0 even if bM = 0
    • Smart kid raises average class score and has high earnings
  •  Measure class quality using jackknife (leave-out mean) score:
    • How good are your classmates’ scores, compared with the classmates you could have had?
slide53

Peer-Score Measure of Class Quality

  • Regression specification:
  • Class quality Δs-icn captures teacher quality + class-level shocks
    • Good teachers raise peers’ end of year scores
  • Class quality Δs-icn varies randomly within schools
  • Two sources of bias in bLM:
    • Attenuation: Δs-icn is a noisy measure of class quality
    • Reflection: with peer effects, smart kid raises peers’ scores and earns a lot, driving up bLM
      • After presenting results, we bound size of reflection bias and show it is smaller than attenuation bias
slide54

Figure 4a: Causal Effect of Early Childhood Class Quality on Own Score

70

65

60

55

First Observed Test Score Percentile

50

45

40

-20

-10

0

10

20

Class Quality (End-of-Year Peer Scores)

slide55

Figure 4c: Causal Effect of Early Childhood Class Quality on Earnings

$17.0K

$16.5K

$16.0K

Mean WageEarnings, 2005-2007

$15.5K

$15.0K

$14.5K

-20

-10

0

10

20

Class Quality (End-of-Year Peer Scores)

slide56

Figure 5a: Effect of Class Quality on Earnings by Year

$18K

$875

$16K

$14K

Wage Earnings

$12K

$10K

$8K

2000

2002

2004

2006

Year

Above-Average Class Quality

Below-Average Class Quality

slide57

Table 8a: Impacts of Class Quality on Earnings

NOTE--All regressions control for student and parent demographics and school-by-entry-grade fixed effects.

slide59

Figure 6a: Fadeout of Class Effects

Effect of 1 SD of Class Quality on Test Scores by Grade

10

8

6

Test Score Percentile

4

2

0

E

0

2

4

6

8

Grade

1 SD Class Quality Effect on Test Scores

95% CI

slide60

Figure 6b: Fadeout of Class Effects

Effect of 1 SD of Class Quality on Earnings

$1000

$800

$600

Wage Earnings

$400

$200

$0

E

0

2

4

6

8

Grade

1 SD Class Quality Effect on Wage Earnings

95% CI

slide61

Bounding Reflection Bias

  • Small impact of KG class quality on subsequent test scores places a tight upper bound on reflection bias
  • Smart kids score high on all tests (test scores highly autocorrelated)
  • To have large reflection bias, smart kid must raise peer scores a lot
    • Large correlation between peer scores and own score in later grades
  • We formalize this intuition in a linear-in-means model and derive a bound on the degree of reflection bias
    • Observed correlation between KG peer scores and 8th grade score places an upper bound on reflection bias of 20%
    • Variance in scores implies attenuation bias of 20% as well, implying preceding estimates are downward biased on net
slide62

Fade-out and Re-emergence: The Role of Non-Cognitive Skills

  • Why do effects of kindergarten class fade out and re-emerge?
  • One explanation: non-cognitive skills (Heckman 2000)
    • Really “social, behavioral & emotional skills”
  • Data on non-cognitive measures (effort, initiative, disruption) collected for random subset of STAR students in 4th and 8th grade
  • Convert mean non-cognitive score to percentile scale as above

Please consider the behavior of Jim Smith over the last 2-3 months. Circle the number that indicates how often the child exhibits the behavior.

#1. Acts restless, is often unable to sit still

#2. Annoys or interferes with peers’ work

Some-

NevertimesAlways

1 2 3 4 5

1 2 3 4 5

slide63

Mean Wage Earnings vs. Grade 4 Non-Cognitive Percentile

$25K

$20K

Mean WagesEarnings, 2005-2007

$15K

$10K

0

20

40

60

80

100

Grade 4 Non-Cognitive Percentile

slide65

Fade-out and Re-emergence: The Role of Non-Cognitive Skills

  • Non-cognitive skills provide a simple explanation of our findings
  • High quality KG teachers raise KG test scores partly through good classroom management
    • Good classroom management instills social skills
    • Social skills not directly measured in standardized tests but have returns in the labor market
    • Rapid fadeout in math and reading tests after KG
    • But significant earnings gains from better KG class
slide66

Part 6: Cost-Benefit Analysis

  • Assume: 3% real discount rate, constant percent income gains, income follows average US income profile
  • One SD increase in KG class quality for a single year
    • Total NPV earnings gain for class of 20 students of $782K
  • 33% reduction in class size
    • $4K-$189K per class (very imprecisely estimated)
  • One SD improvement in teacher quality
    • $170-$214K per class
    • Moving from below-average (25thpctile) to above-average (75thpctile) teacher generates NPV of $320K for a class of 20 students
take away points
Take Away Points
  • KG scores are highly correlated with adult outcomes (earnings, homeownership, college attendance, retirement savings)
  • Small classes improve a variety of adult outcomes
    • Outcomes of blacks, low-income generally improved more
  • Students assigned to a more experienced teacher have higher adult earnings
  • Higher kindergarten class quality improves a variety of adult outcomes
    • Even a single year of a high-quality class has long-term impact
  • Class quality effects fade out on test scores, but persist on non-cognitive measures
more information on star
More information on STAR
  • http://www.heros-inc.org
    • Bibliography
    • public-use dataset
  • “How Does Your Kindergarten Class Affect Your Earnings? Evidence from Project STAR” by Raj Chetty, John Friedman, Nathaniel Hilger, Emmanuel Saez, Diane Schanzenbach, and Danny Yagan. NBER Working Paper #16381. Available at http://www.nber.org