Optimal Preschool
Policies for LowIncome Children
Greg J. Duncan
School of Education
University of California, Irvine
Outline
What skills and behaviors should
preschools be promoting?
Concrete achievement skills, mostly
How good are we at doing that?
So-so, and impacts are smaller now than 40
years ago
Outline (con’t)
What policy levers are available?
Funding + regulating quality and curriculum
What’s the bottom line on them?
Center-based care helps; quality regulation
doesn’t seem to work; and we’re promoting
the wrong curricula in Head Start
Are there successful models out there?
Yes, but only scaled up in one city
What skills and behaviors
matter most for success
in school?
Skills and Behaviors
Problem
Behaviors
Achievement
Engagement
Description:
Concrete math
and reading skills
Ability to control
impulses and
focus on tasks
i) Ability to get
along with
others
ii) Sound mental
health
Example test
areas or
question
wording:
Knowing letters
and numbers;
beginning word
sounds, word
problems
Can’t sit still;
can’t concentrate;
score from a
computer test of
impulse control
i) Cheats or tells
lies, bullies, is
disobedient at
school
ii) Is sad, moody
Duncan and Magnuson, 2011
Skill and behavior gaps between high- and low-income
kindergarteners and fifth graders (SAT scale)
+106
+53
-27
-30
School
engagement
-100
Math (or
~reading)
achievement
Anti-social
behavior
Mental
health
problems
Kindergarten gap
5th grade gap
Source: Early Childhood Longitudinal Study – Kindergarten cohort.
Skill and behavior gaps between high- and low-income
kindergarteners and fifth graders (SAT scale)
+106
+121
+53 +59
-27
-30
-42
School
engagement
-100
Math or
reading
achievement
Anti-social
behavior
Mental
health
problems
Kindergarten gap
5th grade gap
Source: Early Childhood Longitudinal Study – Kindergarten cohort.
-31
Which school-entry academic skills
and behaviors best predict later
school achievement?
Regress later achievement on:
• School-entry math and reading
• School-entry engagement, etc.
Controls for:
• Child IQ, temperament
• Maternal and family measures
Duncan et al. (2007)
Predictive importance for later school
achievement (standardized coefficients)
School-entry:
Grades 1 to 8
achievement:
Reading
.17*
Math
.34*
Engagement/attention
.10*
Anti-social (- expected)
.01 ns
Mental health (- expected)
.01 ns
Duncan et al (2007)’s meta-analysis of six longitudinal data sets, five of
which control for prior IQ
Bottom line for ECE and
school readiness:
Marshmallows be damned!
Concentrate first and foremost on
early math and literacy skills
How well do ECE programs
promote cognitive skills?
• Evidence from strong evaluation
studies published between
1960-2007
• End of treatment effect sizes
(vs. longer-run studies)
Average cognitive impact at end of treatment
2.00
Average effect size in sd units
Perry
1.50
Abecedarian
1.00
0.50
0.00
1955
-0.50
1965
1975
1985
1995
2005
2015
Average cognitive impact at end of treatment
2.00
Head Start
Non Head Start
Average effect size in sd units
Perry
1.50
Y
Abecedarian
1.00
National Head
Start
0.50
0.00
1955
-0.50
1960
1965
1970
1975
1980
1985
1990
1995
2000
2005
2010
2015
Average cognitive impact at end of treatment
2.00
Head Start
Non Head Start
Average effect size in sd units
Perry
1.50
Y
Abecedarian
Boston
pre-K
1.00
National Head
Start
0.50
0.00
1955
-0.50
1960
1965
1970
1975
1980
1985
1990
1995
2000
2005
2010
2015
Why are impacts of programs
from the 60s, 70s and 80s
larger than now?
Counterfactual conditions now are
much more enriching:
• Maternal schooling much higher
• Fewer siblings
• More center-based care
What About Long-Run ECE
Effects?
• Short-term impacts on test scores fade
over time
– Meta-analysis: Decline by .025 standard
deviations each year, or entirely after 8-9 years
• Yet, consistent impacts on adult educational
attainment, earnings and crime across
diverse ECE programs
– Example: Deming (2009) fixed-effect Head Start
study using an index of adult outcomes shows
effect size .23 sd
The Mechanism Puzzle
• We don’t know why there are long-run
effects on human capital when short-run
achievement impacts fade
• BUT evidence suggests that there is not one
explanation for all evaluation study findings
– It’s not only because of “character” or behavior
• Good News, though: Equifinality--a variety
of ECE programs with differing approaches
have positive impacts on adult human
capital through differing pathways
Policy levers
• Funding streams for programs
• Curriculum requirements
• Process quality regulation (QRIS)
ECE Funding & Enrollment
• Two largest funding streams for ECE: Head Start
($8.5 billion) and State Prekindergarten ($5.1
billion)
• In year before Kindergarten about 75% of
children experience ECE in a mix of full- and partday programs
•90% of top income quintile
•65-69% of bottom three income quintiles
• Lower enrollment among Hispanics, Immigrants,
and rural populations
Cost of Expanding ECE Access
• Focus on funding bottom three income
quintiles
~ 1.2 million of these children are not in ECE
(or private ECE)
• Per child cost of program (mix of part and full
day programs): ~$7,500
• New Cost: $9.36 billion (a little more than the
current cost of Head Start)
What is minimal ECE short-run effect
size needed to recoup $7,500?
• Increase of 1% percentile rank in Kindergarten
achievement predicts .5% increase in adult earnings
(Chetty et al., 2011)
• Our estimate of present value of lifetime earnings
(PVLE) at age 5:
– Lower estimate ~$382,392
– Higher estimate ~$681,544
• Break Even if ECE program impacts are :
– Lower PVLE estimate: 4 percentile points (.10-.15 ES)
– Higher PVLE estimate: 2 percentile points (.03-.08 ES)
How to generate large
cognitive impacts?
• Curriculum requirements?
• Process quality regulation (QRIS)?
Types of Curricula
• “Whole-child”
• Content-specific (e.g., math or
literacy)
• “Locally-developed”
Whole-child curricula
• Head Start mandates “whole child”
curricula
• Creative Curriculum is most popular
• HighScope (Perry Preschool) is 2nd
most popular
• No strong evidence on effectiveness
Process Regulation Policy Lever
• All but one state have Quality Rating and
Improvement Systems (QRIS)
• Star-type ratings for quality based on structural
characteristics and classoom observations
(ECERS, CLASS)
• Most run by state family services and not
education departments
• No RCT evidence; value-added evidence
suggests no substantial impacts for stars, ECRS
or CLASS
New RCT Evidence on:
• Which curricula best promote
school readiness?
• Do gains in QRIS-type process
quality match gains in child
outcomes?
The Preschool Curriculum
Evaluation Research (PCER)
Initiative Study
• provided random-assignment evaluations of
14 early childhood education curricula
• 12 grantees; all used common measures of
child outcomes, classroom processes, and
implementation quality
• 2,911 children
Curricula comparisons in PCER
Literacy
Math
I
II
III
Whole-child (Creative
Curriculum and HighScope)
IV
Locallydeveloped
Note: Comparison IV only involves the Creative Curriculum
I. Literacy vs. HighScope and Creative Curriculum
U North Florida n=250
FL Early Literacy and Learning Model
Creative
Florida State n=200
Florida State n=200
Berkeley n=290
FL Literacy Express
FL DLM Early Childhood Express
NJ Ready Set Leap
HighScope
HighScope
HighScope
VA Language Focused
HighScope
University of Virginia
n=200
II. Literacy vs. Locally Developed
UT Houston n=200
TX Doors to Discovery
UT Houston n=200
TX Let’s Begin with the Letter People
Vanderbilt n=210
TN Bright Beginnings
Locally
Developed
Locally
Developed
Locally
Developed
III. Math vs. HighScope and Creative Curriculum
Berkeley and SUNY Buffalo
n=320
CA, Pre-K Math
NY
Creative or
HighScope
IV. Creative Curriculum vs. Locally Developed
UNC Charlotte n=310
Vanderbilt
n=210
NC, Creative Curriculum
GA
TN Creative Curriculum
Locally
Developed
Locally
Developed
Do preschool curricula affect:
• Classroom quality?
• Child school readiness?
Do preschool curricula affect:
• Classroom quality?
• Child school readiness?
Experimental curricula comparisons predicting classroom
observational measures at the end of preschool
ECERS
total
score
TBRS
Math
TBRS
Literacy
Arnett
total
score
I. Literacy v. HighScope and
Creative Curriculum
II. Literacy v. Locally developed
III. Math v. HighScope and
Creative Curriculum
IV. Creative Curriculum v.
Locally developed
Each cell estimate
is from a separate
regression
Experimental curricula comparisons predicting classroom
observational measures at the end of preschool
I. Literacy v. HighScope and
Creative Curriculum
II. Literacy v. Locally developed
III. Math v. HighScope and
Creative Curriculum
IV. Creative Curriculum v.
Locally developed
ECERS
total
score
TBRS
Math
TBRS
Literacy
Arnett
total
score
.25+
-.14
.07
.18
(.15)
(.16)
(.16)
(.16)
.51*
.46
.83*
.38
(.23)
(.32)
(.37)
(.25)
.15
1.16*
.34
.63
(.32)
(.52)
(.31)
(.52)
.61*
.51*
.71**
.99*
(.23)
(.23)
(.17)
(.36)
Do preschool curricula affect:
• Classroom quality?
• Child school readiness?
Experimental curricula comparisons predicting school
readiness skills at the end of preschool
Social
skills
Literacy
Math
Academic
composite composite composite composite
I. Literacy v. HighScope
and Creative Curriculum
II. Literacy v. Locally
developed
III. Math v. HighScope
and Creative Curriculum
IV. Creative Curriculum
v. Locally developed
Experimental curricula comparisons predicting school
readiness skills at the end of preschool
Social
skills
Literacy
Math
Academic
composite composite composite composite
I. Literacy v. HighScope
and Creative Curriculum
.15**
(.05)
-.01
(.05)
.06
(.05)
-.13
(.10)
II. Literacy v. Locally
developed
.15
(.09)
.14+
(.07)
.15+
(.08)
-.18
(.19)
III. Math v. HighScope
and Creative Curriculum
.05
(.10)
.35**
(.11)
.25*
(.11)
.14
(.17)
IV. Creative Curriculum
v. Locally developed
.02
(.08)
.02
(.08)
.02
(.08)
-.03
(.23)
Can’t we do even better than
this?
• What if you built the
curriculum around proven
approaches?
Boston pre-K as a model?
• Curriculum combined proven math &
literacy and behavioral curricula
• Develop “non-cognitive” skills as a
by-product of boosting academic skills
• Strong professional development,
including coaching
• Big impacts, but $12K per child
Boston pre-K
Weiland & Yoshikawa, 2013 Child Development
0.7
.62***
0.6
Effect size
0.5
.59***
.50***
.44***
0.4
0.3
0.2
0.1
0
PPVT-III
W-J LW (early
(vocabulary)
reading)
W-J AP
(numeracy)
REMA Short
(numeracy,
geometry)
39
Positive “Spillover” Effects on All Three
Dimensions of Executive Function Skills
0.7
0.6
0.5
0.4
0.3
.24***
.24***
.21***
.28***
0.2
.11
0.1
0
Backward DS Forward Digit
(working
Span (working
memory)
memory)
Pencil Tap
(inhibitory
control)
DCCS
(inhibitory
control)
TOQ Attention
(att. shifting)
40
What does Boston pre-K look
like?
6-minute video from
restoringopportunity.com
https://www.youtube.com/watch?v=URZkGPwcsn0
Summary
• Focus most on building achievement skills
• Typical ECE programs generate fairly small
impacts, although still may have
Benefits > Costs
• QRIS quality systems aren’t promising
• Mandated “whole-child” curricula aren’t
either
• Experiment with full-monty curricular
approaches
Greg J. Duncan
gduncan@uci.edu
School of Education
University of California, Irvine