the widening academic achievement gap between the rich and the poor… · 2018-09-21 · gap...

the widening academic achievement gap between the rich and the poor:

new evidence and possible explanations

sean f. reardonstanford university

november, 2012

social reproduction

on average, students from families of higher socioeconomic status perform better on academic tests, attain higher levels of schooling, and (as a consequence) attain higher socioeconomic status themselves as adults

however, the extent of social reproduction – the strength of the correlation between parental socioeconomic status and children’s outcomes – is mutable; it may vary across time and place, as a result of social policy, norms, values, and economic conditions.

key questions

how large is the socioeconomic achievement gap in the US?

specifically, how large is the income achievement gap?

how does it compare to other countries?

how has this gap changed over the last 50 years?

what accounts for the evident changes?

data

all available US studies meeting three criteria: nationally-representative sample

standardized achievement test

information on family income

13 studies included Project TALENT, NLS72, HS&B, NLSY79, NELS, Add

Health, Prospects, NLSY97, ELS, SECCYD, ECLS-K, HSLS, ECLS-B.

these include student cohorts born 1943-2001 and tested 1960-2009

measuring achievement gaps

Measuring gap between high- and low-income students is complicated… …because income is continuous, not a binary variable …because income distribution changes over time …because income is reported in categories …because income is reported with error

Comparing gaps across studies is complicated… …because test content differs …because test scales differ …because test reliabilities differ …because samples differ (in age/grade,

representativeness)

Computing income achievement gaps

Solution Standardize test scores within each study Use categorical income data to estimate average

achievement of children in families at 90th and 10th

percentiles of the income distribution Adjust for estimated reliability of income Adjust for estimated reliability of achievement test Use longitudinal studies to assess if/how gaps vary with

age/grade Sensitivity analysis to assess sensitivity to different

sampling designs Weight estimates by inverse of sampling variance

computing income achievement gaps

-0.38

0.37

90/1

0 G

ap

-1.0

-0.5

0.0

0.5

1.0

Avera

ge

Sta

nda

rdiz

ed T

est

Score

$0K$1

K$3

K$5

K

$7.5

K

$10K

$15K

$20K

$25K

$35K

$50K

$75K

$100

K

$200

K0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1

Income (Dollars and Percentiles)

95% CI Within Income Category

Fitted Cubic Curve

Association Between Reading Score and Family Income Percentile,Grade 8 Students, 1988 (NELS data)

computing income achievement gaps

-0.59

0.59

90/1

0 G

ap

-1.0

-0.5

0.0

0.5

1.0

Avera

ge

Sta

nda

rdiz

ed T

est

Score

$5K

$10K

$15K

$20K

$25K

$30K

$35K

$40K

$50K

$75K

$100

K

$200

K0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1

Income (Dollars and Percentiles)

95% CI Within Income Category

Fitted Cubic Curve

Association Between Reading Score and Family Income Percentile,Grade 8 Students, 2006 (ECLSK data)

Adjusting gaps for reliability

findings

How has the achievement gap changed in the last 50 years?

…between high-income (90th percentile) and low-income (10th percentile) children?

…between high-income (90th percentile) and middle-income (50th percentile) children?

…between middle-income (50th percentile) and low-income (10th percentile) children?

1112 910

12

12

10

1110

812

9

710811

1

3

7

10

9

7

103

K K

0.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

Ave

rage

Diffe

ren

ce

in S

tand

ard

ize

d T

est

Sco

res

Betw

een

90th

& 1

0th

Incom

e P

erc

en

tile

Fam

ilies

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

HLS

ECLS-B

95% CI

Quartic Fitted Trend

Study

Trend in 90/10 Income Gap in Reading, 1940-2001 Cohorts

1112 910

12

12

10

11

10

8

1

3

7

10

9710 3

K

9

K

0.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

Ave

rage

Diffe

ren

ce

in S

tand

ard

ize

d T

est

Sco

res

Betw

een

90th

& 1

0th

Incom

e P

erc

en

tile

Fam

ilies

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

HLS

ECLS-B

95% CI


Study

Trend in 90/10 Income Gap in Math, 1940-2001 Cohorts

1112

9

10

12 1210

11

10

8

129

7108

11

1

3

710

9

710

3

K

K

0.00

0.25

0.50

0.75

1.00

1.25

Avera

ge D

iffe

ren

ce in S

tan

da

rdiz

ed T

est

Score

sB

etw

ee

n 9

0th

& 5

0th

In

com

e P

erc

en

tile

Fam

ilies

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

HLS

ECLS-B

95% CI

Fitted Trend1943-2001

Study


1112

9

10

12

12

10

11

10

8 1

3

7

10 97

10

3

K

9

K

0.00

0.25

0.50

0.75

1.00

1.25

Avera

ge D

iffe

ren

ce in S

tan

da

rdiz

ed T

est

Score

sB

etw

ee

n 9

0th

& 5

0th

In

com

e P

erc

en

tile

Fam

ilies

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

HLS

ECLS-B

95% CI


Study


1112

910

12

12

10

11

10

8

12

9 710 8

11

1

3

7

10

97

10

3

K

K

0.00

0.25

0.50

0.75

1.00

1.25

Avera

ge D

iffe

ren

ce in S

tan

da

rdiz

ed T

est

Score

sB

etw

ee

n 5

0th

& 1

0th

In

com

e P

erc

en

tile

Fam

ilies

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

HLS

ECLS-B

95% CI


Study


1112

9

10

12

12

10

11

10

8

1

3

7

10

9

7

10

3

K

9

K

0.00

0.25

0.50

0.75

1.00

1.25

Avera

ge D

iffe

ren

ce in S

tan

da

rdiz

ed T

est

Score

sB

etw

ee

n 5

0th

& 1

0th

In

com

e P

erc

en

tile

Fam

ilies

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

HLS

ECLS-B

95% CI


Study


0.00

0.25

0.50

0.75

1.00

1.25

1.50

Ave

rage

Diffe

ren

ce

in S

tan

da

rdiz

ed

Te

st

Score

s

(90

/50 o

r 50

/10

In

co

me G

ap

)

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

90/50 Income Gap

50/10 Income Gap

Source: Reardon (2011)

Reading, 1943-2001 Birth Cohorts

Income Achievement Gaps (90/50 and 50/10 Gaps)

0.00

0.25

0.50

0.75

1.00

1.25

1.50

Ave

rage

Diffe

ren

ce

in S

tand

ard

ize

d T

est

Sco

res

(90

/50 o

r 5

0/1

0 I

ncom

e G

ap)

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

90/50 Income Gap

50/10 Income Gap


Math, 1943-2001 Birth Cohorts


how large are these gaps?

one standard deviation is the difference between the 31st and 69th percentile

if the gap is one standard deviation, this implies that the average student in a family at the 10th percentile of the income distribution has test scores lower than 84% of students in families at the 90th percentile of the income distribution

one standard deviation is the amount a typical student learns in a year in K-1

3 years in elementary-middle school

6 years in middle-high school

findings

How does the achievement gap change as children progress through school?

development of income achievement gap, by age and subject, all longitudinal studies

0.00

0.25

0.50

0.75

1.00

1.25

1.50

Avera

ge

Diffe

ren

ce

in

Sta

nda

rdiz

ed T

est S

core

sB

etw

ee

n 9

0th

& 1

0th

In

co

me

Perc

en

tile

Fa

mili

es

6 8 10 12 14 16 18

Age

ECLS-K Reading

ECLS-K Math

Prospects G1 Reading

Prospects G1 Math


Prospects G3 Math


Prospects G7 Math

SECCYD Reading

SECCYD Math

NELS Reading

NELS Math

HS&B Reading

HS&B Math

ELS Math

Study

Income Achievement Gaps, by Age and Subject

Fall KFall G1

Fall G3 Fall G5

Fall G8

Spring K Spring G1

Age 4Fall K

0.00

0.25

0.50

0.75

1.00

1.25

1.50

Ave

rage

Diffe

ren

ce

in S

tan

da

rdiz

ed

Te

st

Score

s

(90

/10 I

ncom

e G

ap)

4 6 8 10 12 14

Age

ECLS-K (1992-93 birth cohort)

ECLS-B (2001 birth cohort)

Study


Reading, Ages 4-15

Development of Income Achievement Gap (90/10 Gap)

0.00

0.25

0.50

0.75

1.00

1.25

1.50

Ave

rage

Diffe

ren

ce

in S

tan

da

rdiz

ed

Te

st

Score

s

(90

/50 o

r 50

/10

In

co

me G

ap

)

4 6 8 10 12 14

Age

90/50 Income Gap

50/10 Income Gap


Reading, Ages 4-15


Fall K

Fall G1 Fall G3 Fall G5 Fall G8

Spring K

Spring G1

Age 4 Fall K

0.00

0.25

0.50

0.75

1.00

1.25

1.50

Ave

rage

Diffe

ren

ce

in S

tand

ard

ize

d T

est

Sco

res

(90

/10 I

nco

me G

ap

)

4 6 8 10 12 14

Age

ECLS-K (1992-93 birth cohort)

ECLS-B (2001 birth cohort)

Study


Math, Ages 4-15

Development of Income Achievement Gap (90/10 Gap)

0.00

0.25

0.50

0.75

1.00

1.25

1.50

Ave

rage

Diffe

ren

ce

in S

tand

ard

ize

d T

est

Sco

res

(90

/50 o

r 5

0/1

0 I

ncom

e G

ap)

4 6 8 10 12 14

Age

90/50 Income Gap

50/10 Income Gap


Math, Ages 4-15


findings

How does the achievement gap in the US compare to the gap in other countries?

IcelandNorway

SwedenCanada (Ontario)

NetherlandsCanada (Quebec)New Zealand

SloveniaEngland

GermanySlovak Republic

GreeceUnited States (ECLS-K)

IcelandPoland

DenmarkKorea

New ZealandGermany

LuxembourgPortugal

United States (ECLS-K)

KoreaPolandDenmark

New ZealandHungary

GermanyPortugal

PIS

A 2

009

PIS

A 2

006

PIR

LS

(2001)

0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75

90/10 Income Achievement Gap (95% CI)

Estimated 90/10 Income Achievement Gaps,Reading, PIRLS & PISA

Canada (Ontario)

Canada (Quebec)

Germany

Denmark

England

Greece

Hungary

Iceland

Korea

Luxembourg

Netherlands

Norway

New Zealand

Poland

Portugal

Slovak Republic

Slovenia

Sweden

United States

0.50

0.75

1.00

1.25

1.50

90

/10 I

ncom

e A

chie

vem

en

t G

ap

(sd's

)

-1.0 0.0 1.0 2.0 3.0

Poverty/Inequality Index

Fitted Line (Precision-weighted)

Fitted Line w/o US (Precision-weighted)

Wealthy OECD Countries, 2001-2009 (pooled PIRLS and PISA data)

Association Between Income Achievement Gap and Poverty/Inequality Index

why has the income achievement gap grown?

rising income inequality? changes in family investment patterns rising returns to income? rising investment in children’s cognitive development

(among high-income families)? changing parenting practices?

Increasing correlation between income and other family resources?

increasing socioeconomic segregation? changing social policies (changing social safety net

for the poor)?

income and educational outcomes

income inequality and educational inequality

30%

35%

40%

45%

50%

Sha

re o

f A

ll In

co

me

1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010

YearSource: Piketty & Saez (2012): http://www.econ.berkeley.edu/~saez/TabFig2010.xls

(Includes Capital Gains), 1918-2010

Share of Total Income Accruing to 10% Highest Income Families,

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

11.0

90/1

0 I

ncom

e R

atio

1970 1980 1990 2000 2010

YearSource: Author's calculations from CPS data 1968-2011

Among Families of School-Age Children

Income Inequality (90/10 Income Ratio), 1967-2010

1.5

2.0

2.5

3.0

3.5

4.0

4.5

Incom

e R

atio

1970 1980 1990 2000 2010

Year

50/10 Ratio

90/50 Ratio

Source: Author's calculations from CPS data 1968-2011

Among Families of School-Age Children

Income Inequality (50/10 and 90/50 Income Ratio), 1967-2010

0.00

0.20

0.40

0.60

0.80

Cha

ng

e in S

tand

ard

ize

d T

est

Score

Per

Dou

blin

g o

f Im

com

e

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

HLS

ECLS-B


Study

Trend in Association Between Income and Reading Achievement,Families Below Median Income, 1940-2001 Cohorts

0.00

0.20

0.40

0.60

0.80

Cha

ng

e in S

tand

ard

ize

d T

est

Score

Per

Dou

blin

g o

f Im

com

e

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

HLS

ECLS-B


Study

Trend in Association Between Income and Math Achievement,Families Below Median Income, 1940-2001 Cohorts

0.00

0.20

0.40

0.60

0.80

Cha

ng

e in S

tand

ard

ize

d T

est

Score

Per

Dou

blin

g o

f Im

com

e

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

HLS

ECLS-B


Study

Trend in Association Between Income and Reading Achievement,Families Above Median Income, 1940-2001 Cohorts

0.00

0.20

0.40

0.60

0.80

Cha

ng

e in S

tand

ard

ize

d T

est

Score

Per

Dou

blin

g o

f Im

com

e

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

HLS

ECLS-B


Study

Trend in Association Between Income and Math Achievement,Families Above Median Income, 1940-2001 Cohorts


rising income inequality?

changes in family investment patterns rising investment in children’s cognitive development

(among high-income families)?

changing parenting practices?


increasing socioeconomic segregation?

changing social policies (changing social safety net for the poor)?

8351,264 1,173 1,315

3,536

5,650

6,975

8,872

0

2,500

5,000

7,500

10,000

Exp

end

itu

res (

in 2

00

8 d

olla

rs)

1972-73 1983-84 1994-95 2005-06

Top Income Quintile

Bottom Income Quintile

Source: Duncan & Murnane (2011)

Family Enrichment Expenditures on Children, 1972-2006

changing view of parental role

parental views of their role as parents has changed over twentieth century (wrigley, 1989; schaub, 2010)

increasing focus on importance of parenting for cognitive development

some evidence of social class differences in parenting practices (lareau, 2003)

middle/upper-class: concerted cultivation working-class: accomplishment of natural growth

education policy may play a role, by focusing and legitimating test scores as primary goal of schooling and evidence of success (schaub, 2010)

changing views of parenting, 1900-1985 (wrigley, 1989)

0%

20%

40%

60%

Pro

po

rtio

n o

f M

aga

zin

e A

rtic

les

1900s 1910s 1920s 1930s 1940s 1950s 1960s 1970s 1980s

Source: Wrigley, Julia. (1989). Do Young Children Need Intellectual Stimulation? Experts' Advice to Parents, 1900-1985.History of Education Quarterly 29/1:41-75 (Table 1).

Topics of Expert Advice on Parenting, 1900-1985

Nutrition Intellectual Stimulation

Medical Social/Emotional Development

Fresh Air

relationship between income and other family resources

polarization of families (mclanahan 2004)

increasing returns to college education and cognitive skill (murnane, willett, & levy, 1995)

income more strongly associated with parental education and cognitive skill

increased assortative mating (schwartz & mare, 2005)

high-income families not only have more income, but increasingly also have more of other resources that matter (dual parents, high educational attainment & cognitive skill, smaller families, fewer very young mothers)

adjusted trends in income-achievement and education-achievement associations, reading, 1940-2001

0.00

0.20

0.40

0.60

0.80

1.00

Re

gre

ssio

n-A

dju

ste

d 9

0/1

0 A

chie

vem

en

t G

ap

1940 1950 1960 1970 1980 1990 2000Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

ECLS-B

IncomeCoefficient

ParentalEducationCoefficient

Study

Trends in Adjusted Associations Between Readingand Both Income and Parental Education, 1943-2001 Cohorts

adjusted trends in income-achievement and education-achievement associations, math, 1940-2001

0.00

0.20

0.40

0.60

0.80

1.00

Re

gre

ssio

n-A

dju

ste

d 9

0/1

0 A

chie

vem

en

t G

ap

1940 1950 1960 1970 1980 1990 2000Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

ECLS-B

IncomeCoefficient

ParentalEducationCoefficient

Study

Trends in Adjusted Associations Between Mathand Both Income and Parental Education, 1943-2001 Cohorts

0

20

40

60

80

100

Pro

port

ion

of F

am

ilies

1970 1980 1990 2000 2008

Metropolitan Areas with Population > 500,000, 1970-2008

Proportion of Families Living in High-, Middle-, and Low-Income Neighborhoods

Affluent (>150% of Metro Median)

High Income (125-150% of Metro Median)

High Middle Income (100-125% of Metro Median)

Low Middle Income (80-100% of Metro Median)

Low Income (67-80% of Metro Median)

Poor (<67% of Metro Median)

Neighborhood Type (Based on Median Family Income Level)

summary of trends

income achievement gaps have grown sharply in recent decades (since 1970s birth cohorts, maybe before)

income gaps have grown most rapidly in the top half of the income distribution

income gaps now larger than black-white gap

gaps present when students start school (at least in recent cohorts; no data on earlier cohorts)

inequality and education

differences in inequality, coupled with a stable association between income and educational achievement, seems insufficient to explain the patterns of association between inequality and income achievement gaps

rather, the association between income and achievement has changed as well

but why?

a provisional hypothesis

implications

the link between family income and children’s achievement, coupled with the increasing importance of cognitive skills in determining earnings, produces a feedback cycle that leads to low socioeconomic mobility and growing inequality.

this feedback cycle may operate partly through schooling, though schools (in a narrow, functional sense) do not appear to be a primary cause of this trend

nor is it clear that schools (alone) can reverse this trend, though they may be a helpful mechanism.

policy implications

reduce economic inequality greater investment in early childhood prevent development of gaps (easier than remedying later) most cost-effective developmental age for investment means-targeted programs likely most cost-effective (though

maybe less politically feasible?)

support for low-income families repair/strengthen social safety net programs to develop parenting skills (e.g., Nurse-Family

Partnership)

increase education policy focus on students from low-income families and communities develop and test strategies for improving instruction/learning

for low-income students

income achievement reading gaps, 1940-2001 cohorts, white students

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

Avera

ge

Diffe

ren

ce

in

Sta

nda

rdiz

ed T

est S

core

sB

etw

ee

n 9

0th

& 1

0th

In

co

me

Perc

en

tile

Fa

mili

es

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

ECLS-B


Study

Trend in 90/10 Income Gap in Reading, White Students, 1943-2001 Cohorts

income achievement reading gaps, 1940-2001 cohorts, black students

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

Avera

ge

Diffe

ren

ce

in

Sta

nda

rdiz

ed T

est S

core

sB

etw

ee

n 9

0th

& 1

0th

In

co

me

Perc

en

tile

Fa

mili

es

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

ECLS-B


Study

Trend in 90/10 Income Gap in Reading, Black Students, 1943-2001 Cohorts

income achievement reading gaps, 1940-2001 cohorts, hispanic students

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

Avera

ge

Diffe

ren

ce

in

Sta

nda

rdiz

ed T

est S

core

sB

etw

ee

n 9

0th

& 1

0th

In

co

me

Perc

en

tile

Fa

mili

es

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

ECLS-B


Study

Trend in 90/10 Income Gap in Reading, Hispanic Students, 1943-2001 Cohorts

income achievement reading gaps, 1940-2001 cohorts, male students

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

Avera

ge

Diffe

ren

ce

in

Sta

nda

rdiz

ed T

est S

core

sB

etw

ee

n 9

0th

& 1

0th

In

co

me

Perc

en

tile

Fa

mili

es

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

ECLS-B


Study

Trend in 90/10 Income Gap in Reading, Male Students, 1943-2001 Cohorts

income achievement reading gaps, 1940-2001 cohorts, female students

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

Avera

ge

Diffe

ren

ce

in

Sta

nda

rdiz

ed T

est S

core

sB

etw

ee

n 9

0th

& 1

0th

In

co

me

Perc

en

tile

Fa

mili

es

1940 1950 1960 1970 1980 1990 2000

Cohort Birth Year

TALENT

NLS

HS&B

NLSY79

NELS

Add Health

Prospects

NLSY97

ELS

SECCYD

ECLS-K

ECLS-B


Study

Trend in 90/10 Income Gap in Reading, Female Students, 1943-2001 Cohorts

the widening academic achievement gap between the rich and the poor… · 2018-09-21 · gap...

Documents