Presentation to IFS Poverty Review Workshop 7 ‘Intergenerational Mobility

advertisement
Presentation to IFS
Poverty Review Workshop
7th Sept 2010
‘Intergenerational Mobility
in UK, life chances and the
Role of Inequality and
Education
Paul Gregg
1
Introduction & Background
“Most people are willing to accept wide
inequalities if they are coupled with
equality of opportunities” – The
Economist (Oct 2006)
High profile policy area – previous gmt
addressed both poverty and life
chances and saw them as inherently
linked but is new gmt decoupling them?
2
Introduction & Background
Intergenerational mobility has been
widely studied using Income, Education
and Social Class
But in a wider sense covers Social
Gradients in children‘s life chances –
how life chances differ by a measure of
(permanent) social background
e.g. Marmot commission highlighting
extent of social gradients in physical
and mental health and how these
emerge in childhood
3
Concepts
Original concept of Intergenerational
Earnings mobility is ideally comparison of
life time (permanent) earnings of father
and sons
This is very data intensive so shorter tem
earnings measures used
More recently the question has shifted
towards childhood experience and later life
chances which has shifted emphasis
toward family income in childhood
This also allows for absentee fathers which
varies across cohorts in a non-random way
Methodology – Income based
Intergenerational Income Mobility
ln Yi
son
    ln Yi
parents
r = CorrlnYparents , lnYson   (
Income/earnings

Partial correlation (r)
SD
 i
ln Y parents
SD
ln Y
son
)
NCDS
BCS
0.205 (.026)
0.166 (.021)
0.291 (.025)
0.286 (.025)
Measurement I
Early earnings based research had
highlight high levels of mobility but
concerns raised over biases generated by
measurement error and life cycle stage
So using American data NLSY single period
income = 0.32, average over 3 periods =
0.45 implies true estimate = 0.54
Life cycle bias comes from the age(s) at
which earnings are measured and how
good a proxy they are for lifetime earnings
When earnings is measured early or late in
life course it is a less good proxy for
lifetime earnings (optimal is at about 40)
Life Cycle bias in UK
Age-beta profiles across cohorts
0.4
0.35
beta coefficient
0.3
0.25
0.2
0.15
0.1
0.05
0
23
24
25
26
27
28
29
30
31
32
33
NCDS
34
35
36
BCS
37
38
39
40
41
42
43
44
45
46
Poverty
Income mobility and intergenerational
poverty are not the same thing
Blanden and Gibbons suggest the odds
risk for poor children being poor adults
rose from twice that of non-poor children
to 3.5+ times between the NCDS and BCS
cohorts
All this work asks what happens to poor
children not who become the poor parents
in the next generation – including who has
children from the earnings/employment
distribution
Education/Inequality/Genes
The recent literature has been looking at
the key patterns of mobility and beginning
to look at the drivers
Whether mobility is high or low needs a
benchmark so international comparisons
and changes across time in countries have
been widely investigated
A natural next step was to explore how
these patterns match on to inequality and
education (for example Blanden, 2009)
Estimated Intergenerational Income Persistence
.5
and Income Inequality Across Countries
Italy
France
.3
.4
USA
GBrit
Germany
.2
Sweden
Australia
Canada
Finland
Norway
.1
Denmark
.2
.22
.24
.26
.28
.3
inc_gini
Income beta
Fitted values
10
Estimated Intergenerational Income Persistence
.8
and Education Expenditure Countries
.4
Italy
France
GBrit
Australia
.2
Germany
USA
Sweden Canada
Finland
Norway
Denmark
0
Income beta
.6
Brazil
.03
.04
.05
.06
% GDP on education 1970-1974
Income beta
.07
.08
Fitted values
11
Educational Transmission
Blanden et al. (2007), following Solon, explore the drivers
of intergenerational mobility that are measured at earlier
ages. The process of obtaining β can be thought of in
two stages.
Educi  1   ln Yi parents  1i
InYi
child
  1  Educi  u1i
Cov(u1i , ln Yi parents )
   
Var (ln Yi parents )
12
Cross-cohort decomposition
1970 - BCS
1958 - NCDS
0.32
0.27
Persistence
0.22
0.17
0.12
0.07
0.02
1
2
3
4
1
-0.03
Specification
2
3
4
Educational Transmission
In data with moderately detailed education
records, around 55% of intergenerational
mobility in UK comes through education
Further, around 80% of the rise in
intergenerational income persistence
comes from increased strength of the
relationship between family background
and education (Blanden et al. 2007)
Following Heckman big interest in non-cog
Mood et al. (2010) explore personality
traits as well as education for Sweden and
suggest that that about 45% of IGE is
explained with 2/3 by cognitive/ed and 1/3
by personality measures
Education and Family Background – recent
picture
GCSE year
Birth year
PARENTAL
OCCUPATION (SEG)
Managerial/Professional
Other non-manual
Skilled manual
Semi-skilled manual
Unskilled manual
Top - Bottom
Ratio of top / bottom
PARENTAL
OCCUPATION (NSSEC)
Higher professional
Lower professional
Intermediate
Lower supervisory
Routine
Top - Bottom
Ratio of top / bottom
‘88
‘72
‘90
‘74
‘91
‘75
‘93
‘77
‘95
‘79
‘97
‘81
‘99
‘83
52
42
21
16
12
40
4.3
58
49
27
20
15
43
3.9
60
51
29
23
16
44
3.8
66
58
36
26
16
50
4.1
68
58
36
29
24
44
2.8
69
60
40
32
20
49
3.5
70
59
45
35
30
40
2.3
75
62
49
34
26
49
2.9
‘01
‘85
‘03
‘87
‘06
‘90
77
64
51
34
31
46
2.5
76
65
53
41
33
43
2.3
81
73
59
46
42
39
1.9
Parental Educational and Genetics
Estimates of the impact these drivers is
moving into causal analysis
For instance, looking at increased parental
education on child education/earnings
Results suggest raising a parents
education by 1 year results in increase in
child's education by 0.1-0.25
Parental Educational and Genetics
Studies of genetics using partialling out
variances between identical twins, siblings etc
suggest about 40-50% of IQ is heritable. For
personality it is lower (20%) but
measurement less well developed
Similar approaches being used in IGE
estimation in Sweden (Bjorklund et al. 2006)
Nonadoptees
Biological
father
Adoptees
Adoptee
Father
Adoptees
Biological
father
Years of
schooling
.24
.114
.113
Income
.241
.173
.059
BCS intergenerational test scores
from Claire Crawford, Alissa Goodman and Robert Joyce (IFS)
There is a strong link between the cognitive
skills of parents and their children
 This remains even after controlling for
many detailed environmental factors
 Forms an important reason why children
from poor families do less well at school
than richer ones (which the other studies
could not capture)
 Direct effect alone explains 17% of gap in
rich-poor decomposition – this is an
upward biased estimate of genetic
influence
SES gradients are apparent across a
range of outcomes at age 7 to 9
Income gradients in child outcomes in middle childhood
108
106
Score (mean 100, SD 10)
104
102
IQ (5.85)
100
Key Stage 1 (5.46)
Locus of control (3.30)
98
Self esteem (1.71)
96
Behaviour (2.01)
94
Fat mass (1.34)
92
90
88
30
80 130 180 230 280 330 380 430 480 530 580 630
Equivalised disposable weekly household income
age 3/4
Source: Gregg, Propper and Washbrook (2008) (ALSPAC)
Attainment through childhood
from JRF funded research (CMPO and IFS)
Percentile of the test score distribution
80
70
60
50
40
30
20
Age 3 (M CS)
Highest
Age 5 (M CS)
Age 7 (ALSPAC)
Quintile 4
Age 11
Age 14 (LSYPE) Age 16 (LSYPE)
(ALSPAC/ LSYPE)
Quintile 3
Quintile 2
Lowest
How much of the socio-economic gap in cognitive
outcomes at age 3 is explained by these factors?
0% 1%
Residual Gap
Parental Education
16%
Family Background/Demographics
1%
3%
34%
4%
Family Interactions
Health and Well-Being
Childcare
Home-Learning Environment
25%
16%
Total gap to be
explained:
23 percentile
points
Parenting Style/Rules
Missing Data
© Institute for Fiscal Studies
Evolution of the socio-economic gap in cognitive
outcomes at ages 7 to 11
Residual Gap
7%
Parental Education and
Family Background
13%
4%
6%
63%
6%
-1%
Total gap to be
explained:
31 percentile
points
© Institute for Fiscal Studies
Child's attitudes and
behaviours
Parent's attitudes and
behaviours
Pre-school environments
Schools
Missing Data
Evolution of the socio-economic gap in cognitive
outcomes at ages 11 to 16
Residual Gap
7%
6%
Parental Education and
Family Background
15%
Child's attitudes and
behaviours
Parent's attitudes and
behaviours
Schools
59%
8%
1%
4%
Total gap to be
explained:
33 percentile
points
© Institute for Fiscal Studies
Missing Data
Prior Ability
Key messages
1. Educational achievement persists strongly over
the course of childhood. Those who start a
developmental stage ahead generally finish it
ahead. This implies earlier interventions are likely
to be more effective than later ones.
2. Low SES children exhibit poorer social and
emotional development. These skills matter and
lead low SES children to fall further behind at
school, even conditional on prior academic ability.
3. Parental behaviours and values play an important
role in the transmission of family background to
educational achievement and other outcomes.
Conclusions
Rather speculatively
Societies with higher inequality have lower
mobility for two reasons – 1) higher
inequality gives parents greater
incentives/different resource to invest in
children. 2) the educational inequalities get
higher pay offs in high inequality countries
School environment is more equal than home
environment and tends to generate mobility
but this will depend on the extent of
resources in the schooling system and the
degree of inequality in schooling experience
as the conflict with inequalities in the broad
home learning environment incl. parenting
Additional slides
Permanent Income Decomposition
Components of Permanent Childhood and Current Income in the BHPS
Percentage
share of
variance
Correlation
with
permanent
childhood
income
Permanent childhood income,
components associated with:
Fathers’ social class ( ˆ SC )
15.67
0.431
Other income predictors ( ˆ p X p )
22.26
0.615
Residual permanent income ( ˆ p )
62.07
0.716
7.54
0.398
17.41
0.514
40.55
-0.041
34.52
0.706
75.06
0.487
p
p
Current income, components
associated with:
Fathers’ social class ( ̂ p SC p )
Other income predictors ( ˆp X p )
Transitory and measurement error (
uˆ p  eˆ p )
Residual permanent income (
ˆ p  ˆ p  ˆ p )
Error and residual unmeasured income, (
ˆ p  ˆ p  ˆ p  uˆ p  eˆ p )
Current income ( y p )
(ˆp SC p   p )  (ˆp X p   p )   p  u p  e p
Current income without error =
permanent childhood income
(ˆp SC p   p )  (ˆp X p   p )   p
0.735
1.000
Ed-Income recent evidence
NCDS
1958
BCS
1970
BHPS 1
1975-1980
BHPS 2
1981-1986
BHPS 3
1987-1990
LSYPE
1989/1990
0.7165
1.1315
1.0647
0.7958
0.9880
0.9336
[0.036]***
7841
[0.046]***
5428
[0.155]***
815
[0.258]***
515
[0.249]***
345
[0.035]***
10935
0.0963
0.1360
0.1110
0.0846
0.0885
0.0463
[0.006]***
7196
[0.006]***
6420
[0.019]***
964
[0.031]***
583
[0.029]***
386
[0.005]***
8205
0.1618
0.4164
0.4703
0.4512
[0.010]***
7841
[0.023]***
3769
[0.075]***
638
[0.128]***
373
0.0621
0.1047
0.0697
0.0730
N
[0.004]***
7196
[0.006]***
5529
[0.021]***
946
[0.033]**
568
Degree
0.0553
0.1158
0.0916
0.0884
N
[0.004]***
7949
[0.006]***
5520
[0.017]***
932
[0.033]***
484
-0.0049
-0.0197
-0.0676
[0.002]***
5907
[0.003]***
5546
[0.009]***
949
Variable
Number of Olevels (A*-C)
N
Stay on post –
16
N
Number of Alevels (any)
N
Stay on post –
18
Proportion
time NEET
N
Download