Remarks on The Fading American Dream Alan B. Krueger Princeton University & NBER Conference on Social Mobility and Inequality in Israel June 15, 2017
Chetty, et al. They develop a method of estimating absolute mobility for the 194084 birth cohorts Absolute income mobility defined as fraction of children who earn more than their parents did, when both are observed around age 30:
They estimate mobility by decomposing joint distribution of income into two components: 1.
Marginal income distributions for parents and children, estimated using CPS and Census cross-sections
2.
Joint distribution of parent and child ranks (copula)
Data and Methods
Â
Methods Baseline income measure: pre-tax family income at age 30, deflated using CPI-U-RS Estimate absolute mobility by combining three sets of inputs for each birth cohort c: 1.
Children’s marginal income distributions
2.
Parents’ marginal income distributions
3.
Copula: joint distribution of parent and child income ranks
Calculate absolute mobility for birth cohort c as:
Children’s Income Distributions Estimate income distributions at age 30 for children in each birth cohort born from 1940-84 using CPS data in the years 1970-2014. Sample: individuals born in the U.S. Income defined as sum of family’ personal pre-tax income
Parents’ Income Distributions Constructing parents’ income distributions by child’s birth cohort is more complicated because of lack of panel data Overcome this problem by pooling data from multiple Census cross-sections In essence, find parents age 25-35 with children born in relevant year in the relevant Censuses.
Copula: Joint Distribution of Ranks Estimate copula non-parametrically as a 100 x 100 percentile transition matrix for 1980-82 birth cohorts using IRS data Rank children based on their incomes relative to other children in same birth cohort Rank parents of these children based on their incomes relative to other parents Compute joint probabilities of each rank pair
Copula Stability Chetty et al. (2014) show that copula is stable back to 1971 birth cohort using Statistics of Income 0.1% sample Constant relative mobility (in percentile ranks, not absolute dollars) Baseline: assume copula stability for all cohorts going back to 1940 Later derive bounds for absolute mobility with alternative copulas
Baseline Estimates
Baseline Estimates of Absolute Mobility Consider children in 1940 birth cohort Estimate absolute mobility in four steps: 1.
Estimate parent income distribution using Census data
2.
Obtain distribution of child ranks for each parent rank using copula from tax data for 1980 cohort
3.
Map children’s ranks to incomes at age 30 using 1970 CPS
4.
Calculate fraction of children with incomes exceeding parents by parent income percentile
Pct. of Children Earning more than their Parents
Percent of Children Earning More than their Parents By Parent Income Percentile 100
1940
80
60
40
20
0 0
20 40 60 80 Parent Income Percentile (conditional on positive income)
100
Pct. of Children Earning more than their Parents
Percent of Children Earning More than their Parents By Parent Income Percentile 100
1940
80
1950
60
40
20
0 0
20 40 60 80 Parent Income Percentile (conditional on positive income)
100
Pct. of Children Earning more than their Parents
Percent of Children Earning More than their Parents By Parent Income Percentile 100
1940
80
1950
60
1960
40
20
0 0
20 40 60 80 Parent Income Percentile (conditional on positive income)
100
Pct. of Children Earning more than their Parents
Percent of Children Earning More than their Parents By Parent Income Percentile 100
1940
80
1950
60
1960 1970
40
20
0 0
20 40 60 80 Parent Income Percentile (conditional on positive income)
100
Pct. of Children Earning more than their Parents
Percent of Children Earning More than their Parents By Parent Income Percentile 100
1940
80
1950
60
1960 1970
40
1980 20
0 0
20 40 60 80 Parent Income Percentile (conditional on positive income)
100
Pct. of Children Earning more than their Parents
Mean Rates of Absolute Mobility by Cohort 100
90
80
70
60
50 1940
1950
1960 Child's Birth Cohort
1970
1980
Pct. of Children Earning more than their Parents
Trends in Absolute Mobility: Selected States by Decade 100
Massachusetts New York Ohio Michigan
90 80 70 60 50 40 1940
1950
1960 Child's Birth Cohort
1970
1980
Bounds with Alternative Copulas
Sensitivity to Copula: Bounds on Absolute Mobility Relax assumption that copula stable back to 1940 cohort and derive bounds on absolute mobility under alternative copulas by birth cohort Consider all copulas under which children’s ranks increase with parent ranks (first-order stochastic dominance) Solve for bounds using linear programming subject to constraints
Pct. of Children Earning more than their Parents
Bounds on Absolute Mobility Across All Copulas 100
Upper Bound
80
Copula Observed
60 Baseline Estimates 40
Lower Bound 20 1940
1950
1960 1970 Child's Birth Cohort
1980
Household Income Distributions of Parents and Children at Age 30 For Children in 1940 Birth Cohort
Density
80th percentile of parents distribution
14th percentile of children's distribution
Parents 0
27k
50k 100k Income (Measured in Real 2014$)
Children 150k
Household Income Distributions of Parents and Children at Age 30 For Children in 1980 Birth Cohort
Density
80th percentile of parents distribution
74th percentile of children's distribution Parents
0
50k
80k 100k 150k Income (Measured in Real 2014$)
Children
Additional Sensitivity Analysis
1.
Use alternative price deflators
2.
Use post-tax and transfer incomes
3.
Measure incomes at age 40 instead of 30
4.
Use individual income instead of family income (e.g., sons vs. fathers)
Pct. of Children Earning more than their Parents
Trends in Absolute Mobility: Individual Income, Sons vs. Fathers 100
Baseline Son vs. Father Individual Income
90 80 70 60 50 40 1940
1950
1960 1970 Child's Birth Cohort
1980
Counterfactuals
Counterfactual Scenarios
Two key changes since 1940: lower GDP growth rates and less equal distribution of income Consider two counterfactual scenarios for children born in 1980: 1. Higher growth: GDP growth since birth matching experience of 1940 cohort, with GDP distributed across income percentiles as in 2010 2. More broadly shared growth: Same GDP growth rate, but distribute GDP across income percentiles as in 1940 cohort
Pct. of Children Earning more than their Parents
Counterfactual Rates of Absolute Mobility by Parent Income Percentile 100
Mean AM:91.5%
1940 Empirical
80
60
40
Mean AM:50.0% 1980 Empirical
20
0 0
20 40 60 80 Parent Income Percentile (conditional on positive income)
100
Pct. of Children Earning more than their Parents
Counterfactual Rates of Absolute Mobility by Parent Income Percentile 100
Mean AM:91.5%
1940 Empirical
80
60 Mean AM:61.9%
40
Mean AM:50.0% 1980 Empirical
20
Higher growth: 1940 GDP/family growth rate (2.5%), 1980 shares
0 0
20 40 60 80 Parent Income Percentile (conditional on positive income)
100
Pct. of Children Earning more than their Parents
Counterfactual Rates of Absolute Mobility by Parent Income Percentile 100
Mean AM:91.5%
80
1940 Empirical
Mean AM:79.6%
60 Mean AM:61.9%
40
Mean AM:50.0% 1980 Empirical
20 More broadly shared growth: 1980 GDP/family growth rate (1.5%), 1940 shares Higher growth: 1940 GDP/family growth rate (2.5%), 1980 shares
0 0
20 40 60 80 Parent Income Percentile (conditional on positive income)
100
Conclusions 1.
Rates of absolute mobility have fallen from ~90% for 1940 birth cohort to ~50% for children entering labor market today
2.
Absolute mobility has fallen primarily because of growing inequality in distribution of economic growth Inequality and absolute mobility are tightly linked ďƒ Those who are interested in reviving absolute mobility must be interested in more broadly shared economic growth
Katz and Krueger Comment on Chetty et al. in Science June 8, 2017
Decline in Real Median Income is Key
Absolute Mobility
r=0.995
Parent-Child Median Real Income Diference
Trends in Absolute Income Mobility and the Difference between Children’s and Parents’ Median Real Income Across Birth Cohorts
Key Comments Parent-Children diference in real median income is a sufficient statistic for absolute mobility – and easy to calculate for dashboard. A $10,000 decrease in real median income in children’s generation relative to parents’ is associated with a 9.3 pp decline in absolute mobility. Helps explain why mobility fell before inequality rose for 1945 cohort Correlation between absolute mobility and share of children who earn more than the median of parents’ generation is 0.998. An advantage of these alternative measures is that they can be directly computed from public-use cross-sectional household surveys and do not require longitudinally linked data on children and parents. Another advantage is that they can be calculated for important subgroups, such as African Americans. Note: these high correlations rest on the copula being stable.
Focus on Trends in Median Real Income Also Highlights Causes of Decline in Mobility – I. Human Capital
Slowdown in the growth of human capital investment across successive cohorts, especially for children from low-income families. Children born in the early 1940s had 2 more years of schooling on average at age 30 than their parent’s generation, as compared to only 0.75 more years of schooling for children born in early 1980s (Goldin and Katz 2008). And school inputs, such as class size and the length of the school year, improved at a slower pace for generations of students over time as well (Card and Krueger 1992). Note also that the rise in college completion from the early 1960s birth cohort to the early 1980s birth cohort has been concentrated among children from families in the top half of the income distribution, which has made it even more difficult for children from low-income
Expansion in College Mainly Due to Children from Higher Income Families
Born 1979-82
Born 1961-64
Higher Income Inequality Associated with Lower Intergenerational Mobility .
Less Mobility
g
More Inequality
Inequality (1985 Gini Coefficient) Source: Corak (2011), OECD, CEA estimates
37
Higher Income Inequality Associated with Lower Intergenerational Mobility . g
Less Mobility
United States (2015 Gini)
More Inequality
Inequality (1985 Gini Coefficient) Source: Corak (2011), OECD, CEA estimates
38
Causes II. Labor Demand, Bargaining Power, and Mobility Labor demand shifts in recent decades against low- and middle-skill jobs arising from technological change and globalization have polarized the U.S. labor market and played a role in real earnings declines for non-college workers (Autor, Katz, and Kearney 2006). Fissuring of U.S. workplace with increased domestic outsourcing and use of independent contractors has eroded traditional pathways of upward mobility through stable jobs in high-wage employers (Weil 2014; Katz and Krueger 2016). Decline in union membership has eroded upward mobility possibilities as suggested by the higher rates of upward mobility in communities with higher union density (Freeman et al. 2015). Proliferation of noncompete agreements have restricted opportunities Declining geographic mobility contributed to reduced
Classes of Interventions (1) Policies to foster faster productivity growth -- R&D -- Capital investment (2) Policies to raise human capital, particularly for children from the bottom half of the income distribution -- Universal pre-school, better teachers, greater access to public universities, job training, etc. (3) Policies to raise wages and employment of low-income households -- Minimum wage; collective bargaining; antitrust actions -- Tax subsidies to hire low-employment groups
Labor Force Participation Rate of Men Age 25-54, 1960-2015, Original OECD Members & Israel LaborForceParticipationRateforMenAges25-54: U.S.vs.22Original OECDMember States,1960-2015 100
United States 95
Italy
90
Israel
85
80
1960
1965
1970
1975
1980
1985
1990
1995
2000
2005
2010
2015
Source: Organisation for Economic Cooperation and Development (OECD)
June 2, 2017
41
Labor Force Participation Rate of Women Age 25-54, 1960-2015, Original OECD Members & Israel LaborForceParticipationRateforWomenAges25-54: U.S.vs.22OriginalOECD MemberStates,1960-2015 100
90
80
70
60
50
40
Israel
United States
30
Italy
20
10
0
1960
1965
1970
1975
1980
1985
1990
1995
2000
2005
2010
2015
Source: Organisation for Economic Cooperation and Development (OECD)
June 2, 2017
42
Classes of Interventions (1) Policies to foster faster productivity growth -- R&D -- Capital investment (2) Policies to raise human capital, particularly for children from the bottom half of the income distribution -- Universal pre-school, better teachers, greater access to public universities, job training, etc. (3) Policies to raise wages and employment of low-income households -- Tax subsidies to hire low-employment groups -- Minimum wage; collective bargaining, anti-trust
(4) Tax and transfer policies (E.g., EITC) (5) Place-based and geographic mobility policies (E.g., MTO).
Concluding Comments ďƒ˜ ďƒ˜
ďƒ˜
Raising absolute income mobility is a challenge because of the Gatsby Curve and trends in LFPR. Although most of the items on my list of policy interventions are not a priority of the current U.S. administration and Congress, it is nonetheless important for researchers to document the impact of the changing income distribution on societal well-being and economic opportunity. It is also imperative for researchers to provide scientific evidence on interventions that are likely to be efficacious in raising living standards and enhancing economic mobility. Such research is essential if Nations are to provide more opportunity for all their citizens in the future.