Mobility / Methodology
Mobility Methodology
How the FinObservatory /mobility layer is built, and the concepts it keeps
separate. Every number on the page is computed at build time from the functions
in src/lib/mobility.ts querying the parquet files below; nothing is typed into
the page.
What mobility means here
Distribution measures inequality at a moment. Mobility asks whether that moment is inherited. Two distinct questions run through this page, and they are not interchangeable:
- Absolute mobility: do children end up better off than their parents in absolute terms? Measured for the United States as the share of children who out-earn their parents (income).
- Relative mobility: does a parent's position predict a child's, so that a poor start tends to pass down? Measured within the United States as the adult income rank of children who started poor (the Opportunity Atlas), and across countries as the parent-child correlation of education (the World Bank GDIM).
Income mobility and education mobility need not move together, and a country can rank differently on each. The page never blends the two into one number.
United States: the fading American dream (absolute mobility)
Source: Opportunity Insights, "The Fading American Dream: Trends in Absolute Income Mobility Since 1940" (Chetty, Grusky, Hell, Hendren, Manduca, Narang, Science 356(6336): 398-406, 2017), Online Data Table 1 (national baseline estimates by parent income percentile and child birth cohort).
The estimate is the fraction of children who, at age 30, earn more in real terms
than their parents did at the same age. The headline series (abs_mob_mean, the
OI Table 1 column cohort_mean) averages over the parent income distribution; the page
also shows the series conditional on parents at the 25th percentile
(abs_mob_p25). The estimates use linked, de-identified tax records covering the
full population rather than a survey sample, so their sampling error is
negligible; the paper's robustness table (Online Table 4) documents the
alternative price indices and income definitions behind them. Cohorts run
1940-1984; the series ends there because both generations must be observed as
adults.
United States: the Opportunity Atlas (relative mobility for poor children)
Source: Opportunity Insights, "The Opportunity Atlas: Mapping the Childhood Roots of Social Mobility" (Chetty, Friedman, Hendren, Jones, Porter), the county outcomes table.
The measure is kfr_pooled_pooled_p25: the mean household income rank (on
the 0-1 national scale) reached in adulthood by children whose parents were at
the 25th percentile of the national parent-income distribution, pooled over race
and gender. A rank of 0.50 is the national middle; higher is more upward
mobility. The Atlas publishes a standard error with every estimate
(kfr_pooled_pooled_p25_se), and this page carries it: on the county table
directly, and in the source note as the national median standard error.
State figures are a rollup, not an Atlas number. The Atlas publishes tract,
county and commuting-zone estimates, not states. The state value shown here is
the mean of the county kfr_p25 values weighted by kid_pooled_pooled_n (the
number of children per county). This is stated on the page; it is a population
weighting, not an official Opportunity Insights state statistic. Counties in the
US territories are dropped (the 50 states and the District of Columbia are kept).
Across countries: the World Bank GDIM (education mobility)
Source: World Bank, Global Database on Intergenerational Mobility (GDIM), 2023 March release, education file. 153 economies, ten-year birth cohorts 1940-1989 (labelled by the decade start, so the "1980" cohort is people born 1980-1989), covering about 97 percent of world population.
GDIM reports several measures; the page uses:
cor(GDIMCOR): the parent-child correlation of years of schooling, standardised to 0-1. Higher means more persistence, hence less mobility. Its published standard error iscor_se(CORstd). This is the measure on the cohort chart and the Great Gatsby curve.mu050(MU050): upward mobility, the share of children born to bottom-half-educated parents who reach the top half. Carried in the parquet with GDIM's published lower and upper bounds.
The parent definition is max, GDIM's primary choice: the schooling of the
higher-educated of the two parents. Children are pooled over gender. The
status flag travels with each row: "Co-residents only" estimates (Bangladesh
among them) are computed only from parent-child pairs living in the same
household, which is known to understate mobility for older cohorts; the flag is
preserved rather than dropped.
The Great Gatsby curve
The in-house Great Gatsby curve pairs, for each economy, GDIM education
persistence (cor, 1980 cohort, parent = max) on the vertical axis against
income inequality on the horizontal axis. The horizontal axis is the World Bank
Poverty and Inequality Platform (PIP) Gini index, latest survey per economy,
already in the estate. The two axes' sources are stated per axis on the chart.
Three honesty notes belong to this construction:
- The Gini mixes concepts. PIP reports Ginis from income-based surveys in some countries and consumption-based surveys in others (of the economies on the curve, the majority are consumption-based). Consumption Ginis run systematically lower than income Ginis, adding noise to the horizontal axis.
- The eras do not match. Mobility is for the 1980s birth cohort; the Gini is contemporary. The original Great Gatsby curve (Corak, Krueger) pairs a cohort's mobility with the inequality of its parents' generation. Using the latest Gini is a common simplification and is flagged here, not hidden.
- The relationship is a tendency, not a law. The correlation and the ordinary-least-squares slope shown are computed from the points, and the scatter is wide. The page reports the correlation and lets the reader see the spread; it does not claim causation in either direction.
Bangladesh
Bangladesh is present wherever the sources cover it: GDIM for the 1980 cohort (with the co-resident caveat above) and PIP for the Gini, so it appears on both the cohort discussion and the Great Gatsby curve, where it is highlighted. It is absent from the United States exhibits by construction (those are US-only). Where a source does not cover Bangladesh, its absence is stated rather than silently passed over.
The generation lag
Both mobility concepts require observing two adult generations, so every series here ends a generation ago: absolute mobility with the 1984 US birth cohort, cross-country education mobility with the 1980 cohort. Any statement about children born since would be a projection, and this page makes none.
Verification anchors (quoted at build)
The build scripts assert the publishers' own displayed numbers and fail loudly if a vintage shifts:
- Fading American Dream (
cohort_mean): 1940 cohort 0.915, 1980 cohort 0.500, 1984 cohort 0.503, matching the paper's abstract ("from approximately 90% for children born in 1940 to 50% for children born in the 1980s"). - Opportunity Atlas: at least 3,100 US-state counties carry
kfr_pooled_pooled_p25in the range 0.12 to 0.69. - GDIM: Bangladesh, 1980 cohort, parent = max:
COR0.576,MU05035.76; at least 140 economies present.
Licenses
- Opportunity Insights: publicly and freely available for use with citation (opportunityinsights.org/data). Cited on the page as Chetty et al. 2017 (fading American dream) and the Opportunity Atlas team.
- World Bank GDIM: Creative Commons Attribution 4.0. Required citation carried: "GDIM. 2023. Global Database on Intergenerational Mobility. Development Research Group, World Bank. Washington, D.C.: World Bank Group."
- World Bank PIP: Creative Commons Attribution 4.0.
All three are displayed with attribution only. Consistent with the site's no-redistribution ruling, FinObservatory shows these data with citation and does not re-serve the underlying files, rows, or bulk data.