Written and maintained by CASRAI Editorial Board
Last updated
The Opportunity Atlas is a dataset and interactive map of how adults’ outcomes depend on the Census tract where they grew up. Researchers use it to compare neighborhoods, to link tract identifiers to their own data, and to cite a stable, downloadable source. This guide covers what the data contain, which files exist at which geographic levels, how to cite them, and where the estimates stop being reliable.
What the Opportunity Atlas is
The Atlas reports tract-level estimates of children’s adult outcomes, including household income and incarceration, broken out by parental income, race and gender. It was introduced in the paper The Opportunity Atlas: Mapping the Childhood Roots of Social Mobility by Raj Chetty, Nathaniel Hendren, John Friedman, Maggie R. Jones and Sonya R. Porter, listed on the Opportunity Insights site as American Economic Review 116(1): 1-51 (2026), with an original release dated October 2018. The paper’s own headline example: for children of 25th-percentile parents, the standard deviation of mean household income at age 35 across tracts within counties is $5,000.
Each estimate describes children by where they grew up, not where they live as adults. A child who grew up in a Minneapolis tract and moved to New York is still counted in the Minneapolis tract (Opportunity Atlas FAQ, read 2026-11-07).
Who runs it and what it covers
The Atlas is a collaboration between researchers at the U.S. Census Bureau and Opportunity Insights, a research group based at Harvard University. The authors’ opinions do not necessarily reflect the views of the Census Bureau, and the published results were reviewed so that no confidential information is disclosed; the Opportunity Insights paper page cites Census Bureau Disclosure Review Board authorization CBDRB-FY18-319.
- Population: anonymized administrative data on about 20 million children born between 1978 and 1983, linked to their parents and housed at the Census Bureau.
- Assignment to tracts: children who grew up in more than one tract are assigned to tracts in proportion to the time under age 23 they spent in each.
- Model: a statistical model relates parental income to each outcome, estimated separately for each tract and each race-by-gender group.
- Geography: Census tracts hold about 4,000 people on average. The FAQ reports roughly 70,000 neighborhoods with estimates. Tract-level estimates for all children rest on about 500 children on average; race- and gender-specific estimates on about 130.
- Income units: dollar values are inflation-adjusted with CPI-U to 2015 dollars. Parental income is measured at fixed years (1994-1995 and 1998-2000), and children’s outcomes are observed in their mid-30s in 2014-2015.
How to use the interactive map
The map is hosted at opportunityatlas.org; there is a user’s guide and a tutorial linked from the Opportunity Atlas FAQ. These are the features that the FAQ describes:
- Choose an outcome and a group. Estimates are specific to a group defined by parental income, race and gender. The standard view shows Low (25th percentile), Middle (50th) and High (75th) parental income; Advanced Mode adds the Lowest (0th) and Highest (100th) groups.
- Zoom. Zoomed out, the map shows Commuting Zone averages; as you zoom in it shows county and then Census tract estimates.
- Set the color scale. The “everyone”, “selected group” and “on screen” options change only the coloring, never the underlying data.
- Filter in Advanced Mode. Open the menu icon near the upper left, select a characteristic and drag across the range you want, for example average rents below $1,000 a month. A green tile shows each active filter; click its “x” to remove it.
- Overlay your own data. The “Overlay your Data” option accepts latitude and longitude, addresses or tract identifiers, and you can upload your own boundary file.
- Check sample size. Use the “Number of children” variable before relying on any single tract.
- Save or download. The link icon at the lower left, below the neighborhood characteristics, saves map images and downloads the data you are viewing.
Grey areas labelled “Insufficient Data” mean the estimate would rest on fewer than 20 children, and the Atlas does not publish those.
Downloadable files and how to cite them
For reproducible work, use the files rather than screenshots. The Opportunity Insights data page lists them in Stata and Excel/CSV formats, each with a codebook (see the DDI Codebook entry for what a codebook should contain).
| File family | What it holds |
|---|---|
| All Outcomes by Census Tract, Race, Gender and Parental Income Percentile | The full set of tract-level outcome estimates |
| Household Income and Incarceration for Children from Low-Income Households by Census Tract | Two headline outcomes, including a file for the 1984-1989 birth cohorts |
| Neighborhood Characteristics by Census Tract | Covariates to join to the outcome estimates |
| Tract crosswalk, 2010 to 2020 | Maps between 2010 and 2020 tract definitions, with a ReadMe |
| Replication package | Harvard Dataverse, doi:10.7910/DVN/NKCQM1 |
The paper page states that data are available at tract, county, commuting zone and national levels, by race, gender and parental income percentile. The crosswalk matters because tract boundaries changed between 2010 and 2020: join on the right vintage, or convert first.
The Opportunity Insights page gives no formal citation string. A reasonable reference is the paper (Chetty, Hendren, Friedman, Jones and Porter, “The Opportunity Atlas: Mapping the Childhood Roots of Social Mobility,” American Economic Review 116(1), 1-51), plus the Dataverse DOI for the files you used. That follows the data citation principle: cite the dataset you analyzed, not only the article about it. A DOI-based reference also works with the DOI and metadata practices a repository expects.
Strengths and limits
Strengths
- It follows the same children over time, so it shows whether children in an area grew up to be poor or rich, not only who lives there now. The FAQ describes it as the first dataset giving this longitudinal information at a detailed neighborhood level.
- It is documented: codebooks, a methods paper, a replication package and a crosswalk are all public.
Limits
- Historical cohorts. Estimates cover children born 1978-1983. The authors report that places that produced good outcomes in the past typically produce good outcomes a decade later, and that historical outcomes predict better than recent poverty rates or test scores, but they advise combining the data with local knowledge where areas have changed substantially.
- Predicted, not only observed. The Atlas shows model predictions. For a tract with no low-income families, the low-income estimate is extrapolated from the fitted line, which the FAQ says should be interpreted more cautiously.
- Noise and small samples. Small random “noise” is added to protect privacy, and it can matter more in tracts with few children. The FAQ recommends at least 50 children for single-tract conclusions and looking at clusters of neighboring tracts when studying outliers.
- Observational. Place and outcome are associated in the data. The paper page notes the observational estimates can be compared with causal evidence such as Moving to Opportunity and quasi-experimental studies.
- Place names. Neighborhood names on the map come from Google Maps data and may not match local usage.
Alternatives and companions
| Source | Use it when | Basis |
|---|---|---|
| Opportunity Atlas interactive map | You need to explore or screenshot a place | FAQ, 2026-11-07 |
| Opportunity Insights downloadable files | You need citable, joinable tract, county or commuting-zone data | Data page, 2026-11-07 |
| Traditional neighborhood poverty and income statistics | You need a snapshot of residents now rather than outcomes of children who grew up there | FAQ comparison with existing Census Bureau data |
| Moving to Opportunity and quasi-experimental studies | You want causal evidence to set against Atlas estimates | Paper page, 2026-11-07 |
| Longitudinal surveys such as the Panel Study of Income Dynamics | You need follow-up of the same families in survey form | See the CASRAI PSID guide |
Frequently asked questions
What does the Opportunity Atlas measure?
Average adult outcomes of people born 1978-1983, mapped by the Census tract where they grew up, for groups defined by parental income, race and gender.
Which birth years does it cover?
1978 to 1983 for the main estimates; the data page also lists a file for the 1984-1989 cohorts covering household income and incarceration.
Can I download the data?
Yes. Opportunity Insights provides Stata and Excel/CSV files with codebooks, plus a replication package on Harvard Dataverse (doi:10.7910/DVN/NKCQM1).
Why is a tract grey?
The estimate would rest on fewer than 20 children, so it is not published.
Does it show cause and effect?
No. It reports where children’s outcomes differ. Causal questions need designs such as Moving to Opportunity or quasi-experiments.
Sources
- Opportunity Insights, The Opportunity Atlas paper page, read 2026-11-07.
- Opportunity Insights, Data page, read 2026-11-07.
- Opportunity Insights, Opportunity Atlas FAQ (PDF), read 2026-11-07.
- opportunityatlas.org, the interactive tool.
- Harvard Dataverse replication package, doi:10.7910/DVN/NKCQM1.
