GospelGraph data sources: every input, named
Mostly public data you can check yourself, plus two church directories that fill the congregations public data misses.
If a number on GospelGraph looks wrong to you, this page tells you where it came from, so you can check it and tell us.
What data does GospelGraph use?
GospelGraph is built mainly from three public sources: the U.S. Census Bureau’s American Community Survey (ACS), the Census Bureau’s LEHD Origin-Destination Employment Statistics (LODES), and OpenStreetMap. Church locations also draw on the IRS 990 Business Master File and on two directories that are not public government data: ChurchesUSA, a purchased congregation directory, and Overture Places, a web-derived places corpus added in September 2026. The four church sources are deduplicated against each other before they are joined to census tracts.
Everything behind the scores is aggregate and geographic. None of the neighborhood layers uses private data about individual people.
Where do the neighborhood numbers come from?
The American Community Survey 5-year estimates supply poverty, income, unemployment, housing cost burden, vacancy, education, family structure and demographics. They are a major input to the five-domain Shalom composite behind need pressure and the Gospel Impact Metric, and the base the neighborhood personalities are matched on. The five-year window substantially reduces sampling error at tract level. For the Gospel Desert Index the pipeline reads tables B01003 (total population), B17001 (poverty), B25002 (housing vacancy), B23025 (labor force and unemployment) and B03002 (race and ethnicity). The long-baseline GIM v3 reads twelve overlapping ACS windows, from 2009–2013 through 2020–2024.
LODES links employer payroll records with Census demographics to produce block-level counts of jobs by industry, wage and worker characteristics. Two derived fields matter most: a pedestrian index summing retail, food service and healthcare jobs, the sectors most associated with daytime foot traffic, and an activation ratio comparing workers arriving with workers leaving, which separates places that fill up by day from residential places whose sidewalks are empty. LODES covers formal employment only. The informal economy and volunteer life are not in it.
Where do the church locations come from?
Four sources supply congregations, each with its own strengths and gaps.
- OpenStreetMap: places of worship tagged amenity=place_of_worship, pulled nationally from state-level extracts. Coverage is strong in urban areas; denomination tags are inconsistent. OpenStreetMap, with Overture, also supplies the third places, intersections and public rooms used in the built-form layer.
- IRS 990 Business Master File: registered 501(c)(3) organizations with NTEE codes. GospelGraph uses X20 (Christianity) and X21 (Protestant), restricted to foundation code 10, "Church, IRC 170(b)(1)(A)(i)", so that clergy associations, parachurch ministries and think tanks under the same codes do not enter a church count. X22 (Roman Catholic) is out of scope, because most parishes file under a diocesan group ruling.
- ChurchesUSA: a purchased congregation directory. It is not public data.
- Overture Places: a web-derived places corpus, added in September 2026. It contributes congregations the other three miss and backfills websites, addresses and phone numbers onto records that had none. It is filtered to Protestant traditions.
How are the church sources combined?
All four sources are deduplicated against each other before being joined to tract geography. A record is matched to an existing one on distance and name similarity, with the match radius set per source. The newest source gets the lowest priority, so it can add evidence to a church already known but never overwrites a name or coordinate that a surveyed or registered source supplied. Across the whole church layer, Latter-day Saints and Jehovah’s Witness records are excluded.
From the combined layer each tract gets four fields: the count of churches inside it, the distance from its center to the nearest church (which may sit outside the tract), churches per square mile of land, and the number of distinct denominations present. Only distance, density and denominations feed the Gospel Desert Index; the count is reported but not scored. The full GDI methodology gives the weights.
What do these sources miss?
No church dataset is complete. Small, storefront and immigrant churches are systematically under-registered in all four sources, house churches and informal networks rarely appear at all, and some records may describe congregations that have closed. Every church dataset also has a survivorship problem: failed plants and dissolved congregations are absent by definition. Treat the church layer as a visible-access signal, not a census of the church.
Survey data has its own limits. ACS estimates carry margins of error, five-year averages lag fast change, and the 2016–2020 release was collected through the first year of the pandemic. The full methodology states how each of those limits shows up in the scores.
How does GospelGraph handle missing data?
Where a measure is unavailable, GospelGraph either substitutes a named proxy and says which one, or leaves the field empty with its reason attached. It never invents a value, estimates one from neighboring tracts, or carries an older vintage forward to cover a gap in a newer one. A tract the pipeline declines to score carries no score rather than a zero; 1,070 of 84,414 tracts are Unscored in the Gospel Desert Index for that reason.
Ministry data should be falsifiable. If you see something wrong in your own neighborhood, the system should be able to say which source, threshold or rule produced it. The Gospel Desert Index this site serves is published as gdi-index.json, and the national reading of these sources together is in State of the Field.
Common questions
Is all of GospelGraph’s data public?
Where do GospelGraph’s church locations come from?
Does GospelGraph use private personal data?
Are the church locations complete?
Check the data where you live
Run a free field report for an address you know well, and see whether the ground it describes matches the street you walk.