What a denomination's own polity classes say about the ground it holds
Waldron, Arkansas, in Scott County. One census tract holds two Assemblies of God congregations. First Assembly of God meets at 879 South Main Street. Centro Cristiano Hispano Nueva Vida meets at 433 Featherston Street, a few minutes away and inside the same tract boundary.
Every neighborhood measure these two churches stand in is the same number, because it is the same neighborhood. Poverty at 21.3 percent. Need pressure at 68. Twenty-one churches inside the tract. Exurban Working Homeowners, the dominant household type. Access in the Dry tier at 47.
One of those congregations governs itself and the other does not. First Assembly is General Council-affiliated, which in Assemblies of God polity means a sovereign, self-supporting church. Centro Cristiano Hispano Nueva Vida is District-affiliated, which means it is held under district oversight because it is not yet, or no longer, able to stand alone.
Six hundred and fifty-seven census tracts in America hold that arrangement: at least one self-governing Assemblies of God congregation and at least one that is not. The neighborhood cannot be what separates them, because the neighborhood is held constant by definition. This paper is about what does.
The Assemblies of God sorts every congregation on its roster into one of three classes, and the roster carries the label in a column. The classes are a ladder of self-support.
Sovereign and self-governing. The church calls its own pastor, holds its own property and answers to the national body directly.
Held under district oversight. Either not yet self-supporting or no longer so. The district carries responsibility the congregation cannot.
A daughter work under a parent church. The newest category, and the one that most closely marks a recent plant.
Fifty-eight percent of the denomination governs itself. Forty-two percent is held by a district or a parent.
Two measures do most of the work below, and they are not the same thing.
Access counts how much evangelical gospel witness is structurally within reach of a place, built from congregations and their reach and never from any judgment about faithfulness. Need pressure counts the weight households in that place are carrying: provision and work, justice and protection, life and health, household and generations, social fabric. A neighborhood can score average on the first and extreme on the second, and many do.
Every congregation in the 2024 roster was placed on the census tract its building stands in and that tract read against fourteen national data layers. Measures are taken on distinct tracts rather than on congregations, because two churches on one street describe one neighborhood. Nothing here counts members, attendance, giving or conversions.
Three other rosters ran through the same pipeline for comparison: the Presbyterian Church in America, the Orthodox Presbyterian Church and the Associate Reformed Presbyterian Church.
Group the roster by that column and the three classes separate on almost every layer, in a consistent direction.
| Measure | General Council | District | Parent | Nation |
|---|---|---|---|---|
| Congregations | 6,428 | 3,107 | 1,473 | — |
| Access (GDI) | 50.0 | 51.1 | 49.5 | 50.0 |
| Need pressure | 51.9 | 54.7 | 52.8 | 50.2 |
| Poverty | 14.0% | 16.3% | 15.0% | 13.2% |
| Churches in tract | 9.8 | 10.4 | 8.5 | 6.8 |
| No church at all | 1.9% | 2.8% | 4.8% | 11.6% |
| Access tier Arid or worse | 18.3% | 21.9% | 16.5% | 20.3% |
| Shalom 2024 | 48.1 | 45.3 | 47.2 | 49.6 |
| Shalom hope pillar | 51.9 | 48.4 | 48.6 | 50.3 |
| Suburban | 57.3% | 47.6% | 49.6% | 42.8% |
| Rural | 2.5% | 3.8% | 1.0% | 4.6% |
| MVI (viability) | 51.0 | 51.1 | 53.2 | 51.2 |
Read the first column and you are looking at a settled suburban denomination. Read the second and you are looking at something else: harder ground on every pressure measure, the only Assemblies of God class whose share of worst-tier access tracts sits above the national rate, and the lowest wellbeing score of the three.
The third column is different again, and the row that separates it is the one about emptiness. Parent-affiliated congregations sit in tracts that hold no other church at a rate of 4.8 percent, two and a half times the General Council rate, and they have the fewest other churches around them. They are also the best connected: the only class whose street-intersection density reaches the national figure.
The denomination's hardest ground is held by the congregations that cannot stand on their own.
Take the gap between General Council and District-affiliated on its own terms. Need pressure runs 2.8 points higher. Poverty runs 2.3 points higher. The Shalom composite runs 2.8 points lower and its hope pillar 3.5 points lower. Twenty-two percent of District-affiliated tracts sit in the two worst access tiers against eighteen percent for self-governing churches.
Each of those gaps is small next to the spread between denominations. Together, and all pointing the same way across four independent layers, they describe something real: within one denomination, the churches on the hardest ground are disproportionately the ones a district is carrying.
There is one row that runs the other way and it is worth naming. The ground under District-affiliated churches is falling more slowly than anywhere else in the denomination. Shalom moved by −0.48 between 2020 and 2024 for that class, against −0.73 for self-governing churches and −0.76 for the country. The hardest ground the Assemblies of God holds is also its most stable.
The Assemblies of God runs thirteen district councils organised around a language or a culture rather than a territory: Spanish Eastern, two Korean, National Slavic, Samoan, Brazilian, four Hispanic districts, Florida Multicultural, a Distrito Central. Those districts hold 1,515 congregations, and they stand on harder ground than the geographic districts do. So the obvious reading of section 3 is that the dependent churches simply are the language-district churches under a different name.
They are not, and the cross-tabulation is worth printing in full.
| General Council | District | Parent | Total | |
|---|---|---|---|---|
| Language and ethnic districts | 573 (37.8%) | 852 (56.2%) | 90 (5.9%) | 1,515 |
| Geographic districts | 5,855 (61.7%) | 2,255 (23.8%) | 1,383 (14.6%) | 9,493 |
| Total | 6,428 | 3,107 | 1,473 | 11,008 |
The two cuts are entangled. A language-district congregation is more than twice as likely to be District-affiliated. But 2,255 of the 3,107 dependent churches, nearly three quarters of them, sit inside ordinary geographic districts, so the dependent population is not a restatement of the language one.
Hold each constant and both effects survive. Inside the geographic districts alone, dependency still adds 2.0 points of need pressure and 1.7 points of poverty. Inside the dependent churches alone, belonging to a language district still adds 3.2 points of need pressure and 2.9 points of poverty. They stack.
| Need | Poverty | Access | No church at all | MVI | |
|---|---|---|---|---|---|
| Language · dependent (908) | 56.5 | 18.1% | 49.8 | 4.5% | 54.7 |
| Language · self-governing (558) | 57.4 | 17.9% | 49.6 | 4.1% | 56.2 |
| Geographic · dependent (3,363) | 53.4 | 15.2% | 50.8 | 3.1% | 51.0 |
| Geographic · self-governing (5,608) | 51.4 | 13.6% | 50.0 | 1.7% | 50.5 |
Six points of need pressure separate the corners of that table. Access separates them by one.
Look again at the first row of that cross-tabulation and a different number comes out of it. Of the 1,515 congregations in a language or ethnic district, 37.8 percent govern themselves. Of the 9,493 in a geographic district, 61.7 percent do.
A congregation in one of those thirteen districts is roughly half as likely to be a self-supporting church. Whatever else the language districts are, they are structurally a denomination of dependent works in a way the geographic districts are not.
The first thing to test is whether that is a fact about the neighborhoods those churches are in. Language districts sit in denser, poorer places, and the denomination's dependent churches sit on harder ground, so a gap of this size could be the same finding arriving twice.
It is not, and the next section is the whole argument of this paper.
Split every congregation by the kind of place it stands in and compare the two halves inside each kind. If the gap were a neighborhood effect it would shrink to nothing once the neighborhood is held constant. It does not shrink. In the suburbs it is at its widest.
| Neighborhood type | Language | Geographic | Gap |
|---|---|---|---|
| Suburban | 37.0% | 65.6% | −28.6 |
| Exurban | 42.3% | 61.6% | −19.4 |
| Urban, sparse | 37.2% | 56.0% | −18.8 |
| Urban, dense | 42.2% | 50.4% | −8.2 |
Now hold the difficulty of the ground constant as well. Restrict to the 4,999 congregations standing in tracts with need pressure of 55 or more, the hardest two fifths of the roster, and run the same comparison.
| Neighborhood type | Language | Geographic | Gap |
|---|---|---|---|
| Suburban | 37.6% | 63.0% | −25.4 |
| Urban, sparse | 39.0% | 53.5% | −14.5 |
| Exurban | 48.6% | 57.1% | −8.6 |
| Urban, dense | 46.2% | 48.8% | −2.6 |
Same neighborhood type, same band of hardship, and in suburban tracts a geographic-district congregation is self-governing 63.0 percent of the time while a language-district congregation on ground just as hard is self-governing 37.6 percent of the time. Twenty-five points, with the difficulty of the ground removed from the comparison.
The shape of the table is the surprising part. The gap is widest where conditions are easiest and nearly closes where they are hardest. In dense urban tracts on hard ground the two halves are two and a half points apart, which is nothing. Hardship is not what produces this. Something about how the two kinds of district operate is.
Waldron is not an anecdote. It is the table.
Return to the four-cell table in section 4 and read it down the columns rather than across.
Inside the geographic districts, dependency and hard ground travel together. A dependent congregation there stands at 53.4 need pressure and 15.2 percent poverty; a self-governing one stands at 51.4 and 13.6. That is the pattern you would predict: a church on harder ground has a harder time supporting itself, and the district steps in.
Inside the language districts, the relationship disappears. Dependent congregations stand at 56.5 need pressure and self-governing ones at 57.4, which is slightly harder. Poverty is 18.1 against 17.9. Being self-supporting there has no relationship to how difficult the neighborhood is.
Dependent churches stand on measurably harder ground than self-governing ones. Need pressure 53.4 against 51.4, poverty 15.2% against 13.6%, and tracts with no other church 3.1% against 1.7%.
Dependent and self-governing churches stand on the same ground. Need pressure 56.5 against 57.4, poverty 18.1% against 17.9%. Hard ground is the baseline for the whole district, and self-support does not track it.
Those are two different institutions using one word. In the geographic half, District-affiliated looks like a response to a congregation's circumstances. In the language half, it looks closer to a standing arrangement, closer to a mission order than to a struggling church being carried.
Which of those two readings is correct matters a great deal for what anyone should do about it, and this data cannot settle it. The limits in section 12 say why.
The third class is the one that comes closest to marking a recent plant, and it behaves unlike either of the others.
Parent-affiliated congregations stand on ordinary ground by need: 52.8 against the denomination's own 52.7. What distinguishes them is emptiness. They are in tracts holding no other church at a rate of 4.8 percent, against 1.9 percent for self-governing churches, and they have 8.5 churches around them where the rest of the denomination has ten. Of the three classes they are the only one planting into places with measurably less church in them.
They are also the best-placed for a gathering. Their tracts score 53.2 on viability against the denomination's 51.0, with street-intersection density at 50.1, the only class reaching the national figure, and third-place density at 48.3. One percent of them are rural.
So the Assemblies of God does have a church-planting instrument pointed at under-churched places, and it is the parent-daughter relationship rather than the district. It is aimed at walkable urban and suburban infill, and it is 1,473 congregations, thirteen percent of the roster.
A separate GospelGraph finding, published in The Desert Moved, located the fastest-opening ground in the country: 2,186 tracts that are measurably improving, sit in the worst access tiers and could hold a gathering. Almost none are inner cities. They are new subdivisions, exurban build-out and university towns whose edges filled in.
The Assemblies of God stands in 212 of them, which is 9.7 percent of the frontier and far more than any reformed body reaches. It also means 1,974 of those tracts hold no Assemblies of God church, and 518 hold no church of any kind.
None of the classes in this paper explains that. Self-governing congregations put 2.1 percent of their number on the frontier; dependent ones put 2.3 percent. The denomination overall puts 2.1 percent, the PCA 1.6 and the ARP 2.3. Every roster and every class in this study lands inside a range less than a point wide.
The same flatness shows in the mirror of that set, the tracts that are hardening and already hold a church. Every roster in this study puts roughly twice as many congregations there as on opening ground: 2.0 to 1 for the Assemblies of God, 2.0 for the PCA, 1.7 for the ARP, 3.5 for the OPC.
That is what planting is, and no polity in this study escapes it. A church goes where a core group already lives, where a building is affordable and where a district can carry it. Opening ground changed too recently for any of those conditions to hold. Everyone arrives late because everything that has to line up before a church exists lines up late.
Polity explains who carries a church. It does not explain where the church is.
Each of these looked right at some point in the analysis and the data refused it. They are here because a finding is worth more when you can see what it survived.
The idea was that dense, walkable neighborhoods put enough people within reach to carry a congregation, so self-support would track viability. Across the 4,999 congregations on hard ground, self-governing churches score 52.8 on viability and dependent ones 53.3. The difference runs the wrong way and is half a point. Dead.
Language districts are three tenths of one percent rural and eleven percent dense urban, so they sit where any church would find self-support harder. But holding neighborhood type constant leaves the gap at 28.6 points in the suburbs and 18.8 in sparse urban tracts. Density explains none of it and the gap is largest in the easiest places.
Grouping parent and district affiliation together as a proxy for youth produces a tidy story: the younger cohort sits at 54.0 need pressure against 51.9. Separate the two and it collapses. Parent-affiliated churches, the ones that actually mark recent plants, sit at 52.8, which is the denomination's own average. The signal was coming from dependency, not from age.
Six hundred and fifty-seven tracts hold one Assemblies of God congregation that governs itself and one that does not. In 115 of them the dependent church belongs to a language district and the self-governing one does not. Those 115 are a natural experiment sitting inside the denomination's own records, and someone with access to charter dates and membership can answer in a week what this data cannot answer at all.
A district where dependency tracks hardship is carrying churches through difficulty. A district where dependency and hardship are unrelated is running a different polity. The language districts are the second kind, and whether that is deliberate should be a stated answer rather than an emergent fact. It changes whether 62 percent dependency is a success measure or a warning.
The denomination's presence on its hardest ground runs through 3,107 congregations that a district is carrying. Whatever the reason, that presence is the part of the roster most exposed to a budget decision. Before any district consolidates, the question worth asking is what share of its hard-ground presence sits in the dependent column.
It is the only class reaching under-churched tracts at any rate above the denomination's own, at 4.8 percent against 1.9. It is 13 percent of the roster. If the Assemblies of God wants more presence where there is less church, this is the mechanism already doing it, and it is the smallest of the three classes.
The polity column was on every row of the roster the whole time.
Read the limits before the findings carry any weight.
Tract means, deduplicated. Every figure computed from the national caches keyed by the roster's geoids, so the roster columns and the national column share a definition.
| Measure | General Council | District | Parent | Nation |
|---|---|---|---|---|
| Distinct tracts | 6,101 | 2,936 | 1,374 | 84,414 |
| Access (GDI) | 50.0 | 51.1 | 49.5 | 50.0 |
| Need pressure | 51.9 | 54.7 | 52.8 | 50.2 |
| Poverty | 14.0% | 16.3% | 15.0% | 13.2% |
| Churches in tract | 9.8 | 10.4 | 8.5 | 6.8 |
| Miles to nearest church | 0.96 | 1.29 | 1.02 | 0.96 |
| No church at all | 1.9% | 2.8% | 4.8% | 11.6% |
| Access tier Arid or worse | 18.3% | 21.9% | 16.5% | 20.3% |
| Desert tier | 0.8% | 1.5% | 1.5% | 1.0% |
| Shalom 2024 composite | 48.1 | 45.3 | 47.2 | 49.6 |
| Shalom living | 47.2 | 43.6 | 46.5 | 49.5 |
| Shalom health | 44.2 | 44.4 | 45.3 | 48.2 |
| Shalom hope | 51.9 | 48.4 | 48.6 | 50.3 |
| Shalom change 2020 to 2024 | −0.73 | −0.48 | −0.74 | −0.76 |
| MVI composite | 51.0 | 51.1 | 53.2 | 51.2 |
| Intersection density | 44.7 | 45.6 | 50.1 | 50.0 |
| Third place density | 44.7 | 45.9 | 48.3 | 44.2 |
| Plant posture | 31.6% | 37.1% | 35.9% | — |
| Archetype | General Council | District | Parent | Nation |
|---|---|---|---|---|
| Suburban | 57.3% | 47.6% | 49.6% | 42.8% |
| Urban, sparse | 24.0% | 28.9% | 34.5% | 32.1% |
| Exurban | 11.8% | 12.6% | 8.3% | 12.6% |
| Urban, dense | 4.4% | 7.0% | 6.6% | 7.9% |
| Rural | 2.5% | 3.8% | 1.0% | 4.6% |
Measured per congregation, centroid to centroid, against the 547 tracts nationally that hold no church in the merged church data and carry need pressure above 70.
| General Council | District | Parent | |
|---|---|---|---|
| Median miles | 30.8 | 27.9 | 21.9 |
| Within 5 miles | 14.5% | 19.8% | 19.0% |
| Within 15 miles | 33.6% | 37.0% | 41.6% |
The order is the same one section 8 finds on emptiness. Parent-affiliated churches are nearest to the unreached high-need ground and General Council churches are furthest from it.
Every figure reproduces from the roster.
denom:dump joins all fourteen layers to the roster, one row per congregation, 161 columns.denom:gap appends the nearest churchless high-need tract to every congregation.denom:audit reads every layer against the nation. With --groupby affiliation it produces the three-class tables above.denom:split divides the roster on a column and reads both halves, which produces the language comparison.denom:frontier reads any group against the 2,186 opening deserts.The scripts, the national baselines and the full column dictionary are in design/denominations/README.md and RUNBOOK.md in the GospelGraph repository. The saved runs behind this paper are ag-affiliation-audit.txt and ag-full-layer-audit.txt.
Published alongside this at gospelgraph.com
Every GospelGraph resource · the glossary, the methodology and the rest of the field reports.