How the scores are built
Two scores, four measures each, and one rule: a measure earns its place only if there is a real, cited number behind it. Everything that failed that test was removed rather than estimated.
What goes into each score
Two scores. Four measures each. Every measure has a cited number behind it, and each is set out in full below.
Vulnerability
- 25% Alonemiles to the nearest federation
- 25% SmallJewish population
- 25% Under-resourcedmoney per resident, from filed 990s
- 25% ThreatADL incidents per 100k, by state
Strategic value
- 40% Reachhow many people benefit
- 25% Futuremeasured change, where two figures exist
- 20% StudentsJewish students nearby
- 15% Accesscapital city, or covers a state
What a score of 40 actually means
Less than you would guess from the number, and more than you would guess from the scale. A vulnerability score of 40 sounds like a community two fifths of the way to the worst imaginable case. It is not. It is a community more exposed than about nine in ten of the others on this map.
The scale runs from 0 to 100 because each of the four measures does, and the score is their average. But no community scores near 100, and none ever will, because reaching it would mean being the most isolated, the smallest, the poorest and the most exposed to incidents all at the same time. In practice the field runs from about 5 to about 56, with half of all communities between 24 and 36.
So a score is close to meaningless read alone and clear read against the field, which is why every community page shows where it falls among all of them rather than only its number. The gap between 30 and 40 is not ten points of anything; it is the difference between the middle of the pack and the worst tenth.
- Boston, 16
- Little Rock, 34
- Helena, 52
Rescaling this to fill 0 to 100 was considered and rejected. It would make every score relative to whoever else is in the set, so adding one community would silently rescore all the others, and a community's number would move when nothing about it had changed. The honest fix is to publish the position, not to stretch the number.
How complete the picture is
One square per community. A score needs at least three of the four measures, so the grey squares are communities we decline to score rather than communities we score badly.
- 81scored on all four measures
- 55scored on three of four
- 11not scored: fewer than three available
Vulnerability
How hard is it to sustain Jewish life here? Four measures, weighted equally.
25% each
- 01
Alone
How far is the nearest other federation?
Source. Great-circle distance computed from each federation's coordinates. Exact for all 147.
Note. Measured for all 147, and weighted for those below 25,000 Jews. A community above that runs its own schools, synagogues and JCC and does not depend on a neighbouring federation, so the distance is still shown on its page and simply does not move its score.
- 02
Small
How few Jewish residents are there?
Source. American Jewish Year Book estimates, checked against 2024 county figures from the same authors and against each community's own published study.
Note. Banded rather than linear. The step from 300 residents to 3,000 changes what a community can sustain far more than the step from 100,000 to 300,000.
- 03
Under-resourced
How little money is there per Jewish resident, and how thin is the cushion?
Source. Filed IRS Form 990 returns, retrieved through the ProPublica Nonprofit Explorer. Every figure links to the return it came from.
Note. Operating entities and endowments are read separately and never summed. The cushion is months of operating reserve: net assets WITHOUT donor restrictions, taken from Part X of the filed return, over one month of expenses. Endowment a federation cannot legally spend on operations is excluded. For 32 of 124 filers the return does not break the total down, and those fall back to total net assets, which the panel labels as an upper bound rather than a reserve.
- 04
Threat
How many antisemitic incidents, per 100,000 Jews?
Source. Anti-Defamation League, Audit of Antisemitic Incidents.
Note. ADL publishes by state. Every community in New York carries New York's total, so this measures the state a community sits in rather than that community's own experience. Community-level counts exist on ADL's own map but are not available for programmatic use.
Strategic value
If resources went here, how much would they reach?
Reach 40% · Future 25% · Students 20% · Access 15%
- 01
Reach
How many Jewish people would an investment reach?
Source. American Jewish Year Book estimates.
- 02
Future
Has the community been measured growing or shrinking?
Source. Change between two published figures. US federations use the American Jewish Year Book's federation service area table, with its town-level appendix as a size-checked fallback; the seven Canadian communities use Statistics Canada census tabulations and are labelled as such, because a census count and a survey estimate are not comparable.
Note. Scored for 38 of 147 communities, and left out of the score entirely for the rest, with the remaining weights renormalised. Nobody measures these communities annually, so most have only ever been counted once. A blank here means no second count exists, not that the population held steady.
- 03
Students
How many Jewish students are on campuses nearby?
Source. Hillel International campus estimates, matched to catchments geographically.
- 04
Access
Does this community sit where decisions are made?
Source. State and federal capitals, and federations covering a state or several.
Why the weights are equal
There is no evidence base for saying isolation matters one and a half times as much as antisemitic incidents. An earlier version of this score used uneven weights that were not derived from anything either. Equal weighting is the standard approach where no principled basis for a difference exists, and it is the only version a reader can check without a spreadsheet.
What we left out, and why
Two measures used by the earlier score were removed in August 2026 rather than repaired, because their inputs were not real.
- Territory size. Catchment area was a single placeholder value for all 26 metropolitan federations and for 84 of the 104 regional ones. It is genuinely recorded for the 17 that cover a state or more, which is too few to score a field of 147 on.
- Jewish density. This divided Jewish population by a service-area population that had no recorded source: all 147 values were round thousands, and Little Rock carried the whole of Arkansas against a metro a quarter that size. Those figures have since been retired and replaced with named Census geographies, which the community panels show. Density stays out of the score regardless, because population and isolation already reach for what it was reaching for.
- Capital city and campus presence were counted as vulnerabilities. They are advantages, and they now sit in strategic value, where a community gains from them rather than being marked down.
Why we say measured change and not trend
A trend needs a series. For Jewish communities in the United States, no series exists, because nobody measures them annually. Communities commission a population study every ten to twenty years, and between studies there is nothing to plot.
The American Jewish Year Book publishes a population chapter every year, which sounds like the series we would need, but it is a compilation rather than a measurement: a community's figure moves only when a new local study appears. Phoenix appeared in five consecutive editions at 23,600, carrying an estimate dated 2009, before a 2021 study moved it to 28,600. Across six editions, 905 communities appear consistently and only 146 change at all.
So we publish what can honestly be said: two figures, the years they were published, and the change between them. Twenty-seven communities have that. The rest show nothing, because a number would imply a measurement nobody took.
One source per community, never a mixture. US federations use the Year Book's Table 5, which is published at federation service area level and so matches our unit exactly; its Appendix A was rejected because it lists towns, and a town inside a statewide catchment produces a confident wrong answer. The seven Canadian communities use Statistics Canada census tabulations, and say so on their own pages, because census counts and survey estimates are not comparable with each other.
When we publish no score at all
A community needs at least three of the four measures to receive a score. Averaging whatever happens to be available sounds cautious and is not: it assumes the missing measures would have looked like the present ones. Applied without that rule, three Canadian federations came out as the most vulnerable communities in North America, on half the evidence of every community around them.
Eleven communities are therefore unscored and are drawn in grey. Ten are Canadian: they file a T3010 return with the Canada Revenue Agency rather than a Form 990, and the ADL audit covers US states, so two of the four measures are unavailable. That is a gap in this Atlas, not a statement about those communities.
What these scores are not
They are relative, not absolute. A score of 40 means more exposed than most communities in this set, not 40% of some maximum. They describe conditions a federation operates in, not the quality of its leadership or its work. And they are built on sources that publish on their own schedules, so figures differ in vintage between communities; each community profile shows the year behind its own numbers.
If a figure about your community is wrong, tell us and we will correct it. Corrections with a source move fastest. See Get Involved, and the Terms for how often the data is revised.