Neighborhood Affordability Project · July 2026

Does the city tell you
whether the neighborhood is safe?

Everyone carries a mental map of safe and unsafe cities. So we took the national crime percentile of 264 Orthodox communities, each one ranked against every zip code in the country, and held the mental map up against it. For a family choosing where to live, that map turns out to be close to useless.

The bar below covers a single metro, New York. Its left end is the lowest-ranked of the metro's 92 Orthodox communities and its right end is the highest, on a scale where the 0th percentile means more reported crime than almost any zip code in America and the 100th means less. Red is the bottom of that range, blue the top. Watch how much of the country one metro manages to cover.

Scroll to see why

The short version

Which city you pick barely matters. Which neighborhood you pick is almost everything.

Two facts break the safe-city instinct in half.

Within a single metro, communities span nearly the whole national safety range. In the New York metro alone, communities run from the 6th percentile to the 99th. The metro's name on the map tells you almost nothing about where any one community lands.

Reputation inverts about as often as it holds. The glamorous, "obviously safe" enclaves cluster below the median U.S. zip code. Several "gritty" Rust Belt communities rank in the country's top third.

01 · The city is the wrong unit

One metro can hold both ends of the country

If several Orthodox communities share a metro, do they land anywhere near each other?

"Is this city safe?" has no answer. When several Orthodox communities share a metro, they scatter across the national percentile range, often almost end to end.

Each row in the chart below is one metro with four or more Orthodox communities in it. The bar runs from that metro's lowest-ranked community to its highest, on the same 0th to 100th national scale, so a longer bar means the metro holds a wider spread of places, and further right means safer. The black diamond sitting on each bar is that metro's median community. The dashed vertical line down the middle is the 50th percentile, the median U.S. zip code. Each bar takes its color from its two ends: red where a community falls below that line, blue where it rises above it.

Notice how many of these bars cross the dashed line and keep going.

National safety percentile, by metro

Higher = safer than more U.S. zip codes. Bar spans the lowest to highest community; ◆ is the metro median.

Lower half (below the median U.S. zip) Upper half 50th percentile

Start at the top row. New York's bar begins at 6 and runs to 99, a 93-point spread across 92 communities, with the diamond at 58. The two rows under it barely differ: Washington's 7 communities run from 5 to 96, Miami's 16 from 15 to 95. Now drop to the bottom row. Los Angeles is the one bar on this chart that never reaches the dashed line. All 12 of its communities sit between the 6th and the 42nd percentile, diamond at 27. In most of these metros a family can choose its way to safety. In that one, no neighborhood escapes.

02 · Reputation is a coin-flip

The "obviously safe" places are the low ones

Does a city's reputation at least point in the right direction?

Sort communities by popular reputation, the glamorous coastal enclaves everyone knows are safe against the Rust Belt cities everyone knows are rough, and the national percentiles will not line up. One assumption fails reliably: affluence does not buy national safety. Dense, wealthy, high-traffic districts generate a lot of reported crime.

The chart below puts twelve communities on one scale in two labeled groups. Each dot is one community, placed at its own national percentile, and the scale runs the same way as the last chart: right is safer. The dashed line at the 50th is the median U.S. zip code. A dot to the left of that line is drawn red, a dot to the right of it blue. The label on the left tells you the reputation. The dot tells you the number.

Look at which group ended up on which side of the dashed line.

Reputation against reality, on the national safety percentile

Each dot is a community, placed by its true national percentile. Reputation is on the label; the number is the reality.

In the top group, Beverly Hills and San Francisco sit furthest left at the 6th percentile, and Deal comes closest to the dashed line at the 26th without reaching it. Every dot in that group is red. In the bottom group, Jacksonville sits furthest right at the 91st, Lancaster is the nearest to the line at the 65th, and every dot is blue. Reputation pointed the wrong way in all twelve cases. Read that carefully: it does not make a gritty city secretly safe. It means the name of the city carries almost no information about any specific community inside it, in either direction.

03 · Why the myth survives

The local grade flatters, 86% of the time

Then why is everyone so sure their own neighborhood is fine?

Every community carries two safety readings. A city-relative grade is a letter for how it compares with the rest of its own metro. A national percentile is a rank against every zip code in the country. The first answers "is this a nice part of town?" The second answers "is this a safe place, period?" They routinely disagree, and the gap runs one way.

The three cards below size that disagreement. The local grade is the rosier of the two readings 86% of the time, which is about six communities in every seven.

of the time, the local grade is rosier than the national percentile

letter grades, the median distance between the local grade and the national one

communities graded A-range locally that still fall below the median U.S. zip

That gap does its heaviest work on the "obviously safe" enclaves. Their local reputation says very safe. Their national standing says the opposite.

Each of those six communities gets one row below. The hollow dashed circle pinned near the right-hand end of every row is where reputation puts the place, a stand-in for the word "safe" rather than a measured number. The solid red dot is where the community actually ranks against every U.S. zip code. The orange line joining them is the distance between the two answers, so a longer line means a wider gap. The dashed vertical line at the 50th percentile is the median U.S. zip.

Follow each line from the hollow circle to the red dot and watch how far it has to travel.

Local reputation says "safe." The national number disagrees.

Each line connects a community's reputation (safe, on the right) to where it actually ranks nationally (the dot).

Beverly Hills is the top row and the longest line on the chart: its red dot ends at the 6th percentile, almost the full width of the scale from where its reputation sits. San Francisco travels just as far, to the same 6th. La Jolla stops at the 14th, Miami Beach at the 15th, Mercer Island at the 17th. The shortest line belongs to Deal at the 26th, and even that dot never reaches the dashed 50th. Reputation is not lying. It is answering an easier question, "safe for this city," in place of the one the family asked.

04 · See it on the map

Switch the yardstick and the map changes color

What happens to the same 36 places when you change which yardstick colors them?

Here are 36 communities across the country, each with a real CrimeGrade national percentile. The two buttons above the map choose which crime yardstick colors the dots, and the two rarely agree.

National percentile

How safe a place is compared with every zip code in the country. The honest, comparable number. Red = below the median U.S. zip; blue = above it.

Local grade

How safe a place looks compared with only its own metro, the nice part of town. The story realtors and neighbors tell. Almost everything grades well.

Both scales are built on the same federal source: the FBI's national crime databases, the Uniform Crime Reporting (UCR) program and the National Incident-Based Reporting System (NIBRS), supplemented by direct reports from state and local police. CrimeGrade and DoorProfit both start from that FBI data. They just frame it differently (one against the whole country, one against the local metro), which is exactly why the two lenses can disagree about the same place.

Every dot on the map is one community, sitting at its own location. In the national view the color runs from deep red at the bottom of the country, through pale shades in the middle, to deep blue at the top, and the turn from red to blue is the median U.S. zip. In the local view there are only two colors, both green: dark green for a community graded A inside its own metro, pale green for a B.

Press one button, then the other, and watch the red disappear.

Hover (or tap) a dot for the community's reading. Dots sit at each community's location; several Northeast communities overlap closely.

Start in the national view. The darkest red dot on the map is East Memphis at the 2nd percentile, with University City in St. Louis just behind it at the 5th, and the pair on the California coast, Beverly Hills and San Francisco at the 6th, nearly as dark. The deepest blue anywhere is Jacksonville at the 91st. Now press local, and every dot on the map turns green: these really are the nicer parts of their metros. Two of those local grades are confirmed, Beverly Hills and East Memphis, each graded A within its own metro by DoorProfit despite ranking near the national bottom at the 6th and 2nd percentile. All 36 national percentiles are real CrimeGrade figures: 16 cited in the source papers, 20 more looked up from CrimeGrade (the same instrument) in July 2026, keyed to each community's Orthodox neighborhood. The local-grade layer illustrates the papers' measured flattering pattern. DoorProfit does not publish per-neighborhood grades to look up, so sourcing the local grade for every community needs the master Heshbon database.

The bottom line

Drop the city label. Check the specific place.

So: does the city tell you whether the neighborhood is safe?

No, and it points the wrong way about as often as the right way. The city-level label is the wrong unit of analysis, and within a single metro the safety of individual communities ranges from the bottom of the country to the top.

The only reliable read is the specific community on a national scale. With the property-crime caveat below firmly in mind, that is the one number worth checking before anyone concludes a place is safe or unsafe.

Two cards close the ledger. The left one is the middle community of all 264, the one with as many communities above it as below. The right one is the share of those 264 that fall below the median U.S. zip code.

42nd

median national percentile across all 264 communities

56%

sit below the median U.S. zip nationally

What this measures, and what to read before quoting a number

Where the crime numbers come from: the FBI

Both scales rest on the same federal foundation: the FBI's national crime databases, the Uniform Crime Reporting (UCR) program and the National Incident-Based Reporting System (NIBRS), supplemented by direct reports from state and local law enforcement. CrimeGrade builds its national percentile on that FBI data; DoorProfit calibrates its A–F local grades to FBI UCR national averages (total crime ≈ 2,213 per 100k, violent ≈ 381, property ≈ 1,832). The two verdicts differ because each frames the same FBI numbers against a different yardstick, the whole country versus the local metro, and never because the underlying crime counts differ. Because the FBI's UCR is a voluntary program with reporting gaps, both providers backfill with state and local agency data.

What the percentile actually captures

The national percentile is CrimeGrade's zip-level score, weighted heavily toward total (largely property) crime. Read it as overall reported-crime density rather than as a violent-crime index. Dense, wealthy, high-footfall areas (luxury shopping districts, tourist zones) post high reported counts and therefore low percentiles even where violent crime is moderate. Some of the single-digit scores for affluent enclaves reflect that, and not street-level danger.

Two different sources

Both instruments draw on the same FBI crime data (UCR / NIBRS); they differ only in how they score it. The national percentile (CrimeGrade) ranks a place against every U.S. zip code; the city-relative grade (DoorProfit) grades it against its own metro. This essay uses the national percentile as the comparable yardstick and treats the city-relative grade as the local narrative.

Grain and coverage

Figures are keyed to each community's most common zip code, which can span several miles rather than the specific blocks around a given shul. National percentiles exist for 264 of 319 communities; the within-metro analysis uses metros with four or more communities.

The reputation groups are illustrative

The "posh" and "gritty" lists are hand-selected to represent common perceptions, not a random sample. They demonstrate that reputation is unreliable. They are not a precise effect size.

Source: The Safe-City Myth and the Heshbon crime and real-estate database, Neighborhood Affordability Project, July 2026. Underlying crime data: the FBI's Uniform Crime Reporting (UCR) program and NIBRS, plus state and local police reports, via CrimeGrade and DoorProfit. A visual essay.

Does the city tell youwhether the neighborhood is safe?