How our scores work

Every number on this site traces back to data schools are federally required to report. Nothing is crowdsourced, estimated from news coverage, or guessed by an algorithm trained on reviews. Here is exactly what we do with that data — including what it can't tell you.

The safety score, step by step

  1. Count what's hard to hide. Serious incidents — weapons, physical attacks, arrests, law-enforcement referrals — are weighted 6×. Moderate incidents 2×. Minor conduct issues 0.5×. A school with 200 dress-code writeups is not the same as a school with 20 assaults.
  2. Normalize per 1,000 students, so big schools aren't penalized for being big.
  3. Protect small schools from bad luck. Rates are statistically shrunk toward the rate of similar-sized schools, so one incident at a 90-student school doesn't brand it dangerous.
  4. Rank nationally. The final score is a percentile against every school we were able to score — higher is safer.

The transparency grade — and why most schools have no score

The most dangerous number in school data is a zero that never happened. So before scoring a school we ask a narrower, answerable question: does this school's own report support any claim about its safety?

For every school we build a comparison group of the roughly 1,500 American schools closest to it in enrollment that teach the same grade levels. Within that group we look only at the schools that reported something, and take the 25th percentile of their incident rate — deliberately a low bar, the rate of a low-reporting peer, not a typical one. Multiplied by the school's own enrollment, that gives the number of incidents a low-reporting school its size would have filed. A school that reported less than a quarter of that number has not given us evidence of safety, and we cap its grade accordingly.

We do not adjust this comparison for state. The share of schools reporting nothing ranges from 13% in Georgia to 96% in Puerto Rico. Those are differences in reporting behavior, not in child safety, and hiding them behind a state-level baseline would be the opposite of what this site is for.

GRADEMEANS
AReported every federal category, at a rate in line with peers.
BReported in line with peers, but one or more categories are missing.
CReported almost nothing — but at this school's size a genuine zero is common enough that we can't call it implausible. Not evidence of safety.
DReported almost nothing, and schools its size normally report at least 3.
EReported almost nothing, and schools its size normally report 10 or more.
FNo usable record at all.

A safety score is published only for A and B. A school that filed nothing scores near-zero on incident rate, so scoring it would rank it among the safest in the country. About 42% of schools therefore have no score — that is the method working, not a gap in it.

The bar, in actual numbers

These are the reference rates the grades above were computed from — incidents per 1,000 students, at the 25th percentile of reporting peers. Anyone can reproduce them from the public CRDC file. A school is under-reporting when it files less than a quarter of what its own row here implies for its enrollment.

SCHOOL TYPE1003005008001,2001,800
studentsstudentsstudentsstudentsstudentsstudents
Elementary15.26.85.93.33.43.4
Middle18.210.49.914.012.211.5
High18.711.010.811.012.013.9

Read across: a 1,200-student high school whose peers report 12.0 per 1,000 would be expected to file about 14 incidents. Filing 3 or fewer puts it under the quarter mark; the grade then depends on how many incidents are missing in absolute terms — 3 or more short is a D, 10 or more short is an E.

City and state rankings

City, district, and state pages aggregate the raw serious-incident rates of their schools (never averaging percentiles, which is statistically invalid) and rank each geography against its own peers. A city is only ranked when at least 5 of its schools have trustworthy data — we show "why no ranking" honestly instead of ranking on thin evidence. City pages also show Zillow's typical home value, so safety can be weighed against price — see the Affordable & Safe guide.

Honest limitations

The fuller version of this — including the strongest arguments against publishing school safety data at all — is on the limitations and objections page.

What does the safety score mean?

It's a percentile from 0 to 100 — higher is safer. A 78 means this school's severity-weighted incident rate is lower than 78% of scored U.S. schools. It is not a letter grade or a pass/fail line.

Where does the data come from?

The U.S. Department of Education's Civil Rights Data Collection (CRDC) — mandatory reporting by nearly every public school — plus NCES Common Core of Data for school directory and enrollment information, and Zillow's Home Value Index for the home-price cross-reference on city pages.

Why does a school show no safety score?

Because its record doesn't support one. We publish a score only for schools graded A or B for transparency. A school that reported almost nothing has told us nothing about how safe it is, and scoring it anyway would rank it among the safest in the country — which is exactly backwards. About 42% of schools have no score for this reason.

What is the transparency grade?

An A–F measure of how much a school's own reporting supports a claim about its safety, separate from how many incidents it had. A school with many reported incidents can earn an A: it told us what happened. A school that reported nothing cannot, however clean the row looks.

My child's school reported zero incidents. Are you calling it a liar?

No. For a small school a genuine zero is common, and those schools are graded C — not evidence of safety, but not implausible either. A D or an E means schools of the same size and grade range across the country normally report several incidents, or ten or more, and this one reported almost none. That is a statement about the record, not an accusation about the school.

Can I use this to pick a home or advise clients?

It's built for exactly that research — comparing schools, cities, and districts on the same footing, with home values alongside. But it reflects what schools reported in a federal collection year, not everything that happened. Treat it as one strong input next to school visits, local knowledge, and your own judgment — not a verdict.