What the 25% burglary drop measures

The source is a four-page Bureau of Justice Statistics release, Preliminary Victimization Estimates from the NCVS, 2025 (NCJ 312234, October 2026), by BJS statisticians Jennifer Truman and Susannah Tapp. The NCVS is the National Crime Victimization Survey: the Census Bureau interviews a nationally representative sample of households about crimes they experienced, "including those that are reported and not reported to police." That last part is why it's the better national gauge for burglary than police counts. Plenty of break-ins never get reported.

What the 25% is made of

The headline category is "burglary/trespassing," which BJS notes was "called household burglary in prior reports." It covers unlawful or forcible entry, or attempted entry, of "a permanent residence, other residence (e.g., a hotel room or vacation residence), or other structure (e.g., a garage or shed)." Table 2 splits it into two pieces:

• Burglary/trespassing combined: 14.0 per 1,000 households in 2024, 10.5 in 2025. That's the 25%.
• Burglary, meaning entries where the offender "committed or attempted a theft": 9.7 to 7.6 per 1,000. By my arithmetic, about a 22% drop.
• Trespassing, entries with no theft or attempted theft: 4.4 to 2.9 per 1,000. About a 34% drop, again my arithmetic.

All three changes are flagged as significant at the 95% level. So the decline is real by the survey's own test. But the part most homeowners mean when they say "burglary," someone getting in to take something, fell by roughly a fifth, and the bigger percentage came from entries where nothing was taken. In raw numbers, BJS estimates 1,012,480 burglaries with theft or attempted theft in 2025, down from 1,287,180.

Why you can't line it up with an older chart

This is the part that will trip people up for years. BJS "fully transitioned" to a redesigned survey instrument, the NCVS-R, in January 2025, and the release says plainly that "the NCVS-R estimates are not comparable to legacy NCVS estimates." The old instrument ran from 1992 to 2024.

The 2025-versus-2024 comparison is fair, because in 2024 BJS ran both instruments side by side, half the sample each, and the 2024 number in this release comes from the new one. The companion report, Victimization Estimates from the Redesigned NCVS Instrument, 2024 (NCJ 312200), shows how much the instrument matters. The redesign uses "expanded screener probes and more behaviorally specific language," and among its listed changes: "Requirement removed that location be 'at or near home' to ask about break-ins." In the 2024 split sample, the new instrument measured burglary at 9.7 per 1,000 households versus 8.2 on the old one. BJS's table doesn't flag that particular gap as significant, but across property crime overall the new instrument came in 31% higher.

The practical rule: if you see the 2025 figure plotted as the next point on a decades-long burglary line, someone has mixed two different rulers. A fair comparison going forward starts at 2024 on the new instrument. I've cited older BJS burglary data here before, for instance how often someone is home during a burglary; those numbers come from the legacy instrument and belong on the old ruler.

Three more cautions in the fine print

• It's preliminary. BJS says the estimates "may change" when it publishes final figures in Criminal Victimization, 2025, forthcoming in 2027.
• Two months are missing. Interviews weren't collected in October and November 2025 "during the lapse in federal government appropriations." To compensate, BJS multiplied base weights by 1.5 so the four months it did collect (July, August, September, December) stand in for the second half of the year. That's a reasonable fix, and it's also an assumption: that those four months look like the six.
• Response is about half. 53% of eligible households completed interviews in 2025 (101,980 household interviews). Surveys adjust for nonresponse with weighting, but it's worth knowing the base.

For a sense of precision: the appendix gives a standard error of 0.60 on the 2025 burglary rate of 7.6. A rough 95% range is about 6.4 to 8.8 per 1,000 households (my arithmetic, using 1.96 standard errors).

A side note on the press release, since it's where most people met these numbers. Its headline says violent crime "plummeted" and its subhead says it "dropped 25%." That 25% is violent crime excluding simple assault. In the BJS table, total violent victimization went from 28.7 to 25.7 per 1,000 and simple assault from 18.8 to 18.3, and neither change carries the significance flag. The serious-violence decline is significant. The broad "violent crime" total, by the table's own markings, is not.

What a national rate says about one house

Not much, and the release doesn't pretend otherwise. It has two tables. There's no breakdown by income, home value, region, or whether anyone was home. A falling national average is good news, and it's also an average across 134 million households. It says nothing about a targeted crew working one neighborhood, or about the vacation property that sits empty for months, which, notably, the survey's definition does include.

I help design the AI security systems for a veteran-owned (SDVOSB) home-security company run by fellow veterans. I do not own that company and earn nothing from this link. Full policy here. From that seat, here's how I'd use a number like this. It's useful context for the base rate, not a reason to change a plan. The questions that matter for a specific property are local and specific: what entry points exist, who knows the house is empty and when, and how fast an alarm turns into a response. The national rate can't answer any of them. My background in Army counter-IED electronic warfare taught me the same lesson in a different form: the average threat level in a region never told you what was waiting on a particular road.

How to read the next burglary headline

• "Burglary fell 25%." Ask: burglary alone, or burglary plus trespassing? In this release, burglary with theft fell about 22%.
• "Lowest burglary rate in decades." Ask which instrument. A claim that compares NCVS-R to legacy NCVS years is comparing different measurements.
• "Burglary is down, so security matters less." The survey measures a national average. It doesn't measure risk at your address.

What I could not confirm

Final numbers. These are preliminary; the 2027 final report could revise them.

Why it fell. The release reports rates, not causes. I've made no claim about why burglary dropped, and the release doesn't either.

High-value homes specifically. I found no BJS breakdown of 2025 burglary by home value or household income in either report I read.

Police-reported data. I haven't compared these figures with FBI counts for 2025, which measure a different thing (crimes known to police).

Not my research. I had no part in this survey. My published work is in microwave spectroscopy. The percentages labeled as mine are simple arithmetic on BJS's published rates.

The signal

The 25% drop is real and statistically significant, but it's burglary and trespassing combined. Burglary with theft fell about 22%, from 9.7 to 7.6 per 1,000 households. The figure comes from a redesigned survey that BJS says can't be compared with the 1992 to 2024 series, it's preliminary, and two months of 2025 interviews were filled in by weighting. Take it as good news about the national base rate. Don't take it as a statement about any one home, and don't let anyone plot it on an old chart.