Philly Car Break-Ins Have Fallen 41% From the 2022 Peak — and 2025 Was a Record Low

My car windows were smashed three times in one summer — near Eastern State Penitentiary, on blocks where I had parked a hundred times before and never thought twice about it. The last time was a few days before my son was born. I had to replace the infant car seat before we could even bring him home from the hospital. After that I stopped parking on the street when I had any other option. That is my own small sample of the same crime this article is about.

I went looking at the Philadelphia Police Department’s incident-level data to see whether the summer of 2025 felt meaningfully different than the summers before it. It did. Through August 2026, the city has logged 5,986 theft-from-vehicle incidents — a 41.0% decline from the same January–August window in 2022 (10,142). That is the cleanest single-number statement of the trend. The 2022 spike of 15,546 incidents for the full year was the recent peak; 2025 came in at 9,771 — the lowest annual total in the PPD dataset going back to 2006, and a 37.1% decline from the 2022 peak on a full-year basis.

For the first time since the dataset begins in 2006, the city logged fewer than ten thousand theft-from-vehicle incidents in a calendar year. The previous record low in this series was 2024. The previous high was 2007 at 17,667 — before the current decline, the recent peak in 2022 was still well below the late-2000s level. The drop from 2022 is not a return to the historical norm; it is a new low. The city has fewer car break-ins than at any point in two decades of public reporting.

Other larcenies are up — so this is not a citywide reporting collapse

When one crime category falls sharply, the first question I want to answer is whether the data is real or whether something downstream is suppressing the counts. The most common explanations for a sudden reported decline in a specific UCR category are reclassification (the same incidents getting coded under a different label), procedural changes (the agency stopped taking certain reports), and overall reporting volume dropping (which would drag every category down together).

To test the third explanation, I indexed three UCR categories to their own 2022 level and compared their trajectories. The chart below is the result. Theft from vehicle is the only one of the three that falls. General Thefts (the catch-all larceny category for shoplifting, package theft, snatching) ran about 11% higher than its 2022 level through Jan–Aug 2026. Motor Vehicle Theft (the stealing of the vehicle itself) ran about 31% higher. The Philadelphia Police Department’s overall reporting pipeline is fine — two of the three larceny categories are rising, and they rise faster than theft from vehicle is falling. The decline in this one category is not the whole crime picture contracting.

That rules out the “they just stopped counting” story. It does not rule out the reclassification story. The text_general_code field in the PPD table does separate Theft from Vehicle from plain Thefts, but the two share a UCR code (600). I cannot tell from the public data whether an officer in 2025 who used to write a Theft from Vehicle report is now writing something else. What I can see is that the absolute volume in the plain Thefts bucket grew substantially faster than the theft-from-vehicle bucket shrank — so even if some incidents migrated across the boundary, the net effect across the two categories would still be growth, not decline. The decline here is real, in either direction.

Every hour of the day and every district with a real baseline is down

I sliced the same dataset a few different ways to make sure the decline was broad rather than concentrated. Two cuts stand out.

First, by hour of day. Theft-from-vehicle incidents are not evenly distributed across the 24-hour clock — the bulk land in the afternoon and evening, peaking around 1pm (1,326 incidents in 2022) and again around 8pm (1,127). I compared 2022 against the LTM as of Aug 2026 (September 2025 through August 2026) at every hour. The decline is broad: every single hour of the day shows fewer incidents, with overnight hours (8pm to 4am) down about 46% and midday hours (11am to 3pm) down about 31%. If the decline were driven by, say, a single enforcement action that targeted only evening street robberies, the morning hours would look unchanged. They do not. The decline is citywide and all-day.

Second, by police district. There are 21 numbered patrol districts in Philadelphia (the airport district, District 77, is excluded for low volume). I looked at every district that had at least 100 incidents in 2022 (the threshold that drops small-sample noise). Twenty districts qualify. All twenty show declines. The steepest drops are in District 8 (-60%), the merged 06+09 unit (-57%), District 24 (-55%), District 7 (-53%), and District 18 (-51%). The shallowest drop is in District 25, at about -8% — a meaningful decline but the smallest in the comparison. No district in Philadelphia with a measurable baseline has seen an increase in theft-from-vehicle incidents. The decline is not a story about certain neighborhoods getting safer at the expense of others.

One district caveat. District 6 was folded into District 9 during a 2024 consolidation of district reporting; District 6 reported 7,338 incidents in 2023, 2,193 in 2024, and zero after. The two are reported here as a single 06+09 unit, because treating them separately would show District 6 down 100% (it just stopped existing as a reporting bucket) and District 9 spuriously flat. The 06+09 combined decline of -57% is the meaningful figure. Every other district in the comparison is unaffected.

I cannot decompose the drivers — but here are the candidates the data points at

The data can tell me that the decline is real, broad, and not a measurement artifact. It cannot tell me why. Here is what I think the candidates are — I am not confident in the relative weights, and the data cannot break them out.

The Philadelphia Police Department has, by several accounts, prioritized car-break-in enforcement over the past few years, including focused efforts against the small rings that account for a disproportionate share of incidents. A handful of prolific offenders working a single neighborhood on a single night can move the citywide numbers; taking even one or two of them off the street produces a noticeable shift in the trend. I cannot quantify this from the public data; PPD does not publish clearance or arrest counts in the incidents table.

Automatic License Plate Reader (ALPR) deployment has scaled up in Philadelphia and in surrounding jurisdictions. ALPRs change the risk calculation for a vehicle used in a string of break-ins: driving a stolen plate, or even a freshly purchased plate, becomes a recoverable mistake. To the extent that any meaningful share of theft-from-vehicle incidents are committed by offenders driving between targets in personal vehicles, ALPR coverage makes those targets more expensive.

The vehicles themselves have changed. A May 2024 Highway Loss Data Institute bulletin flagged Kia and Hyundai models without engine immobilizers as dramatically more theft-prone than comparable vehicles — the social-media-driven “Kia Boyz” trend exploited a specific missing anti-theft feature. As those vehicles age out of the fleet and as the affected model years receive software and hardware updates, the share of vehicles on Philadelphia streets that are easy to steal from is shrinking. Newer vehicles are harder to break into in general — laminated side glass, factory alarm systems, and immobilizer standardization all raise the cost and the risk of a quick smash and grab.

I also want to flag a countervailing concern about the integrity of the data itself. A January 2024 Philadelphia Inquirer investigation — a Pulitzer Prize finalist for local reporting — documented years of manipulation of crime-data figures inside PPD, including downgrading serious offenses to lesser ones. A one-year follow-up in January 2025 found that PPD’s data was still failing to count accurately more than a year after the original reporting. The 2025 decline in theft from vehicle would be a tempting target for the same kind of suppression — a high-volume, low-priority category where an officer who wants to clear a case quickly can reclassify a Theft from Vehicle as a non-UCR infraction or decline to take the report at all. I cannot rule that out from the public data, and the article’s headline should be read with that caveat in mind.

What I’d say if you asked me

The 41% drop is real, broad across the city, and not a reporting artifact — that is the strongest thing I can say about it. The weakest thing I can say is that I cannot tell you which combination of enforcement, ALPRs, vehicle hardware, and resident behavior drove it; the public data does not expose the underlying incidents well enough to decompose the contribution. The Inquirer investigation is a real caveat: the same kind of pressure that suppressed other crime categories could be suppressing this one too, and the public data cannot rule that out.

What this article is not: it is not a story about Philadelphia getting safer. Thefts and motor vehicle thefts are both up double digits over the same window. The category that is down is the one that hits people who park on the street and walk away. That is a real, victim-visible thing. It is also one thing.

So given that not all crime categories are declining — in fact, it feels like porch piracy has gotten worse in my neighborhood — you might ask why I focused on car break-ins specifically. I think it comes down to three things.

First, this is one I have personally felt. I had my car broken into several times in 2024, and it has not happened once since November 2024 — nearly two years now. I notice that. I notice it more than I notice the larceny numbers going up.

Second, car break-ins are particularly expensive. A porch pirate might cost you twenty dollars. A car break-in runs around five hundred dollars, and you lose the car for a week while the window is replaced. That is a meaningfully different category of victimization than most of the other property crimes in this dataset.

Third, it is stressful in a way the numbers do not capture. The last time it happened to me, my son was about to be born. I had three days to get the car fixed or I would not have been able to bring him home from the hospital. Cars are critical to how we get around, and having one broken into stops you from going home, going to work, or — as was almost the case for me — bringing your newborn home.

Notes, Sources, and Methodology

All findings come from the Philadelphia Police Department incident-level data on OpenDataPhilly, queried via the phl.carto.com SQL endpoint. Theft from Vehicle is filtered on text_general_code = ‘Theft from Vehicle’. The UCR code (600) is shared between Theft from Vehicle and plain Thefts, so filtering on UCR alone would merge two series that move in opposite directions — text_general_code is the only field that separates them.

Methodology. The headline figure is a like-for-like Jan–Aug comparison: 5,986 (2026) against 10,142 (2022), a 41.0% decline. The full-year comparison is 9,771 (CY2025) against 15,546 (CY2022), a 37.1% decline. The LTM-as-of-Aug-2026 frame is 9,465 (Sep 2025 – Aug 2026) against 15,546 (CY2022), a 39.1% decline. All three framings land in the same band; the headline uses the like-for-like month window because it is the cleanest apples-to-apples comparison and includes the most recent data. The hero chart uses calendar-year totals (CY2018–CY2025). District cuts use CY2022 against LTM as of Aug 2026. Districts below the 100-incident baseline in 2022 are excluded; District 6 and District 9 are merged into a single 06+09 unit because of a 2024 consolidation of district reporting.

Limitations. The incidents table exposes no arrest, clearance, or property-recovery fields, so no outcome breakdown is produced. The article does not decompose the 41% decline into the relative contribution of enforcement, ALPRs, vehicle hardware, or behavioral adaptation. The January 2024 Inquirer investigation of PPD crime-data manipulation is the load-bearing caveat on the headline number; the public data cannot rule out category suppression as a partial driver. The 41% figure is extrapolating from eight months of 2026.