Report 215 · Luxury Home Security
When a house on your street gets broken into, the instinct is to feel lucky it wasn't yours. Criminologists who have mapped tens of thousands of burglaries say the opposite: for a short time, homes near a burglary are more likely to be hit next. The effect is real and well replicated. It's also short, local, and easy to misquote.
The name for this is the near-repeat effect. "Repeat victimization" is the long-established finding that a burgled home is at higher risk of being burgled again. "Near repeat" extends that to the neighbors. The paper that tested whether it holds everywhere is a 2007 study in the Journal of Quantitative Criminology by Shane Johnson of UCL and six co-authors from the UK, the Netherlands, the US and Australia. I read it in full, and it's the backbone of this report.
The method is borrowed from epidemiology. It's called the Knox test, and it was first used to ask whether childhood leukemia cases clustered in a way that suggested contagion. The logic is simple. Take every pair of burglaries in an area and record how far apart they were in distance and in days. Then shuffle the dates randomly, keeping the locations fixed, and count again. Do that 999 times. If real burglaries pair up close in space and time more often than almost any shuffled version, the clustering isn't chance.
The shuffle matters for anyone reading these results. Because the comparison uses each area's own burglaries, an area's overall crime rate or its seasonal pattern can't explain the result. What's left is the specific tendency for one burglary to be followed quickly by another nearby.
The team used police-recorded burglaries from two areas in each of five countries: Canberra and Beenleigh in Australia, The Hague and Zoetermeer in the Netherlands, Auckland and Palmerston North in New Zealand, the Wirral and Bournemouth in the UK, and Philadelphia and Pompano Beach in the US. Philadelphia alone contributed 8,079 offences. They looked in 100-meter and 14-day steps out to 2 kilometers and six months, and they deliberately excluded repeat burglaries of the same address so the result is about neighbors only.
The headline is in the conclusion: "for every data set analyzed, more burglaries occurred close to each other in space and time than would be expected on the basis of chance." In every area, the strongest effect was in the closest cell, within 100 meters and two weeks. Across all of them, homes within 200 meters of a burgled home were at elevated risk for at least two weeks. Risk was highest closest to the first burglary and faded with both distance and time.
The size of the window varied a lot by place. The authors summarize it area by area: about 200 meters and two weeks in Auckland and Pompano Beach, 500 meters and two weeks on the Wirral, 600 meters and four weeks in The Hague, 200 meters and eight weeks in Philadelphia, and up to 1,200 meters and four weeks in Canberra. Housing density explained some of that variation in three of the five countries, but not in the US or the Netherlands.
Here's the part I'd flag for anyone who repeats this research. The paper's effect size is a "Knox ratio": observed burglary pairs divided by the median count from the shuffles. For the Wirral, in the 100-meter, two-week cell, there were 366 real pairs against a median of 296 expected, a ratio of 1.24. That is 24 percent more close pairs than chance would produce. It is not the same as saying your home's risk goes up 24 percent, and it's certainly not "your risk doubles." It describes a pattern across an area's burglaries, not a personal probability. Your absolute risk in any two-week window is still low. It's just measurably less low right after a neighbor is hit.
For a high-value home, the most relevant finding comes from an earlier paper by Kate Bowers and Shane Johnson in the European Journal of Criminology (2005), which I could only read in abstract. In their data, "whereas repeat victimization proper tended to occur in more deprived areas, space-time clustering was more evident in affluent areas." They also found that the house next door was at "substantially heightened risk," especially within one week, and that homes on the same side of the street as the burgled house were at greater risk than those opposite, even after correcting for distance.
The 2007 paper offers two explanations for the affluent-area pattern, both labeled as hypotheses. The first is displacement: an affluent owner who has just been burgled tends to upgrade security quickly, which makes the neighbors the easier target. The second is planning: burglaries in wealthier areas are more likely to be planned trips, so the offender comes back to the area they've already scouted rather than the exact house. The authors' broader explanation is offender "foraging." Once a burglar has learned the access routes, sightlines and layout of one house, much of that knowledge transfers to the houses beside it. A 2024 study of solved burglaries in Edmonton, Canada, which I read only in abstract, found that repeat and near-repeat burglaries there were more likely than distant ones to be committed by the same offender, which fits that picture.
My reading, from designing these systems: the near-repeat effect is useful precisely because it's short. You don't need to rebuild your security because a neighbor was burgled. You need to tighten it for a couple of weeks, which is cheap and easy to do. In practice that means arming the alarm every time, including when you're home, closing every window and door the way you'd lock up for a long trip, checking that cameras covering the side and rear actually record, and paying attention to the same side of the street. If you can, find out how the neighbor's house was entered. The same access route may exist at yours, and that's the knowledge the burglar now has. The 2007 paper frames its findings mainly around police deployment, and the Bowers and Johnson abstract suggests prioritizing attention within two months and about 400 meters of a prior burglary.
A monitored system can make this automatic: a known break-in nearby is exactly the kind of signal that should raise the sensitivity and the response priority on surrounding homes for a set window, then relax again. 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.
For the bigger picture on what burglars say they look for, see Report 209 and Report 029.
Your area's window. The distance and duration vary widely between places, from 200 meters to more than a kilometer and from two to eight weeks in the 2007 data. No single number transfers to your neighborhood.
Whether reacting cuts burglaries. These studies show where and when risk rises. They don't test whether homeowner precautions during that window prevent the next burglary. It's a reasonable inference, not a measured result.
The full 2005 and 2024 papers. I read Bowers and Johnson (2005) and Emeno and colleagues (2024) only in abstract. Their specific numbers are not used here.
Data age. The 2007 study uses police data collected before its publication (mostly 12-month windows; the paper doesn't give calendar years for every area), well before video doorbells and neighborhood camera apps were common. Whether the pattern has shifted since is open.
Burglaries cluster in space and time. A study of police data from 10 areas in five countries found that in every one, homes near a burgled house faced elevated risk for a short window, at minimum within 200 meters for two weeks, fading with distance and time. Earlier work found the effect is stronger in affluent areas and on the same side of the street. The effect size describes area-level patterns, not a personal probability, so don't panic and don't misquote it. When a neighbor is burgled, tighten up for a couple of weeks and learn how they got in.
U.S. Army combat veteran (Counter-IED / Electronic Warfare), peer-reviewed researcher in microwave spectroscopy, and founder & CEO of Shroombiosis. Consults on laboratory operations, AI, and supplement formulation.