Near-Repeat Analysis
Also known as: Near Repeat Calculator Method, Space-Time Near-Repeat Analysis, Near-Repeat Victimization, Contagion Crime Pattern Analysis
Near-repeat analysis tests whether crimes cluster in space and time beyond chance: after a crime occurs, are nearby locations at elevated risk for a short period? Developed in the early 2000s by Townsley, Johnson, Bowers and colleagues for burglary, it formalizes the 'contagion' or 'communicable disease' pattern of crime using a Knox space-time test against a Monte Carlo reference distribution.
Key highlights
- Detects short-lived, localized elevations in risk that static hot-spot maps miss, enabling timely prevention.
- Uses a permutation reference that automatically controls for the underlying spatial and temporal density of crime.
- Produces an interpretable matrix of space–time bands showing exactly how far and how long the elevated risk extends.
- Underpins prospective and predictive policing by forecasting where the next crimes in a series are likely.
- Applicable across crime types — burglary, shootings, vehicle theft, robbery — wherever spatial diffusion is plausible.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use near-repeat analysis when you have geocoded crime events with reliable dates and want to detect and quantify space-time contagion — the elevated short-term risk to locations near a recent crime. It is especially informative for burglary, shootings, robbery, and other crimes thought to spread by offender foraging or retaliation, and it directly supports predictive and prospective hot-spotting. It is less suitable when event dates are imprecise (only month or report date known), when crime counts are too sparse to populate the contingency table, or for crimes with no plausible spatial-diffusion mechanism. Results describe a pattern consistent with contagion but do not by themselves prove the causal mechanism.
Strengths & limitations
- Detects short-lived, localized elevations in risk that static hot-spot maps miss, enabling timely prevention.
- Uses a permutation reference that automatically controls for the underlying spatial and temporal density of crime.
- Produces an interpretable matrix of space–time bands showing exactly how far and how long the elevated risk extends.
- Underpins prospective and predictive policing by forecasting where the next crimes in a series are likely.
- Applicable across crime types — burglary, shootings, vehicle theft, robbery — wherever spatial diffusion is plausible.
- Establishes a space-time pattern consistent with contagion but cannot, alone, confirm the causal mechanism (same offender, learning, retaliation).
- Sensitive to the choice of spatial and temporal band widths, which can change which cells appear significant.
- Requires accurate event dates and coordinates; date uncertainty (e.g., burglaries discovered days later) blurs the temporal signal.
- Needs sufficient crime volume; sparse data leave contingency-table cells too small for stable inference.
- Multiple cells are tested simultaneously, so naive p-values risk false positives without correction.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
What is the difference between repeat and near-repeat victimization?
Repeat victimization is the elevated short-term risk that the same target (the same house, person, or business) is victimized again after an initial offense. Near-repeat victimization extends this to spatial neighbors: properties close to a victimized target also face elevated short-term risk. Near-repeat analysis quantifies how that excess risk decays with distance and time around the originating event.
Does a significant near-repeat pattern prove the same offender did it?
No. A significant Knox ratio shows that crimes cluster in space and time more than chance, which is consistent with several mechanisms: the same offender returning, offenders learning that an area is rewarding ('boost' accounts), or stable attractive features that draw different offenders ('flag' accounts). Distinguishing these requires additional evidence such as modus operandi, forensic links, or offender data.
How are the space and time bands chosen?
Bands should be set a priori from substantive knowledge and the data's resolution — for burglary, commonly 100–400 m spatial bands and 1–2 week temporal bands. Because results can shift with band width, good practice fixes the bands before analysis, reports sensitivity to alternative widths, and uses enough events that each cell has a stable expected count under the permutation reference.
Sources
- 1.Townsley, M., Homel, R., & Chaseling, J. (2003). Infectious burglaries: A test of the near repeat hypothesis. British Journal of Criminology, 43(3), 615–633.
- 2.Johnson, S. D., & Bowers, K. J. (2004). The stability of space-time clusters of burglary. British Journal of Criminology, 44(1), 55–65.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 22). Near-Repeat Analysis. ScholarGate. https://scholargate.app/criminology/near-repeat-analysis