Place-Based Crime Analysis Tool

Prioritise high-crime places, test whether earlier hotspots predict future crime, or evaluate a place-based intervention. The tool reports crime capture, PAI, PPAI, RRI, controlled before and after effects and Weighted Displacement Difference (WDD).

CSV data are processed in your browser and are not uploaded to a server by this page.

Prioritise places

Upload one row per spatial unit with unit_id, area and crime_count.

Upload your CSV

Required: unit_id, area, crime_count

Built-in sample dataset

Choose a sample for this analysis mode.

CSV template

Download a small template showing the required columns.

Priority definition

Units are ranked by crime density.

Unit PAI threshold

Default: PAI ≥ 10.

PPAI α: 0.50

0 = hit rate; 1 = PAI.

Evaluate an intervention

Use simple before and after crime counts to estimate change relative to a comparison area, or test whether a place-based response was accompanied by displacement or diffusion of benefits.

Upload evaluation CSV

Required rows: intervention + control, or intervention + study_area_total.

Built-in sample dataset

Loads counts into the fields below; you can then edit them.

CSV template

Templates use area_type,before,after.

Comparison type

Use an untreated area that is broadly comparable with the intervention area.

The before and after periods should be the same duration. A simple two-period comparison does not test the parallel-trends assumption.

Crime counts

Intervention before

Intervention after

Control before

Control after

Important interpretation note. This is a simple two-period Difference-in-Differences-style comparison. A more credible causal interpretation requires a comparison area that would plausibly have followed the same trend as the intervention area without the intervention. With only one before and one after period, the tool cannot test that parallel-trends assumption. Results may also reflect regression to the mean, wider policy changes, seasonality, reporting changes or other contemporaneous events.

Results summary

Intervention scenarios

These are simple arithmetic scenarios, not causal forecasts. They assume the stated reduction occurs in the selected priority places and no displacement or diffusion elsewhere.

Reduction in selected places	Estimated crimes prevented	Study-area reduction	Remaining study-area crime

Crime concentration curve

Priority-place stability

The same priority rule is also applied independently to T2 for descriptive comparison. T2 selection is not used to calculate the predictive hit rate.

Ranked units

Showing the first 50 ranked units. The download contains all rows and calculated fields.

Evaluation summary

Before and after breakdown

Area	Before	After	Change	% change

Before and after crime counts

How the measures are used

Hit / capture rate

Hit rate = crime in selected places ÷ total crime

In the prediction workflow, this is the proportion of T2 crime occurring in places selected from T1.

Predictive Accuracy Index (PAI)

PAI = (n / N) ÷ (a / A)

n is crime in the selected area, N is all crime, a is selected area and A is total study area. PAI = 1 represents study-area-average crime density; PAI = 10 represents ten times that average density.

Penalized Predictive Accuracy Index (PPAI)

PPAI = (n / N) ÷ (a / A)^α, 0 ≤ α ≤ 1

α controls how strongly spatial coverage is weighted. At α = 0, PPAI equals hit rate. At α = 1, it equals PAI.

Recapture Rate Index (RRI)

RRI = PAI_T2 ÷ PAI_T1

For the same T1-selected places, RRI compares their relative crime concentration in T2 with T1. A value of 1 means the relative concentration is unchanged; above 1 means it increased; below 1 means it decreased.

Controlled before and after

Absolute DiD = (T_after − T_before) − (C_after − C_before)

Relative effect = (T_after / T_before) ÷ (C_after / C_before)

The relative effect is generally more interpretable when treatment and comparison areas have different crime volumes. A value below 1 means the intervention area improved relative to the comparison trend. The tool also estimates the T1-equivalent expected after count by applying the control proportional change to the intervention baseline.

Important interpretation note: controlled before and after

Treat this as a simple Difference-in-Differences-style diagnostic rather than automatic causal proof. The key assumption is that, without the intervention, the intervention and comparison areas would have followed similar trends. One before and one after period cannot test that assumption. Prefer a deliberately selected untreated comparison with similar pre-intervention crime levels and trends; using the remainder of the study area is a weaker non-equivalent comparison.

Weighted Displacement Difference (WDD)

WDD = (ΔT − ΔC_t) + (ΔD − ΔC_d)

WDD combines change in the intervention area with change in its displacement zone, each relative to a matched control. Negative values indicate net crime reduction; a positive displacement component suggests possible spatial displacement, while a negative component suggests diffusion of benefits.

Important interpretation note: displacement and diffusion

Define the potential displacement zone before examining the outcome where possible. A larger or differently shaped zone can materially change the result. The matched control and its displacement zone should represent credible untreated comparisons. WDD separates treatment-area and surrounding-area changes, but it does not rule out other causes of change or identify movement beyond the chosen displacement boundary.

Important interpretation note: PAI

Classical PAI evaluates a selected hotspot area. In the prioritisation workflow this tool also calculates a unit concentration PAI for each individual spatial unit: its crime share divided by its area share. This is useful for ranking and for a rule such as PAI ≥ 10, but it should be interpreted as relative crime concentration rather than evidence of future predictive accuracy. The selected-area summary reports the standard set-level PAI and PPAI.

References

Chainey, Tompson & Uhlig (2008), Predictive Accuracy Index.
Penalized Predictive Accuracy Index (PPAI).
PAI and Recapture Rate Index comparison.
Additional discussion of prediction evaluation measures.
Wheeler & Ratcliffe (2018), Weighted Displacement Difference.
CRIME De-Coder WDD Tool.
ArcGIS Solutions WDD implementation and interpretation guidance.