Objectives: Introduce a systematic method to identify areas with similar levels of disorder (from serene, to “tipping,” to crime-ridden) which is crucial for a valid empirical test of Broken Windows Theory (BWT).
Methods: Systematic social observation data are used of almost 2,000 locations in the city of Amsterdam, the Netherlands. Spatially constrained hierarchical agglomerative clustering is used to aggregate individual observation locations to form homogeneous areas. Davies–Bouldin index and intraclass correlation coefficient are used to objectively identify the optimum number of clusters.
Results: The newly identified areas differ from administrative neighborhoods as well as hot spots of disorder. The regionalization method provides a tentative solution to both the “zonation” and “aggregation” problems of the modifiable areal unit problem liable to affect empirical studies of BWT.
Conclusions: Hot spot analysis fails to identify areas with moderate levels of disorder, which impedes testing the basic precept of BWT. Our results may partly explain why the evidence on the effectiveness of order maintenance policing remains inconclusive. We suggest that randomized controlled trials of order maintenance policing should be performed on these new areas rather than in hot spots of disorder.