A model for predicting burglaries with forced entry in Catalonia

At the University of Girona on 22 November 2021, Dr Pere Boqué Busquet, an officer in the Catalan Police Force (Mossos d’Esquadra), read his doctoral thesis, entitled Mathematical models for the prediction of burglaries with forced entry in Catalonia. Directed by Dr Marc Saez Zafra and Dr Laura Serra Saurina, the thesis analyses the current situation with regard to so-called “predictive policing” and, as its title indicates, proposes a mathematical model for predicting which areas of Catalonia are likeliest to suffer from this type of crime at any given time.

One of the first aspects that Dr Boqué highlights is that predictive models applied in other parts of the world (and particularly in the US) are not valid in Catalonia. Among other questions, the regional and urban configuration of the territory means that repeat victimization patterns (part of the basis for these models) cannot be transferred directly to the Catalan context. Nevertheless, by dividing the territory into cells 5km in length on each side, he succeeded in showing how burglaries with forced entry follow time patterns in the form of waves or streaks, and can therefore be predicted. If the police detected the starting point, they could therefore prevent the pattern of replications or repeat crimes from developing. This conclusion is reported in an article published in the European Journal of Criminology, “‘Surfing’ burglaries with forced entry in Catalonia: Large-scale testing of near repeat victimization theory”, by Pere Boqué, Laura Serra and Marc Saez (November 2020).

The other results of the thesis and the construction of the model are presented in two further articles. In the second article a Log-Gaussian Cox model is applied to explore the possibility of making predictions on a smaller scale, in cells with a length of 500m, or even 250m or 100m, on each side. The conclusion is reached that a small-scale repeated victimization pattern, although it can also be detected in Catalonia, is not fit for purpose for the modelling of the overall dynamics of the pattern of burglaries.   

The third article proposes overcoming this limitation through a “new space-time victimization pattern that extends the concept of near repetition to that of repetition in different areas that have broadly similar characteristics but may be further apart in geographical terms.” These “groupings of areas that are often victimized at the same time” are described as “constellations of burglaries” and form a fixed body of reduced dimensions that is stable in time and on the basis of which predictive models can be created.

These last two articles are still awaiting publication. For this reason, the date of publication for the doctoral thesis has not yet been decided.

Over and above the effectiveness of the mathematical model, and as Dr Boqué emphasised at the end of his reading of his thesis, the difficulty of “predictive policing” does not so much lie in the predictive aspect, i.e, in the possibility or otherwise of predicting crime, as in the “policing” aspect: which preventive actions can be carried out to reduce the probability that the predicted crimes will finally be committed. Only when this model is applied will it be possible to discover the real potential for having a preventive effect on crime rates. Be that as it may, it is clear that a knowledge of the dynamics of space-time patterns of criminality offers the police an advantage that needs to be put to good use.


Aquest apunt en català / Esta entrada en español / Post en français

Deixa un comentari