Can Rose’s paradox be useful in crime prevention?

Geoffrey Rose’s prevention paradox obtains when the majority of cases with an adverse outcome come from a population of low or moderate risk, and only a few from a minority ‘high risk’ group. Preventive treatment is then better targeted widely than on the ‘high risk’ minority. This study tests whether the prevention paradox applies to the initiation of criminal behaviour, as recorded in longitudinal administrative data from Denmark. Children born in 1984 are followed from birth to early adulthood. A discrete-time Cox model allows for changing covariates over time. The initiation of criminal behaviour is proxied as getting a police record over ages 15-22 on a criminal matter. This outcome was predicted, more accurately than by chance, by a combination of over twenty risk factors, reflecting the major crime reduction paradigms. However, it seems impossible to identify a minor group (<5%) in the population from whom criminals are exclusively recruited. Our example illustrates how the applicability of Rose’s prevention strategy, population based, rather than targeted, depends on how narrowly’ high-risk group’ is defined, for a given distribution of estimated risk, and allows for the possible complementarity of population and targeted measures.