The objective of this study is to contribute to the growing literature on cargo theft by empirically testing four specific hypotheses of its causes – the space-time dynamics hypothesis, the economic attractiveness hypothesis, the social structure hypothesis, and the deterrence hypothesis. This study investigates the case of the economic core of one of the most severe regions regarding cargo theft worldwide – São Paulo state. As novelty in crime studies, we estimate Autorrregressive Distributed Lag models (ARDL). We found that the number of cargo thefts of a geographic area can be predicted by itself and that of neigbouring areas. This is an unprescendeted empirical evidence that cargo theft time series are autoregressive and cointegrated. Regarding economic attractiveness and social structure, the results are inconclusive. However, police activity reduces cargo theft in the large metropolitan area and inland municipalities of São Paulo state.