In this paper we aim at clustering Italian contracting authorities in terms of their attitude over time (2015-2017) in managing public procurements, a field particularly prone to the occurrence of corrupt acts. For our purpose we rely on an approach based on red flag indicators, which considers public procurement data and points out possible anomalies in order to alert the system to the possible risk of corruption. As such, this approach allows us to perform the analysis at every moment of the procedure, from the call for tender until the final realisation of the work. By exploiting the richness of information contained in the Italian Banca Data Nazionale dei Contratti Pubblici, a number of red flag indicators proposed by the international literature are computed for the three-year period. By means of a la- tent Markov model for multivariate continuous responses, we aim at: i. identifying clusters of contracting bodies and ii. quantifying the probability for a contracting authority belonging to a certain cluster to move to a different cluster (or to persist in the same cluster) over time. First results show that several clusters of administrations may be highlighted. Among them, one profile draws attention, as it includes administrations with extreme values for all the red flags.
