UC Irvine

Based on 2019 procurement data composed of almost 1.5 million observations, this study employs Latent Class Analysis (LCA) to investigate how latent profiles are defined to represent behavior in the Price Registration minutes based on a set of selected variables. LCA approach brought the possibility of identifying some characteristics that can suggest or direct the efforts of the control bodies in identifying illegal cases. From the seventeen-profile model with binary variables, we identified a latent class (N=2589) that could be the object of a more detailed analysis by the control bodies. From the LCA with binary and continuous variables, the seven-class solution provided the optimal fit for our data, with the lowest BIC value and a high entropy value. The seven relevant factors that increase the risks of renting price registration drafts are the: (i) number of Non-participant Bodies (NPB) per tender; (ii) relative item value; (iii) amount of NPB per supplier per tender; (iv) distribution of winners; (v) proportion of approved quantities per item for NPB per tender; and (vi) tender with only Managing Body (MB) and NPB. The latent class with a high NPB presence with few suppliers (N=749) was the most salient to be more favorable for the sale of price registration minutes, intended to be carried out in future studies. Our results highlight that such an organization-centered approach can provide insights not apparent in the more common variable-centered practice. Although our goal was to unpack further insights within Price Registration System, future work may wish to include more of the measures from the models in the clustering routine.