This paper analyses, through a Random Utility Model (RUM), how a public institution selects among competitive agricultural R&D projects on the basis of observable distinctive features. In particular, we aim at verifying if, which and how other criteria, beyond the pure scientific value, are decisive for selection. From such information, like cost, duration, etc., the institution must infer about the unobservable actual ability, effort and reliability of the scientists themselves. Such analytical framework is empirically applied to a real case, the agricultural R&D activity funded by the Emilia-Romagna Region (Italy) between 2001 and 2006.