In this paper we verify the degree of reliability of brokerage analysts’ recommendations, with reference to Italian IPOs and measure their long-term performance, distinguishing among affiliated and non-affiliated analysts, to test the conflict of interests hypothesis against an alternative ‘superior information hypothesis’. The empirical evidence shows that IPOs recommended by affiliated analysts have a long-run performance that is worse than firms recommended by unaffiliated ones by a relevant amout. This result supports the conflict of interest hypothesis, while it seems to be inconsistent with the hypothesis that underwriter analysts have superior information