In this paper, we examine whether the transparency of the central bank's preferences is desirable. We make two major points. First, in the literature on preference transparency variance-reduction frameworks are often adopted. As a consequence a change in the degree of transparency affects the magnitude of information asymmetries, but at the same time it implies a rather arbitrary effect on the distribution of preferences. We present a clean framework without this problem. Second, using a very general specification of shocks to the central bank's preferences, we show that society prefers transparency if it sufficiently values the employment target, whereas it prefers opacity if it estimates inflation as sufficiently important.