The article presents a survey of theory concerning the "credibility" and "reputation" issues until the beginning of the 1990s, when new concept of "central bank independence" emerged. The "credibility problem" described by Kydland and Prescott means, that optimal policies become sub-optimal with the passage of time, which is called "time inconsistency". When the government or monetary authority change formerly optimal policy rules, the "forward - looking" private sector can anticipate it so that time inconsistent policies lack credibility. This problem can be analysed in terms of a complete-information game between the policy-maker and the private sector. According to Barro and Gordon the policy-maker's concern for his/her reputation for precommitment can sustain the ex ante optimal policy. Canzoneri introduces imperfect monitoring of monetary policy, using a model with stochastic disturbances, which gives almost the same conclusions. Studies by Backus and Driffill, Barro, Vickers or Cukierman and Meltzer represent more or less departure from the extreme assumption of complete-information, and solve finite-time-horizon games. The main result is that uncertainity of the private sector regarding the behaviour of government or monetary authority can be sufficient to sustain the optimal policy for much of the finite time for which the game is played.
