Time-inconsistency, Democracy and Optimal Contingent Rules

Following Kydland and Prescott's (1977) seminal paper on time-inconsistency, a large literature has explored possible frameworks within which monetary policy could overcome this problem -- neatly illustrated in Barro and Gordon's (1983) model. In a stochastic world there appears to be a trade-off between the necessary `tying of hands' to conquer the effects of time-inconsistency and the desirability of flexible response. It is in principle possible to achieve an optimal outcome by use of a discriminatory punishment, however, with a large punishment (sufficient to deter) for using policy to exploit the Phillips curve to reduce unemployment below the natural rate, but no punishment for contingent response to shocks using the same Phillips curve. This paper sets out a model of democratic elections in which floating voters may find it optimal to follow this strategy. The significance of this possibility is that regimes which permit contingent macroeconomic policy responses, while enabling prior targets to be set and policed, are superior to those which do not. This has relevance to the debate over the Exchange Rate Mechanism and the European Monetary Union.