Virtually all of life is a dynamic decision problem. It is therefore appropriate that economics has sophisticated models of how people should take decisions in a dynamic context. In the spirit of modern economics, the word .should' is usually interpreted normatively, and the models inevitably prescribe some restrictions on behaviour implied by certain rationality norms. In the dynamic context, these norms relate to dynamic consistency in one form or another, and it is to these that this research project is directed. We want to investigate the empirical plausibility of these norms and the theoretical and policy implications of their violation.

A dynamic decision problem is an inter-woven sequence of decisions made by the decision maker and moves by some other agent - either Nature in decision problems under risk or uncertainty or some other human agent in game decision problems (or, indeed, both in risky games). As the decision maker proceeds through the decision problem, that is, as time unfolds, the decision maker has to evaluate the payoffs at different points in time and to consider the possible moves by the other agent. In the light of these evaluations the decision maker proceeds to (sequential) decisions. Consistency considerations come into play when the decision maker compares decisions that he or she would make as viewed from different points of time. Economics imposes strong requirements: in terms of relatively evaluating payoffs received at different points in time, the usual consistency requirement is that the same relative weight is given irrespective of the point in time from which the valuation is being made; in terms of processing the risk, the usual consistency requirement is that conditional decisions at particular points in time do not depend upon the point in time from which the decision is being considered. In more familiar words, the first requirement is that discounting is exponential; the second is that people not only plan but that they implement their plans.

Much empirical evidence (both experimental and non-experimental) suggests that these strong consistency requirements are often not met. This evidence raises important issues and it is to the evidence itself, and the issues that it raises, that this project is devoted. We want to look at the key issues: is discounting exponential? do people plan? If they do not, then the whole issue of dynamic inconsistency arises - with further questions: do people understand that they are inconsistent? what do they do about it? We plan to investigate these issues at a variety of levels and using a variety of tools. At one extreme, one local unit will examine a popular television game, Affari Tuoi, to see if contestants display a consistency in their playing of the game. At another extreme, another local unit will use brain scanning technology to see what happens inside the brains of decision makers as they take dynamic decisions. In between, we plan to look at the problem of dynamic inconsistency in a variety of contexts, ranging from simple individual decision-problem experiments, through simple market experiments and games where different people play the different selves of dynamically inconsistent people, to very specific experiments in which the .frame' of the decision problem is deliberately manipulated to test certain issues concerned with perceptions of payoffs at different points in time and with beliefs about future moves by the other agent.

Our starting point is that violations of these strong consistency requirements are commonplace, and hence that present economic theory is deficient. Our plan is to try and identify in what circumstances these requirements are violated, and in what way. The ultimate objective is then to try and reconstruct the damaged parts of economic theory in the light of our findings. This reconstruction is not only important for economic theory itself but also for the prescription of policy measures based upon the theory.