...
On this view then equilibrium dynamics is possible only when learning has ceased and this in turn must mean that no systematic forecasting errors occur so that one is speaking of Rational Expectations Equilibrium. I have elsewhere argued that such an equilibrium is of interest to theorists for various reasons and I do not much care €or dismissing it on bluff “common sense grounds”. But there is a serious flaw and it lends support for the approach I am advocating today.
This flaw is that in general there are far too many candidates for a Rational Expectations Equilibrium. The literature on this is now very large and I shall have to be somewhat brief and dogmatic. The situation is that in almost all cases which have been studied there is a continuum of such equilibria. That is for any one such equilibrium there is another one arbitrarily close to it in the appropriate sense of close. Put more vividly, there is a continuum of theories that agents can hold and act on without ever encountering events which lead them to change their theories. We can have Rational Expectations Equilibrium cycles and indeed Rational Expectations Equilibrium “chaos” in the in the technical sense of that term.
...
Hahn F.H., 1987. “Information, Dynamics and Equilibrium”. Scottish Journal of Political Economy, 34: 321–334, 324.