Uncertainty and Learning in Stochastic Macro Models

Article excerpt

Abstract Limits on information have deep economic impact and affect the conduct of economic policy. In the present paper we explore the effect of substantive uncertainty in a macro model, from both an analytical and methodological point of view. Agents are boundedly rational and make their forecasts according to different techniques and try to learn the values of the various parameters. In this context, a Markov regime switching rule, a VAR system, and recursive least square are considered and compared. As a result, we obtain a model which is mostly keynesian in nature that can be compared with the new neoclassical synthesis models. Simulations are carried out and show the possible appearence of endogenous and persistent fluctuations.

Keywords Endogenous fluctuations * Micro-macro relationships * Uncertainty * Bounded rationality * Learning


According to Leijonhufvud (1968), a macroeconomic model can be considered according to three methodological criteria: (1) the nature of aggregation; (2) the transaction process; and (3) the dynamic properties. This logical repartition is still useful and can be tested, for instance, by referring to Real Business Cycle (RBC) models. Here, the dynamics are driven by exogenous shocks, transactions are in real terms and aggregation is solved by the representative agent device, inside the new neoclassical synthesis, one substantially changes only the second aspect, adding monetary frictions while leaving unaltered the other properties.

Although useful, this Standard of evaluation has become insufficient in time. In fact, two further issues have become relevant--microfoundation and the amount of information that agents are supposed to be equipped with. Limited information is essential in the processes of expectation's formation and learning. As a result, with respect to these two additional criteria, both RBC and the new neoclassical synthesis models appear to be strictly microfounded and grounded on rational expectations, as an implication of the assumption of symmetric information and risk. The alternative paradigms facing a virtually countless set of possible imperfections build a bridge towards a more complex theoretical framework at the cost of being qualified as "ad hockeries."

In the present paper, we focus on the role of uncertainty as a neglected endogenous component in the dynamic process depicted by neoclassical mainstream paradigms. In the new neoclassical synthesis models the system is always in equilibrium, while the dynamics are driven by exogenous shocks. Our thesis is that the introduction of uncertainty may change the nature of the relationships considered so far. In fact uncertainty is: (1) compatible with a situation of disequilibrium that may be persistent; (2) it can justify why in a monetary economy of production (according to Keynes (1936) definition) money is an essential feature; and (3) it implies heterogeneity and therefore it is misleading to refer to a representative agent technique of aggregation.

We focus on substantive uncertainty in order to show that even such a narrower definition of uncertainty is compatible with a Keynesian structure of the model, capable to generate endogenous dynamics grounded in the presence of non-rational expectations. More specifically, agents do not know the model; hence, they are assumed to be bounded rational forecasters. While agents do not possess all the information required by the assumption of rational expectations, the process of learning is quite sophisticated. People behave like econometricians and try to learn the values of the parameters by running regressions. Furthermore, expectations are restricted to be consistent with outcomes.

This approach innovates the current literature in three ways. First, we employ a nonlinear approach, as in Flaschel et al. (2001), rather than a log-linearized model. One advantage of this approach is that the model can have a greater variety of attractors (periodic or strange) as explained by Benhabib et al. …