STATISTICAL LEARNING WITH TIME-VARYING PARAMETERS

Bruce McGough
2003 Macroeconomic Dynamics  
In their landmark paper [1], Bray and Savin note that the constant parameters model used by their agents to form expectations is misspecified and that, using standard econometric techniques, agents may be able to determine the time-varying nature of the model's parameters. Here, we consider the same type of model as employed by Bray and Savin except that our agents form expectations using a perceived model with parameters which vary with time. We assume agents use the Kalman filter to form
more » ... filter to form estimates of these time-varying parameters. We find that, under certain restrictions on the structure of the stochastic process and on the value of the stability parameter, the model will converge to its rational expectations equilibrium. Further, the restrictions on the stability parameter required for convergence are identical to those found by Bray and Savin. * I would like to thank George Evans for many valuable discussions. All errors are mine.
doi:10.1017/s1365100502010325 fatcat:vh533tpscbdwll3nciokk46oqq