Monetary policy rules and welfare in dynamic stochastic models

This dissertation studies the role of monetary policy rules in a closed economy by using an accurate method to compute welfare measures to compare different policy configurations. The modelling framework is represented by a large scale Dynamic Stochastic General Equilibrium model with nominal and real rigidities. The new welfare evaluation approach is based on an accurate second order solution of the full model. Welfare measures based on a second order solution produce more robust results, free of spurious welfare reversal phenomena occurring with first order methods. Two models are presented, solved and discussed. The first model includes price stickiness modelled via quadratic adjustment costs. A variant of this model is represented by the inclusion of distortionary taxation on labor income. A second model includes capital accumulation and a real rigidity inserted via cost of capital adjustment together with a nominal wage quadratic adjustment cost. In both models, several monetary rules with parameters suggested by the empirical literature are compared in terms of welfare. In the model without capital accumulation, a Taylor rule with a small parameter of interest rate smoothing turns out to be optimal, with the inflation targeting coefficient close to the empirical estimates, but with an output targeting coefficient lower. In the model with capital accumulation, the inclusion of a term describing a measure of the inflation coming from nominal wage turns out to be welfare improving with respect to a situation where only the inflation (together with output and a lagged value for the interest rate) is the price measure to be targeted. The optimality is reached when the coefficients on the inflation rate and the measure of wage inflation are set to be the same and when both targets are included in expectation in the monetary policy reaction function. Both models show to have a good fit, if compared with second order moments of data from US economy. For both models, impulse response functions based on second order accurate solution show to have a non-explosive behavior for T-ahead periods of forecasted horizon.