Swiss monetary policy

This thesis considers various aspects of Swiss monetary policy. In the first part of my thesis, I tried to generate a liquidity effect in a DSGE model. The problem with DSGE models is that in response to an expansionary monetary policy, i. e. an increase in the growth rate of money, the interest rate increases due to the anticipated inflation effect. Hence, there is no liquidity effect which would indicate a negative relation between money growth and the nominal rate of interest. In order to create such a liquidity effect, I apply a new-Keynesian model developed by Altig, Christiano, Eichenbaum and Linde (2005) to Switzerland. In addition to the Keynesian elements, i. e. sticky wages and prices, the model incorporates three non-standard features: namely, habit formation in consumption, adjustment costs in investment, and firms’ ability to choose the utilization rate of capital. The model includes two technology and a monetary policy shock. The impulse responses from the model are then compared to the impulse responses from a structural VAR which is to capture the empirical evidence. In the second part, I estimate a Taylor rule with Markov switching regimes for Switzerland. As Switzerland is an open economy, the exchange rate plays an important role. The idea then is that the SNB acts differently when the exchange rate deviates from its trend then when it does not. Inflation, the output and exchange rate gap are included as explanatory variables into the monetary policy rule. Furthermore, I account for interest rate smoothing. Finally, the coefficients are state-dependent in order to see if there really are two different monetary policy regimes followed by the SNB. The two states constitute a Markov chain. The estimation is done ones in a traditional frequentist framework and ones in a Bayesian framework. The former makes use of an EM-algorithm whereas the latter applies a Gibbs sampler. Finally, the last part of my thesis sets up a Gibbs sampler for the estimation of three states Markov switching models with non-constant transition probabilities. While Gibbs samplers for two, three or n-state Markov switching models with constant transition probabilities are widespread in the literature and pretty straightforward, things change when transition probabilities in Markov switching models with more than two states are time-varying. The transition from two to three states is accomplished with help of a multinomial probit model for the latent variable process. Afterwards, the Gibbs sampler is used to analyze the Swiss business cycle. More specifically, the analysis tries to answer the following two questions. First, is the Swiss business cycle subject to switches in regime? Second, is the cycle characterized by duration dependence, i. e. does the probability of a switch in states depend on how long this state already prevailed?