Essays in empirical macroeconomics

Macroeconomic policy makers need to understand the quantitative effects of their policy instruments in order to have any hope of systematically achieving their objectives. In particular, the answers to the following questions are of a critical importance for understanding the efficacy of fiscal and monetary policy interventions: 1) how much does government spending stimulate GDP? 2) how nominally rigid are prices and do these rigidities differ across sectors of the economy? In this dissertation I use state-of-the-art macroeconometric techniques to shed new light on the answers to these fundamental, but unresolved, economic questions. Firstly, I extend the Markov-switching vector autoregressive (MS-VAR) framework to allow for the estimation of models with multiple regime-switching processes governing distinct sets of model parameters. Secondly, using my extensions of the MS-VAR framework, I estimate a model to investigate the extent to which the fiscal multiplier varies over time and what factors cause the time variation. Lastly, I estimate a Dynamic Stochastic General Equilibrium (DSGE) model that has multiple production sectors to assess the extent of heterogeneity in nominal rigidities across sectors and the role these rigidities play in the transmission of monetary policy shocks to real economic variables. Throughout the dissertation I use Bayesian methods for statistical inference, which are based on the likelihood function implied by the macroeconomic model. I find that the government spending multiplier varies over time and, contrary to other recent results in the literature, it varies countercyclically. The results of the MS-VAR estimation suggests that ``expansionary fiscal policy may not expand GDP at the economy most badly needs an expansionary policy intervention and, furthermore, that current DSGE models may suffer from misspecification along dimensions important for evaluating the effects of fiscal policies. With regards to assessing nominal rigidities, I find that the choice of estimation method is key in evaluating the extent of nominal rigidities in multi-sector DSGE models. Likelihood-based estimation methods estimate lower values of nominal rigidities than do moment based approaches