Essays in Applied Time Series Econometrics

Abstract

In the first chapter of this thesis, I estimate Markov-switching models with time-varying transition probabilities to predict the US business cycle regimes. In particular, I evaluate the predictive power of real and financial indicators and find that the slope of the yield curve turns out to be the most reliable indicator for regime predictions. This first chapter paves the way for the next two chapters of this thesis that also use models with Markov-switching for analysing the business cycle. The second chapter (a joint work with Massimiliano Marcellino) combines the Markovswitching model with the MIxed DAta Sampling (MIDAS) model. This new model uses information from variables sampled at different frequencies. We first show in a Monte-Carlo experiment that our estimation method yields accurate estimates. We then apply this new model to the prediction of both the business cycle regimes and GDP growth for the US and the UK. We find that the use of high frequency information and parameter switching performs better than using each of these two features separately. In the third chapter (a joint work with Laurent Maurin and Matthias Mohr), we estimate nine different models of the output gap (univariate, multivariate, linear and non-linear) and compute model-averaged estimates of the output gap. We find some evidence for changes in the slope of the trend of the Euro area output for few periods in 1974 and 2009. Moreover, our model-averages measures of the output gap reduce the uncertainty associated with the output gap estimates and soften the impact of data revisions. We then evaluate the forecasting performance of our output gap estimates for inflation and find that the output gap estimates improve on the forecasting performance of standard AR benchmarks for inflation although the inflation forecasts based on the output gap estimates exhibit a poor forecasting performance since 2008. The last chapter of this thesis (a joint work with Eric Ghysels and Massimiliano Marcellino) is an empirical evaluation of the risk-return relation. We use a MIDAS estimator of the conditional variance and model regime changes in the parameter entering before the conditional variance. We find evidence for a reversed risk-return relation in periods of high volatility, while we uncover the traditional positive risk-return relation in periods of low volatility. In particular, the high volatility regime is interpreted as a flight-to-quality regime. This finding is robust to a large range of specifications.