This thesis consists of three essays in empirical finance and macroeconomics. The first essay proposes a new structural-break vector autoregressive model for predicting real output growth by the nominal yield curve. The model allows for the possibility of both in-sample and out-of-sample breaks in parameter values and uses information in historical regimes to make inference on out-of-sample breaks. A Bayesian estimation and forecasting procedure is developed which accounts for the uncertainty of both structural breaks and model parameters. I discuss dynamic consistency when forecasting recursively and provide a solution. Applied to monthly US data, I find strong evidence of breaks in the predictive relation between the yield curve and output growth. Incorporating the possibility of structural breaks improves out-of-sample forecasts of output growth. The second essay proposes a sequential Monte Carlo method for estimating GARCH models subject to an unknown number of structural breaks. We use particle filtering techniques that allow for fast and efficient updates of posterior quantities and forecasts in real-time. The method conveniently deals with the path dependence problem that arises in these type of models. The performance of the method is shown to work well using simulated data. Applied to daily NASDAQ returns, we find strong evidence of structural breaks in the long-run variance of returns. Models with flexible return distributions such as t-innovations or with jumps indicate fewer breaks than models with normal return innovations and are favored by the data. The third essay proposes a new tilt stochastic volatility model which extends the existing volatility models by modeling the asymmetric correlation between return and volatility innovations in a unified and flexible framework. The Efficient Importance Sampling (EIS) procedure is adapted to estimate the model. Simulation studies show that the Maximum Likelihood (ML)-EIS estimation of the model is accurate. The new model is applied to the CRSP daily returns. I find the extensions are significant and incorporating them improves the accuracy of volatility estimates.

This thesis consists of three essays in empirical finance and macroeconomics. The first essay proposes a new structural-break vector autoregressive model for predicting real output growth by the nominal yield curve. The model allows for the possibility of both in-sample and out-of-sample breaks in parameter values and uses information in historical regimes to make inference on out-of-sample breaks. A Bayesian estimation and forecasting procedure is developed which accounts for the uncertainty of both structural breaks and model parameters. I discuss dynamic consistency when forecasting recursively and provide a solution. Applied to monthly US data, I find strong evidence of breaks in the predictive relation between the yield curve and output growth. Incorporating the possibility of structural breaks improves out-of-sample forecasts of output growth. The second essay proposes a sequential Monte Carlo method for estimating GARCH models subject to an unknown number of structural breaks. We use particle filtering techniques that allow for fast and efficient updates of posterior quantities and forecasts in real-time. The method conveniently deals with the path dependence problem that arises in these type of models. The performance of the method is shown to work well using simulated data. Applied to daily NASDAQ returns, we find strong evidence of structural breaks in the long-run variance of returns. Models with flexible return distributions such as t-innovations or with jumps indicate fewer breaks than models with normal return innovations and are favored by the data. The third essay proposes a new tilt stochastic volatility model which extends the existing volatility models by modeling the asymmetric correlation between return and volatility innovations in a unified and flexible framework. The Efficient Importance Sampling (EIS) procedure is adapted to estimate the model. Simulation studies show that the Maximum Likelihood (ML)-EIS estimation of the model is accurate. The new model is applied to the CRSP daily returns. I find the extensions are significant and incorporating them improves the accuracy of volatility estimates.