Essays on Commodity Price Shocks, Bank Risk and Market Volatility Forecasting

Thumbnail Image
Qiu, Yue
Commodity Price Shocks , Bank Risk , Macro- finance Linkages , Factor Models , VARs , Volatility Forecasting , Heterogeneous Autoregression , OLS Post-Lasso , OLHAR , Model Selection , Tail Risk , Quantile projections , Density forecasts
My dissertation has three main chapters. Chapter 2 develops a structural dynamic factor model that estimates the effects of commodity price shocks on the Canadian macroeconomy, bank lending and bank risk. Unlike most literature treating commodity price changes as exogenous, I identify global structural shocks driving real commodity prices and find that global demand and commodity market–specific shocks are crucial. These two shocks are shown to have quite different implications for bank lending and risk, which confirms the importance of disentangling shocks driving commodity price changes. I also discover that on average, the total loan growth and the noninterest income ratio of large Canadian banks are more responsive to the above two commodity price shocks than those of small Canadian banks. Chapter 3, coauthored with Tian Xie, studies how to better forecast the daily market volatility (VIX) index. We propose utilizing the ordinary least square post–least absolute shrinkage and selection operator (OLS post–Lasso) from Belloni and Chernozhukov (2013) to select the predictors and estimate the coefficients for a heterogeneous autoregressive (HAR) model (Corsi, 2009). In an out–of–sample analysis with the VIX data, our proposed OLS post–Lasso HAR (OLHAR) model generates a different combination of predictors from those ex ante imposed by the standard HAR model. Moreover, the OLHAR model shows its dominance over the standard HAR and other competitor models at forecasting the VIX either weekly, biweekly, or monthly ahead. Chapter 4 develops a tail risk forecasting system tailored to Canada and provides a set of one–quarter ahead forecasts of tail real and financial risk by factor–based conditional quantile projections. My model forecasts the steepest decline in GDP–Oil at 2009Q2, and the peaks of the nonperforming loan (NPL) ratios of Bank of Montreal (BMO) and Canadian Imperial Bank of Commerce (CIBC) at 2010Q2 and 2010Q3, respectively. For BMO and CIBC, the banking factors contribute the most to the peaks of the Value at Risk (VaR) forecasts of the NPL ratios. The commodity price factors also compose a prominent portion of the total contribution from the global factors.
External DOI