Essays on Econometrics and Macro-Finance

Thumbnail Image


Publication or External Link





This dissertation consists of three chapters on empirical macro-finance and the associated econometric methods.

In the first chapter, I develop a semiparametric single-index method for estimating multivariate jump-diffusion processes to model federal funds futures. I find that high-frequency changes in federal funds futures around FOMC announcements, which are the predominant measure of monetary policy shocks in the asset pricing and macroeconomic literature, are strongly forecastable by the estimated models, suggesting that they are not truly exogeneous. In contrast, the unexpected changes in federal funds futures on FOMC announcement days constructed from the semiparametric method are unforecastable by construction, and are strongly correlated with, but not the same as such high-frequency changes around FOMC announcements, suggesting that they are a better measure of monetary shocks.

In the second chapter, I study the predictability of bond yields. I find that federal funds futures, a proxy of monetary shocks, exhibit strong forecasting power on bond yields conditional on information contained in the cross section of the yield curve. Such additional return-forecasting information is effectively summarized by a single factor, and is not captured by unspanned macro factors. By focusing on the return-forecastability of trading strategies that take opposite positions at two different tenors by equal amount and unwind these positions one-day later, I bypass common econometric issues arising from the overlapping nature of bond excess returns.

In the third chapter, I study macro factors in the risk premia of G10 currencies. Motivated by the finding from a structural model with minimalistic assumptions that the predictability of currency risk premium arises from the differences in the market prices of risks between the home and foreign countries, I tackle this problem by identifying return-forecasting macro factors for the G10 currencies. Based on dynamic factor analysis on a large panel of macro variables, it is found that common macro factors possess strong forecasting power on the risk premia of G10 currencies, especially at longer maturities. The single most important factor loads heavily on activities in the US housing market and bond yields, which exhibits uniform and nonlinear forecasting power across all currencies and at a variety of maturities. The strong in-sample forecasting power preserves out-of-sample.