Essays on Financial Forecasting

Abstract

This dissertation investigates forecasting monetary policy, currency strategies, and equity returns. The three chapters are summarized below: 1.) From December 2008 to November 2015, the Federal Reserve’s fed funds target range was 0 to 25 basis points, at the effective zero lower bound for nominal interest rates. Although the fed funds target did not change for seven years, the underlying unobserved dynamics are not constant. Using a dynamic ordered probit model, I estimate the desired fed funds rate in the absence of a zero lower bound. Using real time data, the model has a good track record of predicting date of first rate hike. After the first rate hike, however, the model predicts a front-loaded normalization path. The actual pace of rate normalization was unusually slow in the first couple years of tightening, most likely because the Fed felt asymmetric risk around the low interest rates. In contrast, the model deliberately allows the fed funds rate to move seamlessly in and out of the zero lower bound, neglecting this asymmetric risk. 2.) Carry, momentum, and value are three established and profitable currency trading strategies. However, their performance varies through time such that a timing strategy could increase profits. I use a nonlinear model to combine these three strategies, where time-varying weights are allowed to vary depending on the degree of currency misevaluation. However, I find that even under a nonlinear model, currencies revert to fair value too slowly for a one-month investment horizon timing strategy. Because of this, a factor timing strategy using time-varying weights does not outperform equally weighting the three currency strategies over short horizons. 3.) Buy-and-hold strategies are profitable because the equity market rises on average. However, equity timing could improve risk-adjusted portfolio profits. I generate out-of-sample return forecasts from a range of methods: multivariate regressions, univariate regressions, the average of univariate forecasts, and an expanding window historical average. Then I use these forecasts in a long-short equity strategy. Combined indicators result in 11 percent annualized returns, a 0.73 information ratio, and a 0.59 Sharpe ratio over a 22-year out-of-sample backtest, which outperforms a buy-and-hold strategy.