Forecasting leveraged loan market volatility using GARCH models

This paper compares different GARCH models in terms of their out-of-sample predictive ability of leveraged loan market volatility. The study investigates whether the asymmetric effects of good and bad news on volatility is present and how distributional assumptions affect the selection of GARCH models. Compared to two widely used historical volatility models, the simple moving average and the exponentially weighted moving average, the results suggest that asymmetric GARCH models have marginally better out-of-sample predictive ability. In addition, this study finds that fixed income market volatilities improve the forecasts of loan market volatility. The model comparison involves a regression-based approach, loss functions and statistical tests