Does financial volatility help in explaining and predicting economic activity?

Abstract

Driven by the difficulty to predict the last financial crisis and possible distortion of predictive power of the conventional financial indicators on economic activity, this thesis provides in-sample and out-of-sample analyses whether financial volatility helps in explaining and forecasting economic activity. Several measures of financial volatility were constructed, such as: realized volatility, volatility following a Generalized Autoregressive Conditional Heteroskedasticity (GARCH) process, common long-run component of volatility estimated by Dynamic Factor Model, Principal Component Analysis and cyclical components of financial volatilities filtered out with Baxter-King filter. I find that statistically there are measures of financial volatility that help in explaining economic activity. Moreover, out-of-sample analysis suggests that the model with term-spread and volatility of financial volatility (volatility of value-weighted returns of market portfolio volatility) performs best in forecasting economic activity. The inclusion of a volatility measure reduces the noise in estimated probabilities of expansions and leads to the lowest number of uncertain periods, i.e. periods for which probability of recession is between 16.86% (percentage of recessions in the sample) and 50%, an event that in some studies is already considered as a recession. Thus, a certain financial volatility measure improves forecasts from the conventional financial indicators, especially during less volatile times. Moreover, the most parsimonious measure of volatility predicts business cycles best. On the other hand, industrial production growth seems to be barely affected by financial volatility measures, which tend to be a better predictor for the direction of the future path of the economy than the actual growth rate.