Variance and skewness in density forecasts: assessing world GDP growth

The article explores the critical role of predictive models in policymaking and investment, emphasizing the need for reliable density forecasts that account for both upside and downside risks. It introduces the Bayesian cross-entropy forecasting (BCEF) procedure, a method designed to generate density forecasts by solving an information-theoretical problem under noisy and scattered information. The BCEF procedure combines ex-post and ex-ante knowledge of the variable of interest and associated risk factors, providing a robust framework for generating density forecasts. The empirical application focuses on world GDP growth, using selected risk factors such as term spreads, inflation deviations, energy prices, and the S&P 500 index. The BCEF method is highlighted for its ability to incorporate prior information, recover information under incomplete data, and assess the uncertainty of point forecasts. The article also discusses the decomposition of variance and skewness in density forecasts, offering insights into the contributions of selected risk factors to the predictive distribution. The BCEF method is evaluated against standard symmetric and calibrated asymmetric density forecasts, demonstrating its superior performance in capturing uncertainty around world GDP growth. The findings underscore the reliability and precision of the BCEF approach, justifying its value as a superior tool for economic forecasting.