Finance Cluster Seminar: Dr. Jing Tian
Accurate forecasts are an important source of information for economic and business decision making. Forecasts with long lead times are often more desirable than forecasts with short lead times, as long lead times allow decision makers more time to act on forecast information. Yet forecasts with long lead times often contain more uncertainty than forecasts with short lead times, and forecasts may be revised several times as the target date approaches. In this paper we develop and implement a state space approach to multi-horizon forecast evaluation. The approach allows us to decompose forecast error into measures of horizon specific forecast bias, rational forecast error variability, and implicit forecast error variability. We apply our model based forecast evaluation approach to a real-time dataset of temperature forecasts for Melbourne, Australia. Our results suggest that multi-horizon forecast error can have several unobserved components, and that the contribution of these error components may vary both along the forecast horizon, and between different sub-samples of the series being forecast.
Dr Jing Tian is a lecturer of economics at the Tasmanian School of Business and Economics at University of Tasmania. She received her PhD from the Australian National University in 2010. Her research has focused on econometric forecasting, applied macro econometrics and empirical finance, in particular forecasting under structural breaks, modelling and forecasting macroeconomic variables and real-time data analysis.