A0520
Title: Econometric forecasting of climate change
Authors: David Hendry - University of Oxford (United Kingdom) [presenting]
Abstract: Extreme weather events around the world show that Earths climate is changing rapidly due to human greenhouse gas emissions. The long-term pattern of weather is determined by human behaviour interacting with the physical properties of Earths climate system. To understand climate change and prepare for the future, for both mitigation and adaptation, requires accurate climate forecasts. Human behaviour is non-stationary from both stochastic trends and location shifts, making the interaction of humanity and the climate also non-stationary from distributional shifts, resulting in forecasts that are uncertain and prone to failure. We discuss how climate scientists and climate econometricians produce forecasts, comparing scenario projections of the former based on varying initial conditions with conditional forecasts of econometricians capturing model uncertainty. As climate change is characterised by changes in the changes, the success of all forecasts hinges on ex post handling of unanticipated shifts. Whether those shifts were unanticipated or not, they later become in-sample, so empirical modelling must take account of them to avoid distortions in parameter estimates and forecasts. Modelling location shifts improves the verisimilitude of models and forecasts, and indicator saturation estimators can do so as demonstrated for a system of four key climate variables, atmospheric CO2, global mean surface temperature deviations, ocean heat content deviations and global sea-level rise.
