Long-term forecasting: Meditations of two pitfall collectors

Summary

This article discusses the pitfalls of long-term forecasting. One of the most important is that any model is in fact a subsystem cut from a general societal system. This implies that the distinction between endogenous and exogenous variables is unclear and thus the (sub) model is often unreliable as a tool for forecasting. A second pitfall is that the structure of a model reflects the structure of society. Changing societal structures requires models with changing structures and coefficients. Specific pitfalls are the instability of the functional form of relations, uncertainty about functional forms, variability of parameters. Some attention is paid to the decrease in the risk of wrong forecasts.