| Abstract: | It is argued that the appropriate criteria for evaluating large-scale long-range societal (or “world”) models differ significantly from the criteria that are appropriate for short-range econometric models or for small “toy” models. World models are designed to elucidate major societal problems arising from the interaction of many loosely connected factors—demographic, economic, technological, and environmental—that can be safely neglected in short-range extrapolative models. World models are also intended to explore the implications of long-range policy alternatives. Consequently, it is important that world models be phenomenological, rather than statistical, and they must “capture” the essence of some inherently nonlinear phenomena. This implies achieving qualitative realism without imposing an unwarranted assumption of causal determinism. Unfortunately, the mathematical characteristics of large nonlinear systems may essentially preclude meaningful long-run forecasts. |
