The Scientific Model of Causality

Compares the scientific model of causality developed in economics to methods used in other social sciences, epidemiology, & statistics. It is maintained that the existing literature on causal inference in statistics confuses the distinct tasks of defining counterfactuals & identifying causal models from population distributions & from actual data. Questions of causal inference related to policy evaluation/forecasting problems are discussed to provide a context for comparing alternative approaches. Both individual-level & population-level causal effects are defined & structural econometric models are used to define causality & analyze objective outcomes & subjective evaluations. The identification problem is addressed & the scientific model is applied to the identification of four widely used estimators for causal inference. It is concluded that models of causality used in statistics fail to specify either the mechanisms of external variation that are central to the definition of causality or the sources of randomness producing outcomes. The scientific model approach differentiates between the derivation of a model as an abstract theoretical activity & the problem of identifying models from data. Tables, Figures, Appendixes, References. J. Lindroth