Causal Inference Analysis: A Seductive Process

A test of three causal network models, using the same seven variables in each case, is presented. The variables are taken from research on organization by Pugh et al. Since all three models are supported by the data, the author concludes that causal inference techniques are rarely capable of discriminating among several intuitively plausible networks. They may be able to reject unrealistic networks, but it is probable that these would have been rejected a priori. Although Aldrich's model of organizations is used as an example, it should not be thought this is an attack on his work anymore than any other causal inference approach.'