Abstract
This paper extends the causal modelling technique described in an earlier article (Anderson and Evans, 1974) to nonrecursive causal models that involve feedback and/or reciprocal causation. The problem of identification is discussed and a rule provided that can be used to determine whether or not a unique set of parameter estimates can be found for each equation that makes up the model. Three different procedures are described for estimating the parameters of these equations, namely, ordinary least squares, indirect least squares, and two-stage least squares. A formula is provided for the derivation of the reduced form of the model. The reduced form provides information concerning the total effect of exogenous variables on endogenous variables in the model. Data from an empirical study have been used to illustrate the causal modelling technique that is described.
Get full access to this article
View all access options for this article.