Specification Error and Negatively Correlated Disturbances in "Parallel" Simultaneous-Equation Models

Publisher Website
Google Scholar
Abstract
This article attempts to explain negatively correlated disturbances in "parallel" simultaneous-equation models in which each equation contains essentially the same variables as the others but is written for a different position of the social system under study. The two general methodological explanations presented are: random error in the measures of the endogenous variables and upward bias in the calculation of the (positively signed) reciprocal effects. We show how upward bias in the calculation of these effects can be produced by the following types of specification error: failure to control for random error in the measures of the exogenous variables, the omission of the "cross-effects " of the exogenous variables, the omission of correlated parallel common causes, and contamination of the measures of the exogenous variables b y the endogenous variables. We illustrate the consequences of these last four types of specification error through the analysis of contrived data