The Effects of Causal Feedback On Ordinary Least-Squares Estimators
- 1 November 1974
- journal article
- research article
- Published by SAGE Publications in Sociological Methods & Research
- Vol. 3 (2), 189-208
- https://doi.org/10.1177/004912417400300203
Abstract
When a causal model includes a feedback loop, ordinary least-squares (OLS) is an inappropriate, biased estimation technique. Furthermore, since the system is nonrecursive, a simple application of path analysis or regression is precluded. If the causal feedback is indirect, however, or involves one or more weak causal links, the bias will tend to be small and in some cases may be negligible. If the causal system is correctly specified to include only indirect feedback-i.e., with no direct reciprocal causation between any pair of variables, it is possible to calculate unbiased estimators. Even in other instances, it is often possible to set a reasonable upper limit to the bias encountered by estimators affected only by indirect feedback despite their inclusion in a system in which there is both direct and indirect feedback.Keywords
This publication has 2 references indexed in Scilit:
- Organizational Interdependence and Intra-Organizational StructureAmerican Sociological Review, 1968
- Path Analysis: Sociological ExamplesAmerican Journal of Sociology, 1966