Endogeneity

Abstract

Endogeneity is the condition in which a regressor is correlated with a model's error term, so that its coefficient does not recover the causal effect of interest. This entry generalizes the omitted-variable case of entry 008, presenting omitted variables, simultaneity, selection, and measurement error as four mechanisms producing one condition, and distinguishing endogeneity from imprecision by the fact that additional observations do not reduce it. Appraisal encounters are developed for each mechanism: reverse causation between school quality and neighborhood price levels; renovation as an owner's choice conditioned on unobserved property qualities rather than as a randomly assigned treatment; selection in the comparable set, since sold properties are not a random sample of properties; and condition ratings assigned by an appraiser who already knows the transaction price. The central result is that measuring an omitted variable cures the omission without necessarily curing the endogeneity --- a remodel coefficient estimated from observational data conflates the value the work added with the value of the properties whose owners chose to do it, so an appraiser answering a renovation-payback question from such a coefficient will systematically overstate the return. Since no residual plot or test statistic detects endogeneity from data alone, the entry argues that the defensible response is an explicit argument about mechanism and direction, and that design choices in assembling a comparable set do more for defensibility than any estimator applied afterward.