Property Variables Most Subject to Endogeneity

Valuation Engineer Research Labs has published a new working paper, Property Variables Most Subject to Endogeneity, by William Bert Craytor.

Building on the Foundations entries 008 (Omitted Variable Bias) and 009 (Endogeneity) in Issue 2, the paper asks which variables in a residential sales model are most often endogenous — moving with the part of price the model does not explain — and why. It groups them by cause: features the owner chose, such as remodels and living area; features measured with error, such as living area from MLS and assessor records; transaction variables such as list price, concessions and days on market; and the selection of which houses reach the sales data. A separate section treats school districts and neighborhoods.

Each cause is demonstrated in a simulated market of 40,000 houses whose true price-setting is known, so the direction of each error can be seen. The paper then sets out what this means in practice: when a coefficient can and cannot be read as an adjustment, why a counterfactual value such as "as if remodeled" needs stated assumptions, and how the residual ranking of the Residual Constraint Approach places the subject from MLS data and photographs.

The paper includes the simulation program, a Prolog program that states its logic, and a fifteen-question quiz. It is available on the web and as a PDF:

https://research.valuation-engineer.com/foundations/property-variables-endogeneity/

DOI: https://doi.org/10.67330/3te26280

Research Labs papers are working papers and are not peer reviewed; the Journal remains the version of record.