Hedonic Pricing Model

The coefficient on a characteristic estimates the marginal willingness to pay (implicit price) for a one-unit increase in that characteristic, holding other characteristics constant. For example, if the coefficient on air quality (measured as inverse pollution) is 5000, an improvement in air quality that reduces pollution by 10 units would increase the house price by approximately 50,000. The interpretation assumes other characteristics and market conditions are held constant.

Hedonic pricing infers value from actual market transactions (revealed preferences); contingent valuation asks people hypothetically (stated preferences). Hedonic pricing avoids hypothetical bias but is limited to goods with active markets. Contingent valuation can value any good but is subject to survey biases. Both methods can complement each other: use hedonic pricing for housed goods and contingent valuation for non-market goods.

Omitted variables bias arises if important characteristics are not measured, and these are correlated with measured variables. Include as many measured characteristics as possible. Use spatial fixed effects to control for neighborhood-level unobservables. Consider instrumental variables if characteristics are endogenous (e.g., pollution affected by past development, not just current). Conduct sensitivity analysis: show results are robust to plausible omissions.

Not directly: hedonic regressions show correlations, not causation. To establish causality, use experimental variation (rare) or quasi-experimental methods (e.g., instrumental variables, regression discontinuity, difference-in-differences) exploiting plausibly exogenous changes in characteristics. For example, a zoning change or environmental regulation that affects some properties but not others provides quasi-experimental variation.