Hedonic Models of Real Estate Prices: GAM and Environmental Factors

TL;DR

This paper evaluates the effectiveness of P-spline generalized additive models in predicting real estate prices in U.S. cities, highlighting their ability to incorporate environmental factors and outperform simpler models.

Contribution

It introduces the use of GAM models with environmental factors for real estate valuation and compares their predictive performance against linear models.

Findings

GAM models explain 84-92% of price variance.

Environmental factors have city-dependent significance.

Adding environmental factors slightly improves model fit.

Abstract

We consider the use of P-spline generalized additive hedonic models for real estate prices in large U.S. cities, contrasting their predictive efficiency against linear and polynomial based generalized linear models. Using intrinsic and extrinsic factors available from Redfin, we show that GAM models are capable of describing 84% to 92% of the variance in the expected ln(sales price), based upon 2021 data. As climate change is becoming increasingly important, we utilized the GAM model to examine the significance of environmental factors in two urban centers on the northwest coast. The results indicate city dependent differences in the significance of environmental factors. We find that inclusion of the environmental factors increases the adjusted R-squared of the GAM model by less than one percent.