12 June 2017 in Real Estate Economics
13 July 2021 in The Journal of Real Estate Finance and Economics
This paper couples a traditional hedonic model with architectural style classifications from human experts and machine learning (ML) enabled classifiers to estimate sales price premia over architectural styles, both at the building and the neighborhood-level. We find statistically and economically significant price differences for houses from distinct architectural styles across an array of specifications and modeling assumptions. Comparisons between classifications from ML models and human experts illustrate the conditions under which ML classifiers may perform at least as reliable as human experts in mass appraisal models. Hedonic estimates illustrate that the impact of architectural style on price is attenuated by properties with less well-defined styles and we find no evidence for differential price effects of Revival or Contemporary architecture for new construction.
30 November 2022 in Real Estate Economics
7 August 2021 in The Journal of Real Estate Finance and Economics
Separating urban land and structure values is important for national accounts and for analysis of real estate risk over time. A large part of the literature on urban land valuation uses the land residual method, which relies on the assumption that structures are easily replaced. But urban land value depends on accessibility to nearby land uses, implying that infrastructure and the slowly changing built environment are the most important components of land value. Investments in structures are only slowly reversible, implying that land and structure function as a bundled good whereas land residual theory severs the connection between land value and structure value over time. We develop a simple theoretical model that includes option value and compare to a nested land residual model before and after a shock to values. Cross-sectionally our model shows that land residual theory overestimates structure value. Over time almost all of any change in property value is allocated to land residuals. Data from Maricopa county, AZ, 2012–2018 strongly support option value models when nested within a general model that also includes land residuals. FHFA estimates use entirely different cost estimation methods: our analysis of FHA data suggest that our conclusions generalize to the U.S. as a whole, and that high and rising land value ratios over 50 years (the “hockey stick” pattern found in the literature) are likely an artifact of the residual model.
2 March 2025 in Real Estate Economics
We extend the emerging literature on machine learning empirical asset pricing by analyzing a comprehensive set of return prediction factors for real estate investment trusts (REITs). We show that machine learning models are superior to traditional ordinary least squares models and find that REIT investors experience significant economic gains when using machine learning forecasts. In particular, we show that REITs are more predictable than stocks and that their higher predictability is stable over time and across industries.
30 September 2022 in Journal of Housing Economics