On Balance: Best Practices for Using Hedonic Property Value Models
Hedonic property value regression is a leading technique for estimating how much consumers are willing to pay for nonmarket amenities. The prevailing style of estimation has evolved in recent years to incorporate insights from the “credibility revolution” in applied economics, with high expectations for data quality and econometric transparency. At the same time, recent research has improved our understanding of how parameters identified by quasi-experimental designs relate to welfare measures. This post describes an article summarizing modern best practices for developing credible hedonic research designs and valid welfare interpretations of the estimates. I wrote the article together with Kelly Bishop, Spencer Banzhaf, Kevin Boyle, Kathrine von Gravenitz, Jaren Pope, Kerry Smith, and Christopher Timmins. It was published in the Summer 2020 issue of the Review of Environmental Economics and Policy as part of a symposium on best practices for using revealed preference methods for nonmarket valuation of environmental quality. A 20-minute video summary is posted here.
There have been thousands of hedonic property value studies since the model was formalized in the 1970s and the pace has accelerated due to advances in data, econometrics, and computing power. The model’s enduring popularity is easy to understand. It starts with an intuitive premise that is economically plausible and empirically tractable. The model envisions buyers choosing properties based on housing attributes (e.g., indoor space, bedrooms, bathrooms) and on location-specific amenities (e.g., air quality, park proximity, education, flood risk). In the absence of market frictions, spatial variation in amenities can be expected to be capitalized into housing prices. When buyers face the resulting menu of price-attribute-amenity pairings, their purchase decisions can reveal their marginal willingness to pay (MWTP) for each of the amenities. In principle, estimating MWTP is straightforward. In practice, several key modeling decisions must be made. These include defining the market, choosing appropriate measures of prices and amenities, selecting an econometric specification, and developing a research design that isolates exogenous variation in the amenity of interest.
Defining the Market. Best practices in hedonic estimation start with defining the relevant housing market in a way that satisfies the “law of one price function”. This means that identical houses will sell for the same price throughout that market. The precise spatial and temporal boundaries that satisfy this condition may vary across space and over time as information, institutions, and moving costs change. One common practice is to define the market as a single metro area over a few years. An alternative is to pool data over larger areas and longer periods, and to model the hedonic price function as evolving over space and time.
Data Collection. The gold standard for data is a random sample (or the universe) of housing-transaction prices and characteristics for the relevant study area. Studies that meet this standard often focus exclusively on single-family houses. While data on house prices and attributes are increasingly available for markets around the world, an ongoing challenge is to characterize how households perceive an amenity of interest. This task requires developing an objective measure of spatial variation in the amenity that can be matched to individual houses. As part of this process, it is important to document the information channels that may influence buyers’ beliefs and to assess the sensitivity of MWTP estimates to candidate measures of the amenity.
Functional Form. The price function should be allowed to be flexible. It is hard to justify a linear specification on conceptual grounds and simulation-based studies show that more flexible non-linear specifications provide more accurate estimates of average MWTP. Applications should also rely on robust standard errors and cluster at a spatial-temporal scale of variation in the amenity of interest.
Identification and Interpretation. A best practices research design utilizes a clear source of exogenous variation in an amenity that households can observe. This step is essential to address widespread and well-justified concern about omitted variable biases. Difference-in-differences, matching estimators, spatial dummy variables, instrumental variables, and boundary-discontinuity designs are among the techniques that can help to address this concern. While it is very important to ensure econometric credibility by mitigating omitted-variable problems, it is equally important to consider how econometric modifications to a price function influence how the identified parameters relate to MWTP. The econometric identification strategy can determine whether the identified parameters are best interpreted as point estimates (or bounds) of MWTP for the population of interest (or for a specialized subset of the population).
While the number of issues that must be considered in developing a best practices hedonic study may seem daunting, the effort is justifiable. The modern hedonic property-value model has been refined through more than forty years of intense scrutiny to become one of the premier approaches to valuing changes in environmental amenities in academic research, litigation, and public policy.