Using a Finite Mixture Model of Heterogeneous Households to Delineate Housing Submarkets
Belasco, Eric, Farmer, Michael C., Lipscomb, Clifford A., The Journal of Real Estate Research
We use a finite mixture model to identify latent submarkets from household demographics that estimates a separate hedonic regression equation for each submarket. The method is a relatively robust empirical tool to extract submarkets from demographic information with far less effort than suspected. This method draws from latent class models to group observations in a straightforward data-driven manner. Additionally, the unique information about each submarket is easily derived and summarized. Results are also shown to more convincingly sort submarkets than a prior study in the same area that used more comprehensive data.
(ProQuest: ... denotes formulae omitted.)
One concern in the application of hedonic pricing methods is the reliable detection of housing submarkets. There are several approaches to approximate submarket delineation. Some rely on a well-partitioned geographic area as a submarket indicator or on the type of domicile (apartment or house) to sort households into submarkets. Yet if the analyst is interested in the economic demand properties of hedonic attributes, submarkets need to differentiate households according to different preferences for hedonic amenities. Two recent articles in this journal illustrate the issue. First, Shultz and Schmitz (2009) note the considerable variation in the price effect of golf courses. Further, the authors note that residential composition is not as homogeneous as they initially thought. Second, Farmer and Lipscomb (2010) use within-neighborhood household diversity to explore the competition for different bundles of housing attributes between different types of households. They found households with distinctly different tastes offering similar prices for some houses in a single neighborhood. These two works illustrate the theoretical challenge facing real estate researchers interested in demand-side analyses: household heterogeneity requires matching diverse household tastes to particular houses. Operationally, we define submarkets as subsets of market agents whose preference rankings over the stock of real estate are similar, meaning that a group (submarket) orders houses from most preferred to least preferred in a similar way that is distinct enough from another group (submarket) that has a different ranking from least preferred to most preferred. Empirically, this can be a very time-consuming process; so other strategies have been developed.
In the absence of very specific household level demographic and attitudinal data, researchers such as Osland (2010) use geographically-weighted regression and spatial Durbin models to account for spatial heterogeneity. The current work follows up on prior tests that elicit and analyze a small set of household information directly from a small survey sample (Lipscomb and Farmer, 2005). That work was robust in submarket identification; but unnecessarily data intensive. The current work introduces a more flexible econometric method to detect submarkets that could plausibly be implemented as it greatly reduces the costs to ascribe much more detail in submarket identification at the individual unit level.
The returns to responsible submarket isolation are not trivial. First, explanatory power to predict housing price variation in the dependent variable has been shown to improve dramatically with submarket delineation (Goodman and Thibodeau, 2007). Also, proper aggregability of households into submarkets, which can be characterized by a representative household for each submarket, is required for consistent attribute coefficient estimation (Palmquist, 2004). Put another way, grouping types for hedonic price analysis into those with similar preference rankings is essential, especially if the goal is to conduct welfare analyses of, say, an environmental attribute change using hedonic price estimates (Sieg, Smith, Banzhaf, and Walsh, 2002; Banzhaf and Walsh, 2008).
In a review of the housing submarket literature, we find no clear consensus on how to delineate housing submarkets. …