From a small northwestern observatory…

It’s always nice to be cited in someone else’s research paper! Case in point, a nicely done real estate paper written by three young scholars out of U. Aukland and U. Hong Kong just hit my desk, and much to my enjoyment, they cited my Journal of Housing Research piece on Repeat Sales Analysis. Intriguingly, my original work was on the use of this technique to identify the negative impact of environmental contamination on property prices. Conversely, their work was on the positive impact of water views on property prices. Indeed, this illustrates the fact that, at the far ends of the valuation spectrum, the same methodologies can prove useful.

After exploring repeat sales as a potential tool, the three authors ended up opting for a weighted least squares regression, noting (and I agree with them on this point) that repeat sales models have a “well known error distribution characteristic.” The simple fact is, parametric models, with which most real estate students are familiar, are usually not handy for dealing with real estate data, which is highly non-parametric. However, parametric models (such as regression analysis) require lots of data, and real estate analysis problems are often challenged with thin data sets. There are good non-parametric methodologies out there (such as repeat sales) for such limited data situations, but the statistical properties are not well characterized. However, they were able to incorporate repeat sales into their hedonic regression model. At Greenfield we’ve similarly used nonparametric methods to inform instrumental variables in a method that emulates 2-stage least squares.

As the demands for reliability of valuation models increase, finding ways to better characterize the reliability and confidence of appraisal methods deserves increasing attention. Academic research scholars are used to using large data sets and parametric methods with well-defined statistical characteristics. Practitioners — a category that includes both appraisers and investors — are usually faced with poor data sets and traditional, somewhat heuristic models which work well in practice but have little in the way of statistical characterization. Some academic organizations, such as the American Real Estate Society, strive to span the gulf between research and practice, but more remains to be accomplished in valuation modeling.

The paper, by the way, is “An Empirical Study of Sea View Value by Repeat Sales Method” authored by Edward Chung Yim Yiu (U. Aukland) and Chau Kwong Wing and Siu Kei Wong (U. Hong Kong). Their study was supported by the Research Group on Sustainable Cities and the CRCG Conference Grant for Teaching Staff of the University of Hong Kong, and their findings are available on the Social Sciences Research Network and on Researchgate.

John A. Kilpatrick, Ph.D., MAI — john@greenfieldadvisors.com