Academic journal article Geographical Analysis

Impact of Cliff and Ord on the Housing and Real Estate Literature

Academic journal article Geographical Analysis

Impact of Cliff and Ord on the Housing and Real Estate Literature

Article excerpt

The works of Cliff and Ord have had a major impact on empirical practices in real estate. Cliff and Ord proposed both techniques for detecting as well as modeling spatial dependence. Because the existence of spatial dependence is almost assured in real estate data, their most important contribution was feasible means of estimating spatial models. The full implications of these ideas and the numerous modeling techniques spawned by their seminal works have not been fully explored and provide numerous opportunities for future research.


Spatial ideas have always been fundamental to real estate and housing. However, the need for simplicity in theory and statistical analysis in early work led to distilling the two dimensions of space into a single dimension distance (such as from each home location to the urban center). Ideally, a regression containing a distance variable would yield residuals that show no obvious spatial patterns.

Because often this did not occur in practice, researchers sometimes included distances to other points, regional or neighborhood indicator variables, polynomials in the locational coordinates, and other trend surfaces in an effort to reduce the obvious map patterns in the residuals. Even after controlling for space in this fashion, samples containing a large number of nearby homes usually exhibit spatial clusters of regression residuals with the same sign. This outcome is because pairs of nearby houses lie in the same neighborhoods so that neighborhood indicator variables do not treat these observations differently. In addition, trend surfaces change little over short distances so that these variables provide little gain in explanatory power, and distances to other locations are virtually the same for pairs of neighboring houses. We now know that these standard statistical techniques were based on assumed independence among sample observations, which real estate and housing data violate. In the face of sample data inconsistent with independence, conventional independent statistical methods can at best lead to inefficient model estimates and invalid inference about these parameters.

Against this background, Cliff and Ord (1969) devised a parsimonious specification for the structure of spatial dependence among observations that could be used to quantify the problem of spatial interdependence. Moreover, they proceeded in the corpus of their work (Cliff and Ord 1973, 1981; Ord 1975) to further develop these ideas, and to propose spatial autoregressions that could properly account for spatial dependence in sample data. From a regression perspective, the work by Cliff and Ord led to two strategies. The first involves use of diagnostics to identify spatial dependence in a current model, followed by an associated increase in complexity in the revised model to reduce the dependence. We refer to this as the "spatial detection strategy." As Ripley (1981, p. 98) states,

  Indeed, the philosophy adopted seems to have been that if "spatial
  autocorrelation" is found more explanatory variables should be
  introduced until it disappears!

The second was to incorporate dependence into the estimation model using a spatial autoregression. We refer to this as the "spatial estimation strategy."

In terms of housing and real estate journals, these ideas have been growing in influence over time. An examination of six of the leading journals in this area (Journal of Housing Economics, Journal of Real Estate Finance and Economics, Journal of Real Estate Research, journal of Urban Economics, Real Estate Economics, and Regional Science and Urban Economics) shows a clear pattern. A search of the terms "Cliff and Ord," "spatial autocorrelation," "spatial dependence," and "spatial autoregression" turns up 62 articles published in these journals from 1977 to 2008. Over the 1977-1999 period, 22 of these articles appeared, representing slightly > 1 per year. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.