Improving Land Valuation Models in Sparse Markets: A Comparison of Spatial Interpolation Techniques in Mass Appraisal

The research in this working paper has set up the groundwork for a national AVM of Malawi, Africa, using only secondary data collected by the 2010-2011 Integrated Household Survey. Model variables include physical characteristics of the property, economic variables, as well as location-specific distance and climate variables. This paper additionally helps bridge the current gap in development property tax literature by evaluating response surface analysis (RSA) at a national level, specifically with respect to technical standards of the International Association of Assessing Officers (IAAO). Our initial research shows that variables with positive effects on perception of value include agricultural plot size and estimated annual income (rental) potential. Plots situated further from agrimarkets and auction locations are perceived to be less valuable. Negative effects are associated with sandy soil, higher average annual rainfall, moderate to steep slopes, and plots situated in swamps or marshlands. The ordinary kriging-based predictions, while seemingly less likely to overestimate perceived value, are more regressive. Ordinary kriging achieves superior scores of vertical equity, but inferior scores of uniformity when compared to a current OLS model with location factor adjustment variables derived from RSA