Effectiveness comparison of the
residential property mass
appraisal methodologies
in the USA
Chung Chun Lin and Satish B. Mohan
Department of Civil, Structural and Environmental Engineering,
State University of New York at Buffalo, Buffalo, New York, USA
Abstract
Purpose – Quite a few statistical and artificial neural network (ANN) models have been developed
for the mass appraisal of the real estate by the municipalities. The purpose of this paper is to report the
results of a research conducted to compare the prediction accuracy of the three most used models:
multiple regression model, additive nonparametric regression, and ANN.
Design/methodology/approach – The three models were developed using the housing database of
a town with 33,342 residential houses. In this database, the cutoff point for higher priced homes was
$88 per square foot of living area.
Findings – The research confirmed that using statistical and ANN models are reliable and
cost-effective methods for mass appraisal of residential housing.
Originality/value – It was found that any of the three models can be used, with similar accuracy, for
lower and medium-priced houses, but the ANN is considerably more accurate for higher priced houses.
Keywords United States of America, Neural nets, Residential properties, Statistical methods,
Housing price estimation
Paper type Research paper
1. Introduction
The academia and the real estate industry have attempted to efficiently and accurately
estimate housing prices utilizing various mass appraisal techniques, for example, the
multiple regression (MR) analysis has increasingly been used in the real estate industry
for mass appraisal. Eisenlauer (1968) addresses the concept of using MR analysis as an
appraisal method, and Blettner (1969) extends the concepts introduced by Eisenlauer to
estimate housing prices appraisal also by using MR analysis. The estimates yielded by
MR equation have been used as the basis for the taxation of properties, for assessing
the value of properties for mortgage underwriting, and for the performance analysis of
real estate portfolios (Zheng-Gui, 2006).
Alternatives to MR model (MRM) for the real estate price estimation have increased
in popularity in recent years. An extensive literatures search on mass appraisal
methods, conducted in this research, concluded that the additive nonparametric
regression (ANR) and artificial neural network (ANN) are two of the more popular
examples of such alternatives. These studies are mentioned in the following two
paragraphs. So far, the relative accuracy of these three models has not been compared.
This paper aims to compare the housing price prediction accuracies of these three
models: the MRM, the ANR model, and the ANN model.
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1753-8270.htm
IJHMA
4,3
224
Received 9 November 2010
Revised 3 February 2011
Accepted 14 March 2011
International Journal of Housing
Markets and Analysis
Vol. 4 No. 3, 2011
pp. 224-243
q Emerald Group Publishing Limited
1753-8270
DOI 10.1108/17538271111153013
@Phil_Robichaud
@deepthiw
@JoseServera