A housing price index with the improvement-value adjusted repeated sales (IVARS) method

• DOI: https://doi.org/10.1108/IJHMA-01-2021-0005
• Published date: 06 May 2021
• Date: 06 May 2021
• Pages: 375-391
• Subject Matter: Property management & built environment,Real estate & property,Housing markets
• Author: Chung Yim Edward Yiu,Ka Shing Cheung

A housing price index with the

improvement-value adjusted

repeated sales (IVARS) method

Chung Yim Edward Yiu and Ka Shing Cheung

Department of Property, The University of Auckland, Auckland, New Zealand

Abstract

Purpose –The repeat sales house price index (HPI) has been widely used to measure house price

movements on the assumption that the quality of properties does not change over time. This study aims to

develop a novel improvement-value adjusted repeat sales (IVARS) HPI to remedy the bias owing to the

constant-qualityassumption.

Design/methodology/approach –This study compares the performanceof the IVARS model with the

traditional hedonic price model and the repeat sales model by using half a million repeated sales pairs of

housing transactionsin the Auckland Region of New Zealand,and by a simulation approach.

Findings –The results demonstratethat using the information on improvement values from mass appraisal

can signif‌icantly mitigate the time-varying attribute bias. Simulation analysis further reveals that if the

improvement work done is notconsidered, the repeat sales HPI may be overestimated by2.7% per annum.

The more qualityenhancement a property has, the more likely it is that the property will be resold.

Practical implications –This novel index may have the potential to enable the inclusion of home

conditionreporting in property value assessmentsprior to listing open market sales.

Originality/value –The novel IVARS index can help gaugehouse price movements with housing quality

changes.

Keywords New Zealand, Appraisal, House price index, Home condition report, Improvement value,

Quality-adjusted repeat sales method

Paper type Research paper

1. Introduction

While the hedonic pricing model (HPM) is commonly used as a conventional method to

construct house price indices (HPI), it is almost impossible to control every characteristic of

a property. The method is, therefore, prone to the omitted variablebiases. In this paper, the

repeat sales method is, therefore, developed to overcome the problem. By measuring the

changes in the transaction pricesof an identical house between two sales, the method avoids

the need to control the temporal-invariant housing quality attributes. The constant-quality

assumption is too strong to be realistic.

In contexts ranging from property appraisal to the accounting profession, asset

depreciation is always applied, and there is no exception for a property. Quality

enhancement work, such as renovationand rehabilitation, also exists. Thus, the repeat sales

price index tends to be “over-estimated”ifbetter quality houses are more likely to be resold.

Comparisons betweenthe repeat sales index and the hedonic HPI are also invalid, given that

the omitted variable biases in a hedonic model would cause it to over-estimate price

This research was funded by the University of Auckland Faculty Research Development Funds,

grant number FRDF-3722103.

A housing

price index

375

Received12 January 2021

Revised2 March 2021

Accepted4 March 2021

InternationalJournal of Housing

Marketsand Analysis

Vol.15 No. 2, 2022

pp. 375-391

© Emerald Publishing Limited

1753-8270

DOI 10.1108/IJHMA-01-2021-0005

The current issue and full text archive of this journal is available on Emerald Insight at:

https://www.emerald.com/insight/1753-8270.htm

movements as part of the genuine quality differences captured as temporal changes. This

study aims to develop an improvement-value adjusted repeat sales (IVARS) HPI that

incorporates quality enhancement values(EV).[1] A simulation study is conducted to show

how the quality EV is associated with house prices as cross-validation. Given that the

quality EV are artif‌icially assigned, the hypothesis of an over-estimation bias can be tested

by the simulation study.

So far, there have been only limitedstudies of the quality enhancement effect [2] on HPI.

Because quality data are commonly unavailable, most repeat sales HPI do not consider

quality changes between sales.When a house’s quality is enhanced (or has deteriorated), its

value would appreciate (depreciate)according to the maintenance costs involved. Given that

the quality-change effect holds, then repeat sales price index will always be biased. Using

the appraised values of housing structures in the New Zealand housing market, this study

aims to develop an IVARS HPI that remediesthe constant-quality assumption in traditional

repeat sales methods.

The paper is organised as follows. Section 2 critically reviews the previous attempts to

solve the housing quality changesproblem in the repeat sales method. Section 3 presents the

IVARS models incorporating the use of appraisal information in EV. Section 4 provides

empirical data and results based on the transactions of residential properties in Auckland.

Section 5 describestwo robustness tests. Section 6 concludes.

2. Literature review

HPI are commonly constructed using the HPM (Palmquist,1980;Greenlees, 1982;Mark and

Goldberg, 1984) or the Bailey et al. (1963, BMN)’s repeat sales method. The price index is

derived from the coeff‌icients of time dummies estimated in the specif‌ied regression.

However, the major criticism of HPM is the diff‌iculty of specifying a correctand completed

set of property attributes, which leads to measurement error and omitted variable biases

(Case et al., 1991). The BMN repeat sales method attempts to overcome such errors by

comparing the prices of repeatedly transacted properties. As the method compares the

changes in transacted prices of thesame house, all non-temporal attributes can be cancelled

out, given that the implicit prices of attributes do not change between sales. The simplicity

of the repeat sales methodmakes it one of the most popular methods in constructing HPI.

However, the repeat sales approachhas often been criticised for its ineff‌iciency (e.g. Case

et al., 1991), sample selection bias (Clappand Giaccotto, 1992;Meese and Wallace, 1997), the

incapability of controllingfor quality changes over time, including the quality enhancement

effect (OECD, 2013;Bugden, 2013) and the qualitydeterioration (depreciation) effect (Bailey

et al.,1963).Many studies have examined the effect of qualitydeterioration (Palmquist, 1982;

Goodman and Thibodeau, 1996;Clapp and Giaccotto, 1998a;Cannaday et al.,2005;Chau

et al.,2005). Nevertheless, these studies assume only a linear depreciation and use the

building’s age as a proxyfor such depreciation effects. Even though this practiceis typically

adopted in accounting practices, such an assumption is econometrically not viable. The

exact collinearity between the age and time variables renders the depreciation effect

inestimable. Chau et al. (2005) asserted that in the repeat sales method, the depreciation

effect could only be estimated if external information is available to solve the exact

collinearity problem.

Wong et al. (2018) exploited the structure-to-property value ratio to break the linear

relationship between age and time differences between sales, making it possible to

separate the age effect from a repeat-sales index. Another suggestion is to take age as a

proxy of the depreciation effect in the repeat-sales model. However, we contend that the

collinearity issue is not a problem when we have the data of actual appreciation or

IJHMA

15,2

376