Forecasting house prices in OECD economies

• Author: N. Kundan Kishor,Hardik A. Marfatia
• DOI: http://doi.org/10.1002/for.2483
• Date: 01 March 2018
• Published date: 01 March 2018

Received: 10 June 2016 Revised: 30 January 2017 Accepted: 26 May 2017

DOI: 10.1002/for.2483

RESEARCH ARTICLE

Forecasting house prices in OECD economies

N. Kundan Kishor1Hardik A. Marfatia2

1Department of Economics, University of

Wisconsin–Milwaukee, Milwaukee,

WI, USA

2Department of Economics, Northeastern

Illinois University,C hicago, IL,USA

Correspondence

N. Kundan Kishor, Department of

Economics, University of

Wisconsin–Milwaukee, PO Box 413, Bolton

822, Milwaukee, WI 53201, USA.

Email: kishor@uwm.edu

Abstract

In this paper, we forecast real house price growth of 16 OECD countries using

information from domestic macroeconomic indicators and global measures of the

housing market. Consistent with the findings for the US housing market, we find

that the forecasts from an autoregressive model dominate the forecasts from the ran-

dom walk model for most of the countries in our sample. More importantly, we

find that the forecasts from a bivariate model that includes economically important

domestic macroeconomic variables and two global indicators of the housing market

significantly improve upon the univariate autoregressive model forecasts. Among

all the variables, the mean square forecast error from the model with the coun-

try's domestic interest rates has the best performance for most of the countries. The

country's income, industrial production, and stock markets are also found to have

valuable information about the future movements in real house price growth. There

is also some evidence supporting the influence of the global housing price growth

in out-of-sample forecasting of real house price growth in these OECD countries.

KEYWORDS

forecasting, global housing market, macroeconomic predictors

1INTRODUCTION

The housing market has strong interlinkageswith the business

cycle as well as the financial markets. It is not very surpris-

ing, then, to find significant interest among academicians,

practitioners, and policymakers to understand the behavior of

the housing market. A declining housing market can poten-

tially threaten economic slowdown (Leamer,2007), adversely

affect financial markets as seen in the recent crisis, and could

significantly strain household wealth. Even though the degree

of importance of the housing market may vary across dif-

ferent countries, there is a consensus regarding its overall

crucial role in macroeconomic fluctuations. It is very impor-

tant, therefore, to gain some insight into the future movements

of house prices in developed economies. The objective of

this paper is to examine forecasts of real house price growth

across different OECD countries and examine the forecasting

performance of different macroeconomic variables.1

1The countries considered in this study include Australia, Belgium, Canada,

Denmark, Finland, France, Germany, Italy, Netherlands, Norway, New

Zealand, Spain, Sweden, Switzerland, the UK, and the USA.

Although the housing market is a much bigger portion of

the household's portfolio, research on house price predictabil-

ity is relatively small as compared to the forecasting litera-

ture on stock markets. Few important issues have emerged

from the existing research on forecasting of housing prices.2

Most of the existing studies have focused on in-sample fore-

casting exercises and those that do perform out-of-sample

forecasting exercises are mainly limited to the US housing

market.3Further, in out-of-sample forecasting exercise, it

is generally found that the information contained in house

prices' own past movements is superior to other models

(Crawford & Fratantoni, 2003). This is in spite of strong evi-

dence in the literature suggesting that several macroeconomic

variables such as income and interest rates comove with

2See, for example, Brown, Song, and McGillivray (1997), Crawford and

Fratantoni (2003), Gallin (2006), and Bork and Møller (2015).

3Rapach and Strauss (2009) perform an out-of-sample forecast of state house

prices in the USA. See also Crawford and Fratantoni (2003) and Guirguis,

Giannikos, and Anderson (2005), among others.

170 Copyright © 2017 John Wiley & Sons, Ltd. wileyonlinelibrary.com/journal/for Journal of Forecasting.2018;37:170–190.

KISHOR AND MARFATIA 171

house prices and play a fundamental role in driving the market

dynamics.4

The existing evidence points towards some interestingly

interlinked questions that have not been answered in the

literature. For example, to what extent can one predict

out-of-sample house price movements in the OECD countries

using the information from its past behavior? Toexamine this,

we ought to compare the forecast generated from an autore-

gressive model with that of a model that assumes efficiency

in the housing market. In this case, efficiency may be rep-

resented by a random walk model of real house prices. The

next question is: Can fundamental drivers like income and

interest rates provide additional information that improves

the forecasts over and above house prices' own past move-

ments? These two questions are interlinked because, if the

housing markets are predictable in the OECD economies,

then it is important to understand the out-of-sample forecast-

ing potential of other fundamental variables that are found

to be important in explaining movements in house prices.

Further, given the growing interlinkages in the financial mar-

kets across the globe, we also examine the predictive ability

of aggregate OECD house price and US house prices in

forecasting individual countries' real house price movements.

Our results show that the mean square forecast errors

(MSFEs) of a simple autoregressive model are significantly

lower, on average, than random walk model MSFEs in most

of the countries in our sample.5The results obtained from this

exercise show some evidence of inefficiency in the housing

markets in these OECD countries in terms of its predictabil-

ity. This is consistent with Case and Shiller (1989), (1990)

and Rapach and Strauss (2009), who find that the US hous-

ing market is not efficient, as house price changes can be

predicted using information from economically meaningful

predictors. Not only do the autoregressive model forecasts

outperform the random walk model in almost all economies

by a substantial margin, but also the forecasting performance

of the random walk model deteriorates as the forecast hori-

zon increases. Using Clark and West's (2007) nested forecast

comparison test, we find that MSFEs from the AR models are

significantly lower than those of the random walk model in

different countries.

The evidence from vector autoregressive (VAR) models

suggests that in most of the countries the univariate forecasts

can be improved upon by adding one of the macroeconomic

variables in the model with real house price growth. The

improvement in forecasting performance is particularly sub-

stantial in countries such as the USA, Australia, NewZealand,

Spain, and Italy. For example, 1-year-ahead forecasts of real

4See, for example, Holly and Jones (1997), Hort (1998), Meen (2002),

McCarthy and Peach (2004), Annett (2005), Iacoviello (2002), Himmelberg,

Mayer, and Sinai (2005), and Kishor and Marfatia (2014).

5Our forecast sample runs from 1996:Q1 through 2013:Q4 for most of the

variables. Details of the forecast sample are provided in the Appendix.

house price growth for the USA from a VAR model have

18–23% lower MSFEs than the benchmark autoregressive

model. Using forecast evaluation tests, we also show that

in many of these cases the differences in MSFEs are sta-

tistically significant. Among all the variables, interest rates

provide the best forecast formost of the countr ies in our sam-

ple. This is followed by the other three domestic variables,

namely income, industrial production, and stock markets.

These domestic macroeconomic variables are equally impor-

tant in predicting house price movements, particularly in the

case of the USA, Australia, and a few European countries

such as Italy, Belgium, Norway, and Spain. Global hous-

ing variables that include US house prices and the aggregate

OECD house prices also hold valuable information about

future movements in house prices as compared to the univari-

ate autoregressive model in 13 out of the 16 OECD countries

for at least one of the many forecasting horizons. The dom-

inant role of interest rates in predicting future house price

movement is not surprising. This suggests that house pur-

chasing decisions are more sensitive to the size of monthly

installments than to the size of the loan in relation to house-

hold income and also the state of the real economic activity

and stock markets. These results support the view that the his-

torically low interest rates have been the major contributor to

house price movements in most industrialized countries.

The rest of the paper is organized as follows.Section 2 pro-

vides a brief overview of the related literature. In Section 3 we

lay down the methodology and provide a description of the

data. The results of housing market forecasts obtained from

different models are discussed in Section 4 and conclusions

are provided in Section 5.

2RELATED LITERATURE

Understanding the dynamics of the housing market is impor-

tant, as it plays a very important role in macroeconomic

fluctuations across different countries. An increase in house

prices can potentially cause “overheating” of the economy,

whereas a declining market possibly threatens economic

slowdown (Leamer, 2007). There is equally strong evidence

that shows the causation to run the other way round (Égert

& Mihaljek, 2007; Iacoviello, 2002). It is therefore of great

importance to understand the main determinants of housing

markets and, even more importantly, to forecast house price

movements using the information from these determinants of

house prices.

Since we are interested in forecasting the movement of

house prices, a priori we can use the variables that are found

to have a significant relationship with movements in housing

prices. The housing market behavior can broadlybe explained

by several macroeconomic variables that include income,

inflation, interest rates, money supply, and level of economic