Did bubbles migrate from the stock to the housing market in China between 2005 and 2010?

• DOI: http://doi.org/10.1111/1468-0106.12230
• Author: Yongheng Deng,Eric Girardin,Shuping Shi,Roselyne Joyeux
• Date: 01 August 2017
• Published date: 01 August 2017

bs_bs_banner

SPECIAL ISSUE ARTICLE

Did bubbles migrate from the stock to the

housing market in China between 2005 and

2010?

Yongheng Deng

1

| Eric Girardin

2

| Roselyne Joyeux

3

| Shuping Shi

3

1

National University of Singapore,

Singapore

2

Aix‐Marseille University, Aix‐Marseille

School of Economics, CNRS & EHESS,

Marseille, France

3

Macquarie University, Sydney, Australia

Correspondence

Eric Girardin, Aix‐Marseille University,

Aix‐Marseille School of Economics, 5 Bd

Maurice Bourdet ‐CS 50498 ‐F‐13205

Marseille cedex 1, FRANCE.

Email: eric.girardin@univ‐amu.fr

Abstract

The speculative nature of both stock and housing mar-

kets in China has attracted the attention of observers.

However, while stock market data are easily available,

the low frequency and low quality of publicly available

housing price data hampers the study of the relation-

ship between the two markets. We use original hedonic

weekly resale housing prices of a major Chinese hous-

ing market and study them in conjunction with Shang-

hai's stock market index in the second half of the 2000s.

The use of the Phillips et al. (2015 a,b) recursive explo-

sive‐root test enables us to detect and date speculative

episodes in both markets. We then implement the Gree-

naway‐McGrevy and Phillips (2016) methodology to

detect the presence of migration between the two types

of bubbles. We detect significant migration from the

stock to the housing market bubble in 2009 and a tem-

porary spillover in 2007.

1|INTRODUCTION

It is often argued that ‘Real estate markets are even more prone to bubbles than stock markets

because these markets are dominated by unsophisticated households, short‐sale constraints fre-

quently bind, and often arbitrage is prohibitively costly’(Scherbina & Schlusche, 2014, p. 597).

China's markets provide a unique opportunity to test this hypothesis because its stock market

is characterized by a dominance of inexperienced individual investors, binding short‐sales

constraints (lifted only in 2011), a small asset float (before the split‐share reform of 2005–2006;

see Beltratti, Bortolotti, & Caccavaio, 2009) and heavy share turnover despite high transaction

Received: 26 January 2017 Accepted: 20 April 2017

DOI: 10.1111/1468-0106.12230

276 © 2017 John Wiley & Sons Australia, Ltd Pac Econ Rev. 2017;22:276–292.wileyonlinelibrary.com/journal/paer

costs. According to Bailey, Cai, Cheung, and Wang (2009), (the dominant) individual investors in

China's stock market are less informed and more subject to behavioural biases than institutional

investors. In a similar way, individual investors dominate the residential resale housing market in

China. Another major similarity between the real estate and stock markets in China is that the

link with the prospective income on the respective market is tenuous. Indeed, for many years

Chinese listed firms hardly distributed any dividends, and the Chinese rental market for housing

is very underdeveloped, in such a way that many investors in the housing market adopt a buy‐and‐

hold strategy in which a lot of housing units are reported to be vacant. All these factors make very

likely the presence of active speculative behaviour in these two Chinese asset markets.

Detecting bubbles in asset markets has been an ongoing challenge, which empirical methods

have long been unable to meet. Such a challenge has been magnified in China due to the low

quality of some asset price series. Giglio, Maggiori, and Stroebel (2016) propose a model‐free test

for the no‐bubble condition based on UK and Singapore housing market data. Recently‐devel-

oped recursive explosive‐root (vs random walk) tests (Phillips, Shi, & Yu, 2015a,b (PSY)) enable

researchers to detect the rise and collapse of bubbles on a given asset market.

1

It is shown by

Phillips, Shi, and Yu (2015a) that the PSY method outperforms the recursive method of Phillips,

Wu, and Yu (2011) and the CUSUM strategy of Homm and Breitung (2012). In addition, the

former is much less computationally intensive and more effective than the Markov‐switching

augmented Dickey–Fuller (ADF) test of Hall, Psaradakis, and Sola (1999). We apply the PSY

method in a first step to the detection and dating of bubbles in China both in the stock and real

estate markets. Most existing work using this or other methods (for surveys, see Gurkaynak

(2008), Homm & Breitung (2012) and Breitung (2014)) would only focus on one market in isola-

tion without being able to examine the sequence of bubbles between different markets. However,

in as much as other diversification opportunities for Chinese individual investors are very lim-

ited, it is expected that speculative pressure may move between these two markets. The recently

developed methodology by Greenaway‐McGrevy and Phillips (2016) enables us to examine in a

second step for China the migration of bubbles from the stock market to a first‐tier city real

estate market (or vice versa). In order to conduct a high‐frequency study and to remedy the bias

in official (low frequency) housing prices in China, we use an original weekly hedonic resale real

estate price series of a major housing market in China, covering a sample starting before, and

ending after, the global financial crisis: from 2005 to 2010. The sample starts after a major step

in urban housing reform in China (Yang & Chen, 2014), and ends at the time of a major reform

in the stock market associated with the introduction of stock index futures. Such reforms opened

up diversification opportunities, out of the cash stock market, into the real estate market for the

former and into the futures market for the latter.

A burgeoning literature has attempted to estimate the relationships between the returns or

the volatility of China's government‐released real estate prices and stock prices (see Section 2

below). However, such work (often relying on vector autoregressive systems) typically excludes

the presence of explosive behaviour and simply models the first difference of prices. There is no

existing literature on the relationship between stock and real estate bubbles in China.

The present paper contributes to the existing literature by detecting bubbles both for high‐

frequency original hedonic resale housing market data for a first‐tier city (Beijing) in China

1

The PSY procedure is based on the local explosive characteristic of bubbles (Diba & Grossman, 1988) and designed for

positive bubbles. Diba and Grossman (1988) argue that given free disposal a negative rational‐bubbles component cannot

exist ‘because stockholders cannot rationally expect a stock price to decrease without bound and, hence, to become neg-

ative at a finite future date’.

DENG ET AL.277

bs_bs_banner