urban stud-2011-chen-2049-67

Upload: umar-zaghum

Post on 02-Jun-2018

218 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    1/20

    http://usj.sagepub.com/Urban Studies

    http://usj.sagepub.com/content/48/10/2049The online version of this article can be found at:

    DOI: 10.1177/0042098010379281

    2011 48: 2049 originally published online 16 December 2010Urban StudJie Chen, Feng Guo and Aiyong Zhu

    Provincial Panel DataThe Housing-led Growth Hypothesis Revisited: Evidence from the Chinese

    Published by:

    http://www.sagepublications.com

    On behalf of:

    Urban Studies Journal Foundation

    can be found at:Urban StudiesAdditional services and information for

    http://usj.sagepub.com/cgi/alertsEmail Alerts:

    http://usj.sagepub.com/subscriptionsSubscriptions:

    http://www.sagepub.com/journalsReprints.navReprints:

    http://www.sagepub.com/journalsPermissions.navPermissions:

    http://usj.sagepub.com/content/48/10/2049.refs.htmlCitations:

    What is This?

    - Dec 16, 2010OnlineFirst Version of Record

    - Jul 14, 2011Version of Record>>

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/content/48/10/2049http://usj.sagepub.com/content/48/10/2049http://usj.sagepub.com/content/48/10/2049http://www.sagepublications.com/http://www.sagepublications.com/http://www.urbanstudiesfoundation.org/http://usj.sagepub.com/cgi/alertshttp://usj.sagepub.com/cgi/alertshttp://usj.sagepub.com/subscriptionshttp://www.sagepub.com/journalsReprints.navhttp://www.sagepub.com/journalsReprints.navhttp://www.sagepub.com/journalsPermissions.navhttp://www.sagepub.com/journalsPermissions.navhttp://usj.sagepub.com/content/48/10/2049.refs.htmlhttp://usj.sagepub.com/content/48/10/2049.refs.htmlhttp://online.sagepub.com/site/sphelp/vorhelp.xhtmlhttp://online.sagepub.com/site/sphelp/vorhelp.xhtmlhttp://usj.sagepub.com/content/early/2010/12/14/0042098010379281.full.pdfhttp://usj.sagepub.com/content/early/2010/12/14/0042098010379281.full.pdfhttp://usj.sagepub.com/content/48/10/2049.full.pdfhttp://usj.sagepub.com/content/48/10/2049.full.pdfhttp://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/http://online.sagepub.com/site/sphelp/vorhelp.xhtmlhttp://usj.sagepub.com/content/early/2010/12/14/0042098010379281.full.pdfhttp://usj.sagepub.com/content/48/10/2049.full.pdfhttp://usj.sagepub.com/content/48/10/2049.refs.htmlhttp://www.sagepub.com/journalsPermissions.navhttp://www.sagepub.com/journalsReprints.navhttp://usj.sagepub.com/subscriptionshttp://usj.sagepub.com/cgi/alertshttp://www.urbanstudiesfoundation.org/http://www.sagepublications.com/http://usj.sagepub.com/content/48/10/2049http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    2/20

    48(10) 20492067, August 2011

    0042-0980 Print/1360-063X Online 2010 Urban Studies Journal Limited

    DOI: 10.1177/0042098010379281

    Jie Chen is in the School of Management, Fudan University, Guoshun Road No. 670, Shanghai,200433, China. E-mail: [email protected].

    Feng Guo(corresponding author) is in the The Institute of International Finance, 1333 H St NW,Washington, DC, 20005, USA. E-mail: [email protected].

    Aiyong Zhu is in the Department of Economics, Mannheim University, Mannheim, Germany.E-mail: [email protected].

    The Housing-led Growth HypothesisRevisited: Evidence from the ChineseProvincial Panel DataJie Chen, Feng Guo and Aiyong Zhu

    [Paper first received, November 2009; in final form, June 2010]

    Abstract

    This paper investigates the relationship between housing investment and economicgrowth in China, utilising province-level data via recently developed panel unit roottests and panel co-integration analysis. The empirical results suggest a stable long-run relationship between housing investment, non-housing investment and GDP inChina. In addition, the existence is demonstrated of bi-directional Granger causalitybetween housing investment and GDP at the national level. The panel error correctionmodel supports the hypothesis that housing investment appears to have a dual role,acting as both a driver and follower in Chinas economic fluctuations. However, the

    impacts of housing investment on economic growth in three sub-regions are foundto be strikingly different.

    growth (Weissman, 1955; Harris and Gillies,1963)? Are there any extra returns or disad-vantages of investments in housing comparedwith non-housing investment in the process

    of economic growth (Hendershott, 1989;Harris and Arku, 2006)? Even after decades ofdiscussion and analysis, no conclusive answersto these questions have been reached yet.

    1. Introduction

    The relationship between housing investmentand economic growth has long been a popularissue of debate in the literature of economicdevelopment. Should a less-developed coun-

    try encourage housing improvement as a partof economic development strategy (Turin,1978)? Or is massive-scale housing improve-ment just a necessary outcome of economic

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    3/20

    2050 JIE CHEN ET AL.

    Recent discussions about the relationshipbetween housing investment and economicgrowth have focused on the direction of rela-tionship between each other: whether change

    of housing investment is a leading indicatoror a follow-up of fluctuations in economicgrowth. Empirical studies of this issue includeGreen (1997), Coulson and Kim (2000), Lean(2001), Kim (2004), Wigren and Wilhelmsson(2007) and Leamer (2007). While each mayprovide a partial answer to the question, theirempirical results contradict each other.

    With the global economic recession and

    growing importance of the Chinese economyin the world, Chinas economic stability hasattracted mounting attention. In fact, inChina, much policy attention at the centralgovernment level has been focused recentlyon housing price stability at the national level(Chenet al., 2009). Thus an investigation ofhow Chinas housing investment is related tofluctuations in economic growth could haveimplications for readers. Another attraction of

    this study is that the vast regional disparitiesin China also provide a unique opportunityto examine how the relationship betweenhousing investment and economic growth isrelated to the development levels of regionaleconomies.

    This paper is intended to fill a gap in theburgeoning housing investment literature byexamining recent Chinese data. We employGranger causality tests within a panel co-integration framework to assess the relation-ship between economic output and housinginvestment at the level of quarterly provincialdata. This approach appears well suited toaddress some of the theoretical ambiguitiessurrounding housing investment in develop-ing countries including the direction of causa-tion. Our empirical results demonstrate thatthere is a stable long-run relationship between

    housing investment, non-housing investmentand GDP in China. However, the long-runelasticity of GDP with respect to housinginvestment varies across three sub-regions.

    The rest of the paper is organised as follows.In the next section, we start with the theoreti-cal discussions and then review some relevantliterature. Section 3 provides the method-

    ological issues. Descriptions of the datasetand some preliminary analysis are given insection 4. We provide the empirical resultsand their implications in section 5. The lastsection summarises the main findings anddraws some relevant policy implications.

    2. Theoretical Discussion andLiterature Review

    Policy-makers in many developing coun-tries are often puzzled whether they shouldformulate their policies to boost housinginvestment in order to promote economicgrowth. The housing industry was widelyconsidered to have lower returns comparedwith manufacturing and infrastructure indus-tries for a long time after World War II. Forexample, Weissman (1955) and Harris and

    Gillies (1963) suggested housing investmentas a social expenditure, and this could causea reduction in economic growth. Drewer(1980) argued that, while some componentsof the construction industry may stimulateeconomic growth, some others are just con-sequences of economic growth. Even in theUS, Hendershott (1989) suggested that alarge misallocation of capital towards housinginvestment instead of plant and equipmentinvestment would result in a concomitantdrag in GDP.

    However, since the 1980s, the view thathousing investment could be a contributorto economic growth has been increasinglyaccepted; not only is the housing and construc-tion industry now suggested as a major eco-nomic activity with large multiplier effects, butalso housing development is widely believed

    to be associated with many external social andeconomic benefits (Harris and Arku, 2006). Avariety of authors have looked at the impact ofhousing investment through macroeconomic

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    4/20

    HOUSING-LED GROWTH IN CHINA 2051

    factors such as employment, householdsavings, total investment and labour produc-tivity; examples include Phang (2001) andArku and Harris (2005), among others.

    As shown by Matsuyama (1990), the impactsof housing investment on aggregate economicactivities are fundamentally different fromthe non-housing counterpart. In fact, Leamer(2007) emphasised that residential invest-ments are the first GDP component to softenbefore a recession and provides evidencethat eight of the ten recessions in the USwere preceded by sustained and substantial

    problems in housing, and there was a moreminor problem in housing prior to the 2001recession. He suggested that GDP must pro-vide long-run guidance to both housing andnon-housing investments.

    The housing investment could also besensitive to financial market conditions, inparticular the variability of the supply andcost of credit over the cycle and the roles ofinterest rates and credit controls as instru-

    ments of monetary policy. Iacoviello (2005)found that a financial accelerator effect arisesin the household sector via house prices whenhouseholds ability to borrow depends on thevalue of housing collateral; hence monetarypolicy shock is one of the main causes ofthe downturn in residential investment. Inaddition, if the investment is largely drivenby individual households, and if housingfinance is lacking, housing investment willprobably be pro-cyclical and lag GDP, con-sumption and non-housing investment.1Case and Quigley (2008) summarised threerelated mechanismsi.e. wealth effects,income effects and effects through financialmarketswhich can govern the propagationof changes in the housing market throughoutthe rest of an advanced economy. They con-cluded that the fading boom in the US housing

    market leads to reductions in consumer spend-ing, which in turn affect economic growth.

    Therefore, the empirical evidence investi-gating the direction of relationship between

    housing investment and economic growthremain inconclusive. Green (1997) docu-mented that only residential investment canGranger cause GDP, while non-residential

    investments are Granger caused by GDP.Coulson and Kim (2000) also confirmedGreens result that residential investmentGranger causes consumption expenditureand the impacts of residential investment onGDP are far more pronounced than those ofnon-residential investments.

    On the other hand, Kim (2004) found thathousing investment is not a driver of GDP but

    a follower of the fluctuations of the Koreaneconomy, while non-residential investmentis found to be both a driver and a follower ofmacroeconomic fluctuations. By employingthe data from 14 European countries, Wigrenand Wilhelmsson (2007) reached the conclu-sion that GDP Granger causes total construc-tion in the short run, but not vice versa, andthat public infrastructure policies have aneffect on short-run economic growth but only

    a weak effect in the long run. Furthermore,housing construction does have a long-runeffect on GDP growth. In addition, Lean(2001) found that there exists a bi-directionalcausal relationship between constructionactivities and GDP in Singapore.

    In summary, the literature shows that theleading/lagging linkages between housinginvestment and economic growth vary fromcountry to country. The dynamics of housinginvestment are also affected by national sav-ing rates, since relatively large downpaymentsfrom personal savings have traditionally beenrequired prior to investment, and also byexpectations of future income in relation tomortgage repayment commitments.

    Today, housing investment has become amajor economic activity with far-reachinginfluences in the Chinese economy.2On the

    topic of housing investment and economicgrowth in China, there exist a very smallnumber of works written in Chinese (forexample, Zhen, 2003; Shen and Liu, 2004).3

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    5/20

    2052 JIE CHEN ET AL.

    Furthermore, the existing empirical studiesare all conducted on national-level data, whereaggregation bias could be problematic. DelNegro and Otrok (2007), for example, docu-

    ment that state and regional factors, ratherthan national factors, drive the majority ofthe movement in housing investment. Glaeserand Gyourko (2007) find that time dummiesexplain only about a quarter of the variationin US city-level house price changes, suggest-ing that most of the variation in house pricescomes from city-specific factors. Similarly,for a large country like China, which faces

    regional disparity of economic developmentin both levels of wealth and business oppor-tunities, it is expected that the relationshipbetween housing investment and economicgrowth varies to a certain extent across theregions. The provincial-level data enable us toidentify empirical regularities in the relation-ship between housing and economic growthmore robustly than with national data alone.

    3. Empirical Models

    The estimation of interaction between hous-ing investment and economic growth consistsof three steps. First, we proceed with panelunit root tests for each variable. Two statisticsinvolving the ImPesaranShin (IPS) as wellas the Hadri LM statistics for heterogeneouspanel data are employed. Secondly, we test fora co-integration relationship among the paneldata with the test developed by Pedroni (2000,2004), which allows for different individualeffects and cross-sectional interdependency.4The long-run relationship is also estimatedby using the FMOLS technique for the panel.Thirdly, we perform a Granger-causality testin a panel co-integration framework, whichis important to check the short-run interac-tion effects between housing investment and

    economic growth.

    3.1 Panel Unit Root Tests

    We first employ the IPS panel unit root testproposed by Im et al.(2003) in the multivariate

    framework, which is robust to the presence ofpossible endogeneity among variables, as wellas serial correlation of the estimated errors.

    In addition, Hadri (2000) argued that the

    null should be reversed to be the stationaryhypothesis in order to have a stronger powertest. The Hadri panel unit root test has a nullhypothesis of no unit root in any of the seriesin the panel and is based on the residualsfrom the individual OLS regressions of yit

    on a constant, or on a constant and a trend.Thus, we also employ Hadris Lagrange mul-tiplier (LM) statistic in panel unit root test

    for robustness.3.2 Panel Co-integration Tests

    Pedroni (2000, 2004) extends the EngleGranger (1987) framework to tests involvingpanel data. Consider the following co-inte-grated system for a panel of i = 1, ..., Nmembers

    (1)

    where,yitand xitare the observable variableswith dimension of (NT) 1 and (NT)m respectively. Both are assumed to beintegrated of order onefor example, I(1).The parameters aitand diallow for the pos-sibility of member-specific fixed effects anddeterministic trends respectively. The slopecoefficients biare also permitted to vary byindividuals.

    Pedroni (2000, 2004) develops asymp-totic and finite sample properties of testingstatistics to examine the null hypothesis ofnon-co-integration in the panel. Two typesof test are suggested. The first type is basedon the within-dimension approach, whichincludes four statistics: panel n-statistic,panel r-statistic, panel PP-statistic and panelADF-statistic. The second type is based on

    the between-dimension approach, whichincludes three statistics: group r-statistic,group PP-statistic and group ADF-statistic.

    In this paper, we run the following regressionfor testing co-integration in the panel data

    y t xit it i i it it = + + +

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    6/20

    HOUSING-LED GROWTH IN CHINA 2053

    (2)

    where, GDP is the gross domestic product,

    representing economic growth in China; HI ishousing investment; and NHI is non-housinginvestment.

    This allows for co-integrating vectors ofdifferent magnitudes between the differentprovinces in China, as well as individual (ai)fixed effects. We follow the same procedurein Pedroni (2000) not include the determin-istic trends with the appropriate lag lengthdetermined by the NeweyWest method. Theapproach is to obtain residuals from equation(1) and then to test whether residuals are I(1)by running the auxiliary regression

    (3)

    or

    (4)

    for each cross-section. The Pedroni panel

    co-integration statistic N T, is constructedfrom the residuals from either equation (2)or equation (3). The standardised statistic isasymptotically normally distributed

    N T N

    N,

    ( , )

    0 1 (5)

    where, and n are Monte-Carlo-generatedadjustment terms.

    3.3 Short-run and Long-run Granger-causality Tests

    Granger (1988) pointed out that, if there is aco-integrating vector among variables, theremust be at least one uni-directional Granger-causality among these variables. In addition,when the series are I(1) but co-integrated, the

    Granger causality test should not be appliedin the common VAR specification but mustbe carried out in the framework of ECM(error correction model), as follows (Engleand Granger, 1987)

    (6)

    (7)

    (8)

    where, denotes first difference; kis the laglength optimally chosen; and ECT is the error-correction term derived from the long-runco-integrating relationship.

    Different from the ordinary Granger-causality tests that applied in the VAR speci-fication, the ECM enables us to distinguishbetween short-run and long-run Granger-causality and also provides a weak exogene-ity test of the dependent variable in eachequation. Since the coefficients of laggedfirst-differenced terms capture the short-rundynamics of the system, testing the sum sig-nificance of each explanatory variable condi-tional on the optimum lags in each equation,we can evaluate the short-run Granger-causality in each equation (Toda and Phillips,

    1994). More exactly, we apply the joint WaldF-test on H k0 12 0:h = or H k0 13 0:h = for all k in equation (6), H

    k0 22 0:h = or

    H k0 23 0:h = for all k in equation (7) andH k0 32 0:h = or H k0 33 0:h = for all k in

    ln ln

    ln

    GDP HI

    NHI

    it i i it

    i it it

    = + +

    +

    it i it it

    = +1

    it i it ij it j j

    p

    itei

    = + + =

    11

    = + +

    +

    +

    ln

    ln

    ln

    GDP ECT

    GDP

    HI

    it i it

    k

    k

    it k

    kk

    it k

    1 1 1

    11

    12

    133 1k

    k

    it k it NHI +ln

    = + +

    +

    +

    ln

    ln

    ln

    HI ECT

    HI

    GDP

    it i it

    k

    k

    it k

    k

    k

    it k

    2 2 1

    21

    22

    23kk

    k

    it k it NHI +ln 2

    = + +

    +

    +

    ln

    ln

    ln

    NHI ECT

    NHI

    GDP

    it i it

    k

    k

    it k

    k

    k

    it k

    3 3 1

    31

    32

    333 3k

    k

    it k it HI +ln

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    7/20

    2054 JIE CHEN ET AL.

    equation (8). This cross-equation restrictionis properly tested using the likelihood ratiotest. Under this point of view, causality canbe derived through the asymptotic c2distri-

    bution, with degrees of freedom equal to thenumber of restrictions in the system.

    Meanwhile, since the ECT captures thedeviation from long-run equilibrium betweenco-integrated variables, the t-statistics for theparameters of the lagged ECT in each equa-tion can indicate the speed of adjustmentsof the variables in this long-run relationship.This implies the significant level of long-run

    Granger-causality in each equation (Todaand Phillips, 1994). Therefore, we are able todetermine whether the shocks from housingdevelopment or non-housing investmentGranger-cause changes in Chinas economicgrowth or vice versa.

    Finally, as suggested by a number of recentworks (for example, Asafu-Adjaye, 2000), weimplement the joint Wald F-test for hypoth-eses for the interactive terms between ECT

    term and explanatory variables. Thus, wecan detect whether to incorporate a variableinto the VECM, in order to avoid possiblespurious results due to the omission of somerelevant variables. If, for example, l1togetherwith both h12and h13for all kare insignifi-cant in the GDP equation, we can say thatboth housing and non-housing investmentdo not Granger-cause GDP in the long run.Meanwhile, it also implies that GDP is weaklyexogenous to both housing and non-housinginvestment. This joint test is also aimed toindicate which variable plays the short-runcorrection when the long-run equilibriumof the system is disturbed by a shock (Asafu-Adjaye, 2000).

    4. Data and Preliminary Analysis

    We investigate this postulated frameworkby looking at the data for housing invest-ment and GDP at the Chinese provinciallevel. The quarterly seasonally adjusted data

    series run from 1999q1 through 2007q4 forthree variables representing the logarithmsof housing investment (lnHI), non-housinginvestment (lnNHI) and GDP (lnGDP).5In

    the empirical analysis, all the data are deflatedand measured at the 1999 price to obtainreal variables. All the province-level areas areincluded in the analysis with the exceptionof Tibet. The reason for choosing 1999 as thestarting-point in our study is mainly becauseurban housing was considered as a welfaregood and was allocated by the governmentbefore 1998 (Zhen, 2003). The selection is

    also subject to the availability of data at theprovincial level in China.The choice of housing investment series is

    obviously crucial in the studies. In line withShen and Liu (2004), we employ the grossfixed capital formation in the real estate sectorto proxy for housing investment in this paper.6Non-housing investment is what is left afterhousing investment has been subtracted fromtotal investment in fixed assets.7All data are

    drawn from the China National Bureau ofStatistics. Housing investment as a percentageof GDP rose rapidly from 4.6 per cent in 1999to 9.8 per cent in 2007, reflecting the deepen-ing of market-oriented housing reform andthe fast accumulation of savings in China. Bycomparison, residential investment accountsfor a smaller share of GDP in most OECDcountries, as shown in Table 1. It is also foundthat the shares of housing invesment are largerin some countries like Norway, Portugal andIreland, which may suggest that China stillhas scope for futher growth in the housinginvestment share of GDP.

    In order to avoid the potential aggregate biasin the national-level data, and hence fit withthe actual relationships that have occurredin different parts of China, we also run theempirical tests by clustering the provinces

    which are at a similar development stage. Inline with Chinas National Bureau Statistics,we geographically decompose the nation intothree sub-regions as shown in Table 2east

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    8/20

    HOUSING-LED GROWTH IN CHINA 2055

    (11 provinces), middle (9 provinces) andwest (11 provinces).8With a few exceptions,almost all the provinces in the east region are

    regarded as the leading prosperous region,while nearly all provinces in the western partare among the poorest. We can then proceedto explore how the relationship betweenhousing and economic growth varies withthe degree of economic development acrossthe regions.

    Figure 1 suggests that the provincial-levelratios of housing investment to GDP are cor-related with the levels of GDP per capita in akind of inverted U-shape, except that Beijingand Chongqing appear as outliers. In Table2, we find that Shanghai and Qinghai are theonly two cases to encounter negative growthwith regard to the ratio of housing investmentto GDP during the sample period. WhileQinghai is one of the poorest provinces inChina, with one of the lowest HI/GDP ratios,Shanghai has the highest GDP per capita at the

    province level. This is in line with the findingsof Burns and Grebler (1977), who state thatthe ratio of housing investment to GDP islinked to the stage of economic development

    in an inverted U-shape. Although the BurnsGrebler inverted-U curve is as yet a purelyempirical finding and no theoretical model

    has been proposed to explain it, the ideabehind the phonemenon is quite intuitivelystraightforward. The HI/GDP ratio first riseswith the increase in GDP per capita whenthe economy is taking off and the demandfor larger and better shelter is fuelled by theincreasing per capita income. However, thisratio is expected to reach a peak when theeconomy enters the middle-income periodand then tends to decline when the economybecomes mature and the society has alreadyaccumulated sufficient housing stock.

    Table 2 presents descriptive statistics ofeach variable for each province as well asthe whole country. The growth rate of hous-ing investment is higher than the economicgrowth rate for all the provinces in China.Interestingly, the discrepancy between thetwo growth rates is usually larger for the

    provinces in the middle and western regionsthan those in the eastern region. In addition,we find that the overall standard deviation ofgrowth rates of housing investment is much

    Table 1. Housing investment as a percentage of GDP, 19972006

    1997 2006 Mean (S.D.) Lowest Highest

    Brazil 8.6 6.7 7.79 (0.79) 6.7 8.8

    China 4.0 9.2 6.37 (1.91) 4.0 9.2Denmark 3.9 5.8 4.38 (0.68) 3.8 5.8France 4.1 4.9 4.27 (0.27) 4.1 4.9Germany 7.4 5.4 6.25 (0.84) 5.2 7.4Ireland 6.7 14.6 9.91 (2.81) 6.7 14.6Italy 3.9 4.4 3.93 (0.24) 3.7 4.4Japan 5.0 3.6 4.13 (0.39) 3.6 5.0Korea 6.5 5.6 5.42 (0.65) 4.3 6.5Norway 14.3 13.3 13.09 (1.52) 11.7 16.4Portugal 13.6 10.8 12.86 (1.15) 10.8 14Spain 4.7 9.3 6.93 (1.64) 4.7 9.3Sweden 1.4 3.2 2.06 (0.61) 1.4 3.2

    Turkey 7.6 6.6 5.37 (1.55) 3.2 7.6UK 2.9 4.2 3.30 (0.49) 2.8 4.2US 4.2 5.8 5.06 (0.67) 4.2 6.2

    Sources: China National Statistics Bureau; OECD Factbook 2008: Economic, Environmental and SocialStatistics.

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    9/20

    http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    10/20

  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    11/20

    2058 JIE CHEN ET AL.

    5.2 Panel Co-integration Tests

    For all four panels in Table 4, the group ofpp-statistics clearly rejects the null of noco-integration at the 1 per cent significancelevel. It also reveals that the hypothesis of azero co-integrating vector cannot be rejectedat the 10 per cent significance level by allthe group statistics, except for the group

    r-statistics in the east region sub-panel andall group ADF statistics. Confirmation ofco-integration implies that the variables inthe system move together closely in the long

    run, in spite of drifting individually. Hence,a plausible characterisation of the data is

    that there is a stable long-run relationship

    between GDP, housing investment and non-

    housing investment for the whole of China

    as well as its three separate regions.

    The next step is to gauge numerically the

    long-run relationships. In order to control the

    effect of endogeneity, we apply the FMOLS(fully modified least squares; Philips, 1995)

    technique to estimate the coefficient for

    heterogeneous panels, which corrects the

    Table 3. Panel unit root tests

    Variable

    IPS test

    (p-value)

    HADRI test (p-value)

    Z(MU) Z (TAU)

    Whole country

    lnHI 0.829 0.0000*** 0.2330lnHI 0.000*** 1.0000 0.9982lnGDP 0.310 0.0000*** 0.9984lnGDP 0.000*** 1.0000 0.9965lnNHI 0.999 0.0000*** 0.7200lnNHI 0.000*** 1.0000 0.9982

    East region

    lnHI 0.971 0.0000*** 0.0033***lnHI 0.001*** 0.9889 0.9605lnGDP 0.571 0.0000*** 0.9326lnGDP 0.000*** 0.9936 0.9628lnNHI 0.998 0.0000*** 0.1318lnNHI 0.008*** 0.9960 0.9799

    Middle region

    lnHI 0.969 0.0000*** 0.8465lnHI 0.004*** 0.9902 0.9554lnGDP 0.244 0.0000*** 0.9597lnGDP 0.000*** 0.9839 0.9226lnNHI 1.000 0.0000*** 0.8116lnNHI 0.000*** 0.9853 0.9347

    West region

    lnHI 0.052* 0.0000*** 0.0096***lnHI 0.000*** 0.9729 0.9211lnGDP 0.456 0.0000*** 0.9516lnGDP 0.000*** 0.9894 0.9417lnNHI 0.087* 0.0000*** 0.6629lnNHI 0.000*** 0.9884 0.9508

    Notes: The null hypothesis of the HADRI test is that the panel data are stationary, while the nullhypotheses of IPS tests assume that the data have a panel unit root. denotes first difference ofseries. *** Rejects the null hypothesis at the 1 per cent level; ** rejects the null hypothesis at the 5 percent level; * Rejects the null hypothesis at the 10 per cent level.

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    12/20

    HOUSING-LED GROWTH IN CHINA 2059

    standard OLS for the bias by endogeneityand serial correlation effects in the regres-sors. Table 5 reports the FMOLS results of thewhole-country panel and sub-panels withouttime trends in the manner of equation (8). Thesigns of coefficients of lnHI and lnNHI are allpositive, as expected. Both the coefficients oflnNHI and lnHI are statistically significant at

    the 5 per cent level for all four panels.The elasticity of GDP with respect to hous-

    ing investment is found to be 0.18, which is

    lower than that of non-housing investmentat 0.85 at the country level. However, whenwe examine the regional level, it appears thatthe effects of housing investment in the sub-regions are decreasing with the growth rate ofHI/GDP. That is, the highest elasticity existsin the middle region, where the level of HI/GDP grows much faster than in the other two

    regions. Further, the elasticity of GDP withrespect to housing investment in the middleregion is not only significantly larger than in

    Table 4. Panel co-integration tests

    Whole countryPanel v 2.567 Group r -1.516*Panel r -2.769*** Group pp -11.004***

    Panel pp -9.378*** Group ADF 10.270Panel ADF 6.308

    East regionPanel v -1.376 Group r 0.450Panel r -0.474 Group pp -3.508***Panel pp -3.921*** Group ADF 7.055Panel ADF 4.869

    Middle regionPanel v -1.036 Group r -1.747**Panel r -2.356*** Group pp -8.437***Panel pp -6.728*** Group ADF 4.564

    Panel ADF 3.544West regionPanel v 1.907 Group r -1.440*Panel r -2.202** Group pp -7.375***Panel pp -5.690*** Group ADF 2.278Panel ADF 0.174

    Notes: All reported statistics are asymptotically distributed as standard normal under the nullhypothesis of no co-integration. The v-stat test is right-sided, while the others are left-sided.*** rejects the null of no co-integration at the 1 per cent level; ** rejects at 5 per cent level; * rejects at10 per cent level.

    Table 5. FMOLS estimates (dependent variable is lnGDP)

    lnHI lnNHI

    Country-panel 0.18*** (0.0379) 0.85*** (0.0576)East sub-panel 0.10*** (0.0400) 0.91***(0.0537)Middle sub-panel 0.37* **(0.0681) 0.25** (0.1069)West sub-panel 0.14**(0.0674) 0.94***(0.0985)

    Notes: Standard errors are shown in parentheses. ** indicates statistical significance at the 5 per centlevel; * indicates statistical significance at the 10 per cent level.

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    13/20

    2060 JIE CHEN ET AL.

    the other two regions, but also greater thanthe impacts of non-housing investment inthis region. In contrast, both the eastern andwestern regions have very large elasticity of

    GDP with respect to non-housing investment,which is almost equal to one unit.

    5.3 Panel Granger-causality Tests

    As has already been pointed out, we find astable long-run relationship between housinginvestment and GDP for the whole country.According to Panel A of Table 6, we can rejectthe null of lnGDP not Granger-causing lnHIor lnNHI at the 1 per cent significance level,and the reverse also holds. As these appearto be mutually connected with respect toshort-run relations, it can be clearly denotedas bi-directional causality between housinginvestment and GDP as well as non-housing

    investment and GDP in China. Also in thispanel, the results are different in the casebetween lnHI and lnNHI. The Wald F-teststatistics of the coefficients of lagged changes

    in lnNHI can Granger-cause lnHI at the 1per cent significance level. The reverse effect,however, seems to disappear, which is likelyto be due to the presence of exclusive com-petition between housing investment andnon-housing investment.

    The results for long-run tests show a similarpattern, given that the F-test in the last twocolumns of Panel A suggest that both lnHI

    and lnNHI Granger-cause GDP, and GDP alsoGranger-causes both lnHI and lnNHI. Thesefindings thus indicate that there exists bothshort-run and long-run bi-directional causal-ity between housing investment and GDP aswell as non-housing investment and GDP at

    Table 6. Panel Granger-causality tests

    Dependent

    variable

    Source of causation: explanatory variables

    Short-run Long-run and variables to correct disequilibrium

    (Wald F-statistics) (Wald F-statistics)

    lnGDP

    lnHI

    lnNHI

    ECT only(Coefficient)

    lnGDP,ECT

    lnHI,ECT

    lnNHI,ECT

    Panel A: countrylnGDP 4.85*** 5.03*** -0.82(0.37)** 5.55*** 4.72***lnHI 3.09*** 3.22*** -0.21(0.23) 2.74** 2.77***

    lnNHI 11.64*** 1.84 0.28(0.03) 14.56*** 1.70 Panel B: east regionlnGDP 4.11*** 1.86* -0.89(1.18) 3.81*** 2.41**lnHI 4.30*** 2.66** 1.43(0.64)** 3.98*** 2.32**lnNHI 2.33*** 1.11 0.009(0.43) 3.73*** 0.97

    Panel C: middle regionlnGDP 3.62*** 3.41*** -0.82(0.38)** 5.65*** 5.03***lnHI 0.29 0.14 0.49(3.19) 1.11 0.87lnNHI 2.46** 1.70 1.80(0.89)** 3.63*** 1.54

    Panel D: west regionlnGDP 2.13* 1.78 -0.84(0.51)* 1.80 1.45

    lnHI 1.91 1.24 0.11(0.19) 1.85 1.12lnNHI 8.07*** 1.28 0.25(0.24) 6.96*** 1.25

    Notes: Standard error in parentheses. ***, ** and * indicate statistical significance at the 1 per cent, 5per cent and 10 per cent levels respectively.

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    14/20

    HOUSING-LED GROWTH IN CHINA 2061

    the national level. This finding is in contrastto Green (1997), Coulson and Kim (2000) andLeamer (2007) for the case of US, which allindicate that residential investment appears

    to be a leader of the business cycle. However,it is consistent with what Lean (2001) foundfor Singapore and what Kim (2004) foundfor South Korea. It is noted that these Asianeconomies share a lot of similarity in thestrong government intervention in the hous-ing industry and the premature developmentof the private real estate market. In China,policy decisions can influence the develop-

    ment of the housing market in conjunctionwith market forces, even though governmentpolicies are often reactions to emerging mar-ket conditions. Housing investment has beenlargely influenced by political policy.

    In addition, based on the test statistics of theECT term in each equation, change in GDPresponds to deviation from long-run equi-librium in period t-1 but changes in hous-ing investment or non-housing investment

    do not respond. The ECT term in the GDPequation is negative as theoretically expectedand its value shows that the adjustment speedis fairly fast. Given a deviation of GDP fromthe long-run equilibrium as defined by theco-integration relationship, all three variablesinteract in a dynamic relationship to correctthe disequilibrium. The strong significance ofthe Wald F-statistics in the equations indicatesthat all three variables are endogenous inthe system. Therefore, our empirical resultssuggest that, when GDP deviates from thelong-run equilibrium, both housing invest-ment and non-housing investment jointlybear the burden of the short-run adjustmentto re-establish the long-run equilibrium atthe national level.

    5.4 Regional Disparities

    When the analysis is conducted with respectto the sub-panels in Table 6, the resultsappear that Granger-causality effects for thethree regions behave very differently from

    each other. Starting with the case of the eastregion in Panel B, it can be observed that theresults are consistent with those at the nationallevel. That is, there is a strong bi-directional

    Granger-causality between housing invest-ment and GDP as well as non-housing invest-ment and GDP in both the short run and thelong run. When housing investment deviatesfrom the long-run equilibrium, GDP andnon-housing investment together conduct theshort-run adjustments to re-establish the long-run equilibrium with a high adjustment speed.

    With regard to the case of the middle region,

    we find that only non-housing investmentappears to have strong bi-directional Granger-causality effects with GDP both in the shortrun and the long run. Meanwhile, housinginvestment Granger-causes GDP both in theshort run and the long run, but not vice versa.In this setting, changes in both non-housinginvestment and GDP do respond to deviationfrom long-run equilibrium in period t-1, butchange in housing investment does not. When

    GDP deviates from the long-run equilib-rium, housing investment and non-housinginvestment together conduct the short-runadjustments to re-establish the long-run equi-librium. On the other hand, as shown in panelC, when the non-housing investment deviatesfrom the long-run equilibrium, GDP takes thecorrections to eliminate the disequilibrium.Both adjustment speeds are remarkably high.

    Finally, Panel D in Table 6 for the westregion shows uni-directional causality fromlnGDP to lnNHI in both the short run andthe long run. The results appear not to sup-port any feedback between housing invest-ment and GDP in the long run because thehypothesis of no causality effects cannot berejected at the 10 per cent statistical level inall scenarios. A closer look at the short-termresults, however, reveals a weak uni-directional

    Granger-causality from the housing invest-ment to GDP. In addition, once GDP deviatesfrom the long-run equilibrium, it is the non-housing investment that bears the burden

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    15/20

    2062 JIE CHEN ET AL.

    of the short-run adjustment to re-establishthe long-run equilibrium in the case of thewestern region of China.

    The various degrees of Granger-causality

    relationship between housing investment andGDP across the three regions can be attributedto the roles that housing investment is playingin regional economic growth. Quigley (2008)reviews the linkages between urbanisationand economic development. Since the late1990s, housing investment in China startedto decentralise and became more resource-oriented. Hence, more developed regions have

    more resources and thus make more housinginvestment. The housing investment appearsto Granger-cause GDP only in regions thathave reached a high level of economic devel-opment, often characterised by a high level ofurbanisation (see Figure 2).

    Nevertheless, in regions with a low degreeof urbanisation, such as the west region ofChina, housing investment is less likely todrive economic growth efficiently. This is par-

    tially due to the lack of subsequent upstreamor downstream industry which helps totransmit the multiplying effect of housinginvestment. As a result, housing investmentdoes not generate increased demand in thebackward linkages from construction material

    production or the forward linkages fromhousehold durable goods in the west region,and therefore fails to Granger-cause economicgrowth in this region.

    In addition, Song et al. (2004) find thatforeign direct investment (FDI) has playedan important role in Chinese housing invest-ment. FDI pushes up the demand for urbanhousing by generating many job opportuni-ties and offering employees higher wages. Italso increases housing supply by investingdirectly in the housing sector. Thus, thelinkage between housing investment and

    economic growth is stronger in the eastregion with its larger FDI. Considering thedisparity across regions in China, there arereasons to believe that the causal relationshipbetween housing investment and economicoutput is likely to be dependent on manysocial and economic factors, which mayinclude population growth, income percapita, urbanisation, accumulated housingstock, the homeownership ratio and local

    government intervention in housing devel-opment, among others.9

    To conclude this section, we summarisethe contemporaneous causal patterns inFigure 3 to illustrate the relationshipsamong the three variables at both the

    0%

    10%

    20%

    30%

    40%

    50%

    60%

    70%

    80%

    90%

    100%

    Western part Middle part Eastern part

    Figure 2. The average ratio of urbanisation in the three sub-regions of China.

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    16/20

    HOUSING-LED GROWTH IN CHINA 2063

    national level and the three regional levels.Overall, the figure suggests the existence ofboth short-run and long-run bi-directionalGranger-causality between housing invest-ment and GDP for China as a whole and inthe east region. While both short-run andlong-run uni-directional Granger-causalityfrom housing investment to GDP exist in

    the middle region, only the weak short-rununi-directional spillover effects are found toexist in the west region.

    5.5 Robustness Check

    Demographic factors constitute an impor-tant determinant of demand for housing,especially in the long run. Campbell (1963)identifies the housing sector as being the most

    sensitive to changes in population trends andit is the swings or cycles in population thatlead to swings in housing for the US. As a fur-ther robustness check, we take demographic

    factors into account by specifying the effect ofthe population growth in per capita terms torun panel Granger-causality test to examinethe interaction of housing investment andGDP across regions.

    Overall, we find that the documented rela-tionship between housing investment andGDP as well as non-housing investment and

    GDP persists as shown in Table 7. However,the per capita housing investment appears tohave stronger bi-directional Granger-causalityeffects with per capita GDP both in the shortrun and the long run in the west region.Hence, the relatively small size of popula-tion in the west region would seem to be animportant role in explaining its weak degreeof causal relationship in terms of level between

    housing investment and economic output.The slow population growth has potentiallyreduced the importance of housing invest-ment in the western regional economy.

    Figure 3. Transmitting effects from panel Granger-causality tests.Note: thick and thin lines indicate the long-run and short-run transmission effects respectively.

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    17/20

    2064 JIE CHEN ET AL.

    6. Conclusion

    We investigate whether the massive housing

    investments in recent years have influenced

    the magnitudes of economic output in the

    three different regions having a vast disparityof economic development. By employing

    the data for 30 Chinese provinces, covering

    the period from the first quarter of 1999 to the

    fourth quarter of 2007, the following conclu-

    sions have been derived from our analysis.

    The empirical results from the panelco-integration tests suggest that there is a

    stable long-run relationship among housing

    investment, non-housing investment and

    economic growth for a cross-section ofChinese provinces, even after allowing for a

    province-specific effect. When GDP deviates

    from the long-run equilibrium, both housing

    investment and non-housing investmentjointly bear the burden of the short-runadjustment to re-establish the long-runequilibrium at the national level.

    The panel Granger-causality tests support

    bi-directional Granger-causality betweenhousing investment and GDP as well as non-housing investment and GDP at the nationallevel. It is confirmed that, for China as awhole, housing investment can Granger-causeeconomic growth in both the short run andthe long run. However, more importantly,we find that housing investment has a dualrole, acting as both a driver and a follower inChinas economic fluctuations.

    Finally, we find that the extent of trans-mission between housing investment andeconomic growth behaves substantially differ-ently across the three regions in China. In the

    Table 7. Panel Granger causality tests (all variables are measured in per capita terms)

    Dependent

    variable

    Source of causation: explanatory variables

    Short-run Long-run and variables to correct disequilibrium

    (Wald F-statistics) (Wald F-statistics)

    lnGDP

    lnHI

    lnNHI

    ECT only

    (Coefficient)

    lnGDP,

    ECT

    lnHI,

    ECT

    lnNHI,

    ECT

    Panel A: countrylnGDP 4.33*** 0.91 -0.07(0.02)*** 5.48*** 3.34***lnHI 1.54 1.15 -0.81(3.10) 1.35 1.02lnNHI 1.61 0.30 -0.65(1.12) 1.21 0.52

    Panel B: east regionlnGDP 2.77** 0.98 -0.52(0.51) 2.81** 0.78lnHI 2.78** 4.84*** -7.73(3.43)** 2.44** 3.65***lnNHI 1.20 0.53 1.14(1.93) 0.91 0.92

    Panel C: middle regionlnGDP 2.76** 3.70*** -0.97(0.17)*** 8.42*** 7.97***lnHI 1.49 1.52 0.80(1.00) 1.43 1.36lnNHI 8.04*** 2.66** -1.44(0.82)* 14.46*** 3.14**

    Panel D: west regionlnGDP 4.18*** 1.92 -0.93(0.16)*** 9.76*** 8.15***lnHI 2.37** 1.51 -0.23(0.19) 2.80*** 1.69lnNHI 7.05*** 1.22 -2.26(2.29) 11.63*** 1.31

    Notes: Standard error in parentheses. ***, ** and * indicate statistical significance at the 1 per cent,5 per cent and 10 per cent levels respectively.

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    18/20

    HOUSING-LED GROWTH IN CHINA 2065

    western region, with its low levels of prosperityand urbanisation, housing investment is lesslikely to drive regional economic growth andcontributes much less to growth than does

    non-housing investment. The relatively smallsize of the population in the western regionwould seem partly to explain the weak degreeof causal relationship between housing invest-ment and economic output.

    The evidence presented in our paper sug-gests that housing investment has a pivotalrole in the Chinese economy and that domes-tic policy attentions should be focused on the

    stability of the housing sector. In particular,it cautions that attention should be givento how the vast regional disparities withinChina affect the dynamics of the relationshipbetween housing investment, non-housinginvestment and economic growth.

    Notes

    1. We are grateful to an anonymous referee for

    pointing out this argument.2. According to the estimates of the ChinaNational Bureau of Statistics (2005), in the

    year 2000 every 100 Chinese Yuan increase inhousing investment was expected eventuallyto lead to an additional increase of 315 Yuanin GDP.

    3. Zhen (2003) found that domestic housinginvestment has significant short-run impactson GDP and a co-integration relationshipbetween housing investment and GDP does

    exist. However, Shen and Liu (2004) show thatGDP Granger causes housing investment, butnot vice versa.

    4. Recently, there is an increasing concern inthe econometrics literature that the teststatistics of typical panel unit root tests andco-integration tests may be biased whenthere is a spatial correlation across units (forexample, Breitung and Pesaran, 2008). Byapplying the cross-section dependence testdeveloped by Pesaran (2004), we examine thepresence of spatial dependence in the paneldata but find no significant spatial effects.We greatly appreciate one anonymous refereefor pointing this out. The methodology of a

    panel data model incorporating spatial effectsdeserves our further research effort.

    5. We compare two approaches: one is to let databe seasonally adjusted and another is to use three

    dummy variables to control for seasonal effect(summer, fall, winter and with spring as the baseseason). We find no substantial differences.

    6. Housing investment here includes all types ofbuilding construction (residential, office andshopping), but residential housing constructionconsistently dominates with an average shareof 80 per cent over the study period.

    7. Since there are only annual data availablefor several provinces during certain periods(Anhui, Fujian, Gansu, Guanxi, Guizhou,

    Hannan, Hebei, Henan, InnerMongila,Jiangxi, Ningxia, Qianghai, Yunan, Xinjiang),we apply the widely adopted proportionalDenton method to generate quarterly data(Denton, 1971), where the total fixed capitalinvestment of these provinces is employed asthe quarterly indicator series.

    8. The west region also includes Tibet, but Tibetis excluded in the paper due to data deficiency.

    9. This paper focuses on macroeconomicrelationships and does not consider themicroeconomic aspects of the housingindustry. This issue can be further exploredin future research.

    Acknowledgements

    The first author wishes to acknowledge financialsupport from the China National Social ScienceFoundation (07CJL006) and from the Fudan

    University 211 Project (211XK06). The authors wouldalso like to thank four anonymous referees for theirvaluable comments. The usual disclaimer applies.

    References

    Arku, G. (2006) The housing and economicdevelopment debate revisited: economic sig-nificance of housing in developing countries,

    Journal of Housing and the Built Environment,

    21, pp. 377375.Arku, G. and Harris, R. (2005) Housing as atool of economic development since 1929,International Journal of Urban and RegionalResearch,29(4), pp. 895915.

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    19/20

    2066 JIE CHEN ET AL.

    Asafu-Adjaye, J. (2000) The relationship betweenenergy consumption, energy prices and eco-nomic growth: time series evidence from Asiandeveloping countries, Energy Economics, 22,

    pp. 615625.Ball, M. and Morrison, T. (2000) Housing invest-ment fluctuations: an international comparison,Housing, Theory and Society, 17, pp. 313.

    Breitung, J. and Pesaran, H. (2008) Unit rootsand cointegration in panels, in: L. Matyas andP. Sevestre (Eds) The Econometrics of PanelData: Fundamentals and Recent Developments

    in Theory and Practice, pp. 279322. Dordrecht:Kluwer Academic Publishers.

    Burns, L. and Grebler, L. (1977) The Housing of

    Nations: Advice and Policy in a ComparativeFramework. London: Macmillan.

    Campbell, B. (1963) Long swings in residentialconstruction: the postwar experience, AmericanEconomic Review, 53(2), pp. 508518.

    Case, K. E. and Quigley, J. M. (2008) How hous-ing booms unwind: income effects, wealtheffects, and feedbacks through financialmarkets, European Journal of Housing Policy,8, pp. 161180.

    Chen, J., Guo, F. and Zhu, A. (2009) Housingwealth, financial wealth and consumptionin China, China & World Economy, 17(3),pp. 5774.

    China National Bureau of Statistics (2005) Apreliminary analysis on the effects of housingindustry on national economy, ManagementWorld, 11 [in Chinese].

    Coulson, N. E. and Kim, M. S. (2000) Residentialinvestment, non-residential investment andGDP, Real Estate Economics, 28(2), pp. 233247.

    Davis, M. A. and Heathcote, J. (2005) Housingand the business cycle,International EconomicReview, 46, pp. 751784.

    Del Negro, M. and Otrok, C. (2007) 99 Luftballons:monetary policy and the house price boomacross U.S. states,Journal of Monetary Economics,54, pp. 19621985.

    Denton, F. T. (1971) Adjustment of monthly orquarterly series to annual totals: an approachbased on quadratic minimization,Journal of theAmerican Statistical Association, 66, pp. 99102.

    Drewer, S. (1980) Construction and development:a new perspective,Habitat International, 5,pp. 395428.

    Engle, R. and Granger, C. W. J. (1987) Cointegrationand error correction: representation, estimation,and testing, Econometrica,55, pp. 251276.

    Glaeser, E. L. and Gyourko, J. (2007) Housing

    dynamics. Discussion Paper No. 2137, HarvardInstitute of Economic Research.Granger, C. W. J. (1988) Some recent develop-

    ments in a concept of causality, Journal ofEconometrics, 39, pp. 199211.

    Green, R. (1997) Follow the leader: how changesin housing and non-housing investment pre-dict changes in GDP, Real Estate Economics,25, pp. 253270.

    Hadri, K. (2000) Testing for stationarity in het-erogeneous panel data, Econometric Journal,

    3, pp. 148161.Harris, R. and Arku, G. (2006) Housing and

    economic development: the evolution of anidea since 1945, Habitat International, 30,pp. 10071017.

    Harris, W. D. and Gillies, J. (Eds) (1963) CapitalFormation and Housing in Latin America.Washington, DC: Pan American Union.

    Hendershott, P. (1989) Comments on socialreturns to housing and other fixed capital,

    Journal of the American Real Estate and Urban

    Economics Association, 17, pp. 212217.Iacoviello, M. (2005) House prices, borrowing

    constraints and the monetary policy in thebusiness cycle, American Economic Review,95(3), pp. 739764.

    Im, K. S., Pesaran, M. H. and Shin, Y. (2003) Testingfor unit roots in heterogeneous panels,Journalof Econometrics, 115, pp. 5374.

    Kim, K. H. (2004) Housing and the Koreaneconomy, Journal of Housing Economics, 13,

    pp.321341.Leamer, E. E. (2007) Housing is the business cycle.

    Working Paper No. 13428, NBER, Cambridge,MA.

    Lean, C. S. (2001) Empirical tests to discern link-ages between construction and other economicssectors in Singapore, Construction Managementand Economics, 19(4), pp. 355363.

    Matsuyama, K. (1990) Residential investment andthe current account, Journal of InternationalEconomics, 28, pp. 137153.

    Pedroni, P. (2000) Fully modified OLS for het-erogeneous cointegrated panels, Advances inEconometrics, 15, pp. 93130.

    at University of Groningen on May 17, 2014usj.sagepub.comDownloaded from

    http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/http://usj.sagepub.com/
  • 8/11/2019 Urban Stud-2011-Chen-2049-67

    20/20