城市住房价格波动差异及连锁反应研究

【作者】 张凌；

【导师】 贾生华；

【作者基本信息】 浙江大学 ， 企业管理， 2008， 博士

【摘要】 住房市场的价格波动不仅影响到人们的居住质量和生活水平,对家庭财富总量也有很大影响,并且关系到国民经济发展和社会和谐稳定。近十年来,随着住房分配制度改革的深化,我国住房市场得到快速发展,城市住房价格也同时大幅度上涨。从全国范围看来,一些沿海城市房价水平高,住房价格上涨尤为突出,其波动更容易引起全国人民的广泛关注。而内陆城市总体房价水平低,部分城市近年正呈追赶之势。那么,为何沿海城市房价上涨更快?一些内陆城市近年上涨幅度增大是否受其影响?房价波动的影响因素和原因是什么?住房市场价格波动是否受到心理预期的影响?目前已有学者对我国城市住房价格水平的影响因素,住房价格的投机波动特点等进行了实证研究,但较少关注城市间房价的波动差异。国外住房市场实证研究中对连锁反应的存在与否不同国家和地区有着不一致的结论,国内的新闻舆论中也有各地房价“联动”与“轮动”一说,但目前尚缺乏规范的经验研究。本文试图探讨城市间住房价格波动的差异,并考察我国住房市场是否存在价格的连锁反应。在梳理了国内外学者关于住房价格、住房价格波动相关研究的基础上,本文建立了通过自相关和空间相关两个维度考察城市住房价格波动的分析框架,解释住房价格波动的产生机制。通过收集1995年以来我国35个大中城市住房价格和城市经济数据资料,对我国城市住房价格波动的差异及联系进行了实证研究。在住房价格短期波动差异研究中,在城市房价水平影响因素分析的基础上,通过构建类似和标准误差修正模型,考察房价短期波动的决定因素,分析了房价波动自相关和均值回复表现差异的原因。在房价连锁反应实证研究中,构建了各城市和全国房价的协整和误差修正模型,通过不同城市对全国基本面因素变动的反应差异,各城市相对全国平均房价的偏离趋势,考察是否存在房价连锁现象。并通过几个热点城市的房价协整关系检验和Granger因果检验,考察这些城市间是否存在房价的空间交互作用。为了更好地解释宏观经济现象背后的原因,本文还对三个代表城市进行了住房市场心理及行为因素调研,于2007年10月在杭州、武汉和无锡三个城市各发放调研问卷400份,对有购房意愿者了解其价格预期,购房动机,对住房市场的看法,房价上涨的原因等,在此基础上分析了价格预期的形成过程。本文的主要研究结论如下:(1)我国35个大中城市的住房价格波动受基本面因素波动和短期波动调整过程共同影响。通过房价波动决定因素的多元回归模型和经过严格单根检验的标准误差修正模型两种方法,得到较为一致的结论。从长期看,我国城市住房价格水平由收入、人口、建设成本、人口密度等因素决定。房价的波动,除受这些基本面因素波动影响外,还有短期波动调整过程,包括房价变化的自相关和修正与基本面所决定的均衡房价偏离的均值回复。(2)不同城市的短期波动调整过程存在较大差异。沿海开放城市房价变化自相关现象显著,而内陆城市房价变化自相关不显著,有很强的均值回复倾向。本文通过构建反映基本面因素与短期波动调整交互作用的模型,分析了自相关和均值回复现象的可能原因。收入越高的城市,房价变化的自相关现象越明显;人口密度越大的城市房价变化自相关性越强。更高的收入增长,导致更大的均值回复;建设成本增加越多,房价变化的均值回复现象越小。(3)我国城市间房价存在弱的连锁反应。本文将35个大中城市按照地理位置划分为华东、华南、华北、东北、中部和西部六个区域。除华东地区外,其它各区域的城市房价与全国平均房价比值有较强的回复到长期均衡值的倾向,而在短期各地对于全国因素冲击确实有快慢不同的反应。全国收入水平提高时,华东地区房价产生最大的正反应;利率发生变化,也是华东地区房价产生更大的负效应;货币供应量变化时,华北地区房价有较大反应;建设成本变化时,中部、华南、东北和西部的反应都低于全国平均水平,西部的相对偏离最大。连锁反应呈多“震中”的弱型关联。(4)城市间房价确实存在空间交互作用。本文利用北京、上海、天津、广州、深圳、重庆等六个城市的中房住宅指数季度数据,构建向量误差修正模型,城市房价指数间的协整关系和Granger因果关系表明存在房价空间交互作用。但房价的空间扩散路径与是否邻接没有明显关联。从这六个城市来看,存在热点城市之间,以及从东部沿海城市到西部内陆城市的房价扩散。房价扩散的原因可能有资本流动、信息扩散、预期和地区结构差异,后两者是主要决定因素。(5)房价变化自相关性强的城市在空间上也处于领先波动地位。连锁反应中领先房价周期的华东、华北、华南地区正好都是房价变化自相关性强的沿海城市。这种一致性是因为两种现象的根本原因是相同的,好的经济基础和宜人性是这些地区领先增长的必备条件和长期增长预期的基础,而投资意识和后视预期是短期正自相关和领先增长的主要动力。对杭州、武汉、无锡三个代表城市住房市场微观心理和行为因素的调研表明,价格预期的形成过程反映了房价变化自相关现象的产生,也与连锁反应表现一致。(6)要充分认识城市间房价波动在时间和程度上的差异,讨论房价波动和制定调控政策不能忽视地区差异一概而论。沿海开放城市保持住房市场稳定发展关键就是要杜绝人为炒作,控制投机行为,注重正确的舆论导向,加大市场透明度。内陆城市住房市场要健康培育,既不能无端受牵连打击,也不能盲目看齐发达城市,忽视自身基本面。中央政策制定者应充分考虑其所处的周期发展阶段,处于市场起步阶段时不可与繁荣的沿海住房市场同等对待。地方政府则应该防止市场上升期的过度投机。资本流动和信息扩散在城市房价自相关和空间相关过程中都起重要作用,要防止价格出现激烈波动,应防止投机资金和热钱肆意流动,媒体报道注重风险预防。与已有研究相比,本文的主要学术创新价值有如下几点:(1)通过构建从房价变化自相关和空间相关两个维度考察住房价格波动的研究框架,对住房价格波动的产生机制进行了系统分析,对我国35个大中城市的住房价格波动差异和联系进行了实证研究。国内住房价格研究中有关注具体热点城市的房价波动特点,也有对住房价格影响因素的一般研究,但较少关注城市间波动差异,更缺乏同时关注住房价格波动差异和联系的系统研究。本文采用面板协整和误差修正模型,分析住房价格在自身基本面因素之外的短期调整波动,以及偏离全国房价变动的短期领先波动,发现两者之间确实存在关联性,领先波动的城市往往也是波动自相关性强的沿海城市。对房价波动的系统研究,有利于拓展住房价格异常波动以及住房价格泡沫研究的思路,丰富了我国现有住房价格波动研究,探索性的研究发现为后续相关研究提供了参考和借鉴。(2)采用宏观计量分析和微观心理调研相结合的研究方法。相对于一些研究直接将房价波动的自相关称为投机和泡沫因素,或对房价连锁反应做出一些主观解释,本文作了更细致的工作。首先在价格波动的误差修正模型中,考察了基本面因素与价格波动自相关和均值回复的交互作用,揭示了哪些城市因素影响了自相关和均值回复的强度。在房价连锁反应研究中,对我国住房市场可能的连锁原因进行了定性研究。然后通过对典型城市住房市场购房心理和行为调研,根据对三城市总共1132份有效问卷价格预期形成过程的分析,得出与宏观计量研究一致的结论,发现在我国现阶段预期是影响房价波动的重要原因。这种宏观计量分析基础上,辅以多城市的同期问卷调查对比研究,以及通过开放式问题了解购房者心态的调研方式在我国房地产研究中还鲜有先例。(3)通过结构系数异质和空间交互作用两个维度对我国住房市场的房价连锁反应进行了实证研究。我国近年由于很多城市房价轮番上涨,关于房价“联动”与“轮动”的说法很多,但缺乏规范的实证研究。本文通过构建反映城市房价与全国房价偏离的结构模型,考察基本面因素变动对于房价波动偏离的影响,由此判断是否存在“连锁”反应,另一方面,通过检验典型城市间房价的协整和Granger因果关系,考察房价空间上是否存在交互作用。两方面的证据表明我国住房市场存在弱的连锁反应,连锁的特点与国外的研究发现有所不同,呈多中心的连锁格局,价格的扩散与地理上的邻近没有直接关系,主要是热点城市之间,沿海城市到内陆城市之间的价格扩散。这种研究方法和得出的研究结论对我国学者进一步开展类似研究具有参考价值。受困于数据采集的限制,加上笔者时间和能力有限,本文尚存在一些不足之处,希望在今后的研究中能够得到完善和补充。可能的研究拓展是,随着房价数据序列的延长和面板数据方法的改进,采用更先进有效的检验方法研究面板数据的平稳性和协整关系。同时,房价变化自相关和空间相关之间的联系,以及它们背后的原因,还值得进一步的探索研究。更多还原

【Abstract】 House price dynamics will influence not only the dwelling quality and living level, but also family wealth. It is related to national economy development and social harmonious. In recent ten years, with the reform of housing allotment system, Chinese housing market has gotten rapid development. At the same time, house price has run up sharply. Compared in the whole country, there are higher price levels and more rapid growth in some coastal cities to which people always pay more attention. The house price in inland cities is lower in the mass, but some have shown strong chasing tendency. Then, why the house price rise more rapidly in costal cities? What are the factors and reasons that cause the house dynamics? Weather or not the house dynamics is affected by psychological expectation? At present, some scholars have studied the factors of house price levels in Chinese cities and some have made empirical research about speculative characters of house price. But few focused on the difference of house price dynamics among cities. The empirical research in different countries and districts have inconsistent conclusion about ripple effect. Some reportage in China alleged that there were "linkage" and "by turns" among prices in different cities. But normative empirical research is absent. This paper tries to investigate the difference of house price dynamics among cities and examine the ripple effect in Chinese housing market.After studying abroad and home studies on housing price and price dynamics, this paper constructed an analytical frame to examine house price dynamics through autocorrelation and spatial correlation and explained the causing mechanism about it. Collecting data of house prices and city economy in Chinese 35 metropolitans since 1995, the paper made an empirical study about the difference and relevancy of house price dynamics in our country. In the research of dynamic difference, based on the factors analysis of house price level, constructing similar and normal error correction model, the paper studied the determinant of short-run dynamics, and analyzed the reason of difference in autocorrelation and mean reversion. In the research of ripple effect, the paper constructed national and local error correction model of house price. According to different reaction to national fundamentals, and the departure trend relative to national average house price, the paper investigated if there was ripple effect. With the cointegration and Granger test for several hotspot cities, the paper examined the interaction of house prices. In order to better explain the diversity of the macro- economic phenomenon, the author made a survey for psychology and behavior in three representative housing markets. In October2007, we put out 400 pieces of questionnaire respectively in Hangzhou, Wuhan and Wuxi, trying to know the price expectation, buying motivation, opinion about housing market and the reason of price appreciation, and to analyze the formation of price expectation.The paper research obtains main conclusions as follows: (1)The house price dynamics in Chinese 35 metropolitans is determined together by the dynamics of fundamentals and short-run adjustment dynamics. With the method of direct regress model and normal error correction model, the paper drew consistent conclusion. House price in Chinese cities is determined by income, population, construction cost and density in the long run. Besides the effect of fundamental dynamics, the house price dynamics is influenced by shot-run adjustment dynamics, including house price changes autocorrelation and the mean reversion to the fundamentals.(2)There is obvious intercity difference in shot-run adjustment dynamics. The autocorrelation of house price changes is notable in coastal cities but unapparent in inland cities. There is strong mean reversion tendency in inland. Considering the interaction of fundamentals and adjustment dynamics, the paper examined the possible reasons for autocorrelation and mean reversion. In cities with higher income and bigger density, the autocorrelation of house price dynamics is stronger. The more the income grows, the bigger is the mean reversion; the more the cost grows, the smaller is the mean reversion.(3)There is weak ripple effect in Chinese cities. The paper divided 35 metropolitans into six parts i.e. East, North, South, West, Northeast and the Center. The deviations of regional house price from the national average are stationary—the deviations show no long-run trends-- in five regions except the East. But the short-run coefficients surely exhibit distinct spatial patterns. House price are more responsive to income changes and rate changes in the East than the national average. The money supply has a more positive effect on house price in the North. House price in the Center, South, Northeast and west are less responsive to construction cost than average, and the least is in the West. The ripple effect shows weak linkage with centers more than one.(4)The spatial interactions exist among house price in different cities. The paper used seasonal residential index (China Real Estate Index System) for Beijing, Shanghai, Tianjin, Guangzhou, Shenzhen and Chongqing to construct vector error correction (VEC) model. The results of co integration and Granger test indicated the existence of spatial interaction. But the diffusion is not obviously related to contiguous. To judging with these six cities, the house price diffuses from the east coastal to the west inland and between hotspot cities. The possible reasons for diffusion include capital transfer, information transmits, expectation and differences in regional structure, and the last two are more important.(5)Cities with stronger autocorrelation in house price changes are also those with leading dynamics spatially. The regions with leading dynamics in ripple effect are East, North and South, most are coastal cities with strong autocorrelation in price changes. The consistency is because of their common ground. Good economic fundamentals and amenities are necessary condition for leading growth and bases for long-run expectation. And investment consciousness and backward expectation are main momentum for shot-run positive autocorrelation and leading dynamics. With the survey of micro psychology and behavior for three representative cities i.e. Hangzhou, Wuhan and Wuxi, the paper found that the formation for price expectation reflects the price changes autocorrelation and is consistent with ripple effect.(6)We should fully aware the intercity difference in house price dynamics about time and extent when enacting policies. The key for the coastal to development healthily is to prohibit speculation, to orient public opinion accurately, and to increase market transparence. For the inland, it is both important to not be embroiled in suppress and to not be hoped catching up with leading cities disregarding fund-mental. The policy maker in central government should consider the cyclical phases in different cities and can’t treat underway inland market as same as boom coastal market. Local government should prevent overspectulation in the rising up. Both capital transfer and information transmission are important in autocorrelation and spatial correlation. To prevent drastic dynamics, we should prevent the smart money from flowing wantonly. The news should report the risk warning of housing market.Compared with previous research findings, the main innovation of this paper is manifested in the following aspects:(1) The paper comprehensively studied the causing mechanism of house price dynamics through constructing a research frame with two dimensions—autocorrelation and spatial correlation. Then, it made empirical research for the difference and linkage of house price dynamics in Chinese 35 metropolitans. Some house price research in China focused on dynamics in certain hotspot cities, and some research the general factors of house price. Few paid attention to intercity difference in price dynamics and even fewer studied difference and linkage of house price dynamics simultaneously. Using panel data cointegration and error correction model, this paper analyzed short-run adjustment dynamics beyond fundamentals and leading dynamics deviating from national average, and found that there were really nexus between the two. Leading dynamic cities are usually those coastal cities with strong autocorrelation. The comprehensive study for house price dynamics is useful to rich research of abnormal house price dynamics and house price bubbles, becoming an important complement to present research in our country. The tentative findings will provide reference and experience for future research.(2)This paper used a method of integrating macro econometric model and micro psychology survey in house price dynamics research. Comparing with some studies which directly called autocorrelation as speculation and bubble, and some made subjective explanation for ripple effect, the paper did more detailed. First, it added interaction items of fundamentals and autocorrelation in price dynamics error correction model, detecting the city factors which impact the strength of autocorrelation and mean reversion. In the study of ripple effect, the paper made qualitative analysis for possible reasons. Then, through the survey for house buying psychology and behavior in representative cities, the paper made an anatomy for the formation of price expectation with 1132 effective questionnaires, and drew consistent conclusions with macro econometric analysis. The paper found that expectation is important reason for house price dynamics at present in China. The method of macro econometric analysis accompanied with homochronous survey for several cities and the manner of open question for attitude are unwonted in housing research in China.(3)The paper made empirical research for ripple effect in Chinese housing market through coefficient heterogeneity and spatial interaction. Owing to the price rising in many cities in recent years, there are some sayings about "linkage"and "rise by turns" which haven’t been approved by normal research. The paper examined price deviation from average for the shock of fundamentals through constructing structure model for city’s and national house price and judge the existence of "ripple" with it. On the other hand, the paper did cointegration and Granger test for representative cities to examine spatial interaction. The two dimensions both indicate that there is weak ripple effect in Chinese housing market. The character is to some extent different from foreign countries. Ripple effect has more than one center and the diffusion exists mainly between hotspot cities and from the coastal to inland, not related to contiguity. The method and the conclusion have referenced value for future research.Restricted to time, ability and data collection, this paper has some shortcomings waiting for improvement in future. Possible development is to use more effective test method for stationarity and cointegration with the extension of time series and improvement of panel data methodology. In addition, the relationship of autocorrelation and spatial correlation and the reason behind them are worthy to be further explored.更多还原