• 141.00 KB
  • 2022-04-22 13:44:00 发布

外文参考文献译文及原文本科毕业论文.doc

  • 32页
  • 当前文档由用户上传发布,收益归属用户
  1. 1、本文档共5页,可阅读全部内容。
  2. 2、本文档内容版权归属内容提供方,所产生的收益全部归内容提供方所有。如果您对本文有版权争议,可选择认领,认领后既往收益都归您。
  3. 3、本文档由用户上传,本站不保证质量和数量令人满意,可能有诸多瑕疵,付费之前,请仔细先通过免费阅读内容等途径辨别内容交易风险。如存在严重挂羊头卖狗肉之情形,可联系本站下载客服投诉处理。
  4. 文档侵权举报电话:19940600175。
'外文参考文献译文及原文本科毕业论文1外文文献译文香港公屋内部流动的影响1简要介绍公共房屋是对房屋的成本相当大的补贴,但必须严格竞争用户之间的配给。当配给时,资源可能不被分配到他们最有价值的用途,因为人们无法有效地传达他们的偏好强度。例如,考虑两个完全相同的住房单位,另外有两个准住户分配。在单位A的附近比单位B的附近位置较为理想,所以两个准租户愿意支付的保费为单位A的租金。假设第一人也正好工作在同一街区的单位A的位置,他愿意支付更高的溢价A座给第二个准租户。在一个开放的市场,将分配一个扁平的人,谁更值,即第一个准租户。当租金补贴的水平和盖板固定在非市场手段进行分配,不过,这两个人都将会如实地揭示潜在租户他们更喜欢单位A而不是单位B。没有行政手段来确定这两个哪个是更高价值为较理想的用户单位。在实践中,住房补贴的分配,常常是补贴住房分配偏好的基础上辅以一些彩票陈述或排队机制。本文的目的是研究香港在提供公共房屋的分配使用非价格方法所引起的低效率根据分配的非价格的方法吸引了来自资源配置不当,在经济大和行之有效的文献(例如Cheung,1974;Weitzman,1977;Suen,1989;GlaeserandLuttmer,2003)。当一个人是低于市场价格的分配,销售商须有良好的生产激励不足。由此产生的不足,供应的补贴导致了良好的无谓损失。公认的事实是,对于一般竞争价格补贴导致的价值损耗。巴泽尔(1974)展示了如何通过社会平衡,可通过浪费分配给公共住房方面取得的,准租户可以答应31 他们的劳动供给决策,以免他们的收入超过了资格门槛,而现任租户可以选择留在自己的单位效率低下根据不断变化的情况,以免他们失去了宝贵的补贴。然而,一个效率低下的第三个来源是,当个人无法表达自己的意愿通过自己的喜好来支付力度,分配非价格方法一般不分配稀缺资源,价值最高的用户。如果一个单位价值港币10,000元,每月租金正在为港币2000元租来的,任何准租户的单位,其价值超过港币2000元将在其投标兴趣。不同的分配方法所采纳,平可能最终在它的人的,其价值远低于港币10,000元的手中。Deacon和Sonstelie(1989,1991)和Suen(1989)分析的福利费用从资源配给时被等候的是这样分配不当所引起。在最近的研究中,Glaeser和Luttmer(2003)研究纽约市房屋租金管制下的分配不当。他们发现,对他们的住房单位的特点住户的特征匹配是根据租金管制单位的扭曲。由于租金管制也代表了住房补贴给现有租户,Glaeser和Luttmer的(2003)的见解同样适用于政府资助房屋分析。在香港,公共房屋影响到很大一部分人口。2009年,29%的家庭在政府提供的公屋居住,另外18%股权通过居屋住房补贴和(香港房屋委员会,2010)有关的计划。公共房屋政策的批评往往集中于这些巨额补贴的财政负担强加。但较少人研究是由配给补贴居住。尽管这些费用可能低于成本的直接财政资源配置效率低下引起的有形成本,他们可以对人们的日常生活更直接的影响。在本论文中,我们侧重于对公共房屋的区位选择的扭曲所致。公共房屋福利往往牵制居民的当前位置,在公屋居民预计将不太可能转向更方便的地点或多个合适的单位,即使他们的生活情况发生了变化。这一假设是有关工作锁定在对工作流动性效应(例如,马德里安,1994;Gilleskie和Lutz,2002;亚当斯,2004年;Bansak和拉斐尔,2008)。文学考察了雇主提供的工作流动性健康保险业的影响,这一假设是,工人都不愿意跳槽,唯恐失去他们的保险。同样,我们研究的假设,即公共房屋福利阻止内部的流动性,因为住客不想失去大量住房补贴。31 城市研究在文献中,有大量的工作记录身体内部迁移的实证模式(例如,格林伍德,1985;埃利奥特,1997年;画家,1997年)。博伊尔和沈(1997)研究公众(理事会)房屋内部迁徙的影响,发现英国的公屋住户迁移的可能性较小以外的地方最近的报告中。(2005)表明,运输相关的属性,对居住区位的选择和个人更喜欢一个有重大影响的通勤时间较短的组合,降低运输成本,较低的密度和较高的办学质量的位置。在美国,公共住宅区往往集中在与贫困的高浓度(奎格利,2000年)中央城市的位置。许多人相信,在这样的社区生活限制之一的经济机会(威尔逊,1996)。为了减少经济种族隔离的公共房屋供应projectbased诱导,住房和城市发展部已经尝试过租户的租金补贴,鼓励他们从高贫困转移到低贫困地区(如为机程序移动)。约翰逊等人。(2002)报告说,家庭谁利用这些方案的优势,转移到低贫困地区社会福利的依赖经验,减少和健康状况的改善。还有证据表明,这些措施可以提高儿童的学校教育成果和减少问题行为(见Feins和谢罗德,2005;路德维希等,2005;。罗森鲍姆和Harris,2001)。除了内部的流动性不足,另外在公共房屋的区位选择的扭曲所致体现,是与居住和工作地方不匹配。租赁市场有竞争力的有效分配问题解决了房屋的单位去来的最高值的用户(库普曼斯和贝克曼,1957)。价格systemfails到performthis功能,当公屋单位的租金是固定的,对单位超额需求的配置是由非价格来确定。此外,即使在没有与居住和工作,人们的地方初步情况的变化不匹配。当情况(例如,更改到一个新的工作)订明移动到另一个领域,流动性对公营房屋住户缺乏可能会阻碍资源的优化调配。住宅及工作地点的决定是在充分研究城市经济学中的主题(例如,维克曼,1984年;西格尔,1975;辛普森,1987年;侃,2002年;侃,2003)。我们目前的一些工作的意义也就是在""""的空间错位假说凯因(1968)文献。凯恩的工作最初是涉及到如何影响住宅隔离的黑人在美国的就业机会(见也Holtzer(1991))。虽然种族居住隔离是不是香港的公共屋往往距离远的经济活动中心之间可能有居住和工作的地方不匹配的影响其实主要问题。此外,住房使用权也发挥了在确定住宅的流动性和就业的重要角色(例如,Ioannides及Jean,1996;Kan,2000;Dohmen,2005)。2008年的研究个人的工作和在英国的失业持续时间的住房使用权的影响。他们认为,自置居所,租住公屋就业和失业施加更多的限制,分别为。蒙克等。(2008)认为,自置居所对失业风险和对工资产生积极的影响负面影响。黄登(2006)表明,住房使用权的流动性一致,但对不同的效果比在中国的时间。(2004)认为,有一个拥有住房的居住迁移上给出了在荷兰的住房拥有率迅速上升的影响减少。蒙克等。2006年表明,住房拥有率的倾向阻碍移动作业寻找原因,而是提高了当地就业的机会。作为在香港资助出售单位提供更比私人部门的限制性,众家面临着位置的选择比私人同行更高的失真。这可能也有之间的居住和工作的地方之间的业主不匹配的影响。31 在本论文中,我们使用香港政府取样检验由配给住房补贴和如何分配不当影响这些住户的行为有关的内部流动和旅行到工作的资源配置效率低下引起的。实证结果显示,公屋住户更不动,而且不太可能在同一地区工作或居住地的同行比他们的私人楼宇地区。此外,有条件的移动,公屋住户更比私人楼宇的同行可能要移出原区或住处。那些地区也不太可能在同一个地方工作,他们的生活。我们的结论是基于流动性和交通行为的回归公共入住地区。我们尝试对这类行为的其他因素控制,包括对控制变量的回归在一套广泛。我们的回归模型估计很多使用数百个变量。但是,我们不能排除这种可能性,有一些团体在入住影响。今年的实证结果,本文提出应考虑到这一点需要注意的解释其余未观察到的差异。2简要回顾公共住房分配制度在香港大约有两百万人(约30%的香港人口)住在700,000公屋单位提供的住房部门(香港房屋委员会、2010b)。公有住房申请人或者其家属不得直接或间接地拥有的或者境内的财产。此外,申请者都受到收入及总净资产值范围所定下的住房的权威。房委会维持公屋轮候册申请人。从这个名单,合资格的申请人提供的公屋居所。当申请人的轮到,申请人将获得三个提供公共租住房屋(每提供一个选项)在他的选择。这个地区最多可申请取消,如果申请人不能提供可以接受的理由拒绝全部三个配屋建议(香港房屋委员会,2010)。公共租住屋分为四大区,即市区,扩展市区,新界和离岛。由于没有足够的公众在市区内,可分配给所有申请人的租住公屋单位,申请人只能选择一个由三间非市区为他们未来的住房分配一带地区。截至2010年7月,有超过13万的公屋轮候册(香港房屋委员会,2010b)的申请人。房屋委员会设置一个维持平均轮候时间在三年的目标。在这一过程中,应用程序的公有住房分配处理按照严格的应用序号和申请者的抉择的地区。然而,那些满足资格标准的明确的平的分配方案可能会有机会被分配早些时候租住公屋单位。也有独立的标准和配置方案为受灾人员擅自占用者间隙和公共住房开发价值。根据现有的公共房屋政策,居民可申请,如增加家庭人口,特殊医疗或社会的理由,或重大改进,或由房屋委员会(香港房屋委员会,2010)实施重建计划。根据不同情况调迁每当可能,合资格的住户会被转移到合适的单位在同一个村。住户也可以申请专项转移到另一个村。由于长期租住公屋单位不足,即使租户调迁的资格,轮候时间可以很长,也不能保证,新的单位将匹配租户的喜好。31 除了为租住公屋单位申请,符合条件的家庭也可申请购买经屋计划(居屋)资助单位在订定居屋价格,房委会遵循三条指导原则:(一)抵押贷款到收入的比例不得超过40%;(二)50%的单位应负担的目标群体;及(三)单位应在市场提供的价值30%的折扣。从今年开始第三次入住,可出售居屋单位业主缴付补价后,向房委会在公开市场单位。保险费基本上等于当时的市场价值由于购买力。例如,时间乘以所有人享有折扣,如果一个居屋单位市值为港币2000美元和单位是在所提供的折扣在购买30%的时间,那么保费等于600000港元。居屋单位业主也可以出售其单位对现有公屋租户。在这种情况下,原来的居屋业主将不会被要求支付该单位销售前的地价给房委会,但买家将继承的法律责任支付他或她试图在公开市场出售单位以后。有了这些限制,居屋单位的二手市场相当活跃。3实证框架该数据集,我们在这项研究中使用的是随机子的香港人口普查样本。从人口普查档案资料为依托,我们比较了内部流动公营房屋和私人房屋住户的模式。的公营房屋,在香港(黄,刘,1988)的研究采用了类似的方法,但该项研究的重点在住房消费水平失真。黄(1998)阐述了这个问题和其他扭曲性的公共住房,并提出了私有化的情况。通过引入另一个层面的分配不当的问题(即,流动性和区位选择),我们的研究增加了有关辩论诸如私有化或私人市场的租金补贴不同的政策选择。虽然房屋使用权流动中起着重要的作用,横断面数据集使用了这项研究并没有包含更深入的分析足够的资料。因此,本文件应予以考虑到这一点限制。由于持续的资助房屋短缺的问题,它更适用于公共房屋承租人以获取较私人房屋住户迁往另一单位搬迁到另一个公屋单位困难。此外,它往往是不现实的选项公营房屋住户迁往私人楼宇,因为这样做需要政府高额补贴亏损。这是类似的工作锁定劳动经济学中的作用(马德里安,1994;格鲁伯和马德里安,1994)的想法。在工作锁定效应,人们可能会认为,健康保险由雇主提供的价值是不是因为纯粹的工资差异可能补偿补贴。在政府住房的情况下,确实是一个低租金补贴。因此,我们期望锁房屋内部迁移效果甚至比对工作流动性更强。我们假设,在其他条件相同的情况,公屋住户比私人楼宇住户不动的(假设一)。我们认为,公屋居民较私人楼宇的占用较少流动,因为可供他们选择的范围是有限的。根据同样的逻辑,如果我们比较的人究竟是谁作出的举动,我们希望找到一个在公屋租户是不太可能移动到一个单位,符合他的喜好。因此,在移动的条件,我们预期公营房屋住户更可能比私人楼宇的同行要搬到远离原区或居住(假设二)地区。31 由公屋创造了另一种变形是公共房屋福利住户往往牵制他们的当前位置。当情况发生变化,如更改到一个新的工作或有一个婴儿,一个可以考虑迁移到生活在另一个位置。西格尔(1975)考察了家庭之间的家庭和工作地点在首都地区(另见辛普森,1987)的关系。然而,在公屋居民不太可能转移到更加方便的地点和位置的选择可能无法准确反映其真实位置偏好。在英国,休斯和麦考密克的早期研究(1987)显示,公屋抑制与工作有关的原因局部运动。因此,我们推测这种公营房屋住户不太可能在同一地区或区域的工作作为自己的居住地(假设三)举行。同样,资助出售单位的市场规模比私营部门和这些单位往往位于资助规模较小远离经济活动中心。当一个家庭决定购买资助单位,位置选择可能不能准确反映其真实位置偏好。因此,有条件的移动,我们预计,公共房屋的业主更有可能比他们的私人住宅业主要搬到远离原区或居住(假设四)地区。当与私营部门相比,公共业主不太可能转移到更加方便的地点。因此,我们推测,公共业主不太可能在同一地区或区域的工作作为自己的居住地(假设五)举行。这些假说测试使用在2001年人口普查数据的回归。普查数据集包含有关一个家庭是否已经在过去5年内提出,对过去和现在的居住区和信息。如收入和教育附加变量可以用于控制公营和私营房屋住户的其他差异。不过,可能有公营和私营房屋住户影响我们无法观察结果变量的差异。因为我们并没有令人信服的工具变量,很难梳理出由公营房屋之间的公共房屋占用和不可观测变量的相关性引起的间接影响直接的影响。我们将试图通过引入部分解决了控制变量的广泛设置这个问题。但只要有重要的不可观测组在入住差异影响结果的变量,从我们的回归的估计不能得到一个因果的解释。假设M是一个虚拟变量等于0,如果房客没有感动在过去1年里,和M=1,如果承租人感动。本是M=1,可以通过与M的利益为依。自变量模型为代表的概率是租客是否在公共或私人房屋居住。我们还包括如年龄,性别,教育程度,收入,配偶,家庭规模,婚姻状况,就业状况,为青少年儿童的虚拟变量,以及行业和职业虚拟变量控制变量扩展集。根据我的假设,我们预测为公屋租户的虚拟变量是消极的和重要的系数估计。假设虚拟变量S等于0,如果房客有议员在过去5年,一个新的地区,和S=1,如果租客在同一地区逗留。于是S的概率等于1,过去5年中提出可以通过习题表示有条件(S=1时|M=1时)。这个条件概率可以估算的分,民政事务总署在过去531 years.Hypothesis动议二住户抽样预测,为市民提供房屋承租人本虚拟变量系数的估计是负的和重要的。最后,让我们为W=1,如果工作和居住的地方是在同一地区(或在同一地区),和W=0否则。假设三预测,在一个以W为因变量概率回归,为公屋租户的虚拟变量系数的估计是负的和重要的。4实证结果我们使用的数据集在这项研究是5%,2001年香港人口普查随机子样本。我们包括18岁或以上谁在香港,澳门出生,中国所有住户的头上。根据定义,一个家庭是指房屋的家庭单位,一个家庭的成员谁一起生活以及非亲属组成。因此,一房单位所在地的选择会影响有关家庭allmembers。本文在分析定位的家庭作出的选择,我们主要侧重于户主作出的选择。扩大样本覆盖所有住户,不会改变我们的任何结论。我们的主要目标是比较私人房屋住户的公屋居民的流动性和地点的决定。我们之间作出""公营房屋住户""和""自住业主(公众)的区别。前者是指在公共屋居民出租屋,而后者则主要是指单位资助居屋业主。同样,我们之间作出""私人住宅租""和""业主占用人(私营)"",便出现了四个不同的住房使用权类型总的区别。由于流动性和位置决定由房主(和面临的限制)可能有很大的不同的租户,我们运行回归分别为这两个不同的群体。表一列出了房屋使用权类型的所有住户的摘要统计信息。该数据集由96690意见。其中约32%的家庭在公营租住房屋,另有15%是政府资助房屋单位的业主。该户主的特点进行了比较全面住房使用权类型的不同。平均而言,公屋单位户主年龄较大,教育程度较低,更多的可能是女性,不太可能拥有比年幼的孩子。他们也更可能是失业或经济活动。正如所料,公屋住户家庭收入低于其私人住房同行低。在我们的数据,对公屋租户的家庭平均收入为港币一五二五八美元而私人房屋租户为港币二万四千三百四十四美元。同样,资助房屋业主有港元的平均收入24965美元,而私人房屋的业主为34521港元。31 人口普查文件包含有关当前居住区以及居住区5年前。表1介绍了房屋使用权类型的地理分布,我们香港的总分为三大区域31census普查由政府统计处分类区:市区,新市镇和其他农村地区。城市面积指区议会在香港岛和九龙区。包括十二个新市镇位于新界的新市镇。其他农村地区涵盖新界其他地区。在香港,公营房屋计划在1954年展开以来,公屋租金屋建在市区,特别是在观塘和黄大仙人口不断扩大,它达到了一个点时,市区不能提供足够土地兴建足够公屋单位以满足需求。为了解决这个问题,香港政府开始建设新城镇,以满足对房屋的需求不断增加。如前所述,有足够的公众对可分配到市区所有申请租住公屋单位,只有在新市镇公共租赁住房将分配给新的申请单位。一个典型的新镇是由若干居住区,每个居住区主要是私人和公共屋群组成。2006年,共有12个新城镇和最新的一个是位于北大屿山。从表1我们可以看到,约51%的家庭的样本均位于市区,另外44%的新市镇。在公屋租户,相应的分数分别为48%和51%。对于住房补贴,大约有三分之一的人在城市地区,两个住在新城镇。这个模式是相反的三分之二居住私人楼宇业主:在城市地区有更多的私人房屋的住户比新镇(28990)(16180)。在其他农村地区的家庭只占了所有住户的约5%。更小的分数是在公营房屋住户(0.5%),反映一个事实,即很少有公屋屋建在农村地区。4.1内部流动为了检验假设我,我们估计的虚拟变量M,其中M=1,如果一个家庭已搬迁在过去5年。我们估计为租户和业主的单独一模型,并估算结果显示于表2A和2B。模型1控制了性别,配偶,目前,学校教育,家庭收入,家庭规模,年龄组别假人,假人婚姻状况,活动状态假人,假人年幼的孩子,和区级固定效应。相反系数估计报告,我们报告的概率回归边际效应。由于在观察特色的公营及私营房屋住户重大分歧,我们尝试测试引入到我们的具体的控制变量的回归对我们结果的稳健性。模式2增加了行业和职业指南,取代了在年龄五阶多项式假人的年龄分类,并取代收入和100假人百分收入为10户家庭规模大小的假人线性项。这个模式包括超过400个独立变量的回归和全总的估计结果没有显示,以避免混乱。模式3是一样的模式2,除了进一步控制对配偶的特点相同。据估计,仅在分与配偶目前住户样本。最后模式4,2扩展模型通过引入性别年龄倍倍控制变量的相互作用(我们按年龄组取代傻瓜年龄多项式,以产生相互作用计算)。这最后一个模型是估计使用租户全样本,并能在1100控制变量包含。31 与住房使用权类型的流动性效应是捕获的表中的第一个虚拟变量。例如,负0.250边际效应在第一列的预算为公共房屋承租人在表2A意味着,在本组个人分别为25%不太可能有在过去5比私下向租户年感动楼宇。估计效果的实质性意义,以及在统计意义上显着。请注意,增加了一个控制变量和相互作用方面进行集不重大改变大小或公营房屋租赁内部的流动性估计影响的意义,因为由公共房屋承租人变量的边际效应显示模式2,3和4。这些结果与预测一致的假说一。在这个样本中,29.3公营房屋住户搬到过去5年,而60.2%的私人住户做了相同的。如果我们使用的控制变量(即模型表2A)基本确定我们的Probit模型,意指对公营房屋住户移动的概率,概率的估计原来是29.3%,这是为同一样本的平均水平。假设这些公营房屋住户都成了私人楼宇的住户,该模型预测,他们对移动的概率会改变至50.9%。换言之,公营房屋的作用解释了两间在公营房屋住户和私人房屋住户移动概率观察不同的三分之二,其余的差异可以观察到的特点在他们的其他原因如我们的分歧模型。表2B显示,业主自住的私人住宅不太可能在前面5比业主的公屋住户年感动。买楼是一个高固定成本的重大投资决策。一旦作出决定,它往往牵制业主自住到了自己的当前位置。因此,毫不奇怪,在预计介于业主的公共与私人房屋占用的边际效应比之间的区别是公共与私人房屋住户小。尽管如此,在模型的0.049表甲一估计影响是非常显着差异,是不平凡的物质意义上的业主自住的比例谁在过去5年hadmoved只有36%。这些结果与我们的预测,公共房屋的业主更不动的比是私人房屋的业主一致。这一结论是一样的,我们是否集中在基本模型的控制变量与周围50个或者说,包括几百个观测控制变量等机型。自从加入配偶或相互作用的特点(即,模型3和4)不显示在回归估计产生重大影响,我们将集中在两个基本模式1和2在本文的其余部分。我们可以提供的对移动概率其他控制变量的影响简短评论。该户主的性别被发现有小,但对租房者流动性的负面影响。然而,这对流动性的负面影响变成小业主自住的积极作用。年龄,另一方面,对流动性很大的影响。一般来说,在年轻年龄组的人有更大的比年龄组的人的流动性。例如,估计年龄的人的边际效应在表2A的基本模式18-25为28.5个百分点,比61岁或以上的人说(省略组)高。这对移动性的增加产生积极影响的业主占用45.9%(表2B模式1)。至于婚姻状况和就业状况,我们发现谁是结婚或离婚的人往往小于人们谁是单一的移动,而失业的员工少于移动。我们还发现,教育和家庭收入呈正相关的流动性,而家庭规模呈负相关的。然而,年轻的儿童存在增加了移动的可能性。在模型1的表2B,为5岁或以下的儿童的家庭移动为19.7%,比没有这些小孩的家庭机率较高,估计和为六岁到十儿童的家庭移动基于概率为13.7%高于没有这些孩子的家庭。31 4.2地点的选择在移动的条件,一个家庭可以选择住在一个新地区或留在同一地区。因为华人文化鼓励人民保持密切的家庭联系,它是在香港常见的现象,近亲属生活在同一街区。在许多情况下,一个大家庭的不同家庭住在邻近的屋或即使在同一个块。除了事实,人们更愿意留在熟悉的街道,文化规范表明,在移动的条件,家庭更愿意留在原区居住。由于私人房屋租户在房屋的选择最高的灵活性,他们的行为是他们的真实偏好的良好指标。公营房屋住户及资助房屋另一方面,业主,拥有有限的公屋单位,并选择他们的观察位置的选择,不一定反映其真实的喜好自由。在本小节,我们的样本只包括那些已经在过去5年内提出的家庭。虽然我预测,假设公屋住户不太想搬家,假设二预测,在移动条件,公屋居民更容易移出原来的社区。博伊尔(1995)认为,在英国的长途移民(那些谁动了超过50公里),不太可能会比其他进入房屋类型的住房。这一调查结果是不符合我们的预测不一致的,因为香港的地理面积小。在香港的动作是很少超过50公里。此外,一个用于结果的原因是,地方当局提供的公共房屋以外的地方迁移面积有限的机会。在香港,相比之下,在各区的公共住房分配正由房屋委员会统一管理。在本款的估计,因为我们是制约这一已在过去5年的移动户,样本大小减少到37013。表3A和3B显示地点的选择的模型,其中因变量是家庭是否有在邻里提出的指标变量S。更具体地说,对于前两个表列,我们认为有关是居委会的普查区,我们集合S=1,如果该住户已搬进了新居同一地区,和S=0如果家庭搬到了一个新的居住在不同的地区。回想一下,在我们的数据文件中的香港地区普查分为31个区。最大的区根据这一分类是港岛东区,拥有约198,000户。中位数的地区是湾仔,它拥有大约62,000住户。最详尽的是我们提供数据的地理细分。在第一列的表2B,我们发现,为""公共房屋承租人的边际效应估计""是消极和统计学意义。这一估计意味着,在移动的条件,在同一普查区移动的概率为5%,其中公营房屋住户低于在私人住宅的租户。这一结果与假说二的预测。在列第二,我们添加详细的控制变量(模式2)和估计公共房屋承租人的边际效应越强(从0.052至0.077)。31 如果我们比较了自住业主的行为(见一栏表3B),我们发现,边际效应估计""自住业主(公众)""强于公共房屋承租人中提出的表3A。对在跨区移动的概率为17.6%,比私人楼宇业主公共房屋的业主降低。这一结果与假说四的预测。在列第3和第4表3A和3B,我们检查内使用区域内移动,而不是跨区移动的依赖。这里在城市的三个类别看,一个地区确定了我们的结果的可靠性,新的城镇,农村,以上是我们的解释。请注意,这里定义的普查区域包括一个比较大的,不一定接壤地区。例如,一个家庭,从一个城市移动到另一个新的可能会遇到一个完全不同的街区。出于这个原因,我们预期的结果会比列1和2我们的首选规格的弱。比较栏3栏表3A1,我们发现,估计边际效应""""公共房屋承租人的绝对规模小于0.01,不再统计在1%水平上显着。这种结果的原因之一是,公屋单位可供编配的是大多数是在位于新市镇。我们讨论过,很少有公共房屋的其他农村地区出租屋,而在城镇单位大多已经被占用和转让不可用。因此,在公营房屋住户搬迁,主要是在新的城镇地区。在列第3和表3B,为""自住业主(公众)""仍然负且显着的边际效应估计4。例如,在第4列的预算建议业主占用公共房屋内有一个相同的人口普查相比,业主自住的私人楼宇地区移动14.2%的可能性降低。虽然这个效果比在第2栏的相应估计弱,影响的强度仍然很大。总之,结果列于表3A和3B支持假设二和四。一个用于发现公屋单位再进一步在移动条件的另一种解释是一个简单的阈值模型。由于搬迁费用较高,他们只当它移动真正有价值移动,实在是值得的,如果移动的最佳位置是从很远的当前位置。根据这个方案的解释,公营房屋住户谁动了应该活接近其工作地点,假设同一地点工作和居住的最佳位置是一重要因素。不过,我们没有发现这方面的证据。在公营房屋住户谁提出,在同一地区,因为他们的工作生活7.4%(为公营房屋住户谁不动了相应的比例为21.3%)。在私人住宅住户谁提出,在同一地区,因为他们的工作生活19.0%(为私人房屋住户谁不动的相应百分比为10.2%)。这些数字不支持移动,私人住户正在向更远的,有条件的解释,最佳地点。我们更详细地讨论在后面的小节之间的居住和工作地方的关系。4.3工作地点31 辛普森(1987)开发了其中住宅的位置和工作位置的同时测定模型(西格尔,1975;金,2005)。在本文中,我们推测,由于缺乏选择,公屋居民不太可能找到一个住房单位是位于靠近其工作地点。此外,住宅和工作地点的选择本质上是一个动态的决定(维克曼,1984)。流动性之间的公营房屋住户缺乏阻碍了他们的居住尝试调整位置时,情况发生变化(如更改到一个新的工作),从而增加了一间居住和工作地方错配的机会。在本论文中,我们着眼于户主的决定。由于失业或从事经济活动人士不报告的工作场所,在本款的分析不包括这些人。同为""有效观测样本量造成的工作地点""变量是53334。我们代码中的变量W等于1,如果在同一地区(或地区),因为他或她的作品,和W户主的生活,否则等于0。由此产生的概率型估计显示在表4A和4B。该表4A第二列显示,公营房屋的住户有可能显着减少内工作相比,私人房屋租户的居住同一普查区。边际效应的估计表明,概率为5%的公营房屋住户低于为私人住宅住户参考小组,在同一地区居住,因为他们的工作,控股等特点不变。从表4B,我们发现,估计的边际效应""自住业主(公众)""也是负的(负3%)和重要的。换言之,公共房屋的业主是不太可能在同一地区工作比是私人房屋的业主。请注意,我们已经包含在我们的回归普查区假人。因此,我们的结果是不是驱动的事实,公共屋往往位于远离经济活动中心。换言之,即使我们比较两个家庭在同一普查区,谁住在公营房屋(租用或拥有)更可能持有超过他或她的私人房屋对口跨区就业居住。在表列第3和第4A及4B4,我们定义户主是否拥有跨区域的就业方面的相关变量W。结果是相似的,在早期的规格。即使在控制区域的影响,公屋住户不太可能持有其居住在同一地区的工作比在私人住宅。这些结果是与他们同行的假设三,五,预测一致。辛普森在一份文件(1980年),他认为,一个家庭的居住场所,也是由这家的二级仔工作地点的影响。因此,我们通过改变检查的观察单位是户主的配偶我们的研究结果的稳健性。由于很多户主没有配偶或有配偶谁没有工作,样本大小为进一步减少到27490。对模型2规范为基础,我们发现,公共房屋承租人为和"自置(公众)的边际效应估计"中为同区就业方程分别为0.038和0.057,都是在1%显着水平。如果我们按区域假人代替区假人并专注于在同一地区工作的可能性,估计边际效应成为公共房屋承租人和自置(公众)都分别0.056和0.066顷统计学意义。就像户主,在公共住房二级仔,出租或自住,有在同一地区或区域的普查工作,因为他们生活相比,在私人住房二级仔的可能性降低。5结论31 在香港,公屋赋予了相当大的补贴,很大一部分人口,公共房屋单位都必须相互竞争的家庭配给。公共房屋政策研究往往侧重于这种大规模实行的财政补贴负担,而是由配给住房补贴效率低下引起的资源配置条件较差的文献记载。本研究探讨如何通过检查分配不当影响这些关于内部流动和旅行到工作公屋的从居民的住房单位分配不当的角度隐性成本。这项研究利用了5%的随机子2001年香港人口普查样本测试以下五个假设使用:(一)公屋住户比私人楼宇同行动弹不得。(二)在移动条件,公营房屋住户私人住房超过同行可能要移出原区或居住区域。(三)公营房屋住户不太可能在同一地区工作或居住地的地区。(四)在移动条件,公共业主更比私人楼宇的同行可能要移出原区或居住区域。(五)公共业主不太可能在同一地区或区域的工作作为自己的居住地。我们估计对普查样品检验这些假设的Probit回归。所有的实证结果显着,与我们的预测是一致的。通过确定区位选择的公营房屋引起的扭曲,本研究旨在改善我们的公共房屋政策,在香港的真实成本估计。虽然这些费用低于实际观察到的财政成本,它们对居民的日常生活产生直接影响。政策制定者应探讨引入私人市场的租金补贴的可能性,以实现经济适用住房提供客观经济条件较差的待办,而由于避免了扭曲向政府提供的住房很多。最后,我们想强调的是,我们的结论是基于相关性。虽然我们已经介绍了控制变量(有时共约1000项)广泛集为了控制公营及私营房屋住户细微的差别,我们仍然不能排除在未观察到的特征差异,可能会影响结果变量的可能性。尽管经济理论认为一间公屋居住迁移,减少占用的因果关系,这里的结果不能提供这样的链接确切的证据。在实验数据或有说服力的工具变量的情况下,结果的因果解释应该采取一些注意事项。31 2外文文献原文TheeffectsofpublichousingoninternalmobilityinHongKongbyHon-KwongLui,WingSuen1.IntroductionPublichousingrepresentsasizablesubsidytohousingcostandithastobestrictlyrationedamongcompetingusers.Whenrationingoccurs,resourcesmaynotbeallocatedtotheirmostvaluableusesbecausepeoplecannoteffectivelyconveytheintensityoftheirpreferences.Consider,forexample,twootherwiseidenticalhousingflatsavailableforallocationbetweentwoprospectivetenants.FlatAislocatedinamoredesirableneighborhoodthanFlatB,sobothprospectivetenantsarewillingtopayapremiumrentforFlatA.SupposethefirstpersonalsohappenstoworkinthesameneighborhoodasthelocationofFlatA,andheispreparedtopayahigherpremiumforFlatAthanisthesecondprospectivetenant.Inanopenmarket,FlatAwillbeallocatedtothepersonwhovaluesitmore,namelythefirstprospectivetenant.Whenrentsarefixedatsubsidizedlevelsandflatsareallocatedbynon-marketmeans,however,bothprospectivetenantswillreveal(truthfully)thattheypreferFlatAtoFlatB.Thereisnomeanstoadministrativelydeterminewhichofthetwoisthehigher-valueuserforthemoredesirableflat.Inpractice,subsidizedhousingisoftenallocatedonthebasisofstatedpreferencessupplementedbysomelotteryorqueuingmechanism.Thepurposeofthispaperistostudysomeoftheinefficienciesinducedbytheuseofnon-pricemethodsofallocationintheprovisionofpublichousinginHongKong.Themisallocationofresourcesundernon-pricemethodsofallocationdrawsfromalargeandwell-establishedliteratureineconomics(e.g.,Cheung,1974;Weitzman,1977;Suen,1989;GlaeserandLuttmer,2003).Whenagoodisallocatedatbelow-marketprice,thereisinsufficientincentiveforsellerstoproducethegood.Theresultingunder-supplyofthesubsidizedgoodleadstothefamous‘‘deadweightloss’’illustratedineconomicstextbooks.Lesswellrecognizedisthefactthatcompetitionforthepricesubsidygenerallyleadstothedissipationofvalue.Barzel(1974)showshowequilibriumcanbeachieved31 throughsociallywastefulrationingbywaiting.Inthecontextofpublichousing,prospectivetenantsmaydistorttheirlaborsupplydecisionslesttheirincomeexceedstheeligibilitythreshold,whileincumbenttenantsmaychoosetoinefficientlystayintheirflatsuponchangingcircumstanceslesttheylosethevaluablesubsidy.Yetathirdsourceofinefficiencyisthat,whenindividualscannotexpresstheirintensityofpreferencesthroughtheirwillingnesstopay,non-pricemethodsofallocationgenerallyfailtoallocatescarceresourcestothehighest-valueusers.IfaflatwortharentalofHK$10,000amonthisbeingrentedatHK$2000,anyprospectivetenantwhosevaluefortheflatexceedsHK$2000willbeinterestedinbiddingforit.Dependingontheallocationmethodbeingadopted,theflatmaywellendupinthehandsofsomeonewhosevalueforitiswellbelowHK$10,000.DeaconandSonstelie(1989,1991)andSuen(1989)analyzethewelfarecostsarisingfromsuchmisallocationofresourceswhenrationingisbywaiting.Inamorerecentstudy,GlaeserandLuttmer(2003)studythemisallocationofhousinginNewYorkCityunderrentcontrol.Theyfindthatthematchingofthecharacteristicsoftenantstothecharacteristicsoftheirhousingunitsisdistortedforflatsunderrentcontrol.Sincerentcontrolalsorepresentsahousingsubsidytoexistingtenants,GlaeserandLuttmer’s(2003)insightisequallyapplicabletotheanalysisofgovernment-subsidizedhousing.InHongKong,publichousingaffectsalargefractionothepopulation.In2009,29%ofthehouseholdsweretenantsingovernment-providedhousing,andanother18%ownedsubsidizedhousingthroughtheHomeOwnershipSchemeandrelatedprograms(HongKongHousingAuthority,2010a).Criticsofpublichousingpolicyoftenfocusonthefiscalburdenimposedbysuchmassivesubsidies.Lesswellresearchedarecostsfromtheallocativeinefficienciesinducedbyrationingsubsidizedhousing.Althoughthesecostsmaybelesstangiblethanthedirectfiscalcosts,theycanhaveamoreimmediateimpactonpeople’severydaylives.Inthispaper,wefocusondistortionsonlocationchoicesinducedbypublichousing.Publichousingbenefitstendtotiedowntenantstotheircurrentlocation;residentsinpublichousingareexpectedtobelesslikelytomovetomoreconvenientlocationsormoresuitablehousingunitsevenaftertheirlifecircumstanceshavechanged.Thishypothesisis31 relatedtotheliteratureonthe‘‘joblock’’effectonjobmobility(e.g.,Madrian,1994;GilleskieandLutz,2002;Adams,2004;BansakandRaphael,2008).Thatliteratureexaminestheeffectofemployer-providedhealthinsuranceonjobmobility,thehypothesisbeingthatworkersarereluctanttoswitchtheirjobslesttheylosetheirinsurancecoverage.Inasimilarvein,weexaminethehypothesisthatpublichousingbenefitsdeterinternalmobilitybecauseoccupantsdonotwanttolosetheirsubstantialhousingsubsidy.Intheurbanstudiesliterature,thereisasubstantialbodyofworkdocumentingtheempiricalpatternsofinternalmigration(e.g.,Greenwood,1985;Elliott,1997;Painter,1997).BoyleandShen(1997)studytheeffectofpublic(council)housingoninternalmigrationinBritainandfindthatpublichousingoccupantsarelesslikelytomigratelongdistancesoutsidetheareaofthelocalcouncil.Inarecentpaper,Kimetal.(2005)demonstratethattransportrelatedattributeshavesignificantimpactsonresidentiallocationchoiceandindividualspreferlocationswithacombinationofshortercommutingtime,lowertransportcosts,lowerdensityandhigherqualityofschool.IntheUnitedStates,publichousingcomplexesareoftenclusteredincentral-citylocationswithhighconcentrationofpoverty(Quigley,2000).Manybelievethatlivinginsuchneighborhoodslimitsone’seconomicopportunity(Wilson,1996).Toreduceeconomicsegregationinducedbyprojectbasedpublichousingprovision,theUSDepartmentofHousingandUrbanDevelopmenthasexperimentedwithtenant-basedrentalsubsidiestoencouragethemtomovefromhigh-povertytolow-povertyareas(e.g.,theMovingtoOpportunityProgram).Johnsonetal.(2002)reportthatfamilieswhotakeadvantageoftheseprogramstomovetolow-povertyareasexperiencereductionsinwelfaredependenceandimprovementsinhealthstatus.Thereisalsoevidencethatsuchmovesmayimprovechildren’sschoolingoutcomesandreducetheirproblembehaviors(seealsoFeinsandShroder,2005;Ludwigetal.,2005;RosenbaumandHarris,2001).Inadditiontothelackofinternalmobility,anothermanifestationofthedistortionsinlocationchoicesinducedbypublichousingisthemismatchbetweenplaceofresidenceandplaceofwork.Acompetitiverentalmarketsolvestheefficientassignmentproblemashousingunitsgototheirhighest-valueusers(KoopmansandBeckmann,1957).Thepricesystemfailstoperformthisfunctionwhenrentsforpublichousingunitsarefixedandthe31 allocationofunitsinexcessdemandisdeterminedbynon-pricemeans.Moreover,evenwhenthereisnoinitialmismatchbetweenplaceofresidenceandplaceofwork,people’scircumstanceschange.Whencircumstances(e.g.,changetoanewjob)prescribemovingtoanotherarea,thelackofmobilityonthepartofpublic-housingtenantsmayhindertheoptimalreallocationofresources.Thedeterminationofresidentialandworklocationisawellresearchedtopicinurbaneconomics(e.g.,Vickerman,1984;Siegel,1975;Simpson,1987;Kan,2002;Kan,2003).Alsoofsomerelevancetoourpresentworkistheliteratureonthe‘‘spatialmismatchhypothesis’’ofKain(1968).Kain’sworkwasoriginallyrelatedtohowresidentialsegregationadverselyaffectedtheemploymentopportunitiesofblacksintheUnitedStates(seealsoHoltzer(1991)).AlthoughresidentialsegregationbyraceisnotamajorproblemforHongKong,thefactthatpublichousingestatesareoftenlocatedfarfromcentersofeconomicactivitiesmayhaveimplicationsforthemismatchbetweenplaceofresidenceandplaceofwork.Moreover,housingtenurealsoplaysanimportantroleindeterminingresidentialmobilityandemployment(e.g.,IoannidesandKan,1996;Kan,2000;Dohmen,2005).Battuetal.(2008)studytheeffectsofhousingtenureonindividuals’jobandunemploymentdurationsintheUnitedKingdom.Theyarguethathomeownershipandpublicrentalhousingimposemoreconstraintonemployedandunemployed,respectively.Munchetal.(2008)arguethathomeownershiphasanegativeeffectontheunemploymentriskandapositiveimpactonwages.HuangandDeng(2006)showthathousingtenurehasconsistentbutdifferenteffectsonmobilityovertimeinChina.Heldermanetal.(2004)arguethatthereisadecreaseintheeffectofhomeownershiponresidentialmobilitygivenarapidriseinhomeownershipintheNetherlands.Munchetal.(2006)suggestthathomeownershiphampersthepropensitytomoveforjobreasonsbutimprovesthechancesoffindinglocaljobs.AstheavailabilityofsubsidizedsaleflatsinHongKongismorerestrictivethanthatofprivatesector,publichomeownerswillfaceahigherdistortioninlocationchoicethanprivatecounterparts.Thismayalsohaveimplicationsforthemismatchbetweenplaceofresidenceandplaceofworkamonghomeowners.Inthispaper,weusethecensussampleofHongKongtoexaminetheallocative31 inefficienciesinducedbyrationingsubsidizedhousingandhowthismisallocationaffectsthesetenants’behaviorconcerninginternalmobilityandtravel-to-work.Theempiricalresultssuggestthatpublichousingoccupantsaremoreimmobileandarelesslikelytoworkinthesamedistrictorregionastheirplaceofresidencethantheirprivatehousingcounterparts.Moreover,conditionalonmoving,publichousingoccupantsaremorelikelythantheirprivatehousingcounterpartstohavetomoveoutsidetheiroriginaldistrictorregionofresidence.Theyarealsolesslikelytoworkinthesameplaceastheylive.Ourconclusionsarebasedonprobitregressionsofmobilityandtravelbehaviorsonpublicoccupancystatus.Wetrytocontrolforotherdeterminantsofsuchbehaviorsbyincludinganextensivesetofcontrolvariablesinourregressions.Manyofourregressionmodelsareestimatedusinghundredsofvariables.However,wecannotexcludethepossibilitythattherearesomeremainingunobserveddifferencesacrossoccupancygroupsthataffectmobility.Theempiricalresultspresentedinthispapershouldbeinterpretedwiththiscaveatinmind.AbriefreviewofthepublichousingallocationsystemInHongKongthereareapproximatelytwomillionpeople(about30%ofHongKong’spopulation)livingin700,000publicrentalflatsprovidedbytheHousingAuthority(HongKongHousingAuthority,2010b).Publichousingapplicantsortheirfamilymembersmustnotdirectlyorindirectlyownorco-ownanydomesticproperty.Moreover,applicantsaresubjecttoincomeandtotalnetassetvaluelimitsaslaiddownbytheHousingAuthority.TheHousingAuthoritymaintainsawaitinglistofpublicrentalhousingapplicants.Fromthislist,eligibleapplicantsareofferedaccommodationinpublicrentalhousingestates.Whenanapplicant’sturncomes,theapplicantwillbegivenamaximumofthreeoffers(oneoptionineachoffer)ofpublicrentalhousinginthedistrictofhis/herchoice.Theapplicationwillbecancellediftheapplicantfailstoprovideacceptablereasonforrefusingallthethreehousingoffers(HongKongHousingAuthority,2010b).Publicrentalhousingestatesaregroupedintofourbroaddistricts,namelyUrban,ExtendedUrban,theNewTerritories,andtheIslands.SincethereareinsufficientpublicrentalhousingunitsintheUrbandistrictthatcanbeallocatedtoallapplicants,applicantscanonlychooseonedistrictfromamongthethreenon-urbanareasfortheirfuturehousingallocation.AsofJuly2010,thereweremorethan31 130,000applicantsonthewaitinglist(HongKongHousingAuthority,2010b).TheHousingAuthoritysetsatargetofmaintainingtheaveragewaitingtimeat3years.Intheprocessofpublichousingallocation,applicationsareprocessedstrictlyinaccordancewiththeapplicationsequencenumbersandtheapplicants’choicesofdistrict.However,thosewhofulfilltheeligibilitycriteriaoftheExpressFlatAllocationSchememayhaveanearlierchancetobeallocatedpublicrentalhousingunits.Therearealsoseparatecriteriaandallocationschemesforpeopleaffectedbysquatterclearanceandpublichousingredevelopment.Accordingtoexistingpublichousingpolicies,publichousingtenantsmayapplyforflattransferundervarioussituationssuchasincreasedfamilysize,specialmedicalorsocialgrounds,ormajorimprovementorredevelopmentprogramsimplementedbytheHousingAuthority(HongKongHousingAuthority,2010c).Wheneverpossible,eligibletenantswillbetransferredtosuitableflatswithinthesameestate.Tenantsmayalsoapplyforspecialtransfertoanotherestate.Duetopersistentshortageofpublicrentalhousingunits,evenifatenantiseligibleforflattransfer,thewaitingtimecanbequitelongandthereisnoguaranteethatthenewflatwillmatchthetenant’sPreferences.Besidesapplyingforapublicrentalhousingunit,eligiblehouseholdscanalsoapplytobuysubsidizedflatsthroughtheHomeOwnershipScheme(HOS).InsettingthepriceofHOSflats,theHousingAuthorityfollowsthreeguidingprinciples:(i)themortgage-to-incomeratioshouldnotbemorethan40%;(ii)50%oftheflatsshouldbeaffordabletothetargetgroup;and(iii)flatsshouldbeofferedata30%discountofmarketvalue.Startingfromthe3rdyearofoccupancy,HOSflatownersmayselltheirflatsintheopenmarketafterpayingapremiumtotheHousingAuthority.Thepremiumisbasicallyequaltotheprevailingmarketvaluemultipliedbythediscountenjoyedbytheowneratthetimeofpurchase.Forexample,iftheprevailingmarketvalueofaHOSflatisHK$2000,000andtheflatwasofferedatadiscountof30%atthetimeofpurchase,thenthepremiumisequaltoHK$600,000.OwnersofHOSflatscanalsoselltheirflatstoexistingpublicrentalhousingtenants.Inthiscase,theoriginalHOSownerwillnotberequiredtopaythepremiumtotheHousingAuthoritybeforesellingtheflat,butthepurchaserwillinherittheliabilitytopaythe31 premiumshouldheorsheseektoselltheflatintheopenmarketlateron.Withtheserestrictionsinplace,thesecondarymarketforHOSflatsisquiteinactive.EmpiricalframeworkThedatasetweuseinthisstudyisarandomsub-sampleoftheHongKongPopulationCensus.Relyingoninformationfromthepopulationcensusfile,wecomparethepatternofinternalmobilitybetweenpublichousingandprivatehousingoccupants.AstudyofpublichousinginHongKong(WongandLiu,1988)adoptsasimilarapproach,butthefocusofthatstudyisonthedistortioninthelevelofhousingconsumption.Wong(1998)elaboratesthisandotherdistortionsinducedbypublichousing,andmakesacaseforprivatization.Byintroducinganotherdimensiontothemisallocationproblem(namely,mobilityandlocationchoice),ourstudyaddstothedebateconcerningvariouspolicyalternativessuchasprivatizationorprivate-marketrentsubsidy.Althoughhousingtenureplaysanimportantroleinmobility,thecross-sectionaldatasetusedinthisstudydoesnotcontainsadequateinformationformorein-depthanalysis.Hence,thispapershouldbereadwiththislimitationinmind.Becauseofapersistentshortageofsubsidizedhousing,itismuchmoredifficultforapublichousingtenanttoobtainrelocationtoanotherpublichousingunitthanforaprivatehousingoccupanttomovetoanotherflat.Moreover,itisoftenanunrealisticoptionforpublic-housingtenantstomovetoaprivateflatbecausedoingsoentailslosingasubstantialgovernmentsubsidy.Thisissimilartotheideaofthe‘‘joblock’’effectinlaboreconomics(Madrian,1994;GruberandMadrian,1994).Inthe‘‘joblock’’effect,onemayarguethatthevalueofhealthinsuranceprovidedbytheemployerisnotapuresubsidybecauseofpossiblecompensatingwagedifferences.Inthecaseofgovernmenthousing,thelowrentalisindeedasubsidy.Wethereforeexpectthe‘‘housinglock’’effectoninternalmigrationtobeevenstrongerthanthe‘‘joblock’’effectonjobmobility.Wehypothesizethat,otherthingsequal,publichousingoccupantsaremoreimmobilethanprivatehousingoccupants(HypothesisI).Wearguethatpublichousingoccupantsarelessmobilethanprivatehousingoccupantsbecausetherangeofoptionsavailabletothemismorelimited.Bythesamelogic,ifwe31 comparepeoplewhoactuallymadeamove,weexpecttofindthatatenantinpublichousingislesslikelytomovetoaunitthatmatcheshispreferences.Therefore,conditionalonmoving,weexpectthatpublic-housingtenantsaremorelikelythantheirprivatehousingcounterpartstohavetomovefartherawayfromtheiroriginaldistrictorregionofresidence(HypothesisII).Anotherdistortioncreatedbypublichousingisthatpublichousingbenefitstendtotiedowntenantstotheircurrentlocation.Whencircumstanceschange,suchaschangetoanewjoborhavingababy,onemayconsidermovingtoliveinanotherlocation.Siegel(1975)examinestherelationshipbetweenthehomeandjoblocationofthehouseholdwithinametropolitanarea(seealsoSimpson,1987).However,tenantsinpublichousingarelesslikelytomovetomoreconvenientlocationsandtheirlocationchoicesmaynotaccuratelyreflecttheirtruelocationpreferences.InanearlierstudyintheUnitedKingdom,HughesandMcCormick(1987)showthatpublichousinginhibitslocalmovementforjob-relatedreasons.Wethereforehypothesizethatpublic-housingtenantsarelesslikelytoworkinthesamedistrictorregionastheirplaceofresidence(HypothesisIII).Similarly,themarketsizeforsubsidizedsaleflatismuchsmallerthanthatofprivatesectorandthesesubsidizedflatsareoftenlocatedfarfromcentersofeconomicactivities.Whenahouseholddeterminestobuyasubsidizedflat,thelocationchoicemaynotaccuratelyreflecttheirtruelocationpreference.Therefore,conditionalonmoving,weexpectthatownersofpublichousingaremorelikelythantheirprivatehomeownerstohavetomovefartherawayfromtheiroriginaldistrictorregionofresidence(HypothesisIV).Whencomparewiththeprivatesector,publichomeownersarelesslikelytomovetomoreconvenientlocations.Wethereforehypothesizethatpublichomeownersarelesslikelytoworkinthesamedistrictorregionastheirplaceofresidence(HypothesisV).Thesehypothesesaretestedusingprobitregressionsoncensusdatain2001.Thecensusdatasetcontainsinformationaboutwhetherahouseholdhasmovedintheprevious5years,andinformationaboutthepastandpresentdistrictsofresidence.Additionalvariablessuchasincomeandeducationcanbeusedtocontrolforotherdifferencesbetweenpublicandprivatehousingoccupants.However,theremaybeunobservabledifferencesbetweenpublicand31 privatehousingoccupantsthataffectouroutcomevariables.Becausewedonothaveconvincinginstrumentalvariables,itisdifficulttoteaseoutthedirecteffectofpublichousingfromtheindirecteffectsinducedbythecorrelationbetweenpublichousingoccupancyandunobservablevariables.Wewillattempttopartiallyaddressthisissuebyintroducinganextensivesetofcontrolvariables.Buttotheextentthatthereareimportantunobserveddifferencesacrossoccupancygroupsthataffecttheoutcomevariables,theestimatesfromourprobitregressionscannotbegivenacausalinterpretation.SupposeadummyvariableMisequalto0ifatenanthasnotmovedintheprevious1years,andM=1ifthetenanthasmoved.TheprobabilitythatM=1canberepresentedbyaprobitmodelwithMasdependentvariable.Theindependentvariableofinterestiswhetherthetenantresidesinpublicorprivatehousing.Wealsoincludeanextensivesetofcontrolvariablessuchasage,gender,education,income,spouse,householdsize,maritalstatus,employmentstatus,dummyvariablesforyoungchildren,andindustryandoccupationaldummyvariables.BasedonHypothesisI,wepredictthecoefficientestimateforthepublichousingtenantdummyvariabletobenegativeandsignificant.SupposethedummyvariableSisequalto0ifatenanthasmovedtoanewdistrictintheprevious5years,andS=1ifthetenantstayedinthesamedistrict.ThentheprobabilityofS=1conditionalonhavingmovedinthepast5yearscanbedenotedbyProb(S=1|M=1).Thisconditionalprobabilitycanbeestimatedusingthesub-sampleofhouseholdsthathadmovedintheprevious5years.HypothesisIIpredictsthatthecoefficientestimateforthepublichousingtenantdummyvariableofthisprobitmodelisnegativeandsignificant.Finally,letW=1ifplaceofworkandplaceofresidenceareinthesamedistrict(orinthesameregion),andW=0otherwise.HypothesisIIIpredictsthatinaprobitregressionwithWasthedependentvariable,thecoefficientestimateforthepublichousingtenantdummyvariableisnegativeandsignificant.EmpiricalresultsThedatasetweuseinthisstudyisthe5%randomsubsampleofthe2001HongKongPopulationCensus.Weincludeallheadsofhouseholdsaged18orabovewhowereborninHongKong,MacauorMainlandChina.Bydefinition,ahouseholdreferstoahousingdomestic31 unitconsistingofthemembersofafamilywholivetogetheralongwithnon-relatives.Thus,thelocationchoiceofahousingunitaffectsallmembersofthehouseholdconcerned.Inthispaper,whenanalyzingthelocationchoicesmadebyahousehold,weprimarilyfocusonthechoicesmadebythehouseholdhead.Extendingthesampletocoverallhouseholdmembersdoesnotchangeanyofourconclusions.Ourprimaryobjectiveistocomparethemobilityandlocationdecisionsofpublichousingoccupantswiththoseofprivatehousingoccupants.Wemakeadistinctionbetween‘‘Public-HousingTenants’’and‘‘Owner-Occupiers(Public).’’Theformerreferstotenantsinpublichousingrentalestates,whereasthelatterprimarilyreferstoownersofsubsidizedHOSflats.Similarly,wemakeadistinctionbetween‘‘Private-HousingTenants’’and‘‘OwnerOccupiers(Private),’’givingrisetoatotaloffourdifferenthousing-tenuretypes.Sincemobilityandlocationdecisionsmade(andconstraintsfaced)byhomeownersmaybeverydifferentfromthoseofrentaltenants,werunregressionsseparatelyforthesetwodistinctgroups.Table1presentsthesummarystatisticsofallhouseholdsbyhousing-tenuretype.Thedatasetconsistsof96,690observations.About32%ofthesehouseholdswereinpublicrentalhousing,andafurther15%wereownersofgovernmentsubsidizedflats.Thecharacteristicsofthehouseholdheadswerequitedifferentacrosshousing-tenuretypes.Onaverage,householdheadsofpublichousingrentalunitswereolder,lesseducated,morelikelytobefemale,andlesslikelytohaveyoungchildrenthantheothergroups.Theywerealsomorelikelytobeunemployedoreconomicallyinactive.Asexpected,householdincomeforpublichousingoccupantswaslowerthanthatoftheirprivatehousingcounterparts.Inourdata,theaveragehouseholdincomeforpublichousingrentaltenantswasHK$15,258whilethatforprivatehousingrentaltenantswasHK$24,344.Similarly,ownersofsubsidizedhousinghadanaverageincomeofHK$24,965,comparedtoHK$34,521forownersofprivatehousing.Thepopulationcensusfilecontainsinformationaboutdistrictofcurrentresidenceaswellasthedistrictofresidence5yearsago.Topresentthegeographicaldistributionofhousing-tenuretypesinTable1,weaggregatethe31censusdistrictsofHongKongintothreebroadcensusregionsasclassifiedbytheCensusandStatisticsDepartment:UrbanAreas,31 NewTowns,andOtherRuralAreas.UrbanAreasrefertoDistrictCouncildistrictsonHongKongIslandandinKowloon.NewTownsincludetwelvenewtownsintheNewTerritories.OtherRuralAreascovertheremainingareasintheNewTerritories.InHongKong,thepublichousingprogrammewascommencedin1954.Atthebeginning,publichousingrentalestateswerebuiltinurbanareas,notablyinKwunTongandWongTaiSin.Asthepopulationcontinuedtoexpand,itreachedapointwhenurbanareascouldnotprovideenoughlandtobuildadequatehousingunitstomeetdemand.Tosolvethisproblem,theHongKonggovernmentstartedbuildingnewtownstoaccommodatetheeverincreasingdemandforhousing.Asdiscussedearlier,thereareinsufficientpublicrentalhousingunitsintheurbanareasthatcanbeallocatedtoallapplicants;onlypublicrentalhousingunitsinnewtownswillbeallocatedtonewapplicants.Atypicalnewtownismadeupofseveralresidentialareasandeachresidentialareamainlyconsistsofbothprivateandpublichousingestateclusters.In2006,therearealtogethertwelvenewtownsandthelatestoneislocatedinNorthLantau.FromTable1,wecanseethatabout51%ofthesamplehouseholdswerelocatedinUrbanAreasandanother44%wereinNewTowns.Amongpublichousingrentaltenants,thecorrespondingfractionswere48%and51%.Forownersofsubsidizedhousing,approximatelyone-thirdofthemlivedinUrbanAreasandtwo-thirdslivedinNewTowns.Thispatternisreversedforprivatehousing:therewerefarmoreprivatehousinghouseholdsinUrbanAreas(28,990)thaninNewTowns(16,180).HouseholdsinOtherRuralAreasonlymadeupabout5%ofallhouseholds.Thefractionwasevensmalleramongpublic-housingtenants(0.5%),reflectingthefactthattherewereveryfewpublichousingrentalestatesconstructedinruralplaces.4.1.InternalmobilityInordertotestHypothesisI,weestimateaprobitmodelforthedummyvariableM,whereM=1ifahouseholdhadmovedwithinthelast5years.Weestimatethemodelseparatelyfortenantsandowners,andtheestimationresultsareshowninTables2Aand2B.Model1controlsforgender,spouse-present,yearsofschooling,householdincome,householdsize,agegroupdummies,maritalstatusdummies,activitystatusdummies,youngchildrendummies,anddistrictlevelfixedeffects.Insteadofreportingcoefficientestimates,wereport31 marginaleffectsoftheprobitRegression.Becauseofthesubstantialdifferencesinobservablecharacteristicsbetweenpublicandprivate-housingtenants,wetrytotesttherobustnessofourresultsbyintroducingdetailedcontrolvariablesintoourregression.Model2addsindustryandoccupationaldummies,replacestheagecategorydummiesbyafifth-orderpolynomialinage,andreplacesthelineartermsforincomeandhouseholdsizeby100dummiesforincomepercentileand10dummiesforhouseholdsize.Thismodelincludesatotalofover400independentvariablesintheregressionandthefullestimationresultsarenotshowntoavoidclutter.Model3isthesameasModel2exceptthatitfurthercontrolsforspousalcharacteristics.Itisestimatedonlyonthesub-sampleoftenantswithspousepresent.FinallyModel4expandsonModel2byintroducingagetimesgendertimescontrolvariableinteractions(wereplacetheagepolynomialbyagegroupdummiesinordertoproduceinteractionterms).Thislastmodelisestimatedusingthefullsampleoftenantsanditcontainsover1100controlvariables.Theeffectofhousing-tenuretypeonmobilityiscapturedbythefirstdummyvariableintheTables.Forexample,thenegativemarginaleffectestimateof0.250inthefirstcolumnfor‘‘PublicHousingTenant’’inTable2Ameansthatindividualsinthisgroupwere25%lesslikelytohavemovedinthepast5yearscomparedtorentaltenantsinprivatesectorhousing.Theestimatedeffectissignificantinthesubstantivesenseaswellasinthestatisticalsense.Noticethataddinganextensivesetofcontrolvariablesandinteractiontermsdoesnotmateriallychangethesizeorthesignificanceoftheestimatedeffectofpublichousingtenancyoninternalmobility,asshownbythemarginaleffectsofthe‘‘PublicHousingTenant’’variableinModels2,3and4.TheseresultsareconsistentwiththepredictionofHypothesisI.Inthesample,29.3%ofpublic-housingtenantshadmovedinthepast5years,while60.2%oftheprivatetenantshaddonethesame.Ifweuseourprobitmodelwiththebasicsetofcontrolvariables(i.e.,Model1ofTable2A)toimputetheprobabilityofmovingforpublic-housingtenants,theestimatedprobabilityturnsouttobe29.3%,whichisthesameasthesampleaverage.Assumingthesepublic-housingtenantsallbecameprivate-housingtenants,themodelpredictsthattheirprobabilityofmovingwouldchangeto50.9%.Inotherwords,theeffectofpublichousingexplainstwo-thirdsoftheobserveddifferenceinthe31 probabilityofmovingbetweenpublic-housingtenantsandprivate-housingtenants;theremainingdifferencecanbeattributabletodifferencesintheirotherobservablecharacteristicsasspecifiedinourprobitmodel.Table2Bshowsthatowner-occupiersofprivatehousingarelesslikelytohavemovedintheprevious5yearscomparedtoowner-occupiersofpublichousing.Buyingaflatisamajorinvestmentdecisionwithhighfixedcosts.Onceadecisionhasbeenmade,ittendstotiedownowner-occupierstotheircurrentlocation.Itisthereforenotsurprisingthatthedifferenceinestimatedmarginaleffectbetweenowner-occupiersofpublicversusprivatehousingissmallerthanthatbetweentenantsofpublicversusprivatehousing.Neverthelesstheestimatedeffectof0.049inModel1ofTable2Aishighlysignificantstatisticallyandisnon-trivialinthesubstancesense,astheproportionofowner-occupierswhohadmovedinthepast5yearsisonly36%.Theseresultsareconsistentwithourpredictionthatownersofpublichousingaremoreimmobilethanareownersofprivatehousing.Thisconclusionremainsthesamewhetherwefocusonthebasicmodelwitharound50controlvariablesorothermodelsthatincludeafewhundredobservablecontrolvariables.Sinceaddingspousalcharacteristicsorinteractions(i.e.,Models3and4)donotshowsubstantialimpactontheregressionestimates,wewillfocusonthetwobasicModels1and2intherestofthispaper.Wecanprovidebriefcommentsontheeffectsoftheothercontrolvariablesontheprobabilityofmoving.Thegenderofthehouseholdheadisfoundtohavesmallbutnegativeeffectonmobilityforrenters.However,thisnegativeeffectonmobilityturnstobesmallpositiveeffectforowneroccupiers.Age,ontheotherhand,hasstrongeffectonmobility.Generally,peopleinyoungeragegroupshavegreatermobilitythanpeopleinolderagegroups.Forexample,theestimatedmarginaleffectforpeopleaged18–25ofthebasicmodelinTable2Ais28.5percentagepointshigherthanthatforpeopleaged61orabove(theomittedgroup).Thispositiveeffectonmobilityincreasesto45.9%forowneroccupiers(Model1ofTable2B).Withregardtomaritalstatusandemploymentstatus,wefindthatpeoplewhoaremarriedordivorcedtendtobelessmobilethanpeoplewhoaresingle,andthattheunemployedarelessmobilethanemployees.Wealsofindthateducationandhouseholdincomearepositivelyrelatedtomobility,31 whilehouseholdsizeisnegativelyrelatedtoit.However,thepresenceofyoungchildrenincreasesthelikelihoodofmoving.BasedontheestimatesinModel1ofTable2B,theprobabilityofmovingforhouseholdswithchildrenagedfiveorbelowis19.7%higherthanhouseholdswithouttheseyoungchildren,andtheprobabilityofmovingforhouseholdswithchildrenagedsixtotenis13.7%higherthanhouseholdswithoutsuchchildren.4.2.LocationchoiceConditionalonmoving,ahouseholdmaychoosetoliveinanewdistrictorstayinthesamedistrict.SinceChinesecultureencouragespeopletomaintainclosefamilylinks,itisacommonphenomenoninHongKongthatcloserelativesliveinthesameneighborhood.Inmanycases,differenthouseholdsofanextendedfamilyliveinadjacenthousingestatesoreveninthesameblock.Inadditiontothefactthatpeopleprefertostayinfamiliarneighborhoods,theculturalnormsuggeststhat,conditionalonmoving,householdsprefertostayinthesamedistrictofresidence.Sinceprivate-housingtenantshavethehighestflexibilityinhousingchoice,theirbehaviorisagoodindicatoroftheirtruepreferences.Public-housingtenantsandownersofsubsidizedhousing,ontheotherhand,havelimitedfreedominchoosingahousingunitandtheirobservedlocationchoicesmaynotnecessarilyreflecttheirtruepreferences.Inthissubsection,weonlyincludeinthesamplethosehouseholdsthathadmovedintheprevious5years.WhileHypothesisIpredictsthatpublichousingoccupantsarelessliketomove,HypothesisIIpredictsthat,conditionalonmoving,publichousingoccupantsaremorelikelytomoveoutsidetheiroriginalneighborhood.Boyle(1995)findsthatintheUnitedKingdomlong-distancemigrants(thosewhomovedover50km)werelesslikelytomoveintocouncilhousingthanotherhousingtypes.ThisfindingisnotinconsistentwithourpredictionbecausethegeographicalsizeofHongKongissmallandmovesinHongKongarerarelyover50km.Moreover,onereasonforBoyle’sresultisthatlocalauthoritiesofferlimitedopportunitiesforpublichousingtomigratefromoutsidethelocalarea.InHongKong,bycontrast,theallocationofpublichousinginalldistrictsisunderunifiedcontrolbytheHousingAuthority.Intheestimationofthissubsection,sincewearerestrictingtohouseholdsthathadmade31 amoveintheprevious5years,thesamplesizereducesto37,013.Tables3Aand3BshowaprobitmodeloflocationchoicewherethedependentvariableisanindicatorvariableSforwhetherthehouseholdhadmovedwithintheneighborhood.Morespecifically,forthefirsttwocolumnsoftheTable,weassumethattherelevant‘‘neighborhood’’isacensusdistrictandwesetS=1ifthehouseholdhadmovedtoanewresidenceinthesamedistrict,andS=0ifthehouseholdhadmovedtoanewresidenceinadifferentdistrict.RecallthatinourdatafiletheterritoryofHongKongisdividedinto31censusdistricts.ThelargestdistrictaccordingtothisclassificationisEasterndistrictonHongKongIsland,withatotalofapproximately198,000households.ThemediandistrictisWanChai,andithasabout62,000households.Censusdistrictsarethemostdetailedgeographicalsubdivisionsthatareavailableinourdata.InthefirstcolumnofTable2B,wefindthatthemarginaleffectestimatefor‘‘PublicHousingTenant’’isnegativeandstatisticallysignificant.Theestimateimpliesthat,conditionalonmoving,theprobabilityofmovingwithinthesamecensusdistrictis5%loweramongpublic-housingtenantsthanamongprivate-housingtenants.ThisresultisconsistentwiththepredictionofHypothesisII.Incolumntwo,weadddetailedcontrolvariables(Model2)andtheestimatedmarginaleffectfor‘‘PublicHousingTenant’’becomesstronger(from0.052to0.077).Ifwecomparethebehaviorofowner-occupiers(seecolumnoneofTable3B),wefindthatthemarginaleffectestimatefor‘‘Owner-Occupier(Public)’’strongerthanthatfor‘‘PublichousingTenant’’aspresentedinTable3A.Theprobabilityofwithin-districtmoveis17.6%lowerforownersofpublichousingthanforownersofprivatehousing.ThisresultisconsistentwiththepredictionofHypothesisIV.Incolumns3and4ofTable3Aand3B,wechecktherobustnessofourresultsbyusingwithin-regionmoveinsteadofwithin-districtmoveasthedependentvariable.Here,aregionisdefinedintermsofthethreecategoriesofUrban,NewTowns,andRuralthatweexplainabove.Notethatacensusregionasdefinedhereencompassesarelativelylargeandnotnecessarilycontiguousarea.Forexample,ahouseholdthatmovedfromonenewtowntoanothermayencounteracompletelydifferentneighborhood.Forthisreason,weexpecttheresultstobeweakerthanthoseinour31 preferredspecificationsofcolumns1and2.Comparingcolumn3tocolumn1ofTable3A,wefindthattheabsolutesizeoftheestimatemarginaleffect‘‘PublicHousingTenant’’islessthan0.01andisnolongerstatisticallysignificantatthe1%level.OnereasonforthisresultisthatmostofthepublichousingrentalunitsthatareavailableforallocationarelocatedinNewTowns.WediscussearlierthatthereareveryfewpublichousingrentalestatesinOtherRuralAreas,whileunitsinUrbanAreasaremostlyalreadyoccupiedandunavailableforassignment.Asaresult,therelocationofpublic-housingtenantsisprimarilywithintheNewTownsregion.Incolumns3and4ofTable3B,themarginaleffectestimatefor‘‘Owner-Occupier(Public)’’remainsnegativeandsignificant.Forexample,theestimatesincolumn4suggesttheowner-occupiersofpublichousinghada14.2%lowerprobabilityofmovingwithinthesamecensusregioncomparedtoowner-occupiersofprivatehousing.Althoughthiseffectisweakerthanthecorrespondingestimateincolumn2,themagnitudeoftheeffectisstillsubstantial.Inshort,theresultspresentedinTables3Aand3BsupportHypothesesIIandIV.Onealternativeinterpretationforthefindingthatpublichousingoccupantsmovefurtherconditionalonmovingisasimplethresholdmodel.Sincethecostofmovingishigher,theyonlymovewhenitisreallyworthwhiletomove,anditisreallyworthwhiletomoveiftheoptimallocationisveryfarfromthecurrentlocation.Accordingtothisalternativeinterpretation,public-housingtenantswhomovedshouldliveclosertotheirplaceofwork,assumingco-locationofworkandresidenceisoneimportantdeterminantofoptimallocation.Howeverwedonotfindevidenceforthis.Amongpublic-housingtenantswhomoved,7.4%liveinthesamedistrictastheywork(thecorrespondingpercentageforpublic-housingtenantswhodidnotmoveis21.3%).Amongprivate-housingtenantswhomoved,19.0%liveinthesamedistrictastheywork(thecorrespondingpercentageforprivate-housingtenantswhodidnotmoveis10.2%).Thesefiguresdonotsupporttheinterpretationthatconditionalonmoving,privatetenantsaremovingtofartherbut‘‘optimal’’locations.WediscusstherelationshipbetweenplaceofresidenceandplaceofworkingreaterdetailinthesubsectionthatFollows.4.3.Placeofwork31 Simpson(1987)developsamodelinwhichresidentiallocationandworkplacelocationaresimultaneouslydetermined(seealsoSiegel,1975;Kimetal.,2005).Inthispaper,wehypothesizethat,duetolackofchoice,publichousingtenantsarelesslikelytofindahousingflatthatislocatedclosetotheirplaceofwork.Moreover,thechoiceofresidentialandworklocationisinherentlyadynamicdecision(Vickerman,1984).Thelackofmobilityamongpublic-housingtenantshinderstheattempttoadjusttheirresidentiallocationwhencircumstanceschange(suchaschangingtoanewjob),thusincreasingthechanceofamismatchbetweenplaceofresidenceandplaceofwork.Inthispaper,wefocusonthedecisionofthehouseholdhead.Sinceunemployedoreconomicallyinactivepersonsdonotreportaplaceofwork,theanalysisinthissubsectionexcludessuchindividuals.Theresultingsamplesizewithvalidobservationsforthe‘‘PlaceofWork’’variableis53,334.WecodethevariableWequals1ifthehouseholdheadlivesinthesamedistrict(orregion)asheorsheworks,andWequals0otherwise.TheresultingprobitestimatesaredisplayedinTables4Aand4B.ThesecondcolumnofTable4Ashowsthatpublic-housingtenantsaresignificantlylesslikelytoworkwithinthesamecensusdistrictoftheirresidencecomparedtoprivate-housingtenants.Theestimatedmarginaleffectshowsthattheprobabilityis5%lowerforpublic-housingtenantsthanforthereferencegroupofprivate-housingtenantstoresideinthesamedistrictastheywork,holdingothercharacteristicsconstant.FromTable4B,wefindthattheestimatedmarginaleffectfor‘‘Owner-Occupier(Public)’’isalsonegative(minus3%)andsignificant.Inotherwords,ownersofpublichousingarelesslikelytoworkwithinthesamedistrictthanareownersofprivatehousing.Notethatwealreadyincludecensusdistrictdummiesinourregressions.Thusourresultisnotdrivenbythefactthatpublichousingestatestendtobelocatedfartherawayfromcentersofeconomicactivities.Inotherwords,evenifwecomparebetweentwohouseholdsresidinginthesamecensusdistrict,theonewholivesinpublichousing(rentedorowned)ismorelikelytoholdcross-districtemploymentthanhisorhercounterpartinprivatehousing.Incolumns3and4ofTables4Aand4B,wedefinethedependentvariableWintermsofwhetherthehouseholdheadholdscross-regionemployment.Theresultsaresimilartothose31 intheearlierspecifications.Evenaftercontrollingforregionaleffects,occupantsofpublichousingarelesslikelytoholdajobwithinthesameregionoftheirresidencethanaretheircounterpartsinprivatehousing.TheseresultsareconsistentwiththepredictionofHypothesesIIIandV.InapaperbySimpson(1980),hearguesthatahouseholdresidentiallocationisalsoinfluencedbytheworklocationofthesecondaryearnersofthehousehold.Wethereforechecktherobustnessofourresultsbychangingtheunitofobservationtobethespouseofthehouseholdhead.Sincemanyhouseholdheadsdidnothavespousesorhadspouseswhodidnotwork,thesamplesizeisfurtherreducedto27,490.BasedonthespecificationofModel2,wefindthatthemarginaleffectestimatesfor‘‘PublicHousingTenant’’and‘‘OwnerOccupier(Public)’’intheequationforsame-districtemploymentare0.038and0.057respectively,andbotharesignificantatthe1%level.Ifwereplacedistrictdummiesbyregiondummiesandfocusonthelikelihoodofworkinginthesameregion,theestimatedmarginaleffectsbecome0.056and0.066for‘‘PublicHousingTenant’’and‘‘OwnerOccupier(Public)’’respectivelyandbotharestatisticallysignificant.Justlikehouseholdheads,secondaryearnersinpublichousing,rentedorowneroccupied,havealowerprobabilityofworkinginthesamecensusdistrictorregionastheylivewhencomparedtosecondaryearnersinprivatehousing.5.ConclusionsInHongKong,publichousingconfersasizablesubsidytoalargefractionofthepopulation,andpublichousingunitshavetoberationedamongcompetinghouseholds.Studiesonpublichousingpolicyoftenfocusonthefiscalburdenimposedbysuchmassivesubsidies,buttheallocativeinefficienciesinducedbyrationingsubsidizedhousingarelesswelldocumentedintheliterature.Thisstudyinvestigatesthehiddencostsofpublichousingfromtheperspectiveofthemisallocationofhousingunitstotenantsbyexamininghowthismisallocationaffectsthesetenants’behaviorconcerninginternalmobilityandtravel-to-work.Thisstudymakesuseofthe5%randomsub-sampleofthe2001HongKongPopulationcensustotestthefollowingfivehypotheses:(i)Publichousingoccupantsaremoreimmobilethanprivatehousingcounterparts.31 (ii)Conditionalonmoving,public-housingtenantsaremorelikelythantheirprivatehousingcounterpartstohavetomoveoutsidetheiroriginaldistrictorregionofresidence.(iii)Public-housingtenantsarelesslikelytoworkinthesamedistrictorregionastheirplaceofresidence.(iv)Conditionalonmoving,publichomeownersaremorelikelythantheirprivatehousingcounterpartstohavetomoveoutsidetheiroriginaldistrictorregionofresidence.(v)Publichomeownersarelesslikelytoworkinthesamedistrictorregionastheirplaceofresidence.Weestimateprobitregressionsonthecensussampletotestthesehypotheses.Allempiricalresultsarestatisticallysignificantandconsistentwithourpredictions.Byidentifyingthedistortionsinlocationchoicesinducedbypublichousing,thisstudyservestoimproveourestimateofthetruecostsofpublichousingpoliciesinHongKong.Eventhoughthesecostsarelesstangiblethantheobservablefiscalcosts,theyhaveadirectimpactontenants’everydaylives.Policymakersshouldexplorethepossibilityofintroducingprivate-marketrentsubsidies(see,forexample,Gruisetal.,2005andJohnsonetal.,2002)toachievetheobjectiveofprovidingaffordablehousingtothelesswell-to-dowhileavoidingmanyofthedistortionsduetogovernment-providedhousing.Lastly,wewouldliketoemphasizethatourconclusionsarebasedoncorrelations.Whilewehaveintroducedextensivesetsofcontrolvariables(sometimestotalingtoaround1000variables)inordertocontrolforobservabledifferencesbetweenpublicandprivatehousingoccupants,westillcannotexcludethepossibilityofdifferencesinunobservedcharacteristicsthatmayaffecttheoutcomevariables.Eventhougheconomictheorysuggestsacausallinkbetweenpublichousingoccupancyandreducedresidentialmobility,theresultspresentedherecannotprovidedefinitiveevidenceforsuchalink.Intheabsenceofexperimentaldataorconvincinginstrumentalvariables,acausalinterpretationoftheresultsshouldbeadoptedwithsomeCaution.31'