Conference PaperPDF Available

The impact of China's Housing Provident Fund on Homeownership and Housing Quality: Evidence from two surveys

Authors:

Abstract and Figures

In the absence of well-functioning credit and mortgage markets, the Chinese government has implemented several reforms to encourage home purchases. Among the most prominent of these is the Housing Provident Fund, which requires employers and employee contributions to a pool which is later used to make mortgage loans to participants. We use the 2011 Chinese Household Finance survey and a smaller survey from Jinan, Shandong province to examine the extent to which the Fund encourages the acquisition of owner-occupied housing. We find that fund participants, depending on the size of the contributions and length of time in the program, display a very modest increase in homeownership. Fund utilizers also purchase smaller properties than they otherwise might have, due to the down payment restrictions on mortgage loans. These results are somewhat stronger for younger households. The HPF actually seems to encourage investment in real estate more than owner-occupation, per se. 2
Content may be subject to copyright.
1
The impact of China’s Housing Provident Fund on Homeownership and Housing Quality:
Evidence from two surveys
Mingzhe Tang
Department of Economics
Shandong University
N. Edward Coulson
Department of Economics and Lied Institute for Real Estate Studies
University of Nevada, Las Vegas
This paper is preliminary, comments welcome at mingzhe.tang@gmail.com or
n.edward.coulson@gmail.com
ACKNOWLEDGEMENTS: Mingzhe would like to thank Mengyao Pan for excellent research
assistance, and Shandong Social Science Planned Research Office for grant support (grant no.
13CJRJ06). Special thanks for comments byJie Chen, Charles Leung, and participants at the
following conference presentations: the 8th Hong Kong Economic Association Biannual Conference,
the 4th International Workshop on Regional, Urban and Spatial Economics in Tsinghua University,
and the 2015 AREUEA International/Global Real Estate Summit in Washington DC.
ABSTRACT: In the absence of well-functioning credit and mortgage markets, the Chinese
government has implemented several reforms to encourage home purchases. Among the most
prominent of these is the Housing Provident Fund, which requires employers and employee
contributions to a pool which is later used to make mortgage loans to participants. We use the 2011
Chinese Household Finance survey and a smaller survey from Jinan, Shandong province to examine
the extent to which the Fund encourages the acquisition of owner-occupied housing. We find that
fund participants, depending on the size of the contributions and length of time in the program,
display a very modest increase in homeownership. Fund utilizers also purchase smaller properties
than they otherwise might have, due to the down payment restrictions on mortgage loans. These
results are somewhat stronger for younger households. The HPF actually seems to encourage
investment in real estate more than owner-occupation, per se.
2
1. Introduction
Before 1988, questions about the housing tenure choices of Chinese households were
unaskable. Housing, like much of the economy, was tightly controlled and characterized by a strong
link between housing allocation and place of employment (and status within the workplace). The
reforms which began in 1988 changed all that. No longer was it necessary to find housing through
one’s employer; instead, Chinese households could find housing (if they chose) in the private market,
and homeownership (although not always with the full set of property rights enjoyed by households in
the west) became possible1. Fu, Tse and Zhou (2000) , Huang and Clark (2002), Li (2000), Meng,
Gregory and Wang (2005)) and Logan, Bian and Bian, (2002), among others, investigated household
behavior in the wake of these reforms and several stylized facts were uncovered. Among these were
that many of the household characteristics that motivated owner-occupation in the West also did so in
China; that nevertheless the take-up of homeownership was very low due to high prices associated
with the private housing market; and that employer housing was still preferred for that reason.
Affordability was, and remains, a major stumbling block. Contributing to this is the lack of a
well-developed credit market. Mortgage financing of the type found in, say, the US, with long-term
fully-amortizing payback structures, is still quite uncommon.
In the absence of such institutions, the Chinese government has turned to savings augmentation
plans—specifically the Housing Provident Fund (HPF), to help Chinese households become
homeowners2. In the next section we provide more details about the HPF; it suffices here to say that
the fund is created through mandatory contributions by employees and employers, and gives
contributing employees access to the fund for the purpose of purchasing a home. The research in
this paper seeks to explain the extent to which the HPF helped in this regard both in the sense of
1 In this paper, we use the term homeownership in the Western sense of a household owning and
occupying a housing unit, and will sometimes use the term owner-occupation instead. In China,
households often invest in real estate without residing in one of the units they own, perhaps because they
still receive subsidized rental housing from their employer. While such households own homes, and are in
the strict sense homeowners, this activity is referred to as real estate investment.
2Housing provident funds are not unique to China. Singapore’s Central Provident Fund, which started in
1955, can used for housing purposes. HPFs have been a policy tool in a number of developing countries
including Mexico, Brazil, Nigeria, Jamaica and The Philippines although the details differ across countries
(Chiquier, 2009).
3
achieving homeownership and in achieving higher quality housing. We do so mainly through the use
of the Chinese Household Finance Survey. This new survey of Chinese households contains detailed
information about real estate holdings, living arrangements, credit market experiences, and
sociodemographic characteristics of the respondents. Among the questions on credit market
experiences (which essentially inquire as to the method of financing a home purchase) are questions
about the use of the Housing Provident Fund, including length of time in the fund, monthly
contribution rates and the like. We use this data to construct econometric models of
owner-occupation by Chinese households. We examine three decisions. First, we model the binary
decision to become owner-occupiers, as a function of demographic and locational characteristics as
well as HPF access. Next, we investigate how the HPF affects housing quality choice. HPF
regulations discourage the purchase of large units by imposing greater burdens on purchases of such
units. Thus we estimate a probit model that examines whether a large or small house (defined below) is
purchased conditional on purchasing a home at all. Given these two probit models we finally estimate
a model that predicts the precise size of the purchase.
To preview our results, we find that factors that measure a household’s access to HPF do have a
positive influence on the decision to become homeowner-occupiers. However that influence does
not appear to be strong. Moreover the extra burden, in the form of larger down payments,imposed
on households who want to use the HPF to buy larger units does appear to push such households
toward smaller properties. These results are somewhat stronger for younger households (those whose
heads are less than 40 years of age), a result of their greater exposure to the Fund.
We conclude with a final model which estimates not whether HPF encourages homeownership but
rather simply real estate investment. As is well understood, Chinese households often view real estate
not as a consumption good as much as an investment opportunity, in part because of the lack of other
asset markets (Coulson and Tang, 2013). Our results indicate that the HPF seems to encourage real
estate investment somewhat more than its stated goal of homeowner-occupation.
In the next section we describe the HPF in more detail, in order to set the stage for our empirical
analysis. In section 3 we describe the survey, and in Section 4 we present the empirical model and
regression results. Section 5 concludes.
4
2. The Housing Provident Fund
The Housing Provident Fund was established in Shanghai in 1991 and has since been implemented in
most large Chinese cities3. The operation of the fund proceeds from mandatory contributions by
workers and employers into the fund, for later withdrawal for the express purpose of home
purchase4. In the initial implementation, the Shanghai government required both public work units
and employees to reserve 5% of wages for the fund. The interest paid on this fund was set by the
Shanghai government at a rate r that corresponded to the extant risk-free rate of return (Buttimer, Gu
and Yang, 2004). Given the employer contributions, the rate of return on the employee’s fund
contribution was 100+r percent, obviously well above market rate. The funds are pooled and then
lent out to qualified contributors for the purpose of purchasing a dwelling. The loan may be at
subsidized rates (Yang and Chen, 2014).
These contribution requirements were under the control of the local authority, and so the exact
administration of the fund varied across municipalities5. Normally the local authority decides the
minimum and maximum contribution rates from an employee’s monthly salary by the employer, and
all work units by law are bound by this minimum rate. Yang and Chen (2014) note the broad
variation in rates over time and across municipalities. In the 1990s the rates were around 4% in
Jinan, which soon increased to 5% (Zhao and Bourassa, 2003). Yang and Chen (2014) observe
similar increases in other cities; for example in Tianjin the rate was 8% from 2003 through 2005,
and then increased steadily over the next several years until it reached 11% in 2008 and beyond. In
Hangzhou the rate in 2003 could be as low as 8%, but was 12% in 2012.
The actual contribution rate could depend on other factors even within cites. In Jinan the
rate quoted in the previous paragraph was higher—8%-- for foreign employers. The rate in
3 Previous studies of the HPF include Yeung and Howes (2006), Zhao and Bourassa. (2003), and Deng, Shen,
and Wang (2011). An interesting theoretical simulation of the HPF is in Buttimer, Gu, and Yang (2004).
Empirical analyses similar in flavor to our own include Xu (2012) and Ying, Luo and Chen (2013). But these
papers do not have the level of detail and/or the geographic scope of our data.
4 While some larger cities such as Beijing and Shanghai allow qualified individuals to withdraw their HPF
balances to pay the rents, as of 2013 this is not a wide practice. We assume in this sample that housing
purchases are the sole use of the fund.
5 Though the regulations on contribution are determined by the local municipalities, namely the local housing
provident fund management centers, other regulations on the usage of the HPF fund, such as the interest rates
charged on the HPF loan and down payment requirement and so on, are determined and implemented at the
nationwide level as other financial sectors regulated by the Peoples Bank of China, and local fund management
centers usually do not change them.
5
Guangzhou can apparently be as high as 20% (for reasons that are not given). Cities such as Shanghai,
allow employers in addition to the required HPF account, to establish a supplementary housing
provident fund account for their employees, which serves as an additional implicit tax-free housing
benefit to reward their employees. Thus the adoption of the HPF, even within cities, differed across
sectors. It was initially mandated in state and state-supported sectors, and as time progressed was
mandated for foreign-owned firms and other sectors. By 2003, almost all sectors have been
mandated to contribute to housing provident funds. However, even today it has not been taken up
by the less formal sectors of the Chinese economy. The importance of this is that the ability to use
the fund for its intended purpose will depend on the particular employment circumstances (including
location) of the particular household. Thus any econometric analysis of the effects of the HPF will
need to control for these characteristics.
3. Data
Our data is drawn from the Chinese Household Finance Survey of 2011 (CHFS, hereafter)6. The
CHFS adopted the popular stratified probability proportional to size (PPS) sampling strategy, which
proceeds in three stages. First, primary sampling units (PSU) were established to include 2,585
counties from all provinces and large special municipalities throughout the whole of mainland China
except for Tibet, Xinjiang, Inner-Mongolia, Hong Kong and Macao. Thus the sample is
representative of the diverse geographic regions of mainland China, covering over a population of
one billion. From these, 80 PSUs were chosen. The secondary stage involves sampling from those
80 chosen counties/cities about 4 residential committees/villages per county/city and the last stage is
to randomly sample about 20-50 individual households per committee/village from those selected at
the previous stage. Then a household representative will be interviewed in person over all survey
questions. There are a total of 8438 households interviewed in the survey, including both urban and
rural households. Around 60 percent are from the eastern region of the country, the most
economically-advanced. About 25 percent are from the middle region, and around 15 percent from
the relatively less-developed west.
6 The survey is funded by the People’s Bank of China and conducted through Survey and Research Center
for China Household Finance at the Southwestern University of Finance and Economics(SWUFE), China.
The detail about the survey can be found at http://www.chfsdata.org/detail-14,15.html and also Gan
et.al.(2013).
6
Tables 1 through 3 provide some summary statistics on, respectively, the housing market experiences
of the respondents, their use of the HPF, and their socio-demographic characteristics7. In each of
these tables we report two tabulations. The first is for the sample as a whole. The second is for
those households whose head is less than 40 years of age. The Fund is, as noted, relatively new, and
its biggest impact is on younger households, who have, roughly speaking entered the workforce after
its introduction. Therefore in the analysis that follows, we expect the extent and importance of the
HPF to be stronger for this group.
Table 1 shows that the vast majority, 90%, of the respondents own property. Ownership of
multiple properties is common. 73 percent of the respondents did own a single unit, but over 13
percent owned two units and over two percent owned more than that. The average number of
property owned by all those homeowners is 1.21, with the maximum reaching 15 units8. Of the 5686
property owners in the survey, 90 percent were owner-occupiers. The owner-occupation rate in
the survey is therefore about 80% which is in line with data from other Chinese data (of urban
households). The average size of the unit is approximately 112 square meters9; the reduced sample
size (about 100 missing obs.) for this variable is due to non-reporting. We believe the cause of the
non-reporting is lack of knowledge; in the modeling below we do not treat the sample selection
arising from non-reporting as endogenous. Table 1 also reports the year when the households
bought their property. Around 50% of these homeowners owned their property after 2000, one year
after China officially abolished the housing welfare allocation system. The data also shows that since
2000 about 300 households had become homeowners every year until 2009.
Table 1 also reports the homeownership situation for the 40-and-under-aged group. There is not
much significant difference from the all-age-inclusive group. The homeownership rate is 84.5%, 3
7 Note for the purpose of this study, we exclude 2289 households with strictly rurual hukou who are not
covered by the housing provident fund. See the later paragraph on the hukou system in China and the
variable ruralp.
8 The Survey Center at SWUFE does not release information on those very wealthy families in the CHFS
data public available, thus we do not have the data for those households in our sample.
9 According to the official statistics released by the Ministry of Housing and Urban-rural Development of
China, in 2012, the building area per capita is 32.7 square meters for urban areas and 36.2 for rural areas(see
the Chinese Central TV’s report: http://news.cntv.cn/18da/20121112/105864.shtml). The average
household size in China is 3.10 in the 2010 census. According to the Chinese Family Panel Studies by Peking
University, the average dwelling size for Chinese households is 116.4 square meters in 2012.
(http://www.isss.edu.cn/cfps/EN/) We also remove 9 observations whose unit size is less than 10 square
meters.
7
percentage points lower but the occupation rate is only 67%, much lower than the all-aged group.
Households of the younger group generally buy their property three years later than the all-aged
group and therefore the average size is a little larger since the floor areas of new residential
constructions in China has increased consistently over the years after the reform.
Table 2 displays some summary statistics for HPF participation. Despite the government’s mandatory
contribution requirement for all work units since 2003, Table 2 shows that only 18 percent of the
households had at least one household member contributing to a HPF account in the sample by
2011. Some of this is due to self-employment or other employment circumstance that precludes
HPF participation, but it is worth noting that the lack of penetration of HPF accounts will preclude
the HPF program from making a significant difference in Chinese mortgage markets10. We also
observe a tremendous variation in the household’s monthly deposit as well as in the HPF
participation duration. While the mean monthly contribution is 470 RMB per HPF-holder
household, the 95th percentile is 1700 RMB, and the maximum is 11,920 RMB. The average
duration of contribution per HPF-holder household is around 129 months with a standard deviation
of over 123 months. Compared to the all-aged group, the 40-and-younger-aged group has much
higher housing provident fund penetration rate, greater monthly contribution deposits with larger
variation as well, though shorter duration of participation into the program.
Table 3 reports some important demographic characteristics of Chinese households relevant to our
study. The average age in our sample is 48.5 years and the proportion of households with a married
couple is around 87%11. The average education level is about middle school and a little over - 20% of
the population has associate and above education. Due to increased mobility in the labor market, the
average number of people currently living together in the same household is 3.49, less than 3.77, the
mean of the reported family number of the household. The average number of children per
household is only 0.76 and most families have only one child, as a result of the one child policy.
Compared to the all-aged group, households of the younger group have higher education levels on
average, over 30% obtaining some college education and above. There are more people holding
advanced job positions or owning non-agricultural businesses. They have higher income as well as
10 Housing provident fund regulations give waiver rights to businesses that make a loss in a particular year,
and to individuals who emigrate, go bankrupt, are laid off or endure similar events.
11 We remove households whose representatives’ age is less than 16 years old, because the minimal age for
an individual with full civil rights is 16 years of age.
8
greater wealth on average, but the greater standard deviations in both measures, 30% larger than the
all-aged group.
In China, there has been a long history of separating urban population from rural population and the
government has used it as an important policy instrument to achieve the country’s economic
development goals (Coulson & Tang, 2013). Every household in China is required to have a
registered residency with a local government authority, either urban (or “non-agricultural”) hukou or
rural (or “agricultural”) hukou. Historically, rural hukou holders had their farmland and conducted
agricultural and related businesses, they were accordingly allocated a certain amount of land and
allowed to build their own dwellings12. As a result, the housing provident fund is designed only to
cover urban hukou holders who need to purchase their housing from the market. However, in
concert with market economic reform, more and more rural households have migrated into cities, and
some have successfully obtained an urban hukou (Chen and Coulson (2002), Chan (2010), Wu and
Trieman (2004)). In the CHFS, when a household is surveyed, in addition to the household head or
the survey respondent, other members of the household are listed, and some of these family members
may have a hukou status different from the head/respondent13. Therefore to understand the effect
of housing provident fund policy, we construct the variable “ruralp”, the fraction of rural hukou
holders in the surveyed household, to reflect this characteristic. The mean is 0.31, which means a
large proportion of the surveyed households have more family members holding urban hukou. Since
only urban-hukou holders are qualified for the housing provident fund, we remove all pure rural
households for the purpose of our study. This results a sample size of 6322 households. The
correlation between ruralp and the HPF dummy (“hpfdy”) is significantly negative, around -0.338.
Interestingly, the size of the mixed household where family members have different hukou status is
4.6, much bigger than that of the pure urban households. One possible explanation for this
somewhat curious result is the rural-to-urban migration of one family member, who then obtains
urban hukou, and is then joined by multiple family members (who retain rural hukou). Average
12 Yet this land ownership right is called collective land ownership right. Farmers as individual households
do not own the land, the whole village does. As a result, individual farmers do not have full ownership
rights for these farmlands, yet they are able to build dwellings in accordance with household size.
13 In fact, in CHFS, two measures of “rural” status are used. One measure is location-related, meaning the
household is urban or rural when interviewed if it is located in urban or rural area defined by the National
Statistics Bureau. The other measure is hukou-related. Since the second definition is closely related to HPF,
we adopt the second measure in our study and construct “ruralp” accordingly.
9
household annual labor income is 68,553 RMB in 2010, and the average household financial wealth
excluding real estate holdings is 78,375 RMB. Both measures display large variation and are highly
skewed to the right. The standard deviation of income is 180,749 RMB, and the standard deviation of
wealth is 302,594 RMB14.
Because of China’s gradual economic reform (Liew (1995), Feltenstein, Nsouli(2003)),
households with certain employment or occupations could enjoy privileges in the housing market that
cannot be measured by the observed wealth or income. To capture this heterogeneity, we use
information such as the number of people in the household that hold leadership position in
government, party organizations, enterprises and other entities, the number of people employed in
government agencies, the number of people employed in the state-owned enterprises, or whether the
household owns any business, either agricultural or industrial. There are around 28 percent
households owning an agricultural business, and only 15 percent owns a non-agricultural business.
Households owning a business have a greater homeownership rate than those not-owning any
business, and agricultural business owners have a higher homeownership rate than none-agricultural
business owners. But non-business-owning households have a higher HPF coverage rate than
business-owning households.
Finally we also have data on household purchases of real estate. This is of importance
because of the implicit goal of the HPF program to discourage “overinvestment”. In particular
those who want to purchase units under 90 square meters are only required to make a 20% down
payment, while those who wish to purchase larger units must make a 30% down payment if they wish
to utilize HPF15. The CHFS contains information on the size of the purchased units; we construct a
binary variable which indicates whether the unit was bigger or smaller than 90 sq.m. as a function of
demographic and HPF measures. There are almost equal number of small-unit homeowners (50.7%)
and large-unit homeowners (49.3%) in the data.
4. Econometric models
14 The average exchange rate between Chinese yuan (or RMB) and US dollar in 2011 is around 6.76:1. So
the above numbers would be $8,595, $9,108 , $21,120, $33,730.
15 This regulation was issued by the People’s Bank of China in the early 2000s and has been implemented
throughout the country until early 2015. We consider that this specific-90-sq.m.-threshold was derived
based on the then-desirable-housing-size of 30 sq. m. per capita and the standard family size of three
people (due to the one-child policy).
10
Our empirical strategy centers on our desire to investigate two related questions: First, does
participation in the Housing Provident Fund increase the probability of owner-occupation in Chinese
households; and second, does participation in the fund encourage the purchase of smaller units due to
the differing down payment requirements. The first question, in isolation, could be answered with a
simple probit model. The second part is somewhat more intricate. A simple regression of unit size
on a set of variables that included measures of HPF exposure is inadequate since the use of the HPF
is more problematic when households purchase large units. Put another way, the regression
parameters differ when large and small units are chosen. However we cannot simply split the sample
and estimate separate regressions for large and small choices of size. A split of a regression function’s
dependent variable necessarily raises issues about the truncated distribution of the error term. Our
resolution of this is to estimate a bivariate probit model for the choice first to become a homeowner,
and conditional on an affirmative answer to that choice, to then choose a large or small unit. In a
subsequent pair of regressions of actual size (in square meters) on variables of interest (including HPF
measures) bivariate Mills ratios that account for the truncation in the subsequent stage are included.
The estimation of two probits need to be joint because the unobservable household characteristics
that lead households to become owner-occupiers would probably be correlated (one might assume
positively) with the characteristics that drive the size decision.
To be specific, consider the bivariate probit of the owner-occupation decision and the “large or
small” decision, bearing in mind that only owner-occupiers will make the second decision. We
specify a random utility model of these two decisions:
=+
(1)
=+
(2)
Where = 1 if the household becomes an owner-occupier, and 0 otherwise, and (conditional on
= 1), = 1 if a small (less than 90 square meters) unit is purchased. and ’s are vectors
of conditioning variables. The likelihood function is written as:
11
(,,|,)=ln()

+
 (ln(,,)

 (ln(,,)
 )
(3)
where is the correlation coefficient of the and , and , is the -dimensional normal c.d.f.
Now actual choice of square footage is modeled. The estimation method follows that of Lahiri and
Song (2000) and Hotchkiss and Pitts (2005); these authors treat the problem of estimating a linear
model where the dependent variable is subject to a two-stage selection procedure. The parameters of
that linear model can be consistently estimated with the addition of two-dimensional Mills ratios that
can be calculated using the parameters of the bivariate probit model above.
So consider the linear regression model
=+
(4)
where represents small () or large unit (), is the size of the th owner-occupier’s residence,
is a set of covariates and is a residual. The issue is that when the sample of
owner-occupiers is self-selected, the expected value of is not zero. This is corrected by inserting
the conditional mean of into the model (Hotchkiss and Pitts, 2005) where are the
aforementioned bivariate Mills ratios. To be specific, for the small group,
=(|= 1, = 1) + =[| >,> ]+.
(5)
For the large group,
=(|= 1, = 0) + = >,< +.
(6)
Inserting the bivariate Mills ratios to obtain the following
[|>,> ]=+
(7)
12
>,< =+
(8)
where
=()[ 
(1)/ ]
(,;)
=()[ 
(1)/ ]
(,;)
=()[1 
(1)
]
(,;)
=()1 
(1
)

(,;)
.
These ’s are therefore added regressors in (4) and consistent estimates of the parameters of (4) are
thereby obtained. Because they are generated regressors, the standard errors of the coefficients are
bootstrapped.
5. Results
We estimate the model described in equations (1) through (8) using data from the CHFS. The data
vectors W and X are roughly as presented in Tables 2 and 3, with special attention paid to the three
housing provident fund parameters: hpfdy, the dummy variable to indicate whether any family
member in the household has an active HPF, hpflength, the total participation duration in the HPF
by all family members in the household measured by months, and hpfmonth, the total monthly
deposit into the HPF accounts by all family members.
13
We first consider the owner-occupation decision. Table 4 presents the results of the probit model for
homeownership. We present two specifications, for the sample as a whole and for the under-40
group. These specifications include the demographic variables discussed in section 3, along with
province fixed effects. Marginal effects at the sample means are presented in subsequent columns.
Column 1 shows generally that the same variables that influence homeownership in the west also
have an impact in China. Age of householder and size of household, including number of kids,
influence this decision in the expected positive direction. One puzzling result is that compared to
single people, married couples are less likely to become owner-occupiers. One possible explanation
comes from the results of Wei, Zhang and Liu (2012) who claim young single men bought properties
to increase their value in the marriage market. Also, increases in income do not cause a rise in
ownership probabilities, except at relatively high incomes. This is not incongruent with US results,
when the distinction between permanent and transitory income is taken into account (Goodman,
1988; Coulson, 1999), the first being much more important than the second. Increases in wealth are
far more important for ownership attainment. Also important is self-employment, which is of
interest in that such households do not contribute to HPF. Any effect of other employment
conditions cannot be precisely estimated and are in any case exceedingly small in general. All of
these results basically hold with the younger sample.
Importantly, in both the broader and younger samples, each of the three HPF parameters has a
positive coefficient, which is consistent with our prior expectation that the fund has helped families
obtain homeownership. For the full sample the effect of participation itself seems rather large;
enrollment raises ownership by three percentage points, although the effect is not precisely measured
(t=1.45). Moreover, the other two measures of HPF participation are very small and very
imprecisely measured. The test that all three HPF coefficients are zero cannot be rejected at usual
significance levels.
We regard the results for the younger sample as being slightly more encouraging. The participation
coefficient is smaller in magnitude, but the coefficients on contribution size and length of
participation are larger, and in the latter case, measured with more precision (t=2.25). This latter
coefficient indicates that 12 month longer participation raises the ownership probability by 0.6
percentage points. In this younger sample the joint hypothesis on the three HPF coefficients has a
14
p-value of 0.02, broadly indicating that the Fund has a positive influence on the owner-occupation of
younger participants.
Next we investigate the size of housing consumption. Table 5 reports the results from the size binary
probit regression with selection bias corrected using the model from Table 4 as the first stage model).
The first thing to note is that the selection matters. The Wald test of no correlation between the
two error terms is rejected at any usual significance level. The coefficients generally make sense, and
often signal large impacts of the demographic variables in both samples, more so than the ownership
model in Table 4. For example, households with higher incomes or wealth levels own larger homes.
So do households with family members being an entrepreneur. Employment in state-owned
enterprises however does not result in larger housing consumption, but this negative effect can be
alleviated to a large extent if some family member holds a leadership positions in their work unit.
Government employment gives the household a sizeable advantage in housing consumption which
can be further strengthened if a family member has advanced rank in the government. This yields a
vivid picture of current Chinese housing consumption stamped with features of typical Chinese
gradual economic reform. The gradualistic economic reform in China results in dualism in
economic development, i.e. some industries are open to the market, others are not. And the ratio
between the two is under the control of the government and planned and implemented according to
economic and political goals. This process necessarily has created huge room for rent-seeking
behaviors. Sectors not open to the market are state-owned and nowadays are mostly in heavy
manufacturing industry, energy industry, telecommunications industry, banking industry and so on.
State-owned enterprises in these industries even have bureaucratic administrative ranks comparable to
those of governments of different levels. Because these SOEs have a very strong bond with the
government which holds vast resources, the government often provides many benefits to them, such
as cheaper loans, government procurement contracts, cheaper land, and so on. For example,
accompanied with the housing market reform, the government has implemented various policies to
provide affordable housing for qualified middle and low income households that is much less
expensive than purchase from the private marketHuang, 2012. Because government is the only
supplier of the land, employment with government or state-owned enterprises or related sectors thus
offers some advantages in obtaining such cheaper housing.(because the average wages in these sectors
are nominally lower than other sectors.) Because this economic-comfortable housing supply is limited
and its allocation is not by market, employees in leadership position in aforementioned sectors
15
normally are able to buy one and more likely to buy a larger unit than ordinary employees. This
housing may even be built by some subsidiaries of these SOEs that obtained the land parcel at
cheaper price or free from the government16.
Returning to the variables of primary interest, in the full sample, all three coefficients of the HPF
parameters in probit model indicate that increased exposure to the housing provident fund in any of
the three dimensions is associated with buying smaller houses. (Again recall that this result controls for
income and wealth.) These results are statistically and substantively significant. Other things equal,
those with HPF accounts are 3.9 percentage points more likely to buy a smaller unit (although this
effect is not precisely estimated). A 500 RMB increase in the monthly contribution is associated
with a half-percentage point increase in that same probability (this effect, though somewhat small, is
statistically significant). One year longer in the program is also associated with a greater propensity
to purchase a small house (by less than a fifth percentage point). The effects are jointly significant
with a prob-value of 0.019.
At first glance, these results are counter-intuitive, since enrollment in, and exposure to, the program is
nominally associated with greater resources for purchasing and occupying a greater amount of real
estate. As noted above, we speculate that this is due to the more onerous downpayment requirements
that arise when the purchase is of a large unit, one with more than 90 square meters. Households face
a tradeoff: greater exposure to the fund increases their financial means to buy a house, but the wealth
requirements (in the form of the downpayment constraint) to buy a bigger house cause them to
purchase less housing when they choose to avail themselves of the fund.17 The results indicate that
generally the latter path is chosen.
16 The Economic-Affordable Housing also called economic comfortable housingwas initiated in 1998
with the goal to help middle and low income households to become homeowners. However, it has now
become “latent benefits and privileges” for government officials or some SOEs’ high-ranked
employees.( You, J. (2013) , International Business Times, Feb.2013, see the link:
http://www.ibtimes.com/chinas-crazy-real-estate-market-bubble-bursting-1021560, as well as various news
report in China. (http://finance.qq.com/a/20130623/000120.htm,
http://house.sina.com.cn/focus/jjsyf/))
17 In considering these results two things should be remembered. The first is that the sample in the size
probit regression is only of homeowners so that the decision to buy a home has already been made (and
controlled for). The decision tree we imagine is one in which the household first decides to become an
owner, then decides how big a house to buy, which seems to be influenced by the household exposure to
HPF. Therefore, we are not considering those who do not buy a house because they only will consider a
large unit but cannot afford the larger downpayment (even with the help of HPF). The second is that the
16
There is some variation in these results for the younger sample. In this group the enrollment effect
is stronger. The probability of buying a smaller unit rises by 5.7 percentage points (as opposed to
3.9) although the smaller sample leads to a smaller t-ratio for this parameter. As in the broader
sample, households under 40 with higher contributions are also more likely to buy a smaller unit. The
effect is much larger in this group (with a t-ratio of 1.73). However in this sample the effect of
enrollment length is estimated to be negative (although very imprecisely negative.) This imprecision
causes the joint test to have a much larger p-value than in the broader sample. But in general we
re-emphasize the conclusion that engagement with the HPF does cause smaller houses to be
purchased, even controlling for household resources and demographics, and that we provisionally
assign a role to the downpayment constraints attached to larger houses when HPF is used.
To empirically verify this explanation, we next examined data collected from another smaller survey
conducted in Jinan, that contained detailed information from about 671 urban households on their
housing consumption and investment18. As in the CHFS, we found the HPF takeup rate to be low;
over 60 percent of households who have active HPF accounts do not use it when they purchase their
homes19.
In this smaller survey we divide the sample in Table 6.a according to primary financing method:
HPF loan, commercial bank mortgage, mixed loan, and cash payment. The mean square footage
measures are distinctively different among these three. For households that take HPF loans the
regression controls for income, wealth and occupation, so that the differences between HPF and non-HPF
users is largely controlled for as well.
18 We use this survey because it contains information on the primary source of financing for their place of
residence, a datum which is not contained in the CHFS. About 40% of the survey was conducted as
random on-street interviews and 60% were done as random client interviews through the central bank
branch system in the city that oversees more than 100 commercial banks that serve over 90% urban
population in Jinan. Though the survey has high response rate (over 90%) and asked detailed questions on
HPF, due to the sample limitation both in geographic scope and in number, we are unable to conduct solid
regression analysis for the research questions under study. Here we use it as an external check for our
empirical investigation of the national survey data.
19 The homeownership rate in Jinan survey is 57%, much lower than the CHFS, because the survey was
collected from eight main urban districts of Jinan where housing prices are much higher than the suburban
areas. But this figure is comparable to the one found in Coulson and Tang (2013). The homeownership rate
is always lower in main urban districts than the outskirt areas and the rural areas. The HPF coverage rate
here is 57%, higher than CHFS, but lower than desired since all types of work-units are bound by the law
to pay the HPF tax for their employees.
17
average home area is 91.58 square meters, 7 square meters smaller than that for those who have HPF
but take commercial bank loans for their home purchase. Over thirty percent of homeowners with
active HPF accounts use one-time cash payment for their home purchase and the average area is 10
to 12 square meters larger than the previous two groups. The sharp discrepancy in size between
households using HPF loans and those without using HPF loans, even though both have active HPF
accounts suggests that tightness of credit constraint differs between these two groups, and so does
the effectiveness of HPF on their home purchase. It suggests also that the size constraints are
binding.
Thus we return to the CHFS data to further investigate. Table 6.b. shows the distributions of home
size, HPF monthly deposits and HPF participation length by small-sized homeowners and large-sized
homeowners. It is very clear that these two groups differ significantly in all these measures. There
are 3123 small-sized urban homeowners in CHFS whose average home unit size is only 62.7 square
meters, and 2366 large-sized urban homeowners whose average home unit size is over 166 square
meters. Small-sized homeowners and large-sized homeowners also differ significantly in their HPF
exposure. The small-sized homeowners have much higher HPF participation rate than the large-sized
homeowners. Nearly 24% of small-sized homeowners have HPF accounts, and this number drops to
less than 15% when we consider large-sized homeowners. The second panel in Table 6.b. also shows
that small-sized homeowners enjoy much higher monthly HPF contribution and longer participation
duration in HPF than their large-sized counterparts(both nearly 20% more). There are tremendous
variations in these measures across two size groups. In particular, the variations in the monthly
deposits are over two times greater in the small-sized group than in the large-sized group. These
patterns are preserved when examining the 40-and-under aged group.
.
As a result, we estimate a regression model for choice of house size for both small-sized homeowners
and large-sized homeowners in the CHFS that include inverse Mills ratios from the estimates in Table
5, expecting to see the signs of estimates for HPF parameters differ between these two groups. These
are contained in Table 7. For small owners the coefficient estimates for the background variables are
not particularly instructive. Most are of the expected sign but have large standard errors. The HPF
parameters are of somewhat more interest. In our original modeling effort, we observed a large
negative coefficient on the household’s HPF duration and we presumed that this was due to changes
18
in housing supply over the past decade or so. In China the average floor area has been increasing
substantially in the years since the housing reform. In response to this, in Table 7 we include the
purchase year (date), as well as the interaction between that variable and HPF length. The coefficient
of date indicates that even within the category of small homes, average size has been increasing at
about a half-percent per year, and the coefficients of HPF length and the interaction term indicate
that the effect of HPF duration is in fact positive after the year 2000 in both samplesthat is, for
much of the time the Fund has been in full-fledged nationwide existence. The other two HPF
coefficients are positive, as expected, although the enrollment binary is very imprecise. In the large
purchaser regression wealth plays a much more significant and important role in the consumption
decision.20 This is expected, not only because of the greater resources required, but also to
overcome the regulations of the program. Most noticeably, the parameters related to the HPF
(including the interaction term) have signs opposite to those in the small buyers model in both
samples. In particular, the enrollment binary is negative: those with access to the fund who choose
to buy larger units (with the fund) are hamstrung by the down payment requirement and therefor
purchase smaller units (within the larger category). The HPF length parameter, along with the
interaction term indicate that after 2006 (2000 for young households), length is associated with
smaller houses as well. Higher fund contributions are also associated with smaller units (again, within
the large category)
Putting all of the terms together, using the full sample, consider a HPF enrollee, with the program
since 2005, who deposits 500 RMB in the HPF account and purchases a small housing unit in the
beginning of 2014 (i.e. was enrolled for 96 months). That household’s housing purchase is 6.7%
larger than it would have been without the exposure to the HPF. This is somewhat small but
nevertheless measurable increase. By contrast, a person buying a large unit, but with the same HPF
experience would buy a 22.5% smaller unit (within the large category) than a person without any HPF
exposure. Our guess is that the HPF person actually eschews the use of HPF when making the
purchase and so is quite constrained financially.
20 These households may not use their HPF funds at all in their housing investment.
19
The fact that users of the HPF fund are steered toward smaller properties indicates that Chinese
housing policymakers are worried about the use of the HPF fund for mere speculation. In order to
investigate this, we use the CHFS to estimate the propensity of fund participants to engage in
speculative behavior. We do this by (1) estimating a probit model for the propensity to invest in real
estate (as a function of HPF exposure and other variables) without necessarily occupying said property; (2)
estimating a model of quantity of real estate investment (conditional on investing in the first place,
again as a function of HPF), using an ordered probit. The ordered probit is necessary in the second
stage because the CHFS reports total household real estate holdings in categorical form. The
dependent variable (own_hdg) describes the number of properties held by the household, ranging from
1 to 5. The results are contained in Table 8. To save space we report only those coefficients
pertaining to the fund,21
We first consider the real estate investment equation. The selection equation, which asks whether or
not the household invests in residential real estate at all, is seen to be influenced by a number of the
demographic variables. In the broader sample the existence of HPF employment does raise the
probability of investment (and by about the same amount as it does ownership. While the other
HPF parameters are somewhat imprecise, they are all positive, and larger than the comparable
parameter is the owner-occupation model. We reach the conclusion that HPF does aid in the
acquisition of real estate investment, rather more convincingly than we could for owner-occupation
(the joint test that all three coefficients are zero is rejected at the 10% level).
Turning to the model of real estate investment quantity, we find older households are likely to
buy more properties, that educated households also purchase more, and somewhat more surprisingly,
that larger households purchase more properties. This belies the idea that the reason larger
households purchase of real estate is to owner-occupy, because one property in most cases would be
sufficient for that. For the HPF parameters, mere entry into HPF does not cause a greater number
of properties to be purchased, however, greater monthly contributions, as well as longer duration in
the program both are statistically significant. This is quite sensible, as greater exposure to the HPF
through larger contributions and/or greater duration, will create greater capability to take advantage
21 The results on the background variables are as expected, though not always precise. Wealth is of
particularly strong importance, as might be expected. See also Coulson and Tang (2013).
20
of the program for investment purposes. However this is not the case for younger households, who
could invest in multiple properties without drawing on resources from the fund22.
6. Conclusion
In the absence of fully operational mortgage market, the encouragement of homeownership in China
must entail some amount of credit market subsidization. One program in this vein is the savings
program of the Housing Provident Fund. We use the China Household Finance Survey to evaluate
the use of the fund by Chinese families. Our results indicate that the effect of exposure to the HPF
has, as of the time of the survey very slightly increases the probability of homeownership, and this
result is very imprecise. This, combined with the lack of enrollment so far in the fund, combine to
make the HPF’s effect on the homeownership rate minimal. Users of the fund seem to be somewhat
constrained; they more often buy smaller units, but interestingly, among those who purchase small
(<90 sq m) properties, users of the fund actually increase the size of their properties. This does
not happen with large unit purchasers. The conclusion is that those households who are most
limited are able to use the fund to their advantage. For other users, our final results suggest that the
HPF induces housing investment, rather than owner-occupation as such.
22 The estimates of demographics such as education, income, wealth, business owners and so on are all
positively significant in both all-aged group and the younger group in the quantity regression.
21
References
Buttimer Jr, Richard J., Anthony Yanxiang Gu, and Tyler T. Yang. (2003)"The Chinese housing provident fund."
International Real Estate Review 7,1, 1-30.
Chan, Wing Kam (2010) “The Household Registration System and Migrant Labor in China: Notes on a Debate.”
Population and Development Review vol.36 (2):357-364.
Chen, Aimin, and Edward Coulson, (2002) “Determinants of urban migration: evidence from Chinese cities.
Urban Studies 39 (12), 2189-2197
Chiquier, Loic (2009) “Housing Provident Funds” in Loic Chiquier and Michael J. Lea, eds. Housing finance policy
in emerging markets. World Bank Publications, 2009
Coulson, N. Edward (1999). Why are Hispanic-and Asian-American homeownership rates so low?:
Immigration and other factors. Journal of Urban Economics,45(2), 209-227.
Coulson, Edward, and Mingzhe Tang.(2013) “Institutional and Demographic Influences on the Presence, Scale
and Geographic Scope of Individual Chinese Real Estate Investment.Regional Science and Urban Economics,
vol.43(2),p. 187-196.
Deng, Lan, Qingyun Shen, and Lin Wang. (2011) "The emerging housing policy framework in China." Journal of
Planning Literature 26, 2
Julie Hotchkiss and Melinda Pitts (2005) “Female labour force intermittency and current earnings: switching
regression model with unknown sample selection” Applied Economics, 37, 545-560
Deng, Lan, Qingyun Shen, and Lin Wang. "The emerging housing policy framework in China." Journal of
Planning Literature 26.2 (2011)
Feltenstein, Andrew, Nsouli, Saleh M.,(2003). Big Bang versus gradualism in economic reforms: an
intertemporal analysis with an application to China.IMF Staff Papers 50, 458480.
Gan Li, Zhichao Yin, Nan Jia, Shu Xu, Shuang Ma, and Lu Zheng (2013). Data you need to know about
China: Research Report of China Household Finance Survey 2012. Springer
Goodman, Allen C. (1988). An econometric model of housing price, permanent income, tenure choice, and
housing demand Journal of Urban Economics,23(3), 327-353.
Huang, Youqin. (2012). “Low-income Housing in Chinese Cities: Policies and Practices”. The China Quarterly,
212,pp 941-964 doi:10.1017/S0305741012001270
Kajal Lahiri and Jae Song (2000) “The Effect of Smoking on Health Using a Sequential Self-Selection Model”
Health Economics, 9, 491-511
22
Liew, Leong H., 1995. Gradualism in China's economic reform and the role for a strong central state. Journal of
Economic Issues 29, 883895.
Wei, Shangjin, Xiaobo Zhang, and Yin Liu. (2012). “Status Competition and Housing Prices.” NBER Working
Paper No. 18000.
Wu, Xiaogang, and Donald J. Treiman (2004) “The Household Registration System and Social Stratification in
China: 1955-1996.” Demography Vol. 41, No. 2 (May, 2004), pp. 363-384
Xu, Yilan. (2012) "Credit Access and Housing Consumption: The Case of the Housing Provident Fund."
available at SSRN 2144497
Yeung, Stanley Chi-Wai, and Rodney Howes. (2006) "The role of the housing provident fund in financing
affordable housing development in China." Habitat International 30.2
Ying, Qianwei, Danglun Luo, and Jie Chen. (2013) "The Determinants of Homeownership Affordability
among the ‘Sandwich Class’: Empirical Findings from Guangzhou, China." Urban Studies 50.9
You, J. (2013). The provision of affordable housing in a transitional economy: evidence from Nanjing, China.
PhD thesis, Faculty of Architecture, Building and Planning, The University of Melbourne.
Zhao, Yingshun, and Steven C. Bourassa. (2003) "China's urban housing reform: recent achievements and new
inequities." Housing Studies 18, 721-744
23
Tables
Table 1. Homeownership In CHFS 2011*
Variables
Mean
Obs.
Mean40
Std.40
Obs40
definition
own
0.88
6322
0.845
0.362
1872
owns a unit
own_hdg
+
1.21
5686
1.23
0.53
1610
number of units owned (1,…,5)
units_1
0.728
0.68
fraction of single unit owner
units_2
0.133
0.136
fraction of two units owner
units_3
0.022
0.023
fraction of three units owner
occupier
0.80
6322
0.66
0.47
1872
Dummy on owner occupier:1-yes, 0-no.
size
112.4
5490
114.8
87.89
1548
floor area for the first unit in sq.meters
small
0.52
5490
0.50
0.50
1548
Dummy if the unit sq.m not greater than 90
date
++
2000
6147
2003
1563
Year when the property was owned
*: the first three columns are for all-aged groups, and the next three columns are for under-40-aged group.
+: the last category includes households that own more than 5 units, and the maximum is 15. There are only 5
observations in this group.
++: median of date is reported.
Table 2. Summary Statistics on Housing Provident Fund from CHFS*
Variable
mean
std.
obs.
mean40
std40
obs40.
definition
hpfdy
0.183
0.387
6145
+
0.292
0.454
1844
dummy for having a housing provident fund
account, 1=yes, 0--no.
hpfmon
469.62
820.87
1304
++
544.68
944.6
539
total monthly deposit by the household into
their HPF in 2010 (in RMB) conditioning on
having HPF access.
hpflengh
129.53
123.53
1304
102.81
96.23
539
total months of contribution by the household
to the HPF ( in months) conditioning on
having HPF access
*: the first three columns are for all-aged groups, and the next three columns are for under-40-aged group.
+: households that at least one family member holds urban hukou. And the mean is weight adjusted. The mean will
increase slightly to 0.26 if only “state-ruled” urban households are considered.
++: only consider households that have access to housing provident funds.
24
Table 3. Summary Statistics on Demographics from CHFS*
Variable
Mean
std
mean40
std40
definition
age
48.46
14.34
33
4.97
age of the household heads.
married=1
.87
0.84
Fraction of married households in the sample. Marriage status:
0-single, 1-married, 2-other (divorced, separated)
edu
3.66
1.79
4.42
1.88
categories for education: 1-no school,
2-elementary,3-middle-school,4-high school, 5-junior skill
school,6-associate degree,7-4-year college,8-master
graduate,9-doctoral graduate
kids
0.76
0.78
1.03
0.75
number of kids
familys
3.77
1.59
3.66
1.40
number of people in the household
livh
3.49
1.52
3.48
1.42
number of people living together
emph
1.87
1.29
1.89
0.96
number of people employed in the household
leader
0.048
0.25
0.069
0.31
number of people in the household that hold leadership position
in government, party organizations, enterprises and other entities
bossagr
0.29
0.45
0.23
0.42
owner of an agricultural business
bossother
0.154
0.36
0.21
0.41
owner of an industrial business
localp
0.705
0.30
0.53
0.33
fraction of people in the household having local registered hukou
ruralp
0.312
0.33
0.33
0.31
fraction of people in the household having rural registered hukou
wusoeh
0.09
0.33
0.14
0.40
number of people in the household employed by a state-owned
enterprises
wugovh
0.041
0.216
0.047
0.23
number of people in the household employed by the government
hhincw
6.7888
17.89
8.8311
22.54
total household income in 2010 (in 10,000RMB)
wealth
7.7835
30.29
10.1609
38.9257
total household financial assets excluding real estate in 2010 (in
10,000 RMB)
*all weight-adjusted means or medians, so are the standard devisions. Households whose representatives’ age less than
16 and whose family members holding rural hukou are excluded. The first two columns give the summaries for all-aged
group, and the next two columns are summaries for the under-40-aged group.
25
Table 4. Probit Estimation of Owner-occupation.
Dependent Variable: occupier. Model (1) includes households of all ages, and Model (2) excludes those older
than 40.
Variables
1
M.E.of (1)
2Age40
M.E. of(2)
1.hpfdy
0.159
0.029
0.107
0.022
(1.405)
(1.45)
(0.60)
(0.60)
hpfmonh
2.37e-5
4.5e-6
0.00008
0.00002
(0.566)
(0.57)
(0.74)
(0.74)
hpflengh
4.97e-4
9.44e-5
0.0027**
0.0005
**
(0.926)
(0.92)
(2.25)
(2.23)
age
0.0638
***
0.004
-0.036
0.004
(4.062)
(6.25)
(-0.30)
(1.45)
age2
-4.57e-4***
0.008
(-3.011)
(0.44)
1.married
-0.256
-0.044
-0.074
-0.150
(-1.440)
(-1.58)
(-0.31)
(-0.32)
2.married
-0.374*
-0.067
-0.129
-0.026
(-1.890)
(-2.03)
(-0.42)
(-0.42)
kidsm
0.297
***
0.056
0.210**
0.043
**
(4.392)
(4.48)
(2.01)
(2.04)
livh
0.118
***
0.022
0.201***
0.041
***
(3.329)
(3.33)
(3.61)
(3.57)
edu
-0.0105
-0.002
-0.113***
-0.02***
(-0.404)
(-0.40)
(-2.73)
(-2.72)
hhincw
-0.0117
***
-0.0021
-0.0066
-0.001
(-2.988)
(-3.02)
(-1.22)
(-1.21)
hhincw2
3.75e-5**
2.4e-5
(2.426)
(1.24)
lgwealth
0.140
***
0.027
0.125***
0.026
***
(6.009)
(6.28)
(3.24)
(3.28)
emph
0.117
***
0.022
0.087
0.018
(2.875)
(2.90)
(1.26)
(1.27)
wusoeh
0.0443
0.008
-0.168
-0.035
(0.456)
(0.46)
(-1.25)
(-1.25)
wugovh
-0.0145
-0.002
-0.255
-0.053
(-0.098)
(-0.10)
(-0.98)
(-0.97)
leaderh
-0.0226
-0.0043
-0.063
-0.013
(-0.159)
(-0.16)
(-0.29)
(-0.29)
1.bossagr
0.750
***
0.125
0.668***
0.134
***
(6.708)
(7.83)
(3.99)
(4.14)
1.bossother
-0.138
-0.027
-0.198
-0.042
(-1.435)
(-1.39)
(-1.58)
(-1.53)
localp
1.746***
0.331
1.965***
0.405***
(12.27)
(13.85)
(9.66)
(12.23)
ruralp
-0.130
-0.025
-0.934***
-0.192
***
(-0.830)
(-0.83)
(-4.10)
(-4.13)
_cons
-3.885***
-1.578
(-8.045)
(-0.86)
26
Province FE
Province FE
N
6145
1844
pseudo R-sq
0.316
0.423
Tests of Provincial FE
0.000
0.000
HPF parameters:
hpfdy,hpfmonh,hpfleng
h
0.163 0.024
Social status variables:
wusoeh,wugovh,leaderh
0.958 0.502
Self-employment:
bossagr, bossother
0.000 0.000
t statistics in parentheses. * p<0.10, ** p<0.05, *** p<0.01
27
Table 5: binary size choice regression with owner-occupiers selected.(probit with heckman selection.)
Dep.var: small
(1)
M.E. (1)
Selection
(2)age40
M.E.
Selection
hpfmonh
0.000044*
0.00001
0.000016
0.0003*
0.00008
0.00001
(1.936)
(0.315)
(1.73)
(0.09)
hpflengh
0.000514
.0002
0.0004
-0.003
-0.0009
0.0039**
(1.22)
(0.579)
(-0.28)
(2.34)
1.hpfdy
0.125
.039
0.122
0.200
0.057
-0.194
(1.16)
(0.812)
(0.93)
(-0.91)
date
-0.0206***
-.006
-0.026***
-0.007
(-6.25)
(-3.09)
age
-0.0039
-.001
0.016***
-0.009
-0.002
0.0023
(-1.50)
(4.295)
(-0.55)
(0.01)
1.married
-0.051
-.016
-0.333
0.165
0.046
-0.198
(-0.32)
(-1.429)
(0.69)
(-0.67)
2.married
0.151
.047
-0.264
0.737*
0.20
-0.039
(0.75)
(-1.011)
(1.67)
(-0.11)
kidsm
-0.065
-.020
0.249**
-0.321**
-0.090
0.337**
(-1.13)
(2.809)
(-2.37)
(2.53)
livh+
-0.059**
-.018
0.213***
-0.072
-0.020
0.244***
(-2.17)
(4.531)
(-1.21)
(3.03)
edu
-0.009
-.003
0.004
-0.032
-0.009
-0.048
(-0.41)
(0.132)
(-0.72)
(-0.87)
hhincw
-0.008***
-.003
-0.004**
-0.0076**
-0.002
-0.0039*
(-3.80)
(-2.283)
(-2.20)
(-1.85)
lgwealth
-0.114***
-.035
0.011
-0.216***
-0.06
0.0213
(-5.77)
(0.365)
(-5.17)
(0.41)
emph
-0.084***
-.026
-0.075
-0.0835
-0.02
-0.136
(-2.74)
(-1.288)
(-1.18)
(-1.42)
wusoeh
0.224***
.070
0.134
0.240*
0.067
-0.136**
(2.61)
(0.926)
(1.72)
(-2.07)
wugovh
-0.318***
-.10
-0.014
-0.135
-0.038
-0.392
(-2.97)
(-0.075)
(-0.70)
(-1.22)
leaderh
-0.137
-.043
-0.170
-0.513***
-0.14
-0.259
(-1.35)
(-1.022)
(-2.88)
(-1.00)
1.bossagr
-0.110
-.034
0.537***
-0.205
-0.058
0.569***
(-1.27)
(3.915)
(-1.23)
(2.80)
1.bossother
-0.191**
-.059
-0.120
-0.113
-0.032
-0.312**
(-2.300)
(-1.092)
(-0.81)
(-2.07)
localp
-0.008
-.002
2.153***
-0.342
-0.096
2.721***
(-0.041)
(13.97)
(-0.72)
(12.02)
ruralp
-0.719***
-.222
-0.410**
-0.730***
-0.204
-1.124***
(-5.472)
(-2.148)
(-2.77)
(-4.02)
_cons
43.85***
-1.418***
56.429***
(6.606)
(-3.011)
(3.29)
athrho
0.989***
0.497
(5.017)
(1.35)
N
5380
1524
Wald test of
indep. Eqns.
0.000
0.177
Test of hpf
parameters
Owoc eqn.
0.604
0.113
Small eqn.
0.019
0.206
Test of
wusoeh,wugov
h
Owoc eqn.
0.015
0.605
Small eqn.
0.0001
0.142
28
Test of bossagr,
bossother
Owoc eqn.
0.000
0.000
Small eqn.
0.080
0.000
t statistics in parentheses; * p<0.10, ** p<0.05, *** p<0.01
Note: + in selection equation, variable “familys”( family size) is used. There is no qualitative difference between “familys” and
“livh”. Provincial effects are controlled for both equations in all regressions.
Table 6.a. The Average Floor Area by Payment Method and HPF account
HPF account
Payment Method
Cash one-time
Commercial loan
HPF loan
Mixed Loan*
No
87
92.08
NA
NA
Yes
103.12
98.56
91.58
111.05
Note: Mixed loan is a combination of HPF and commercial loans.
Source: Jinan Survey in 2013. Survey covers 6 main urban districts of Jinan, through random sampling and on-street
interviews. Total number of valid survey is 671 and homeownership rate is 57.5% and HPF coverage rate is over 57%.
Table 6.b. Distributions of home size, HPF monthly deposits and HPF participation length
Size
90
Size>90
All-aged
40
All-aged
40
Obs.
3123
832
2366
777
Mean size
62.74
65.4
166.5
164
Median size
64
68
130
125
Have HPF(fraction)
0.236
0.32
0.141
0.18
hpfmonh(monthly deposits)
hpflengh(total months)
Size
90
*
Size>90
Size
90
Size>90
Mean
545.12
623.8
458.3
471.1
128.1
93.0
106.7
76.4
Std.
2178
999
701.5
634.9
125.9
89.4
117.5
83.4
Median
226
280
210
250
120
90
79
60
75 percentile
580
666
600
600
216
160
180
144
90 percentile
1200
1500
1200
1100
312
251
288
240
Obs.*
779
301
498
243
779
301
498
243
*:for each size group considers only those having active HPF accounts and the second column is 40
and younger group.
Source: CHFS 2011 data.
29
Table 7: Size Regression
Results from size regression. For (1) and (2), dependent variable is logged floor area for small owners(fl.a.<=90
sq.m.); and for (3) and (4), dependent variable is logged floor area for large owners (fl.a.>90 sq.m.). All regressions have
included mills ratio from corresponding heckprob regressions and provincial effects.
Small-sized Owners
Large-sized Owners
(1)
(2)Age40
(3)
(4)Age40
1.hpfdy
0.0065
0.0242
-0.149***
-0.117
(0.24)
(0.47)
(-3.05)
(-1.44)
hpfmonth
2.28e-5
1.17e-5
-5.2e-5
-3.52e-5
(1.39)
(0.57)
(-1.34)
(-0.56)
hpflengh
-0.0742***
-0.121*
0.133**
0.028
(-3.85)
(-1.70)
(2.52)
(0.17)
date
0.0057***
0.0015
0.0039*
0.00359
(3.64)
(0.65)
(1.77)
(-1.00)
hpflengh*date
3.71e-5***
6.05e-5*
-6.63e-5**
-1.38e-5
(3.84)
(1.70)
(-2.52)
(-0.17)
age
-0.0003
0.0007
-0.0004
-0.011*
(-0.33)
(0.18)
(-0.34)
(-1.73)
Married=1
-0.016
-0.053
0.028
0.062
(-0.46)
(-1.19)
(0.54)
(0.80)
Married=2
-0.037
-0.032
0.075
0.482**
(-0.84)
(-0.23)
(0.91)
(2.27)
livh
0.0197*
0.021
0.034***
0.065***
(1.71)
(1.41)
(3.62)
(2.88)
hhincw
-0.0007
0.0017
0.00017
0.0006
(-0.57)
(1.15)
(0.22)
(0.60)
lgwealth
0.0150**
0.0166
0.045***
-0.018
(2.11)
(1.23)
(2.84)
(-0.56)
own_hdg
-0.015
-0.002
-0.056**
-0.021
(-0.73)
(-0.08)
(-2.12)
(-0.49)
wusoeh
-0.009
-0.046
-0.078
-0.069
(-0.43)
(-1.24)
(-1.46)
(-0.94)
wugovh
0.035
-0.022
0.060
0.014
(1.01)
(-0.46)
(1.07)
(0.19)
leaderh
0.016
0.0260
-0.0048
-0.083
(0.52)
(0.57)
(-0.11)
(-0.98)
1.bossagr
0.032
-0.013
0.0224
-0.092
(0.85)
(-0.21)
(0.58)
(-1.16)
1.bossother
0.023
0.0241
0.077
-0.0198
(0.74)
(0.60)
(1.47)
(-0.24)
ruralp
0.021
0.0049
0.214**
0.112
(0.36)
(0.05)
(2.46)
(0.74)
mills1
-0.011
-0.024
-0.157
2.66e-4
(-0.13)
(-0.26)
(-1.30)
(0.00)
mills2
0.109**
0.074
-0.138
0.231
30
(2.30)
(1.54)
(-0.75)
(0.96)
_cons
-7.447**
1.010
-3.629
12.626*
(-2.35)
(0.22)
(-0.77)
(1.75)
N
3033
803
2303
689
adj. R-sq
0.160
0.097
0.189
0.262
Province FEs
0.000
0.000
0.000
0.000
HPFparameters:
hpfdy, hpflength,
hpfmonh
0.0007 0.284 0.000 0.34
Social status:
wusoeh,wugovh,
leaderh
0.670 0.64 0.44 0.49
Business owner:
bossagr, bossother 0.627 0.77 0.32 0.47
Mills ratios
0.017
0.24 0.033 0.030
t statistics in parentheses, computed by bootstrapped standard errors with sample weight adjusted.
* p<0.10, ** p<0.05, *** p<0.01
Table 8. Regression on housing units decisions with selection.
Model (1) considers only households whose family members hold at least one urban registered hukou, i.e.
dropping off those pure rural households, since by initial design housing provident fund is only made available
to urban households with registered urban hukou. Model (2) considers households whose representatives’ age
is less than 40 years old. In both regressions, all provincial effects and household heterogeneous effects are
controlled such as wealth, age, family structures, employment condition and so on.
(1)
(2) Age40
(1)selection
(2)selection
own_hdg
own
own
hpfdy
0.123
0.217
0.137
0.327*
(1.24)
(1.27)
(1.10)
(1.76)
hpfmonh
0.0002*
0.0001
0.00007
0.0001
(2.17)
(1.34)
(0.19)
(1.18)
hpflengh
0.216**
-0.174
0.0008
0.0009
(2.66)
(-0.66)
(1.52)
(0.74)
date
-.00008
-0.010
(-0.02)
(-1.61)
hpflengh#date
-0.0001***
8.65e-5
(-2.66)
(0.66)
_cons
-1.531***
-1.287*
(-3.74)
(-1.65)
atanhrho_12
-0.106
0.133
(-0.46)
N
6223
1850
t statistics in parentheses;* p<0.05, ** p<0.01, *** p<0.001
Test of housing provident fund parameters:hpfdy,hpfmonh,hpflengh,hpflengh#date
[own]
0.088
0.336
[ownhdg]
0.015
0.487
Both
0.009
0.443
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
This paper argues that the low-income housing programme in China has so far failed to provide adequate housing for the poor for three main reasons: the central government's failure to define a clear mission; a lack of commitment from local governments; and an exclusionary policy towards migrants. A systematic review of low-income housing policy in China shows that the central government juggles its economic and socio-political goals thereby causing constant changes in low-income housing policy. Meanwhile, the existing public finance system, the performance evaluation system and localization in policy implementation have all resulted in a lack of commitment from local governments to low-income housing. Inadequate provision is made worse by problems with allocation. Despite encouraging changes since 2010, many factors underlying the government's failures remain unchanged, thus the fate of low-income housing remains uncertain.
Article
Full-text available
The ‘sandwich class’ is an East Asian term referring to the lower middle class. Using a large and unique micro dataset collected in Guangzhou, this paper examines empirically the potential determinants of housing affordability of the ‘sandwich class’. It is found that an individual’s self-perceived range of housing options is dependent not only on his/her current and predictable economic and financial resources, but also on his/her occupation and socio-demographic characteristics. The empirical findings are found to be robust against potential endogeneity problems and not sensitive to the regression specifications used. The empirical findings reported here help to deepen our understanding of the homeownership affordability determinants of the lower middle class in emerging economies characterised by an underdeveloped housing system.
Article
Full-text available
While in standard housing economics housing is regarded as an asset and a consumption good, we study in this paper the consequences for housing prices if housing is also a status good. More concretely, if a family’s housing wealth relative to others is an important marker for relative status in the marriage market, then competition for marriage partners might motivate people to pursue a bigger and more expensive house/apartment beyond its direct consumption (and financial investment) value. To test the empirical validity of the hypothesis, we have to overcome the usual difficulty of not being able to observe the intensity of status competition. Our innovation is to explore regional variations in the sex ratio for the pre-marital age cohort across China, which likely has triggered variations in the intensity of competition in the marriage market. The empirical evidence appears to support this hypothesis. We estimate that due to the status good feature of housing, a rise in the sex ratio accounts for 30-48% of the rise in real urban housing prices in China during 2003-2009.Institutional subscribers to the NBER working paper series, and residents of developing countries may download this paper without additional charge at www.nber.org.
Article
This article describes the emerging housing policy framework in China, which includes three major affordable housing programs and a heavily regulated housing finance sector. The three programs are the Economical and Comfortable Housing (ECH) program, the Housing Provident Fund (HPF) program, and the Cheap Rental Housing (CRH) program. For each program, the authors examine how it works and whether it has been effective in achieving its policy objectives. They also describe the characteristics of China’s newly developed mortgage market and present some examples of the recent government efforts to regulate mortgage lending and to address concerns about potential real estate bubbles.
Article
We use the results of a survey taken in ten large Chinese cities to investigate several aspects of Chinese real estate investment. The survey provides significant (though not comprehensive) information on the personal and family characteristics and real estate holdings of the survey respondents. In this paper we ask three inter-related questions: 1. What factors are associated with individuals who own property in China? 2. What factors are associated with the extent of such ownership—i.e. what determines the number of properties owned? 3. What factors determine the geographical extent of this ownership; in particular, who owns local property and who owns geographically dispersed?
Article
Several Asian countries have established savings and loan programs called housing provident funds (HPFs), which comprise a voluntary or mandatory savings account and eligibility for discounted mortgage loans. This study evaluates the impact of a mandatory HPF in China on homeownership using the China Health and Nutrition Survey (CHNS) from 1989 to 2009. The empirical results indicate that households enrolled in the program were more likely to own a home since the HPF loans became available in 1998, and such difference was fully explained by the length of the enrollment history which was related to the HPF loan benefits by program design. The success of the the HPF was in part attributable to its program designs that feature behavioral economics theories, such as automatic enrollment, mental accounting, and self-discipline. The empirical findings have implications for designing effective housing policies to promote homeownership.
Article
This study seeks the causes of variation in the amount of migration to Chinese cities over the period of 1995-99. We use a city fixed-effect model with lagged values of 'pull' factors for each city. While wage income shows no significant influence on migration, per capita gross city income does, suggesting that migrants seek returns above and beyond mere wages. Our regression results show that cities with high ratios of the manufacturing and service sectors grow most rapidly. We find that the developmental environment and the job-creation potential of the private sector, indicated by the number of proprietors per capita, make a significant difference in attracting migrants. Except for government fiscal expenditures, we find the attributes of the quality of urban life, such as housing-market conditions and transport, have little explanatory power.
Article
China's urban housing reform started in the early 1980s, as a part of comprehensive economic reform. The old system was dominated by work units that provided substantial in-kind services to their employees, including heavily subsidised housing. It brought three serious problems: housing shortages, corruption and inequities. The goals of housing reform were to solve these problems through urban housing privatisation, commercialisation and socialisation. This study examines China's urban housing reform through a case study of the city of Jinan. It first reviews the history of housing development in Jinan from the 1950s to the 1980s, and then analyses Jinan's reform practices in the 1990s. Its finding is that, after many years' efforts, housing reform in Jinan has made substantial progress, but mostly with respect to privatisation. The problem of housing shortages has been addressed and crowding has lessened considerably. Anecdotal evidence suggests that corruption is less widespread than in the past. The other important housing problem, inequity, still exists and, in some respects, has even worsened. In particular, a new form of horizontal inequity has arisen due to the persistent role of work units in housing provision. This paper suggests that, in the future, the government should take a more positive role rather than leaving the market alone to deal with the problem of housing inequity.
Article
Since the adoption of the open door policy, the economy of China has been growing rapidly. Along with economic achievement, housing problems in major cities in China have remained largely unsolved. Under the traditional planned economy since 1949, all housing production was the responsibility of the State. Due to the lack of finance, new housing production has not been able to keep up with the increasing urban population. Urban housing reform introduced in 1988 proposed a strategy of developing affordable housing for the ordinary households through the private housing market. However the implementation of such a strategy has encountered a number of obstacles including the lack of a second hand housing market, an undeveloped real estate profession, unclear land and property legislation and an absence of proper property management support. Amid these obstacles, the lack of housing finance would appear to be at the core of the problem. In order to generate funding for housing development, the Chinese Government introduced the Housing Provident Fund (HPF) Scheme in Shanghai in 1991. Under the HPF, all employees were required to contribute a proportion of their salaries to HPF and employers contributed a similar amount. Accounts for individual workers were set up in the Construction Bank of China. Today workers are allowed to withdraw their HPF savings when they retire, alternatively they can use their HPF savings to purchase homes in the private housing market, or from the Comfortable Housing Projects. HPF is now implemented in most cities in China, although there are some variations in the operation of Scheme. By using an empirical study of the HPF Scheme in Shanghai, this paper reviews the role of the HPF in financing affordable housing development in China.