Mandatory savings,credit access and home ownership: The case of the housing provident fund

Abstract

Chinese

Several Asian countries have established savings and loan programs called housing provident funds, which comprise of a voluntary or mandatory savings account and eligibility for discounted mortgage loans. This study evaluates the impact of a mandatory housing provident fund in China on home ownership using the China Health and Nutrition Survey from 1989 to 2009. The empirical results indicate that households enrolled in the program were more likely to own a home since the housing provident fund loans became available in 1998, and such difference was fully explained by the length of the enrolment history which was related to the housing provident fund loan benefits by program design. The success of the housing provident fund was in part attributable to its program designs that feature behavioural economics theories, such as automatic enrolment, mental accounting, and self-discipline. The empirical findings have implications for designing effective housing policies to promote home ownership.

Keywords

Credit constraint home ownership housing provident fund mandatory savings behavioural economics

Introduction

Saving is a challenge for many individuals and households due to procrastination (Brown and Previtero, 2014) and a lack of self-control (Thaler and Shefrin, 1981). In Asian countries where the housing prices are too high for the working class to afford, households may be intimidated by the high prices and therefore choose to consume rather than to save (Yoshikawa and Ohtaka, 1989; Engelhardt, 1994). As a result, tax-exempt saving programs have been established in several countries to encourage savings for housing. In addition to allowing employer-matched contributions to the saving accounts, these programs also provide discounted mortgage loans to qualified participants. Such self-sustained savings and loan programs, called housing provident funds (HPFs), are available in China, Thailand, India, Singapore, and Turkey. Based on empirical evidence from a natural experiment in China, this study examines whether such programs for housing can promote home ownership by incentivising savings and providing credit access.

The Chinese HPF is a mandatory, top-down national program. All employees working in the covered sectors are automatically enrolled in the program. Similar to the defined contribution retirement plans in the United States, such as the Section 401(k) employer-sponsored retirement plans, the Chinese HPF defines contributions to individual HPF accounts as a specific percentage of the participant’s gross salary, where the employer matches 100 percent of the contributions to the account (Buttimer Jr et al., 2004). The contributions from both parties are tax-exempt, which provides tax incentives for saving. Unlike 401(k) plans, the HPF participants do not manage the investment in their account, but they can make withdrawals from their individual account to pay for home construction, home improvement, home purchase, home mortgage down-payments, and monthly mortgage payments. Moreover, participants may apply for a HPF loan after their first 12 consecutive monthly contributions to the Chinese HPF savings account (Burell, 2006). Participants may direct their monthly HPF contributions toward monthly mortgage payments for mortgage loans made by commercial banks or HPF mortgages, where the required down-payment and interest are lower. When participants have been enrolled for longer, greater cumulative deposits are available for withdrawal to make down-payments and the maximum loan amount allowed by the HPF is larger.¹

In this study, I empirically examine the effect of HPF benefits on home ownership in light of a natural experiment in China. At present, no direct data are available about the usage of HPF loans by a nationally representative household sample.² Hence, I quantify the Intention-To-Treat (ITT) effect of the HPF for those who were made eligible for the HPF loans rather than the Treatment on the Treated (TOT) for those who actually took the loans (see Heckman et al. (1999) for a discussion about the difference between the ITT and the TOT). The Chinese HPF has gradually mandated coverage in more employment sectors over the years, which has led to substantial variations in the HPF loan benefits across households and across time depending on how many household members are enrolled in the program and when the enrolment started. The empirical evidence from the China Health and Nutrition Survey (CHNS) from 1989 to 2009 suggests that compared with households without the HPF enrolment, households where both the head and the spouse were enrolled in the HPF had a home ownership rate 38 percentage points higher after 1998, whereas those where only one of the couple was enrolled had a home ownership rate 14 percentage points higher. The increase in home ownership can be fully explained by the length of enrolment, where each additional year of membership in the HPF program increased home ownership by 2 percentage points. The results remain robust when additional control variables and alternative model specifications are applied. There is evidence that the treatment effects of the HPF loans were not driven by the housing privatisation for the state sector. Households that were most likely to take HPF loans also showed similar increases in home ownership, suggesting the ITT is likely to be close to the TOT.

My study quantifies the impact of HPF on home ownership using a sound econometric method to draw a causal inference. It adds to the descriptive studies that show increasing housing investment and housing demand after the HPF was implemented (Chen and Deng, 2014; Chen et al., 2006; Yeung and Howes, 2006). The findings provide evidence to address the debate over the effectiveness of the Chinese HPF. In practice, potential home owners typically face a wealth constraint because they need to accumulate the mortgage down-payment, an income constraint in terms of meeting their monthly mortgage payments, and a constraint on access to credit (Engelhardt, 1994; Barakova et al., 2003). The empirical findings of this study suggest that a HPF program could help the participants circumvent these constraints. Participants may withdraw a down-payment from their mandatory savings and direct their monthly HPF contributions toward monthly mortgage payments. Moreover, the HPF provides access to cheaper credit for eligible households, where the HPF loans can reduce the down-payment and the interest rates compared with mortgage loans made by commercial banks.

My findings of this study highlight the importance of designing government housing and mortgage programs that incorporate behavioural economics theories. For instance, the automatic enrolment of workers in the HPF-covered sectors was one of the ‘libertarian-paternalistic’ (Thaler and Sunstein, 2003) interventions introduced to nudge saving, which can help those who lack the ability and information to make saving decisions (Thaler and Benartzi, 2004). The HPF savings account provides intrinsic motivations to self-discipline regular savings, especially for those who experience self-control problems in expense management and budgeting. The HPF program design also provides extrinsic motivation to take advantage of the tax exemption and the matched employer contributions, in the same way as the U.S. consumers use the Individual Retirement Accounts (IRAs) and 401(k) retirement accounts. In addition, the HPF program also helps the participants to create a ‘mental account’ (Thaler, 1990) for money earmarked for housing purposes. Dupas and Robinson (2013b) and Nam et al. (2013) showed that individuals save more if their saving accounts are earmarked for specific purposes and that they tend to have a lower marginal propensity to consume money designated for health, housing, education, and retirement. For instance, several tax-advantaged saving accounts for designated purposes have gained increasing popularity in the United States, such as the Health Savings Account (HSA) and Section 529 savings plan for tuition. In summary, all of those behavioural economics features of the HPF contribute to its success in promoting home ownership in China. Similar applications of the behavioural economics theories should be considered in designing effective government programs for saving and housing.

My study also suggests that the HPF can be a promising alternative to subsidised mortgage programs, such as the Government Sponsored Enterprises (GSEs), the Community Reinvestment Act lending, and Veterans Affairs loans and Federal Housing Administration loans in the United States (Quercia et al., 2003; Bostic and Gabriel, 2006; Gabriel and Rosenthal, 2008; Rosenthal et al., 1991; Duca and Rosenthal, 1991). The HPF provides favourable financing options with low interest rates and down-payment requirements. The credit risk for such loans is relatively low given that participation is employment-based and it depends on the previous deposits made by participants. The nationwide delinquency rate for HPF loans was 0.07% in 2006, 0.06% in 2007, and 0.04% in 2008.³ Moreover, the HPF savings and loan programs are self-sustaining without fiscal input. All of these features make the HPF an appealing alternative program to promoting home ownership.

Backgrounds of the Chinese HPF

Chinese housing reform and the tenure choice

The HPF in China was first initiated in 1994 by the state council as a nationwide mandatory savings and loan program for housing based on employment. The HPF was part of a broader housing reform that privatised public housing, reformed housing subsidies, as well as established affordable housing programs and the housing market (Duda et al., 2005; Deng et al., 2011; Pudney and Wang, 1995). The HPF program was first introduced in 1994 mostly as a saving program. Its loan function was limited to making construction loans to employers. Only four state-owned banks were allowed to make the HPF loans to the employers at the time. Starting in 1998, the HPF loans were made available to individual participants for home purchases. All commercial banks in China were allowed to offer the HPF loans (Gibson, 2009). Table A1 in the Appendix provides a summary of the evolution of the HPF program.

Before the housing privatisation in 1994, workers within the state sector can rent housing units from their employer, i.e. welfare housing. The landlord-renter relationship was not different from one with a private landlord except that the employers usually subsidise the rent. The tenants could not inherit or sell the housing unit. The privatisation in 1994 allowed employees to buy ownership of their public housing units. If employees could not afford to buy full ownership of their housing units, they could purchase partial ownership from their employers at a discounted price. This partial ownership entitles home owners to the rights to use and inherit their property, but not the right to sell the property (Wang and Murie, 2000; Fu et al., 2000; Wang, 2011, 2012). On the contrary, full home ownership means full financial responsibility, clearly defined property rights, and the freedom of disposal. Previous studies show housing reform provided employees in the state sector with some housing endowments (Wang, 2012; Wang et al., 2005), but it had a limited impact on traditional home owners and those outside of the state sector (Wang and Murie, 2000).

Through waves of top-down policy changes between 1994 and 2002, the HPF has expanded its mandatory enrolment to more sectors (see Yeung and Howes (2006) for a review of the development of the program). The state sector and joint ventures were the first sectors mandated by the HPF program after 1994.⁴ The collective sector and private sector joined the program in 1999.⁵ Self-employed, temporary, and home workers are not covered by the program.⁶ The past decades have witnessed substantial growth in HPF deposits and wider usage of HPF loans. By the end of 2008, 77.5 million employees had participated in the HPF program and the total deposits exceeded two trillion Yuan.⁷

The HPF saving incentives and loan benefits

For employees who work in the employment sectors covered by the HPF, the program mandates monthly contributions as a specific proportion of the participant’s salary, with equal matched contributions from their employers. The contribution was initially 5% of the gross salary from both employees and employers with no contribution limit. By 2002, the prevailing contributions rates have increased to 10–12%, a considerable amount of mandatory savings for employees covered by the program. The accumulated fund in the individual account balance can be used for home purchase, home improvement, home construction, mortgage down-payments, and monthly mortgage payments.

The HPF program makes home ownership easier because participants can take advantage of both the savings and loan facilities. Withdrawals can be made from the HPF for mortgage down payments and monthly mortgage payments. If no withdrawals or loans are made during the employment period, the participants will receive a lump sum payment at the time of retirement, which corresponds to their lifetime contributions and interest. The interest rate on HPF deposits is lower than the long-term savings rate.⁸

A HPF loan is the best option if a participant needs to finance the purchase of a home. HPF loans have favourable terms, where the down-payment requirements and interest rates are lower than those for mortgage loans made by commercial banks (Yeung and Howes, 2006). Indeed, HPF loans only require a 20% down-payment, whereas mortgage loans made by commercial banks typically require at least 30%. Table A2 in the Appendix shows the historical interest rates for HPF loans and mortgage loans made by commercial banks,⁹ which clearly demonstrate that the HPF loans rates are more favourable. For example, the interest rate for mortgage loans made by commercial banks with terms of more than 5 years increased from 5.94% at the end of 2008 to 7.05% in 2011, whereas the rate for comparable HPF loans remained at 3.87% over this period.

The function of the HPF has been evolving with the home ownership growth. Since the housing booms in the late 1990’s and early 2000’s, home ownership of urban households has approached almost 90%, and about 20% of households own more than one home according to the China Household Finance Survey (CHFS) data from 2012 (Gan et al., 2014). In recent years, especially after the 2008 housing boom in China, HPF loans have become a financing tool for purchasing investment homes rather than first homes.

The HPF debate

There has been a debate over the effectiveness of the Chinese HPF. Burell (2006) provided a thorough review of the program and summarised its success and limitations. In particular, there is disagreement regarding the usage of HPF loans across different cities. According to statistics provided by the Ministry of Housing and Urban–Rural Development of China, by the end of 2008, 41% of the total HPF deposits were used for HPF withdrawals and 34% for HPF loans. In 2011, Deng et al. (2011) reported that the ratio of loan beneficiaries to contributors was 25% nationwide, 67% for Shanghai, and 17% for Beijing. However, other earlier studies reported lower HPF utilisation rates in Beijing, Shanghai, Guangzhou (Duda et al., 2005; Li, 2010; Li and Yi, 2007). The discrepancy was partly because these studies relied on small-scale surveys to collect information about the HPF loan use in the absence of a nationally representative survey.

Although home ownership was the primary policy goal for the HPF just like for any other housing programs in the world, it is worthwhile to note that the home ownership-driven policies may compromise other important life outcomes which are beyond the scope of this paper. For instance, the strong incentive to the save for home ownership may sacrifice consumption in the short run and therefore undermine the quality of life. The increased demand for housing spurred by the HPF may raise the housing price and therefore limit the housing choices and lower the housing satisfaction.

Other concerns about the HPF include inequality, corruption, and insufficient risk management. Some people argued that the HPF program elevates the wealth inequality because workers with lower income and those in less developed regions and declining sectors contribute lower amount to their HPF accounts because the contributions are related to income. Moreover, the HPF program management is decentralised to local committees, and the investments and risk managements are nontransparent to the public and the participants. This brings up the concerns about the potential corruptions arising from fund use and fund investment. A thorough account of HPF efficiency would need to consider the administration costs related to HPF investment and management, and the potential corruptions at the local committees, all of which are unavailable at the current data capacity.

Empirical method

Identification

In this study, the HPF benefit received by a household is proxied by the employment sectors of the household head and the spouse. Hence, I use a difference-in-difference (DID) specification to identify the ITT, i.e. the heterogeneous treatment effects of the HPF on enrolled households with different levels of benefits. I define the enrolment status based on the employment sectors last observed before 1998, i.e. before the HPF loans became available. It is assumed that the employment sector of household members before 1998 was unrelated to their demand for housing, and that they could not perfectly predict the HPF loan benefits available 4 years after the programs initiation, and thus workers did not select the employment sectors covered by 1998. A worker is considered to be enrolled if s/he worked in the sectors eventually covered by the program, i.e. the state sector and joint ventures (first covered in 1994), and the collective and private sectors (first covered in 1999). A household could have zero, single, or double enrolment in the HPF, and the treatment dosages were time-invariant.

I use the following Linear Probability Model (LPM) with a DID specification to estimate the treatment effect of HPF loan benefit:¹⁰

\begin{matrix} Y_{i t} = α_{d} D_{i t} \times H_{i t} + α_{s} S_{i t} \times H_{i t} + X_{i t} β + θ_{i} + μ_{t} + μ_{t} \times P_{i} + Φ (t) + ϵ_{i t} \end{matrix}

(1)

The dependent variable $Y_{i t}$ equals one if household i has home ownership in year t, and zero otherwise. The HPF enrolment status for household i may be one of three cases: $D_{i t}$ for double enrolment, $S_{i t}$ for single enrolment, and zero enrolment as the reference. The dummy $H_{i t}$ equals one for all years after 1998, when the HPF first provided home mortgage loans to employees; but zero otherwise. The coefficients of interest are the coefficients for the interaction terms, $α_{d}$ and $α_{s}$ , which measure the differential treatment effects (Angrist, 2004) relative to the reference group, i.e. households without HPF enrolment. The coefficient $α_{d}$ quantifies the effects of the HPF for households with double enrolment, and the coefficient $α_{d}$ quantifies the effects for households with single enrolment.

I control for the vector of time-variant household characteristics $X_{i t}$ and the household fixed effects $θ_{i}$ that could affect home ownership. In particular, marriage formation and dissolution are associated closely with the tenure choice (Xu et al., 2015). Specifically, married couples are more likely to own a home (Grinstein-Weiss et al., 2011) than the unmarried counterparts, yet divorce is related with lower likelihood of home ownership (Hendershott et al., 2009). According to Huang and Clark (2002) and Fu (2016), household income, household size, the age, and educational levels of the household head and spouse affect the housing tenure choice during the transitional period in China, hence I control for those factors. In addition, household fixed effects are included to control for unobservable time-invariant characteristics related to home ownership, such as preference, demand for housing, and other time-invariant differences across employment sectors. The year fixed effects $μ_{t}$ are included to control for systematic shocks in a certain year at the national level, and the year-province interaction term $μ_{t} \times P_{i}$ controls for time-variant housing prices and local HPF policies at the province level. A quadratic function of the number of years since 1989, $Φ (t)$ , is included to control for the common trend in economic development and the growing demand for housing in China since the ’open door’ policy of 1979.

In addition to the enrolment status, I also consider the enrolment history, where it is assumed that each additional year of enrolment had the same impact on home ownership. The home ownership is expected to increase with the enrolment history length. The model specification is the following:

\begin{matrix} Y_{i t} = α_{d} D_{i t} \times H_{i t} + α_{s} S_{i t} \times H_{i t} + α_{l} L_{i t} + X_{i t} β + θ_{i} + μ_{t} + μ_{t} \times P_{i} + Φ (t) + ϵ_{i t}, \end{matrix}

(2)

where $L_{i t}$ is the number of total enrolment years for household i, which is defined as the sum of the enrolment years for the household head ( $L_{i 1 t}$ ) and the spouse ( $L_{i 2 t}$ ). For $j = 1, 2$ :

L_{i j t} = {\begin{matrix} t - 1994 if employed in the state sector or joint ventures \\ t - 1999 if employed in the collective sector or private sector \\ 0 otherwise \end{matrix}

(3)

Data

In this study, I use panel household data obtained from the China Health and Nutrition Survey (CHNS). The CHNS employed a multi-stage, random cluster process to draw sample households from nine provinces: Liaoning, Jiangsu, Shandong, Henan, Hubei, Hunan, Guangxi, Guizhou, and Heilongjiang. This study employed eight waves of survey data acquired in 1989, 1991, 1993, 1997, 2000, 2004, 2006, and 2009. The baseline sample comprised of urban households where the head and spouse were aged between 18 and 40 years in 1989. Thus, at the end of the sampling period, the oldest member of the sample did not exceed the retirement age of 60 years, which is when the HPF contributions and loan eligibility terminate. In order to implement a fixed-effect specification, the sample is restricted to urban households surveyed in at least two waves, with at least one survey before and one survey after 1998. Among the 567 households in the unbalanced panel, 82.72% were last observed in 1997 before the HPF treatment, and 59.96% were observed in at least six of the eight survey waves.

Table 1 reports the summary statistics, with 3257 household observations in the sample. In the last survey period before 1998, 37.57% of households had zero enrolment, 16.23% had single enrolment, and 46.21% had double enrolment. In the periods before and after 1998, the zero enrolment group had the highest home ownership, followed by the single enrolment group and the double enrolment group. The gaps in home ownership narrowed after 1998, thereby suggesting a relative increase in home ownership for the latter two groups. When partial ownership was included in broadly defined home ownership, the zero enrolment group was not affected, whereas the other two groups exhibited small increases in home ownership immediately after the privatisation of housing in 1994. Nevertheless, there were still differences in home ownership between the groups when partial ownership was considered, but the gap narrowed after 1998. In terms of demographics, the ages of the household heads and spouses, and the gender and age of the first child were similar. Households with double enrolment were smaller in size and more educated, but the same patterns were evident in the periods before and after 1998, thereby suggesting time-invariant differences across groups, which will be controlled by the household fixed effects in the model described below. In terms of the marital status, overall the zero enrolment group had the highest rate of marriage, i.e. the lowest rate of divorce or separation, followed by the double enrolment group and then the single enrolment group. After 1998, the marriage rate decreased by about 10 percentage points for the zero enrolment group and 9 and 7 percentage points for the single and double enrolment groups respectively.

Table 1.

Summary statistics by enrolment status.

VARIABLE	(1)	(2)	(3)	(4)	(5)	(6)
	Before 1998			After 1998
	Zero	Single	Double	Zero	Single	Double
Home Ownership
Full ownership	0.916	0.725	0.312	0.957	0.885	0.717
Ownership incl. partial	0.918	0.747	0.402	0.959	0.927	0.824
Demographics
Household size	4.257	3.734	3.290	3.935	3.580	3.017
Household head’s age	36.523	35.391	36.423	47.453	46.689	47.383
Spouse’s age	34.888	34.266	34.835	45.834	45.116	45.738
The 1st child is male	0.508	0.500	0.500	0.540	0.465	0.505
Age of the 1st child at 1998	15.742	14.987	14.689	15.245	13.993	13.964
Head high school graduate	0.150	0.318	0.505	0.196	0.430	0.620
Spouse high school graduate	0.132	0.279	0.479	0.156	0.343	0.584
Married	0.993	0.979	0.991	0.895	0.888	0.922
HH income (×1000 Yuan inflated to 2009)	12.955	16.412	15.130	23.904	34.819	38.943
Years of enrolment	0.000	0.425	1.267	0.000	7.790	17.678
Observations	561	233	701	723	286	755

Notes: The sample consists of urban households whose head and spouse were between 18 and 40 years of age in 1989 and were surveyed in at least two of the eight waves of CHNS in 1989, 1991, 1993, 1997, 2000, 2004, 2006, and 2009, with at least one survey before 1998. The means are reported for households with zero, single, and double enrolment, respectively.

Empirical results

Effects of HPF loan benefits on home ownership

I examine whether home ownership increased disproportionately more in HPF-enrolled households when HPF loans became available in 1998. In Figure 1, the solid lines with diamonds, circles, and crosses plot the time trends for full home ownership during the sample period for households with zero, single, and double enrolment, respectively. The two vertical lines denote the initiation of HPF in 1994 and the first availability of HPF loans in 1998. The time trends show that full home ownership was stable in all three groups before 1998, when the zero enrolment group had the highest rate of home ownership and the double enrolment group had the lowest. In the period after 1998, even the home ownership rate of the zero enrolment group continued to grow over time, however home ownership by the two enrolled groups accelerated even more and eventually exceeded 90% in 2009. These time trends provide visual evidence that HPF loan access increased home ownership for households enrolled in the program.

Figure 1.

Time trends for home ownership.

First, I estimate the effect of enrolment, and I then control for member years since first enrolment in the HPF program. Table 2 shows the estimated treatment effects on the full home ownership, where columns 1 and 2 represent the baseline cases with individual household fixed effects and time fixed effects; columns 3 and 4 control for household demographics, such as the ages of the head and spouse, household size, and annual household income; and columns 5 and 6 further control for the national time trend in home ownership since 1989. The inclusion of household demographics (columns 3 and 4) and household demographics and time trend (columns 5 and 6) does not change the estimates, thereby indicating that the increase in home ownership was not driven by changes in household composition or income, or by sample attrition. In the most robust case, HPF loan benefits increased home ownership by 37.3 percentage points for households with double enrolment and by 13.6 percentage points for those with single enrolment (column 5). After controlling for enrolment length (column 6), one additional year of HPF deposits increased home ownership by 2 percentage points for both enrolled groups, and the enrolment length fully explained the variation in home ownership among the three groups.

Table 2.

The treatment effects on full home ownership.

	(1)	(2)	(3)	(4)	(5)	(6)
$D_{i t} \times H_{i t}$	0.382***	0.042	0.373***	0.039	0.373***	0.039
	(0.033)	(0.050)	(0.033)	(0.050)	(0.033)	(0.050)
$S_{i t} \times H_{i t}$	0.141***	−0.014	0.136***	−0.015	0.136***	−0.015
	(0.040)	(0.044)	(0.040)	(0.044)	(0.040)	(0.044)
length		0.021***		0.020***		0.020***
		(0.003)		(0.003)		(0.003)
Observations	3259	3259	3259	3259	3259	3259
Number of HHs	567	567	567	567	567	567
Household FE	YES	YES	YES	YES	YES	YES
Year×province FE	YES	YES	YES	YES	YES	YES
Control	NO	NO	YES	YES	YES	YES
Time trend	NO	NO	NO	NO	YES	YES

Notes: The dependent variable is a dummy for the full home ownership. $D_{i t}$ and $S_{i t}$ are dummies for households with double and single enrolment, respectively; $H_{i t}$ is a dummy for years after 1998. Length is the number of years since the earliest enrolment in the HPF program. *** p<0.01, ** p<0.05, * p<0.1, estimates and clustered standard errors from a Linear Probability Model are reported.

To examine what could have happened to the home ownership rates of the single and double enrolment groups in the absence of the HPF, I estimate the counter-factual home ownership rates for each of the three enrolment groups from equation (1). Specifically, I set the treatment effects of the HPF, i.e. the interaction terms, as zero; and I use household characteristics, household fixed effects, time fixed effects, province-specific time trend and the quadratic form of the years since 1989 to estimate the home ownership. The counter-factual trends are plotted in Figure 2. By 2009, the zero enrolment group could have had a home ownership rate of 97.0 percent, whereas the single enrolment group could have had a home ownership rate of 81.7 percent and the double enrolment group 47.4 percent. The counter-factual analysis suggests that even in the absence of the HPF program, the home ownership rates for the single and double enrolment groups could have grown, but the growth rate would have been much lower. Home ownership in China was low and dispersed in the early 1990s. Without the HPF program, home ownership in China would probably have taken much longer to achieve the current home ownership rate.

Figure 2.

Counterfactual home ownership.

Using a different control group

In following subsections, I run various tests to check the robustness of the baseline findings. The first test addresses the comparability of the treatment and control groups. In Figure 1 where the common trend assumption for the DID identification is validated, the zero enrolment group has a higher home ownership rate to begin with (an average of 91.6% before 1998). This was likely because of the pre-reform self-selection into employment sectors, where those who have inherited homes from their family or have bought their homes were less likely to enter the state sector for the benefits of housing. The pre-existing selection does not invalidate the DID identification though, because household fixed effects are included to account for the unobservable characteristics that led to such self-selection. Moreover, the home ownership rate of the zero enrolment group continued to grow after 1998, which makes them a valid control group.

The more serious threat is that the growth of home ownership for the control group would have been constrained by 100%, which could cap the maximum growth of home ownership for the treatment groups. To address this concern, I delete the zero enrolment group and instead use the single enrolment group as the control group. Table 3 reports the estimation results. The empirical results suggest that the double enrolment group had a home ownership rate 21.8 to 22.9 percentage points higher than the single enrolment group after the HPF loans became available. This magnitude is consistent with that in the baseline regression where the HPF program increased the home ownership by 37.1 to 38.2 percentage points for the double enrolment group and 13.6 to 14.1 percentage points for the single enrolment group, as compared to the zero enrolment group. In addition, each additional year of HPF enrolment explained 2.6 percentage points of the increased home ownership rate in the robustness case. This finding suggests that the ease of access to the HPF loans and the enrolment history both affected the participants’ home ownership decisions.

Table 3.

The treatment effects using single enrolment as the control.

	(1)	(2)	(3)	(4)	(5)	(6)
$D_{i t} \times H_{i t}$	0.229***	−0.012	0.218***	−0.019	0.218***	−0.019
	(0.046)	(0.055)	(0.047)	(0.055)	(0.047)	(0.055)
length		0.026***		0.026***		0.026***
		(0.005)		(0.005)		(0.005)
Observations	1975	1975	1975	1975	1975	1975
Number of HHs	354	354	354	354	354	354
Household FE	YES	YES	YES	YES	YES	YES
Year×province FE	YES	YES	YES	YES	YES	YES
Control	NO	NO	YES	YES	YES	YES
Time trend	NO	NO	NO	NO	YES	YES

Notes: See Table 2 for notes.

Excluding state employees

In the second robustness test, I address the concern that the housing privatisation for the state sector may have driven the treatment effects found in the baseline case. The housing privatisation during 1994 granted partial home ownership to state employees who had been enrolled in the program since 1994. This is demonstrated by the dashed lines in Figure 1, where broadly defined home ownership including partial ownership is plotted for the three groups. The double enrolment group benefited the most from housing privatisation, where their home ownership rate increased immediately after 1994 when partial home ownership was considered. However, the broadly defined home ownership continued to grow even after the privatisation, which provides suggestive evidence that the HPF loans helped the transition from partial home ownership to full home ownership.

To further bolster the confidence that the housing privatisation was not the main driving force for the treatment effects found in the baseline case, I re-estimate the treatment effects, excluding the households of which either head or spouse were employed in the state sector before 1998, when the HPF became available. The number of households in the sample reduces from 567 to 309. The results are shown in Table 4. When the control variables and the time trend are included, the HPF increased the home ownership by 14.4 percentage points for the double enrolment households and by 7.9 percentage points for the single enrolment households (column 5). In addition, the difference in home ownership is fully explained by the one percentage point increase for each additional year of enrolment (column 6). Although the magnitudes of the point estimates are generally smaller than in the baseline case, the same patterns of treatment effects exist. This is strong evidence that the baseline HPF treatment effects were not entirely driven by the housing privatisation for the state sector.

Table 4.

The treatment effects for non-state employees.

VARIABLES	(1)	(2)	(3)	(4)	(5)	(6)
VARIABLES	Owner	Owner	Owner	Owner	Owner	Owner
$D_{i t} \times H_{i t}$	0.145**	0.034	0.144**	0.034	0.144**	0.034
	(0.066)	(0.081)	(0.066)	(0.081)	(0.066)	(0.081)
$S_{i t} \times H_{i t}$	0.080*	0.022	0.079*	0.020	0.079*	0.020
	(0.042)	(0.045)	(0.042)	(0.046)	(0.042)	(0.046)
length		0.010**		0.010**		0.010**
		(0.004)		(0.004)		(0.004)
Observations	1834	1834	1834	1834	1834	1834
Number of HHs	309	309	309	309	309	309
Household FE	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
Control	NO	NO	YES	YES	YES	YES
Time trend	NO	NO	NO	NO	YES	YES

Notes: See Table 2 for notes. The sample excludes households where either the head or the spouse were employed in the state sector by 1998.

Examining high-demand households

In the third robustness check, I address the concern that the baseline case only identifies the ITT of the HPF without observing the households that actually used the loan. Although the actual loan use is not available in the data, I examine the impact of the HPF for households that were more likely to have actually used HPF loans, i.e. the households with a higher demand for housing. These are households that have more than three household members, those with more than one child, those with a son as the first child, and those with a child of marriage age during the treatment period. The rationale for defining the latter two groups is the status competition (Wei et al., 2012) in the marriage market in China.

Table 5 presents the results of this analysis, where column 1 shows that for the 279 households with more than three household members, the HPF loan benefit increased home ownership by 31.6 percentage points for households with double enrolment and by 7.1 percentage points for those with single enrolment although the latter is less precisely measured. Moreover, each additional year of HPF enrolment explained 1.3 percentage points of the differences in home ownership (column 2). Among the 252 households with more than one child (columns 3 and 4), the HPF loan benefit increased home ownership by 31.2 percentage points for households with double enrolment and by 7.9 percentage points for those with single enrolment. Among the 274 households for which the first child was a son (columns 5 and 6), the HPF loan benefit increased home ownership by 41.7 percentage points for double enrolment and by 15.9 percentage points for single enrolment. Among the 106 households with a child of marriage age in the treatment period (columns 7 and 8), the HPF loan benefit increased home ownership by 46.6 percentage points for double enrolment and by 28.7 percentage points for single enrolment. This is evidence that the findings in the baseline case are robust to the restrictions to high-demand households and that the TOT is likely to be close to the ITT.

Table 5.

The treatment effects for households with high housing demand.

VARIABLE	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Household size>3		Children>1		First son		Marriage children
$D_{i t} \times H_{i t}$	0.316***	0.108	0.312***	0.090	0.417***	0.057	0.466***	0.272**
	(0.056)	(0.072)	(0.060)	(0.083)	(0.046)	(0.068)	(0.082)	(0.137)
$S_{i t} \times H_{i t}$	0.071	−0.020	0.079*	−0.021	0.159***	−0.004	0.287**	0.205*
	(0.044)	(0.048)	(0.046)	(0.051)	(0.060)	(0.066)	(0.116)	(0.121)
length		0.013***		0.013***		0.022***		0.011*
		(0.004)		(0.004)		(0.004)		(0.006)
Observations	1737	1737	1557	1557	1592	1592	616	616
Number of HHs	279	279	252	252	273	273	106	106
Household FE	YES	YES	YES	YES	YES	YES	YES	YES
Year×province	YES	YES	YES	YES	YES	YES	YES	YES
Control	YES	YES	YES	YES	YES	YES	YES	YES
Time trend	YES	YES	YES	YES	YES	YES	YES	YES

Notes: See Table 2 for notes. Columns 1–2 analyse households with more than three household members. Columns 3 and 4 analyse households with more than one child. Columns 5 and 6 analyse households with a son as the first child. Columns 7 and 8 analyse households with children older than 20 years in 1998.

Conclusion

Home ownership has been pursued as a desirable policy goal in most countries because of the social benefits associated with home ownership (see Engelhardt et al. (2010) for a thorough review). Despite the benefits of home ownership, saving for housing has been a challenge for households in many countries. In this study, I examine the impact of the HPF on home ownership. The HPF programs typically comprises of mandatory or voluntary savings with tax-exempt employer-matched contributions and discounted home mortgages. Empirical evidence based on a mandatory program in China suggests that since HPF loans became available in 1998, home ownership increased substantially more among households where both the household head and the spouse enrolled in the HPF compared with households where there was only one enrolment, and relative to those with zero enrolment in the program. The length of HPF enrolment fully explains the differences in the increases of home ownership, which is consistent with the program design that associates the loan benefits with the enrolment history.

The HPF is a savings and loan program, of which both functions are important to promoting home ownership. According to the literature on financial accessibility, consumers are likely to increase their savings once they have access to a savings account (Aportela, 1998; Bruhn and Love, 2009; Dupas and Robinson, 2013a). Consumers are also less likely to consume the savings for housing because of mental accounting (Thaler, 1990). For these reasons, the HPF savings account was likely to incentivise participants to save more than the HPF required amount so that they could access the HPF loans earlier and take advantage of the low down payment requirement and the low interest rates. Burell (2006) estimated a housing value to household annual income ratio of 13.4 in 2002 for a typical Chinese urban household with two manufacturing employees. In this case, it will take 22.33 years for the household to accumulate a 20% down payment with a 12% contribution rate to the HPF. However, the majority of households in my sample had completed the transition from renter to home owner by 2009, i.e., 15 years after the HPF was first established. This provides suggestive evidence that voluntary savings beyond the HPF requirement was very likely under the HPF incentive scheme.

The HPF can be a promising housing program alternative to the commercial credit expansions due to its low credit risk and self-sustained funding. Its success in promoting home ownership was in part attributable to its design that incorporates several behavioural economics theories. The HPF automatic enrolment guarantees 100 percent participation rate, the HPF saving accounts provides a self-discipline device for saving, and the earmarked use for housing creates a mental account where the marginal propensity to consume is lower. The HPF matching contributions and the tax exemption also incentivises saving. There have been increasing applications of behavioural economics theories in policy designs regarding saving for retirement and education. More applications along this line can be explored to promote home ownership.

Footnotes

Appendix

Table A2.

Interest rate adjustments by the Central Bank.

	HPF rates (%)		Commercial rates (%)
	<5 yrs	>5 yrs	1–3 yrs	3–5 yrs	>5 yrs
21 Apr 1991			9.00	9.54	9.72
15 May 1993			10.80	12.06	12.24
11 Jul 1993			12.24	13.86	14.04
1 Jan 1995			12.96	14.58	14.76
1 Jul 1995			13.50	15.12	15.30
1 May 1996			13.14	14.94	15.12
23 Aug 1996			10.98	11.70	12.42
Jan 1997	7.65	8.10
23 Oct 1997			9.36	9.90	10.53
25 Mar 1998			9.00	9.72	10.35
1 Jul 1998			7.11	7.65	8.01
7 Dec 1998			6.66	7.20	7.56
10 Jun 1999			5.94	6.03	6.21
21 Sep 1999	4.14	4.59
21 Feb 2002	3.60	4.05	5.49	5.58	5.76
29 Oct 2004	3.78	4.23	5.76	5.85	6.12
17 Mar 2005	3.96	4.41
28 Apr 2006	4.10	4.59	6.03	6.12	6.39
19 Aug 2006			6.30	6.48	6.84
18 Mar 2007	4.32	4.77	6.57	6.75	7.11
19 May 2007	4.41	4.86	6.75	6.93	7.20
21 Jul 2007	4.50	4.95	7.02	7.20	7.38
22 Aug 2007	4.59	5.04	7.20	7.38	7.56
15 Sep 2007	4.77	5.22	7.47	7.65	7.83
21 Dec 2007			7.56	7.74	7.83
16 Sep 2008	4.59	5.13	7.29	7.56	7.74
8 Oct 2008	4.32	4.86	7.02	7.29	7.47
22 Oct 2008	4.05	4.59	6.75	7.02	7.20
27 Nov 2008	3.51	4.05	5.67	5.94	6.12
23 Dec 2008	3.33	3.87	5.40	5.76	5.94
20 Oct 2010			5.60	5.96	6.14
9 Feb 2011			6.10	6.45	6.60
5 Apr 2011			6.40	6.65	6.80
7 Jul 2011			6.65	6.90	7.05

Notes: ‘Zhongguo Renmin Yinhang Guanyu Shangtiao Jinrong Jigou Renminbi Cundaikuan Jizhun Lilv de Tongzhi (A Notification about Adjusting the Renminbi Interest Rates for Financial Institutes)’ for listed years. In years where the cells are empty, there was no adjustment of the interest rates.

Source: People’s Bank of China.

Acknowledgements

I am very grateful to Amy Ando, Daniel Berkowitz, Maria Ferreyra, Mark Hoekstra, Werner Troesken and Randall Walsh for their invaluable guidance and comments. I thank the editor and the three reviewers for their constructive comments.

Funding

This work was partially funded by the National Institute of Food and Agriculture at the United States Department of Agriculture (Hatch project number ILLU-470-367). The data used in this study are from the China Health and Nutrition Survey (CHNS), a survey project sponsored by the National Institute of Nutrition and Food Safety, China Center for Disease Control and Prevention; the Carolina Population Center, University of North Carolina at Chapel Hill; the National Institutes of Health (NIH; R01-HD30880, DK056350, and R01-HD38700); and the Fogarty International Center, NIH, the China-Japan Friendship Hospital, and the Ministry of Health.

Notes

References

Angrist

(2004) Treatment effect heterogeneity in theory and practice. The Economic Journal114(494): C52–C83.

Aportela

(1998) Effects of financial access on savings by low-income people. Technical report, MIT, USA.

Barakova

Bostic

Calem

et al . (2003) Does credit quality matter for homeownership?Journal of Housing Economics12(4): 318–336.

Bostic

Gabriel

(2006) Do the GSEs matter to low-income housing markets? An assessment of the effects of the GSE loan purchase goals on California housing outcomes. Journal of Urban Economics59(3): 458–475.

Brown

Previtero

(2014) Procrastination, present-biased preferences, and financial behaviors. Unpublished Manuscript, University of Illinois at Urbana-Champaign and University of Western Ontario, USA.

Bruhn

Love

(2009) The economic impact of banking the unbanked: Evidence from Mexico. World Bank Policy Research Working Paper Series, No. 4981.

Burell

(2006) China’s housing provident fund: Its success and limitations. Housing Finance International20(3): 38.

Buttimer

Jr Gu

Yang

(2004) The Chinese housing provident fund. International Real Estate Review7(1): 1–30.

Chen

Deng

(2014) Financing affordable housing through compulsory saving: The two-decade experience of housing provident fund in China. Housing Studies29(7): 937–958.

10.

Chen

Hao

Turner

(2006) Housing finance, economic development and housing affordability in China: A case study of Shanghai. Paper for SSE-LSECCER joint conference, November 2006, Stockholm, Sweden.

11.

Deng

Shen

Wang

(2011) The emerging housing policy framework in China. Journal of Planning Literature26(2): 168–183.

12.

Duca

Rosenthal

(1991) An empirical test of credit rationing in the mortgage market. Journal of Urban Economics29(2): 218–234.

13.

Duda

Zhang

Dong

(2005) China’s homeownership-oriented housing policy: An examination of two programs using survey data from Beijing. Report, Joint Center for Housing Studies, Graduate School of Design and John F Kennedy School of Government, Harvard University, USA.

14.

Dupas

Robinson

(2013a) Savings constraints and microenterprise development: Evidence from a field experiment in Kenya. American Economic Journal: Applied Economics 5(1): 163–192.

15.

Dupas

Robinson

(2013b) Why don’t the poor save more? Evidence from health savings experiments. The American Economic Review 103(4): 1138–1171.

16.

Engelhardt

(1994) House prices and the decision to save for down payments. Journal of Urban Economics36(2): 209–237.

17.

Engelhardt

Eriksen

Gale

et al . (2010) What are the social benefits of homeownership? Experimental evidence for low-income households. Journal of Urban Economics67(3): 249–258.

18.

(2016) The persistence of power despite the changing meaning of homeownership: An age-period-cohort analysis of urban housing tenure in China, 1989–2011. Urban Studies 53(6): 1225–1243.

19.

Tse

Zhou

(2000) Housing choice behavior of urban workers in china’s transition to a housing market. Journal of Urban Economics47(1): 61–87.

20.

Gabriel

Rosenthal

(2008) The GSEs, CRA, and homeownership in targeted undeserved neighborhoods. Paper in Conference on Built Environment: Access, Finance, and Policy. June 2008.

21.

Gan

Yin

Jia

et al . (2013) Data you need to know about China: Research report of China household finance survey 2012. New York: Springer.

22.

Gibson

(2009) The privatization of urban housing in China and its contribution to financial system development. Journal of Contemporary China18(58): 175–184.

23.

Grinstein–Weiss

Charles

Guo

et al . (2011) The effect of marital status on home ownership among low-income households. Social Service Review85(3): 475–503.

24.

Heckman

LaLonde

Smith

(1999) The economics and econometrics of active labor market programs. Handbook of Labor Economics3: 1865–2097.

25.

Hendershott

Ong

Wood

et al . (2009) Marital history and home ownership: Evidence from Australia. Journal of Housing Economics18(1): 13–24.

26.

Huang

Clark

(2002) Housing tenure choice in transitional urban China: A multilevel analysis. Urban Studies39(1): 7.

27.

(2010) Mortgage loan as a means of home finance in urban China: A comparative study of Guangzhou and Shanghai. Housing Studies25(6): 857–876.

28.

(2007) Financing home purchase in China, with special reference to Guangzhou. Housing Studies22(3): 409–425.

29.

Nam

Kim

Clancy

et al . (2013) Do child development accounts promote account holding, saving, and asset accumulation for children’s future? Evidence from a statewide randomized experiment. Journal of Policy Analysis and Management32(1): 6–33.

30.

Pudney

Wang

(1995) Housing reform in urban china: Efficiency, distribution and the implications for social security. Economica 62(246): 141–159.

31.

Quercia

McCarthy

Wachter

(2003) The impacts of affordable lending efforts on homeownership rates. Journal of Housing Economics12(1): 29–59.

32.

Rosenthal

Duca

Gabriel

(1991) Credit rationing and the demand for owner-occupied housing. Journal of Urban Economics30(1): 48–63.

33.

Thaler

Sunstein

(2003) Libertarian paternalism. American Economic Review 93(2): 175–179.

34.

Thaler

(1990) Anomalies: Saving, fundibility, and mental accounts. The Journal of Economic Perspectives 4(1): 193–205.

35.

Thaler

Benartzi

(2004) Save more tomorrow? Using behavioral economics to increase employee saving. Journal of Political Economy112(1): 164–187.

36.

Thaler

Shefrin

(1981) An economic theory of self-control. The Journal of Political Economy89(2): 392–406.

37.

Wang

(2011) State misallocation and housing prices: Theory and evidence from China. The American Economic Review101(5): 2081–2107.

38.

Wang

(2012) Credit constraints, job mobility, and entrepreneurship: Evidence from a property reform in China. Review of Economics and Statistics94(2): 532–551.

39.

Wang

Murie

(2000) Social and spatial implications of housing reform in China. International Journal of Urban and Regional Research24(2): 397–417.

40.

Wang

Bramley

(2005) Chinese housing reform in state-owned enterprises and its impacts on different social groups. Urban Studies42(10): 1859–1878.

41.

Wei

Zhang

Liu

(2012) Status competition and housing prices. Technical report, National Bureau of Economic Research, No. w18000.

42.

Johnson

Bartholomae

et al . (2015) Homeownership among millennials: The deferred American dream?Family and Consumer Sciences Research Journal44(2): 201–212.

43.

Yeung

Howes

(2006) The role of the housing provident fund in financing affordable housing development in China. Habitat International30(2): 343–356.

44.

Yoshikawa

Ohtaka

(1989) An analysis of female labor supply, housing demand and the saving rate in Japan. European Economic Review33(5): 997–1023.