The document analyzes the effect of cancelling home purchase restrictions on housing prices in Zhuhai, China. It uses a differences-in-differences analysis with control cities to estimate the policy effect. The results show that cancelling the restrictions increased Zhuhai's average housing price growth rate by 2.043 percentage points relative to the control cities. Micro-level housing transaction data was not available, so the analysis uses housing price index data from real estate agencies. The counterfactual model finds that without the policy change, Zhuhai's housing prices would have grown 0.1016 log points less from March to October 2016.

1. The Effect of Cancellation of Home-Purchase
Restriction on the Housing Price in Zhuhai1
Siyu Peng2
UW-Madison
Abstract
In this paper, I evaluate the effects of the cancellation of home-purchase restriction policy on
housing prices in China using a counterfactual analysis. I construct the counterfactual log
housing price of Zhuhai using the selected control cities. I then calculate the policy effect by
conducting difference-in-differences (DID) and regression analysis. The results indicate that
the cancellation of house-purchase restriction increased the average growth rate of housing
price by 2.043 percentage points and for cities in control group, the increase in housing prices
comes more from the increase in real GDP.
Keywords: home-purchase restriction, counterfactual, housing price
1 Introduction
Facing the downward trend of the current economy, Chinese government carried out a series of
policies in recent years to either relax or cancel the home-purchase restriction policy aiming at
stimulating the slowing economy. Home-purchase restriction was first started in Beijing in May
2010 and then progressively implemented in most major cities in China. It led to restrictions
on the home-purchasing rights of both resident households and non-resident households.
The restriction policy in Zhuhai targeted the second and above purchase of the non-resident
households and the above second purchase of the resident households for purchasing houses
1
I am greatful for the help and useful suggestions provided by Professor Chris Taber.
2
speng35@wisc.edu
1

2. with the overall area less than 144 square meters. On March 16, 2016, Zhuhai, abolished its
home-purchase restriction completely, leaving Beijing, Shanghai, Guangzhou and Shenzhen
the only four cities with the enforcement of the restriction policy.
The cancellation of the purchase restriction aroused heated discussion on whether it could
effectively stimulate the local housing price and furthermore propelling the entire downshifting
economy.
Exploring the answer to this question involves overcoming several difficulties. First,
detailed micro-level data is required in order to conduct the evaluation of the policy effect,
which, however, is hard to obtain in general. Second, it is a comparatively complicated task to
model how and why housing prices in a city have changed over time, as well as the factors
that may have deep impacts upon the changes. Third, confronting with confounding policies
taking effects, for example, the property tax policies and the policy of indemnification housing
etc., it is not easy to differentiate the effect of the purchase restriction policy from the rest.
A subtle problem is the optimal choice of the control cities to construct the counterfactuals.
Since most major cities have already canceled the purchase restriction ahead of Zhuhai, our
candidate control cities for Zhuhai are mainly concentrated in the major cities with policy
restrictions going on within our period of study.
The cancellation of the home-purchase restriction policy in Zhuhai may significantly
lead to increased housing price through the expanding demand. Resident households with
sufficient purchase power will have the tendency to buy extra houses, with the forecast of the
future rising house price and the traditional Chinese belief of “preparing houses for offspring”.
Non-resident households either for the purpose of real estate investment or the first step
to reside in this city, are much more possible to be attracted to purchasing based on the
ideal geographical location and the current development status in Zhuhai. With the relatively
stable supply, the increasing of the housing demand from both the resident households and
non-resident households will result in the growth of the housing price.
Our results indicate the plausibility of our analysis. By comparing the changes before and
after the cancellation of the policy in both Zhuhai and the control cities, I found that the
difference of the growth rate of unit price of Zhuhai before and after the implementation of the
cancellation policy is 2.762, while the difference of the growth rate of the control group before
and after is only 0.719. Therefore, the cancellation of home-purchase restrictions increased
the growth rate of the unit price by 2.043. I also explore the contribution to this rising of the
house price that may come from the implementation of the abolish of the restriction policy.
Finally, I relate our studies to some existing literature. The effect of the cancellation
of the home-purchase restrictions on housing prices has not been systematically studied in
2

3. the existing literature since this order was implemented within the recent few months. In
contrast, Du and Zhang (2015) point out that the implementation of the home-purchasing
policy reduced the annual growth rate of the housing price. Sun et al. (2016) find that Beijing’s
house-purchasing restriction policy triggered decrease in resale price, price-to-rent ratio and
a reduction in the transaction volume of the for-sale market.
The goals of our study are to calculate the effect of the cancellation of the policy on the
housing price in Zhuhai in comparison with control cities and to figure out the contributions
of the policy. The structure of the paper is as follows: section 2 describes the data that I
applied and provides brief descriptive statistics about the variables. Section 3 presents the
basic ideas of the difference in differences analysis and the empirical evidence that supports
the proposed positive relationship between the cancellation of the restriction policy and the
price. Then I provide evidence that the control cities I selected could effectively be used to
predict the price trend of Zhuhai.
2 Data
The unit housing price level in a city level can be explored mainly by the unit housing
price of each transaction appeared within that period in the city, and is determined by the
regional GDP, the population, and the average income. In order to estimate the effect of the
cancellation of the policy, micro-level data (the dealing price of each transaction), are required
for precisely computing the average dealing price and the average log dealing price.
As a general overview of the real estate development, I get the city-level real estate
developing and investment data including the monthly cumulated real estate investment, the
cumulated real estate percentage of increment, the total constructed real estate area and
percentage of increment from National Bureau of Statistics of China. These characteristics
provide a general description of the housing environment from the view of government policy
and the supply side.
Due to the fact that I get no access to the dealing price of each transaction for cities and
no publicly available data sets meet our requirements of the micro-level data , I begin our
empirical analysis by grabbing the data from the websites of two of the major housing price
agencies, Lianjia and Fangtianxia.
3

4. 99.51010.511
LogPrice
015
03.2015
04.2015
05.2015
06.2015
07.2015
08.2015
09.2015
10.2015
11.2015
12.2015
01.2016
02.2016
03.2016
04.2016
05.2016
06.2016
07.2016
08.2016
09.2016
10.2016
Time Period
id = Beijing id = Guanzhou
id = Shanghai id = Shenzhen
id = Zhuhai
Figure 1: Log Price (01.2015-10.2016)
01.2015
02.2
Table 1 provides the average growth rate of the housing price before and after the
implementation of the cancellation of the home-purchase restriction policy in five cities. As
shown in the second column of the table, the average growth rate of the log price in Zhuhai is
comparatively slow among these 5 cities. After March 2016, the average growth rate of the log
price in Zhuhai jumps to the highest among all cities. Here I see the evidence of the impact of
the policy implemented since March 2016 in Zhuhai. House price significantly growth after
the policy taking effect relative to the rest of the control cities.
Table 1: THE TREND OF HOUSE PRICE IN ZHUHAI VERSUS THE CONTROL CITIES
Percentage change in house price over the period
City 01.2015-02.2016 03.2016.10.2016 Difference
Zhuhai 0.84 3.6 2.76
Beijing 0.63 1.93 1.30
Shanghai -0.11 1.39 1.50
Shenzhen 1.09 2.56 1.47
Guangzhou 3.35 1.96 -1.39
The detailed summary information is shown in Table 2. It contains the mean, standard
deviation and the number of observables of the unit price, log housing price and the growth
rate of log price in both Zhuhai and control cities.
4

5. Table 2: SUMMARY STATISTICS
Variable Mean Std. Dev. N
beijing 35217.591 2656.392 22
shanghai 37236.864 4120.591 22
shenzhen 42864.091 8810.591 22
guangzhou 17511.455 575.327 22
zhuhai 14757.5 2039.997 22
control price 33207.5 3959.989 22
log(beijing price) 10.467 0.074 22
log(guangzhou price) 9.77 0.032 22
log(shanghai price) 10.519 0.108 22
log(shenzhen price) 10.645 0.211 22
log(zhuhai price) 9.591 0.131 22
growth control 0.017 0.009 21
growth treatment 0.019 0.021 21
3 Model
In this section, I evaluate the impact of the cancellation of the home-purchase restrictions
on housing prices in Zhuhai by applying the difference-in-differences model(DID) analysis.
By estimating the coefficient of the control cities being used to construct the control group
and comparing the growth rate of the log price in Zhuhai with that of the control group, I
find that the control group that I construct is proper. Then I explore the contributes of the
cancellation of the restriction policy to the growth of the housing price.
3.1 Empirical methodology
I evaluate the policy effects of the cancellation of the home-purchase restriction order by
conducting Difference-in-Differences(DID) analysis. Let
g0
it = bi ft + αi + it, i = 1, ..., T (1)
5

6. where βi denotes the K × 1 vector of factor loading for city i, αi and γi represent city and
month fixed effects. it is the idiosyncratic term with E( it) = 0. Stacking N × 1 g0
it into a
vector yields
g0
t = Bft + α + t (2)
where gt = (g0
1t, ..., g0
Nt) , α = (α1, ..., αN ) , = ( 1t, ..., Nt) , and B = (β1, ..., βN ) is the
N × K factor loading matrix.
Let g1
it denotes the growth rate of unit housing price of city at time under purchase
restrictions. Often I do not simultaneously observe housing price with the effect of policy and
without the effect of policy. The observed data
git = ditg1
it + (1 − dit)g0
it (3)
where dit = 1 if the restriction policy is canceled at city i at time t, and dit = 0 otherwise.
At time T1 + 1, here March 2015, the cancellation of purchase restriction took effect in
Zhuhai. Therefore,
git = g1
it, for t = T1 + 1, ..., T.
For cities not subject to the cancellation of the policy, I have
git = g0
it, for t = 1, ..., T.
The effect of the cancellation of purchase restrictions policy in Zhuhai at time t will be
δit = g1
it − g0
it (4)
3.2 Control Group
For the purpose of measuring the influence of the cancellation of the home-purchase restriction
policy in Zhuhai, I select cities and construct the counterfactuals. Since the order took effect
in March 2016 in Zhuhai, the ideal control cities should contain characteristics: first, with
the purchase restriction policy implemented throughout the study period Jan 2015-Oct 2016;
second, without any home-purchasing related policy changes during the period of studying,
otherwise may be difficult to identify the impact of certain policy; third, either geographically
close to Zhuhai or be the major cities in China. In that case, both the early-canceling cities
and the over small size cities are excluded from the set of the control city group. For instance,
in Aug 2014, Hangzhou, one of the major first-tier cities of China, canceled the home-purchase
restriction policy, but two years after, in Sep 2016, under the threat of the local real estate
bubbles, the restriction policy was re-implemented. Given that the impact of policy changes
6

7. during the study period, it is difficult to disentangle the separate time effects and the policy
effects.
Bearing in mind these important selection rules, I select Beijing, Shanghai, Guangzhou
and Shenzhen as the control cities for Zhuhai.
Table 3: WEIGHTS OF CONTROL GROUP FOR ZHUHAI
(01.2015-10.2016)
City Coefficient St.Error T-stat
constant -13.531 1.585 -8.54
Beijing -0.034 0.413 -0.08
Guangzhou 1.608 0.285 5.63
Shanghai 0.635 0.352 1.81
Shenzhen 0.102 0.095 1.07
R2
= 0.9886
The OLS weights based on Jan 2015-Oct 2016 are reported in Table 3. Figure 2 plots the
actual and constructed growth paths for the period Jan 2015-Feb 2016.
9.459.59.559.6
LogHousingPrice
0 05.2015 10.2015 03.2016
Time
Predict Path Actual Path
Figure 2: ACTUAL AND PREDICTED PATH OF HOUSING PRICE IN ZHUHAI
Figure 2 shows that the counterfactual path, produced by the control groups, behaves
the similar trend with the actual path of Zhuhai’s housing price before the cancellation of
purchase restriction with an R2
of 0.99.
Next, I construct the counterfactual log housing price of Zhuhai, without purchase restric-
tions from Mar 2016-Oct 2016. The actual housing price of Zhuhai and the counterfactuals
7

8. constructed based on the control groups are shown in Table 4. Then the estimate policy effect
is the difference of the two.
Table 4: TREATMENT EFFECTS OF ZHUHAI CANCELLATION OF HOME-PURCHASE
RESTRICTION(03.2016-10.2016)
Time Actual Control Treatment
03.2016 9.6203 9.5770 0.0432
04.2016 9.6560 9.5994 0.0605
05.2016 9.6868 9.6196 0.0673
06.2016 9.7070 9.6339 0.0730
07.2016 9.7318 9.6358 0.0959
08.2016 9.7938 9.6466 0.1472
09.2016 9.8516 9.6868 0.1648
10.2016 9.8758 9.7152 0.1606
Average 9.7409 9.6393 0.1016
According to Table 4, the estimated average treatment effect Mar 2016-Oct 2016 is 0.1016
(log price). Specifically, the average actual log housing price without the cancellation of
purchase restriction is 9.6393 estimated by the selected control group.
9.69.79.89.9
LogHousingPrice
02.2016 04.2016 06.2016 08.2016 10.2016
Time
Predicted Path Actual Path
Figure 3: ACTUAL AND PREDICTED PATH OF HOUSING PRICE IN ZHUHAI
The difference between the average actual and control log housing price indicates that
the estimated average treatment effect is 0.1016 (log housing price). In another word, log
housing prices of Zhuhai increased by more than 0.1016 compared with the log housing price
had there been no cancellation of home-purchase restrictions. Figure 3 plots the actual and
8

9. constructed paths within Mar 2016-Oct 2016, showing that the effect of cancellation of the
purchase restriction on housing price at each month is positive.
4 Empirical Analysis
In order to get the difference in differences, I calculate the growth rates of house prices as the
average house price in this month minus the average house price in the previous month and
then divided by the average house price in this month.
4.1 Difference-in-differences (DID) Analysis
Table 5 shows the growth rates of house prices for the treatment group and control group in
the months before and after the cancellation of home-purchase restriction policy.
The growth rate of house prices of the treatment group (Zhuhai) increased significantly
by 2.762 percentage points (from 0.843 percentage to 3.604 percentage). There was a 0.719
percentage points of increase in growth rate of house price for the control group (from 1.240
percentage to 1.959 percentage). Taken together, these figures suggest a price response of
2.043 percentage points, with a standard error of 0.9. Therefore, I can reach a conclusion
that cancellation of home-purchase restriction caused an increase of growth rate of the house
price.
Table 5: THE GROWTH RATE OF HOUSE PRICE OF CITIES
(01.2015-10.2016)
Pre-Cancel Post-Cancel Differ DID
TreatmentGroup :
With cancellation 0.843 3.604 2.762
(0.420) (0.598) (0.755)
UntreatmentGroup :
Without cancellation 1.240 1.959 0.719 2.043
(0.279) (0.217) (0.387) (0.866)
4.2 Regression Analysis
I begin by testing how the cancellation of home-purchase restriction affects the house price.
By taking the log of house prices and creating a dummy variable of policy, I build our first
9

10. model. The results show that the cancellation of home-purchase restriction has a significant
impact on the house price, with an R-square of 0.129. Meanwhile, 0.235 captures the effect of
cancellation on the house prices. Therefore, I can reach a conclusion that the cancellation of
home-purchase restriction did increase the Zhuhai’s price house.
In the second model, I add a new variable called real GDP, which is calculated by dividing
GDP by CPI. Since real GDP is one of the most crucial economic indicator, the increase of
real GDP may cause an improvement of real estate investment and development of real estate
property, which will increase the house price eventually. As the results show, real GDP is also
highly significantly to increase the house price and it also reduces the coefficient of policy.
And this new model markedly increase R-square (from 0.129 to 0.344), which means for cities
in control group, the increase in house prices comes more from the increase of real GDP.
Table 6: EFFECT OF HOUSE-PURCHASE CANCELLATION AND REAL GDP ON HOUSE
PRICE
(1) (2)
unit log(price) unit log(price)
policy 0.235∗∗∗
0.229∗∗∗
(0.051) (0.044)
real GDP 0.034∗∗∗
(0.006)
constant 10.181∗∗∗
9.701∗∗∗
(0.012) (0.082)
adjusted R2
0.129 0.344
N 110 110
Standard errors in parentheses
∗
p < 0.10, ∗∗
p < 0.05, ∗∗∗
p < 0.01
5 Conclusion
In this paper, I evaluated the effects of the cancellation of the home-purchase restrictions
on the housing prices using a counterfactual analysis. I exploit the dependence of housing
10

11. prices among different cities, which is taken as the time effect separated with the policy effect,
and construct the counterfactuals using data from cities with home purchase restriction keep
implemented during the entire study period.
After taking considerations of the characteristics that the control cities should hold, I
select Beijing, Shanghai, Shenzhen and Guangzhou as the control group of Zhuhai. I find that
the cancellation of the home-purchase restriction increased the growth rate of the log housing
price in Zhuhai by 2.043 in comparison with the control group without policy impacts. The
results suggest that in order to control the soaring housing price, at least in the short run, the
home-purchasing restriction policy should not be abolished.
References
Bai, C., Li, Q., and Ouyang, M. (2014). Property taxes and home prices: A tale of two cities.
Journal of Econometrics, 180(1):1–15.
Du, Z. and Zhang, L. (2015). Home-purchase restriction, property tax and housing price in
china: A counterfactual analysis. Journal of Econometrics, 188(2):558–568.
Dynarski, S. (2004). The new merit aid. College Choices: The Economics of Where to Go, When
to Go, and How to Pay For It, pages 63–100.
Sun, W., Zheng, S., Geltner, D. M., and Wang, R. (2016). The housing market effects of local
home purchase restrictions: Evidence from beijing. The Journal of Real Estate Finance and
Economics.
11