A Tale of Two Cities: Mainland Chinese Buyers in Hong Kong Housing Market

Full text

A Tale of Two Cities: Mainland Chinese Buyers in Hong Kong
Housing Market
Yi Fan*
, Maggie Rong Hu†
, Wayne Xinwei Wan‡
, Zhenping Wang§
February 21, 2021
Abstract
This paper examines the effect of mainland Chinese buyers’ housing purchases in Hong Kong.
Contrary to media’s allegation on mainland buyers causing huge bubbles in Hong Kong hous-
ing market, we find that mainland buyers only constitute less than 4% of housing transactions in
Hong Kong from 2001 to 2017, and their price premium over locals is only 1.4% on average. The
mainland premium is higher for properties attracting more interests from mainland buyers, such as
luxury units larger in size (3.52%). We also find the price premium varies with hedging demand
for the currency risk over time (safe haven effect). Furthermore, the price premium is higher in
buildings with more existing mainland homeowners (residential sorting). At last, the price pre-
mium is lower if the mainland buyer has stronger bargaining power such as more prior transaction
experience or facing a mainland seller.
Keywords: Hong Kong housing price, Chinese investors, safe haven effect, residential sorting,
bargaining power
JEL Codes: R23, O18, F22
*Department of Real Estate, National University of Singapore. Email: yi.fan@nus.edu.sg
†School of Hotel and Tourism Management and Department of Finance, The Chinese University of Hong Kong.
Email: maggiehu@baf.cuhk.edu.hk
‡Department of Land Economy, The University of Cambridge. Email: xw357@cam.ac.uk
§Booth School of Business, The University of Chicago. Email: zhenping.wang@chicagobooth.edu
Electronic copy available at: https://ssrn.com/abstract=3477421

1 Introduction
Mainland Chinese buyers have been blamed for driving real estate prices in major cities over
the world. In 2017, their purchase of international residential and commercial properties reaches
$119.7 billion, ranked No.1 compared to other foreign investors.1Despite plenty of anecdotal evi-
dence on Chinese buyers driving up housing prices in Hong Kong and other global cities, literature
is scant on quantifying the impact of mainland Chinese buyers.
As Asia’s financial center which features free flow of capital as well as close financial integra-
tion with mainland China, Hong Kong has been a popular destination for investments from main-
land China. Despite being the world’s most expensive housing market for the 10th year (Kwan,
2020), Hong Kong is still popular among mainland Chinese investors given its close geographic
proximity and cultural similarity. Separated only by a strip of water, Hong Kong is a close neigh-
bor to the city of Shenzhen, the first special economic zone of mainland China in its marketization
reforms. Known as a bastion of Cantonese culture, Hong Kong is naturally connected to the Pearl
River Delta megalopolis in mainland China which shares cultural similarity and economic pros-
perity.
Hong Kong is also famous for its competitive low tax rate, open business environment, and
free capital market, which facilitates easy capital flow in and out, and allures investors around the
world, especially from mainland China. Given the relatively low transaction and holding costs,
the Hong Kong real estate market attracts considerable interests especially, making it a popular
destination of mainland investments. Further, Hong Kong’s pegged exchange rate also provides a
natural hedge to the dollar against a weakening Chinese yuan. Given all these advantages, Hong
Kong’s real estate market experienced a significant demand surge from mainland Chinese between
2007 and 2012.
The surge in housing demand from mainland investors inevitably crowds out local buyers to
some extent, given the scarce land supply in Hong Kong. While the revenue department does not
1www.cnbc.com/2018/09/07/china-investors-set-to-investment-more-in-overseas-property-investment.
html
1
Electronic copy available at: https://ssrn.com/abstract=3477421

provide detailed breakdown on buyers’ nationality, industry practitioners and developers are often
quoted as pointing Chinese mainlanders are the biggest buyers, causing mixed sentiment towards
mainlanders. Local media often accuses the influx of mainland Chinese investors for driving up
housing prices and exacerbating the housing affordability issue in Hong Kong, although most of
those claims are without rigorous analysis.2
In this paper, we attempt to address this question by examining the role of mainland buyers
in Hong Kong housing market and documenting empirical evidence on their impact on housing
price. We seek to offer a clarification on mainland Chinese’s alleged influence in pushing up Hong
Kong housing prices. Using a unique data set comprising all housing transactions with official
names of buyers and sellers recorded in the Hong Kong Land Registry, we identify mainland
Chinese buyers from the spellings of their full names, given that mainland China and Hong Kong
use distinctive ways of spellings due to historical reasons and the spellings rarely change with
migration experience. We identify 3.7% out of 687,598 housing transactions are made by mainland
Chinese buyers between 2001 and 2017.
We start with analyzing the price differential between mainland buyers and local buyers. Our
baseline result shows that mainland buyers pay a 1.4% price premium relatively to local buyers,
controlling for the housing characteristics and other fixed effects. Our heterogeneity analysis re-
veals that the price premium paid by mainland buyers reaches 3.5% for luxury units over 80 square
meters, for which more than 6% transactions are made by mainlanders. Further, mainland buyers
pay more for properties located in Hong Kong Island and Kowloon, which are considered as more
central locations than New Territories. The price premium also varies over time, reaching close to
4% in 2011 when the percentage of mainland buyers peaks in Hong Kong housing market.
We then examine the effect of one-year lagged amount of building-level mainland Chinese
buyers on the transaction prices in the same building. To offer robust estimation, we use an instru-
mental variable (IV) strategy (Fischer, 2012; Gonzalez & Ortega, 2013; Ottaviano & Peri, 2006,
2012; S´
a, 2015; Saiz & Wachter, 2011), where the predicted presence of building-level mainland
2See for example
abcnews.go.com/Business/mainland-chinese-hong-kong-luxury-housing/story?id=15578813.
2
Electronic copy available at: https://ssrn.com/abstract=3477421

buyers based on historical settlement patterns is used as the instrumental variable for the actual
amount of building-level mainland buyers. We find that this spillover effect is pretty significant.
In our sample, the average number of mainland buyers at building level in a year is about 0.84.
If the one-year lagged number of mainland buyers increase by 1, the housing price in the same
building will increase by 4.24% in the next year. Thus, although the current price premium is very
low, about 1.4% only on average, further increasing demand of mainland buyers could push up real
estate price much higher given such significant spillover effect.
Next, we seek to understand factors driving the price premium of mainland Chinese buyers
by exploring variations in mainland buyers’ housing investment demand over time and across
buildings. Specifically, we examine the safe haven effect that drives the demand of mainland buyers
at the aggregate level. Also, we check the residential sorting effect which influence the demand
of mainland buyers at building level due to cultural proximity preference. At last, we analyze
how the price premium varies at transaction level driven by bargaining power due to asymmetric
information and statistical discrimination.
The safe haven effect has been documented in recent literature to have a significant impact on
cross-border fund flow and investment (Badarinza & Ramadorai, 2018; Cvijanovi´
c & Spaenjers,
2020; S´
a, 2015). In our context, it posits that when Chinese currency depreciates, mainland in-
vestors seek diversification opportunities and invest in Hong Kong real estate. This is because they
are attracted to Hong Kong’s pegged exchange rate to the US dollar as a hedge against a weakening
yuan. The second factor of residential sorting hinges upon the agglomeration preference of cross-
border home buyers in destination region (Andersen, 2010, 2016; Borjas, 2002; Deng et al., 2020).
It predicts that mainland buyers would prefer to live in regions where earlier mainland Chinese
agglomerate and are willing to pay a higher price premiums for units that offer cultural proximity
(Bajari & Kahn, 2005; Card et al., 2008; Zhang & Zheng, 2015). The third factor of bargaining
power difference is from the rationale that mainland buyers face greater information friction in
accessing investment options and hence weaker bargaining power relative to locals (Ling et al.,
2018; Baryla & Ztanpano, 1995; Lambson et al., 2004; Turnbull & Sirmans, 1993).
3
Electronic copy available at: https://ssrn.com/abstract=3477421

We find supportive evidence that those three factors are concurrently in play in influencing
mainland buyers’ price premium. At the aggregate level, when Chinese Yuan depreciates, we
see a surge in the number of mainland buyers entering Hong Kong’s housing market, to hedge
against potential risk in Chinese currency. As a result, one standard deviation increase in the
HKD/CNY exchange rate is associated with a 1.86% increase in the price premium of mainland
buyers. Regarding residential sorting, we find that if the one-year lagged proportion of mainland
residents increases by one standard deviation, mainland buyers will pay an extra price premium
about 1.3%.
In addition to demand side factors, transaction-level characteristics, such as mainland buyers’
experience in the Hong Kong market (asymmetric information) and the identity of counter parties
(statistical discrimination), could also contribute to the price premium paid by mainland Chinese.
We explore the third factor on mainland buyers’ weaker bargaining power. Mainland buyers face
higher information asymmetry problems in the search process (Ling et al., 2018) and may also
experience greater statistical discrimination from local sellers (Armstrong, 2006; Stole, 2003).
Specifically, we find that mainland buyers with more prior transaction experience in Hong Kong
will enjoy a lower price premium. Furthermore, mainland buyers will only pay a price premium
when the counter party is a local seller.
One potential caveat is that our identification strategy may not be able to distinguish mainland
Chinese migrants residing in Hong Kong from mainland Chinese investors who only invest in
property in Hong Kong remotely. However, by comparing our transaction data with Hong Kong
population census in 2006, 2011, and 2016, which show that the numbers of mainland migrant
homeowners in the past five years, we provide suggestive evidence that the majority of identified
mainland buyers in our sample are migrant buyers.
Our paper has two major contributions. First, we contribute to the thin literature on the impact
of mainland Chinese buyers on real estate market in Hong Kong and other global cities. Pavlov &
Somerville (2020) find that districts with wealthy Chinese migrants have high housing prices than
other districts in Canada. Li et al. (2020) document that the aggregate level of Chinese housing
4
Electronic copy available at: https://ssrn.com/abstract=3477421

purchases in the U.S. drives up housing price in regions with more Chinese migrant population.
Different from these papers, we identify demand from mainland Chinese at the transaction level,
instead of using demand measure at aggregate level or district level. Our sharp identification
of mainland buyers helps to eliminate any confounding effect over time since we can exclude
them using time fixed effects in the regression, such as time variation of economic fundamentals.
Furthermore, despite numerous media reports blaming mainland Chinese buyers in pushing up
housing price in Hong Kong, we find that mainland buyers are still a reasonably small part of
Hong Kong’s overall residential market with an average presence of 3.7%, not exceeding 8% even
at the peak in 2011. Their average price premium is 1.4%, not close to the staggering prices
quoted in newspapers. Even if we only focus on those housing segments favored by mainland
buyers including luxury units or units in central locations, the price premiums do not exceed 4%,
implying the blames laid on mainland buyers for causing the skyrocketing price in Hong Kong is
unwarranted. Hence, our paper documents empirical evidence on mainland buyers’ real impact in
Hong Kong housing market, which helps clarify the misunderstanding towards mainland buyers.
Second, our paper adds to the literature exploring mechanisms through which foreign investors
could impact local real estate markets. A majority of empirical studies focus on safe haven effect
in the setting of financial assets such as gold or securities (Baur & McDermott, 2010; Hood &
Malik, 2013; Klingler & Lando, 2018; Ranaldo & S¨
oderlind, 2010). Few studies have explored
the role of the foreign investors in the destination housing market (Badarinza & Ramadorai, 2018;
Cvijanovi´
c & Spaenjers, 2020; Deng et al., 2020). Our evidence shows that three important factors
are concurrently in play in determining mainland buyer’s price premium in Hong Kong housing
market, including the safe haven effect, residential sorting, and bargaining power of mainland
Chinese buyers.
The remainder of the paper is organized as follows. Section 2 introduces the institutional back-
ground of mainland Chinese migrants in Hong Kong and Hong Kong’s housing market. Section 3
describes the data set. Section 4 describes empirical design and main results. Section 5 presents
our discussions of the factors that drive the variation of mainland buyer’s price premium. Sections
5
Electronic copy available at: https://ssrn.com/abstract=3477421

6 concludes.
2 Institutional Background
Hong Kong is often viewed as a Chinese immigrants’ society, with only about 60% of Hong Kong’s
population born locally, and many Hong Kong residents have family ties in mainland China. Cur-
rently, the Chinese government implements a permit-quota system to restrict the entry of mainland
Chinese citizens into Hong Kong (Lam & Liu, 1998). The number of mainland Chinese migrants
coming to Hong Kong via the “One-way Permit” (OWP) scheme constitutes a major proportion
of cross-border migration. This one-way permit scheme allows up to 150 mainlanders each day
to migrate into the city. These add to a population of 7.4 million crammed into just 1,100 sq km
(425 square miles), of which 40 percent is country parks or nature reserves. In 2018, about 41,000
mainlanders migrated to the city via the OWP scheme (Appendix Figure A1). Over the 12 months
before that, 55,700 mainland Chinese moved to the city via the scheme between mid-2016 and
mid-2017, for an 11-year high. According to a report from South China Morning Post (SCMP),
there were about 950,000 mainlanders migrated to the city via the scheme in total as of the end of
2016, making up about 12.8% of Hong Kong’s population (Ng & Ng, 2018). 3
The prices of residential housing estates in Hong Kong have been rising consecutively over the
last few years (Figure 1). Hong Kong is now ranked as the world’s most expensive city to live in,
as the disparity between the median home price and the median household income continues to
expand (Carozzi et al., 2018). Hong Kong topped the list for the eighth year in a row, with home
prices regarded as being “least affordable”. The city’s apartments cost 18.1 times gross annual
median income, which is much higher than the 5.1 benchmark ratio for “severely unaffordable”.
Limited housing supply and large capital flows from mainland buyers are believed to be the main
driving factors, angering many residents who can’t afford to get on the property ladder.
Mainland Chinese investors consider the Hong Kong housing market to be an attractive location
3In addition to the OWP, Chinese government can issue an unlimited number of “Two-way Permits” (TWP), which
allow holders to enter Hong Kong for the purpose of visiting family or doing business, but require that they return to
China after a designated period.
6
Electronic copy available at: https://ssrn.com/abstract=3477421

for long-term return and profit (Bloomberg, 2016). The performance of the housing market in
China, as well as the exchange rate between CNY and Hong Kong dollars (HKD), play a large
role in Chinese investors’ interests in Hong Kong’s real estate market. Since property markets
in China, especially in tier-one cities, are getting overheated and offer limited growth (Appendix
Figure A2), mainland Chinese investors turn to overseas investment for higher yield and asset
diversification. Given Hong Kong’s pegged exchange rate to the US dollar as a hedge against
the Chinese yuan, mainland property investors invest in Hong Kong’s real estate market as a hedge
against the depreciation of CNY or the slow down of China’s economy. Figure 4 plots the exchange
rate movement of HKD to CNY and the number of mainland home buyers in Hong Kong over
years. A positive correlation is observed since China introduced a floating currency policy in
2006, especially after the government allowed a more flexible floating range since 2010. When
CNY depreciates, more Chinese buyers come to Hong Kong for property investment despite the
weaker purchasing power.
The HKSAR government has acknowledged the unaffordability problem in the housing mar-
ket. It has implemented a series of cooling measures to rein in property prices and provide more
affordable housing, as listed in Figure 1. Many of the cooling measures introduced by the HKSAR
government from 2012 onward are intended to suppress demand from investors and nonresidents.
Take the stamp duty as an example. In Hong Kong, all housing buyers must pay a stamp duty
which is a small percentage of the property’s price. Non-permanent residents (PR) purchasing
a flat after October 2012 are subject to an additional 15% buyer’s stamp duty (BSD). All home
buyers for a second flat are subject to a double stamp duty (DSD), ranging from 1.5% to 8.5%.
After Nov 2016, the DSD increased to 15% for all local home buyers purchasing a second flat and
all nonlocal buyers, which means that the total stamp duty payable by non-PR mainland buyers is
30% of the housing price.
As shown in Figure 1, the market indeed cooled down slightly in 2012. Subsequent to the
additional stamp duty imposed on nonlocal housing buyers, the percentage of mainland buyers
in the entire primary and secondary residential real estate markets drops from 7.3% in 2011 to
7
Electronic copy available at: https://ssrn.com/abstract=3477421

4.1% in 2014, according to our transaction data from the Land Registry. However, the number of
mainland buyers regained its growth in 2015. According to the data from the Centaline Property
(one of the largest local real estate agencies), in 2016 Q2, mainland buyers accounted for 16.3%
of all purchases by value, the highest since the 15% surcharge on nonlocal buyers was imposed in
2012 Q4. Mainland buyers even accounted for 31% of property purchases of homes worth at least
HK$20 million (US$2.58 million) in 2016 Q2. These statistics imply that mainland buyers were
undeterred by the extra taxes they faced when buying properties in Hong Kong.
Although the government’s cooling measures impede mainland Chinese from entering Hong
Kong’s housing market, growing housing demand from mainland Chinese migrants is expected
in the near future. Going forward, it is estimated that around 21,000 working professionals from
mainland China could become permanent residents in Hong Kong by 2019 (Liu, 2018). This
number is likely to increase in future years as the Hong Kong government’s visa program for
mainland Chinese attracts more high-earning young professionals. To prevent the overheating of
the property market, tighter restrictions may be imposed on nonlocal home buyers if necessary to
rein in property prices, even as Hong Kong maintains its open-door policy for mainland Chinese
migrants, as Chief Executive Carrie Lam stated in the July 2018 Policy Address.
3 Data
3.1 Sample Construction
We use data from EPRC Limited,4covering all residential housing transaction records in the Hong
Kong Land Registry. The data set contains a comprehensive set of information on transaction
details, including transaction date and price, as well as housing characteristics, such as address,
district, housing type (e.g., private single building, private estate building, or village house), floor
level, unit number, gross or net unit size, number of bedrooms and living rooms, building comple-
tion year, bay-window size, and community facilities (e.g., swimming pool and clubhouse). We
4See www.eprc.com.hk/index.htm
8
Electronic copy available at: https://ssrn.com/abstract=3477421

also match each transaction with their distances to various local amenities, including Mass Transit
Railway (MTR), bus stop, hospital, school, university, and coastal line. We exclude the sample
before 2000 because there are too few transactions from mainlanders in earlier years. Our final
sample period is from 2001 to 2017.
This data set also provides the full official names of both sellers and buyers in the romanized
spelling of Chinese characters, or “Pinyin”. Hong Kongese, mainland Chinese, and Taiwanese use
distinctive spellings in Chinese Pinyin.5The spelling of an individual’s official name is determined
by the birthplace and is usually unchanged upon later migration experience. Therefore, we are able
to identify where buyers are originally from based on the distinctive spelling of their official names.
We consider Hong Kong buyers as local buyers and others as nonlocal buyers. The majority of the
nonlocal buyers are from mainland China. In our data, nonlocal buyers constitute 5.95% of the
non-institutional buyers from 2001 to 2017, while 60% of the nonlocal buyers are identified as
mainland Chinese.
The identified mainland Chinese buyers include both migrant residents living in Hong Kong
and investors living in mainland China. According to Hong Kong population census data pub-
lished every five years, the number of recent mainland homeowners, defined as mainland-born
residents who relocated to Hong Kong and purchased a home in the past 5 years, is approximately
6,600, 12,100, and 13,500 from 2002 to 2006, from 2007 to 2011 and from 2012 to 2016, respec-
tively. In our transaction data, the number of mainland buyers in the corresponding periods are
6,577, 16,494, and 15,840. This suggests that most of our identified mainland buyers are mainland
migrants instead of investors who do not live in Hong Kong.
We apply the following criteria to construct the main sample of regression. First, we only
include the resale housing transactions since the prices are negotiated by individual buyers and
sellers and thus reflect their private valuations. Second, we exclude the non-arm’s length contracts
(e.g., deeds of gifts or change of names) and the contracts that are not settled finally. Third, we
5We refer to the official Chinese romanization schemes published by government agencies in Hong Kong, main-
land China, and Taiwan. A full list of the spelling used in our classification is available upon request.
9
Electronic copy available at: https://ssrn.com/abstract=3477421

exclude the village houses in Hong Kong, which may be subject to policy restrictions in resales.6
Fourth, since our interest is to study the effect of mainland buyers, we only include transactions
made by local and mainland buyers in the sample.7Since some spelling patterns exist for both
mainland Chinese and Hong Kongese, we exclude about 0.49% transactions for which the buyer’s
origin cannot be identified. Further, transaction records with incomplete information on transaction
date, price, or floor number are deleted. Lastly, we trim the top and bottom 1% of the transaction
price, housing size, and building age to exclude the impact of outliers, either extremely poor or
luxury housing units. Our final regression sample contains 687,598 transactions from 2001 to
2017, 3.67% of which are purchased by mainland Chinese.
3.2 Descriptive Statistics
Table 1 presents summary statistics for housing units bought by mainland buyers and local buyers
in our sample.8We find that compared with local buyers, mainland buyers tend to purchase more
expensive units in terms of total transaction price as well as net unit price (price per square foot).
On average, mainlanders pay 1.30 million HKD more in terms of total price, which is about 1,700
HKD extra per square foot. On the other hand, the characteristics of housing units are very similar,
such as property size, number of bedrooms or living rooms, unit floor, building age and etc. Given
similar housing units with different transaction prices, the price premium paid by mainland buyers
could be driven by some unobserved housing characteristics or their different private valuations
from local buyers. In our regression later, we use estate fixed effects to control for any unobserved
housing characteristics regarding the property neighborhood.
Figure 2 compares total transaction price and net unit price paid by buyers from different origins
over years. It reveals that non-local buyers consistently pay both a higher total price and net unit
price than local buyers. Mainland buyers initially pay less premium than other non-local buyers in
6Appendix Table A1 explains the definitions of different housing types in Hong Kong and Appendix Table A2
summarizes their distributions in different regions.
7Including non-local individual buyers other than mainland Chinese does not impact the conclusions of our base-
line and spillover estimations.
8Appendix Table A1 presents the definition of variables.
10
Electronic copy available at: https://ssrn.com/abstract=3477421

2001, but gradually catch up over time.
Figure 3a and figure 3b compares the time variation of the percentage of mainland buyers
with average total price and aggregate transaction volume in Hong Kong respectively. We can
see that the percentage of mainland buyers shares similar rising trend with both average price and
aggregate volume from 2001 to 2010. The proportion of mainland buyers increases from 1% in
2001 to 5% in 2010, which corresponds to the increasing cross-border communication after the
handover of Hong Kong to China in 1997. The average price continues to rise since 2011, while
the aggregate transaction volume plummets twice in 2011 and 2013,9and then becomes relatively
flat after wards. On the other hand, the percentage of mainland buyers reaches its peak in 2011,
about 8%, and then starts to drop drastically afterward until 2014 due to the tightened real estate
policies aimed at non-permanent residents in 2012.10 It then gradually rises again from 2015 to
2017. In summary, these stylized facts imply that the demand from mainland Chinese buyers
highly correlates with the dynamics of the Hong Kong housing market before 2010, while its
time variation is more independent with the Hong Kong housing market after 2011. In section 5,
we associate the post-2011 time variation of mainland buyers’ demand with the relative currency
strength of mainland China and Hong Kong.
4 Empirical Design and Main Results
4.1 Empirical Design
In our baseline regression, we estimate the price difference of mainland buyers in the hedonic
regression (Rosen, 1974) below:
ln (Pit )=β1MBit +X0
itλ+φq+ρe+it ,(1)
9A series of policies to cool down the real estate market were announced since 2011, such as the special stamp
duty aiming at speculators who buy and sell real estate properties frequently. See Figure 1.
10The introduction of the buyer’s stamp duty for migrants was announced in 2012 Q4. See Figure 1.
11
Electronic copy available at: https://ssrn.com/abstract=3477421

where Pit represents the pretax transaction price of unit iat date t.MBit is a dummy variable in-
dicating whether the transaction involves mainland buyers. Therefore, the coefficient β1estimates
the price difference of mainland buyers compared to local buyers. To exclude any price difference
driven by housing heterogeneity among local and mainland buyers, we follow He et al. (2020) to
include a comprehensive set of categorical control variables Xit. Specifically, we control for the
unit-level physical features, including the number of bedrooms and living rooms, bay-window in-
dicator, bay-window size, net unit size, unit orientation dummies, and floor range dummies.11 φq
captures the year-quarter fixed effects, while ρecontrols for the estate fixed effects. Standard errors
are clustered at the district and year.
We further investigate the spillover effect of mainland buyers on housing prices. Unlike most
analyses in prior literature which studies the average price effect and the percentage of migrant res-
idents at district level (Saiz, 2007; Saiz & Wachter, 2011; Gonzalez & Ortega, 2013; S´
a, 2015), we
conduct our analysis at the individual transaction level to capture more granular effect of mainland
home buyers. To achieve this, we follow the empirical strategy of Campbell et al. (2011), using
the amount of previous mainland buyers at building level as an explanatory variable:
ln (Pit )=β1MBit +β2Nb,t−365,t−1+X0
itλ+φq+ρe+it .(2)
Nb,t−365,t−1measures the presence of mainland buyers within the same building as the housing unit
iduring 1 year before the transaction date t. We use either the percentage of mainland buyers or
the log of total number of mainland buyers. β2captures the spillover effect of previous mainland
buyers in the neighborhood on the subsequent housing transaction price.
The estimation of β2or the spillover effect might be biased if there are some endogenous
factors driving both the previous presence of mainland buyers and the housing price in the same
11The number of bedrooms and living rooms are encoded in categories, with the missing values in an extra category.
The bay-window indicator denotes whether bay-windows are included in housing price. The bay-window size, net unit
size, age of buildings, building completion years, and distances to amenities are encoded in 10 equally sized categories.
The floor range dummies are formed by firstly classifying four building groups, which are VeryLowRise (on or less
than 10 floors), LowRise (11 to 30 floors), MidRise (31 to 60 floors), and HighRise (61 floors or higher), and then
encoding the floor groups for every 5 floors within each building group.
12
Electronic copy available at: https://ssrn.com/abstract=3477421

neighborhood. To address this issue, we employ the instrument variable (IV) estimation proposed
by S´
a (2015) and Saiz & Wachter (2011). Specifically, we use a shift-share prediction of the
number of mainland buyers as an instrumental variable for the actual number of mainland buyers.
According to the historical settlement pattern of mainland buyers, we construct a predicted number
of mainland buyers. We assume that the influx of mainland buyers is highly correlated with its
historical settlement pattern, which affects future housing price only through its association with
the influx of mainland buyers. The predicted number of mainland buyers of building bfrom t−365
to t−1 is then calculated as:
MBS tockb,t−365−1
MBtockt−365−1
×MBFlowt−365,t−1,(3)
where MBFlowt−365,t−1is the total number of mainland buyers in Hong Kong in one year before
transaction date t.MBS tockb,t−365−1is the cumulative number or total stock of mainland buyers in
building bat date t−365 −1. MBS tockt−365−1is the total stock of mainland buyers in Hong Kong
at date t−365 −1. To allow for the construction of a reliable stock, we include only samples from
2011 to 2017 in IV estimations and use the prior transactions in the same building since 2001 as
the stock. The predicted number of mainland buyers is then divided by the total number of buyers
in building bfrom t−365 to t−1 to get the predicted proportion of mainland buyers.
4.2 Baseline Results
Table 2 presents the estimation of mainland buyers’ price premium based on the regression setting
as in equation (1). Columns (1) displays the results with controls for unit-level physical features,
estate fixed effects, and year-quarter fixed effects. We find that on average, mainland buyers pay
1.39% more in price than local buyers. In Column (2) using estate by year-quarter fixed effects
instead, mainland buyers pay a 1.56% higher price than local buyers. In Column (3), we replace
the estate fixed effects with the district fixed effects and include additional control variables for
building-level features, including the age of buildings, building completion years, swimming pool
13
Electronic copy available at: https://ssrn.com/abstract=3477421

indicator, club house indicator, and distance to train station (MTR), bus stop, hospital, school,
university and coastal line. The estimated price premium paid by mainland buyers is 2.0%. Overall,
the estimated price premium is statistically significant at the 1% level for all three settings. These
results reveal that mainland Chinese buyers pay a higher price than local home buyers in Hong
Kong, after controlling for an extensive list of housing features.
We next examine the heterogeneity of the mainland buyers’ price premium in terms of location,
housing size and time period. We expect that the price premium is positively correlated with the
mainland buyer’s demand.
In terms of location, we find that the percentage of mainland buyers is higher in Hong Kong
Island (4.16%) and Kowloon (4.25%) than in New Territories (3.25%). The first two regions are
considered as more centralized location with higher housing price, while New Territories is the
peripheral region of the city. Panel A of Table 3 presents the heterogeneous effects of mainland
buyers on housing prices in the three regions. It reveals that for units in a more premium location
with a higher average unit price, mainland buyers pay a larger price premium than local buyers.
For example, on Hong Kong Island, where the city’s Central Business District (CBD) is located
and the average unit price is the highest among the three regions, mainland buyers pay a 1.55%
higher price than local buyers. In New Territories, mainland buyers only pay a 1.07% higher price
than local buyers.
Following Hu et al. (2020), we classify the units into three categories, namely mini units less
than 40 square meters (sq. m.) or 430 square feet (sq. ft.), luxury units larger than 80 sq. m. or
831 sq. ft., and ordinary units between 40 to 80 sq. m., respectively. In our sample, there are 3%
mainland buyers in mini units, 3.7% in ordinary units, and 6.3% in luxury units. Panel B of Table
3 presents the heterogeneous effects of mainland buyers on housing prices across the three size
groups. For luxury units, the mainland buyers pay 3.52% more than local buyers. For ordinary
units, the price premium 1.53%. In contrast, mainland buyers only pay a small and statistically
insignificant price premium for mini units.
The price premium paid by mainland Chinese buyers also varies across periods, as shown in
14
Electronic copy available at: https://ssrn.com/abstract=3477421

Panel C of Table 3. We do not find that mainland Chinese buyers pay any price premium between
2001 and 2006. However, mainland buyers pay an evidently higher price than local buyers by
1.96% from 2007 to 2012. The premium decreases to 1% in the later sample from 2013 to 2017.
Figure 5 visualizes the time variation of both the percentage of mainland buyers and also their price
premium. After 2006, all estimates of mainland buyers’ housing price premium are positive and
statistically significant at the 5% level. In general, the magnitude of price premium is positively
correlated with the presence of mainland buyers over time.
4.3 Spillover Effect of Mainland Buyers
Our baseline estimation reveals that mainland buyers purchase properties in Hong Kong at a higher
price. In this section, we study whether such price premium paid by mainland buyers has spillover
effect on the subsequent housing market. Thus we turn to equation (2) to estimate the impact of
lagged amount of mainland buyers in the same building on subsequent housing price.
Table 5 presents the OLS and the IV estimation results of equation (2). We include only
transactions from 2011 to 2017 as our regression sample to allow a reliable stock period to build
the IV. In Columns (1) and (2), the independent variable is the proportion of mainland buyers in the
same building over the previous year. The average percentage of mainland buyers at the building
level is 5.11%. Column (1) reports the OLS estimation result, and it reveals that if the lagged
proportion of mainland buyers in the same building increases by 0.01, the subsequent transaction
price will increase by 0.08%, a 3% increase compared to the mainland buyer’s price premium
(2.4%) in this sample. This estimate is statistically significant at the 1% level. Table 4 reports the
corresponding first-stage IV estimation result. As expected, the first-stage results reveal that the
predicted proportion of mainland buyers based on historical settlement patterns is highly correlated
with the actual proportion of mainland buyers. The F statistics of the first-stage IV estimation is
107.24, which indicates a strong instrument. Column (2) of Table 5 reports the second-stage IV
estimation result. It indicates that a 0.01 increase in the lagged percentage of mainland buyers will
result in a higher housing price in the subsequent year by 0.17%, a 7% increase compared to the
15
Electronic copy available at: https://ssrn.com/abstract=3477421

mainland buyer’s price premium (2.4%) in this sample. This estimate is not statistically significant.
In addition, we check the results by alternative explanatory variable, log of the lagged number
of mainland buyers in the same building, and the results are reported in columns (3) and (4) of
Table 5. We find that, for a building with average number of mainland buyer which is 0.84 in this
sample period, if the lagged number of mainland buyers increase by 1 (equivalent to an increase of
0.784 from log of the average number of mainland buyers), the housing price in the same building
will increase by 1.21% (column 3) or 4.24% (column 4) in the next year. In summary, our results
indicate that the influx of mainland buyers increases the housing price in proximate neighborhoods.
Our empirical findings closely relate to a common puzzle in the literature: Will a very small
proportion of home buyers significantly impact the entire housing market? Some studies demon-
strate the contagion effect of a small proportion of foreclosed properties on the entire market
(Anenberg & Kung, 2014; Campbell et al., 2011; Harding et al., 2009). Piazzesi & Schneider
(2009) propose a search model to demonstrate how a small fraction of optimistic investors can
have a large effect on prices without buying a large share of the housing stock. We provide addi-
tional evidence that mainland buyers create an upward price momentum in the Hong Kong housing
market. Although their percentage is only 3.7% in the entire buyer population, the momentum they
create can be quite influential and drive up the market.
5 Further Analysis On the Effect of Mainland Buyers
In the previous section, we have shown that how the price premium of mainland buyers varies for
location, housing size and time period. In this section, we analyze how the price premium change
at the aggregate level driven by the safe haven effect. Furthermore, we also examine the variation
of price premium at neighborhood level driven by the residential sorting effect, and at individual
level driven by asymmetric information and statistical discrimination.
16
Electronic copy available at: https://ssrn.com/abstract=3477421

5.1 Aggregate Level: The Safe Haven Effect
Over time, mainland buyers will be more willing and eager to enter Hong Kong’s housing market
when they expect a weaker domestic currency or a higher currency risk in the near future. This mo-
tivation is termed as the “safe haven effect” (Badarinza & Ramadorai, 2018; Ranaldo & S¨
oderlind,
2010), which means that investors limit their exposure to a risky market by increasing investment
in another safer market. We use the HKD/CNY exchange rate as the identification for the safe
haven effect. Specifically, given Hong Kong’s pegged exchange rate to the US dollar as a hedge
against the Chinese yuan, we expect that mainland Chinese home buyers are more attracted to the
Hong Kong residential property market when they experience a weakening performance for the
Chinese currency.
Figure 4 plots the percentage of mainland buyers versus HKD/CNY over time. Since China
introduced a wider floating currency policy in 2010,12, we only observe a more volatile HKD/CNY
in the later sample. The correlation of the percentage of mainland buyers and the 1-month lagged
exchange rate from 2010 to 2017 is 0.29, suggesting that the demand of mainland buyers is higher
when the Chinese yuan is weaker.
Column (2) of table 7 shows how the price premium of mainland buyers varies with HKD/CNY
from 2010 to 2017. HKD/CNY is standardized to mean 0 and standard deviation 1 to facilitate
easier interpretation. The result indicates that one standard deviation increase of HKD/CNY will
drive a 1.9% increase of the mainland buyer’s price premium.
5.2 Neighborhood Level: Residential Sorting
Following the assimilation theory of immigrants in the literature, we explore the influence of resi-
dential sorting on the mainland buyer’s choice of residential location (Alba & Logan, 1992; Zunz,
1982). It implies that incoming mainland buyers prefer to live where earlier mainland Chinese
12According to Wikipedia, Since 2006, the renminbi exchange rate has been allowed to float in a narrow margin
around a fixed base rate determined with reference to a basket of world currencies.On 19 June 2010, China claims
that they would ”proceed further with reform of the RMB exchange rate regime and increase the RMB exchange rate
flexibility”.
17
Electronic copy available at: https://ssrn.com/abstract=3477421

agglomerate (Andersen, 2010, 2016; Borjas, 2002). As a result, mainland buyers have higher de-
mand thus are willing to pay a higher price premium for housing units in those areas (Bajari &
Kahn, 2005; Card et al., 2008; Zhang & Zheng, 2015).
Specifically, we investigate the effect of lagged proportion or lagged number of mainland buy-
ers on the probability that the subsequent buyer is also a mainland Chinese. Our empirical specifi-
cation is as follows:
MBit =β4Nb,t−365,t−1+X0
itλ+φq+ρe+it ,(4)
where MBit is a dummy variable equal to 1 if the buyer of unit iat date tis a mainland Chinese and
0 otherwise. Nb,t−365,t−1is the measurement of 1-year lagged mainland buyers in the same building.
We use either the lagged proportion of mainland buyers or the lagged number of mainland buyers
(in logarithmic form) as Ni,t−365,t−1in separate regressions. The coefficient β4, therefore, indicates
the effect of previous presence of mainland buyers on the probability that the subsequent buyer is
also from mainland China.
Table 6 presents the logit estimation results of equation (4), with margins at the means re-
ported. Column (1) displays the estimates using the lagged proportion of mainland buyers as the
explanatory variable, while in Column (2) we use the lagged number of mainland buyers. Results
in Column (1) reveal that one standard deviation increase in the lagged proportion of mainland
buyers will lead to an increase of 0.3% in the probability that the subsequent buyer in the same
building is also mainland Chinese. Since mainland buyers constitute 3.67% of the home buyers in
our sample period, this translates to a 8% increase in mainland buyers at the building level. One the
other hand, one standard deviation increase in the log of number of mainland buyer in a building
during the previous year, the probability that the subsequent buyer is also mainland Chinese rises
by 0.9% (Column (2)), which translates to a 25% increase in mainland buyers at the building level.
Both estimates are statistically significant at the 1% level.
Column (3) of table 7 presents how the price premium of mainland buyers varies with residen-
tial sorting. The result indicates that one standard deviation increase of the lagged proportion of
mainland buyers in the same building is associated with a 1.35% increase of the price premium
18
Electronic copy available at: https://ssrn.com/abstract=3477421

paid by mainland buyers. These results thus support the existence of a residential sorting chan-
nel that, at least in part, drives the demand variation of mainland buyers and also drives the price
disparity.
5.3 Individual Level: Bargaining Power
Inferior bargaining power in the local housing market is also a potential factor that explains the
price premium paid by mainland Chinese buyers. Here we mainly discuss the lack of market
information from the perspective of buyers and statistical discrimination from the perspective of
sellers.
5.3.1 Asymmetric Information
Mainland Chinese buyers, a majority of whom are recent migrants to the city, are expected to
have less market information (Ling et al., 2018), higher searching cost (Baryla & Ztanpano, 1995;
Lambson et al., 2004; Turnbull & Sirmans, 1993), and weaker local networks (Tu et al., 2017).
First-time migrant buyers may also face time constraints in their search for housing, while local
buyers do not (Ihlanfeldt & Mayock, 2012). Therefore, migrant buyers normally possess less
bargaining power than local buyers (Wilhelmsson, 2008; Zhou et al., 2015).
Thanks to the abundant records of transactions over 17 years with detailed information on
names, we are able to identify buyers with multiple transactions. For each transaction, we get
the number of prior transactions made by the buyer. Column (4) of table 7 shows how the prior
market transaction experience impacts the price premium paid by mainland buyers. We find that
as the mainland buyer accumulate more experience in real estate transactions, the price premium
will be lower. In summary, the results support our hypothesis that buyers with more knowledge of
the local housing market will pay a lower price, but mainland buyers initially have limited local
market knowledge and will need to accumulate more market exposure to enjoy the same benefit of
market knowledge, as local buyers do.
19
Electronic copy available at: https://ssrn.com/abstract=3477421

5.3.2 Statistical Discrimination
Statistical discrimination from the seller’s side is another possible explanation for the price dis-
parity. Local sellers may consistently charge mainland buyers higher prices, possibly because
they possess a belief in the mainland buyer’s stronger willingness to pay or less knowledge of the
market. An analogy is institutional sellers who impose different pricing strategies based on the
consumer’s background (Armstrong, 2006; Stole, 2003; Holmes, 1989). Specifically, firms recog-
nize the various needs and willingness-to-pay of customers, and therefore charge them differently,
either based on the time the customer makes the purchase (M¨
oller & Watanabe, 2010; Stavins,
2001) or by customers’ geographic segmentation (Bolton & Myers, 2003).
With detailed information on the names of sellers, we identify sellers’ background using the
same methodology with which we classify the home buyers. Then, we use a subsample of trans-
actions made by local and mainland buyers and sellers, and estimate the price differences out of
all combinations of buyers and sellers with different backgrounds. Following Equation (1), the
empirical specification is modified as follows:
ln (Pit )=γ1MBit +γ2MBit ×LS it +γ3MS it +X0
itλ+φq+ρe+it .(5)
MBit is a dummy variable denoting the mainland buyers, LS it is a dummy variable denoting the
local sellers, and MS it is a dummy variable denoting the mainland sellers. Therefore, the base
group contains the transactions made by local sellers and local buyers. γ3captures how the main-
land sellers charge local buyers differently compare with local sellers. γ1+γ3shows the price
difference if the transaction is made by a pair of mainland buyer and seller, compared to the base
group. γ1+γ2captures how local sellers charge mainland buyers differently compared with local
buyers.
Column (5) of table 7 presents the estimation of equation 5. γ3is about -0.45%, indicating
that mainland seller charges a price discount to local buyers compared with local sellers. γ1+γ3
is about -0.81%, which suggests that the price premium of mainland buyers does not exist when
20
Electronic copy available at: https://ssrn.com/abstract=3477421

the counterparty is a mainland seller. γ1+γ2is about 2.54%, which shows that price premium of
mainland buyers when the counterparty is a local seller.
It is noteworthy that the information channel and statistical-discrimination channel are not
mutually exclusive; they can coexist and interact with each other to shape the overall pattern of
housing premiums paid by mainland buyers.
5.4 Horse Racing Analysis
While our previous evidence reveals that both the safe haven effect at the aggregate level and
other channels at the neighborhood or individual level explain the price premium paid by mainland
Chinese home buyers in Hong Kong, it is of great interest to investigate whether those factors have
independent explanatory power in a horse racing analysis. Therefore, we conduct a horse racing
analysis of all previously identified channels as shown in column (6) of table 7. We can see that
for most driving factors, the estimated magnitude is similar with the results in previous sections.
6 Conclusion
This paper examines the housing price premium paid by mainland Chinese buyers in the Hong
Kong housing market, employing a comprehensive data set of housing transaction records in Hong
Kong between 2001 and 2017. We find that mainland buyers pay a 1.4% price premium compared
with local buyers. Such price premium is larger for homes with large sizes or in central locations,
as these homes attract more interest from mainland buyers. Also, the price premium is positively
correlated with the HKD/CNY exchange rate over time, since a weakening Chinese yuan will trig-
ger more demand from mainland buyers due to hedging need for the currency risk. Furthermore,
the price premium is higher in buildings with larger amount of mainland buyers in the previous
year, given that such buildings attract more mainland buyers motivated by culture proximity prefer-
ence. At last, the bargaining power of mainland buyers play an important role in deciding the price
premium. As mainland buyer gains more experience from previous transactions, they will pay less
21
Electronic copy available at: https://ssrn.com/abstract=3477421

premium compared with mainland buyers with no prior experience. Also, mainland buyers only
pay the price premium when the counterparties are local sellers.
Our study contributes to understanding the impact of mainland Chinese buyers on the real estate
market in Hong Kong. Despite considerable media reports blaming mainland Chinese buyers in
pushing up Hong Kong’s housing prices, we show empirical evidence that they pay relatively low
housing premiums on average and present heterogeneous patterns across market segments. It also
contributes to identifying determinants on how much price premium mainlanders pay. Although
using data from the Hong Kong market, the findings offer policy implications on cross-border
housing purchases in other regions with geographic proximity and cultural similarity.
22
Electronic copy available at: https://ssrn.com/abstract=3477421

References
Alba, Richard D, & Logan, John R. 1992. Assimilation and stratification in the homeownership
patterns of racial and ethnic groups. International Migration Review,26(4), 1314–1341.
Andersen, Hans Skifter. 2010. Spatial assimilation in Denmark? Why do immigrants move to and
from multi-ethnic neighbourhoods? Housing Studies,25(3), 281–300.
Andersen, Hans Skifter. 2016. Spatial assimilation? The development in immigrants’ residential
career with duration of stay in Denmark. Journal of Housing and the Built Environment,31(2),
297–320.
Anenberg, Elliot, & Kung, Edward. 2014. Estimates of the size and source of price declines due
to nearby foreclosures. American Economic Review,104(8), 2527–51.
Armstrong, Mark. 2006. Recent developments in the economics of price discrimination. Cam-
bridge University Press.
Badarinza, Cristian, & Ramadorai, Tarun. 2018. Home away from home? Foreign demand and
London house prices. Journal of Financial Economics,130(3), 532–555.
Bajari, Patrick, & Kahn, Matthew E. 2005. Estimating housing demand with an application to
explaining racial segregation in cities. Journal of Business and Economic Statistics,23(1), 20–
33.
Baryla, Edward, & Ztanpano, Leonard. 1995. Buyer search duration in the residential real estate
market: The role of the real estate agent. Journal of Real Estate Research,10(1), 1–13.
Baur, Dirk G, & McDermott, Thomas K. 2010. Is gold a safe haven? International evidence.
Journal of Banking &Finance,34(8), 1886–1898.
Bloomberg. 2016. Mainland buyers of HK homes help drive prices. China Daily, Oct.
23
Electronic copy available at: https://ssrn.com/abstract=3477421

Bolton, Ruth N, & Myers, Matthew B. 2003. Price-based global market segmentation for services.
Journal of Marketing,67(3), 108–128.
Borjas, George J. 2002. Homeownership in the immigrant population. Journal of Urban Eco-
nomics,52(3), 448–476.
Campbell, John Y, Giglio, Stefano, & Pathak, Parag. 2011. Forced sales and house prices. Ameri-
can Economic Review,101(5), 2108–31.
Card, David, Mas, Alexandre, & Rothstein, Jesse. 2008. Tipping and the dynamics of segregation.
The Quarterly Journal of Economics,123(1), 177–218.
Carozzi, Felipe, Cheshire, Paul, & Hilber, Christian AL. 2018. Introduction: Measuring afford-
ability: Alternative perspectives. 14th Annual Demographia International Housing Affordability
Survey: 2018 Rating Middle-Income Housing Affordability, AI.
Cvijanovi´
c, Dragana, & Spaenjers, Christophe. 2020. “We’ll always have Paris”: Out-of-country
buyers in the housing market. Management Science.
Deng, Yongheng, Hu, Maggie R., & Lee, Adrian D. 2020. Melting pot or salad bowl:
Cultural distance and housing investments. Real Estate Economics,n/a(n/a). eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1111/1540-6229.12311.
Fischer, Andreas M. 2012. Immigrant language barriers and house prices. Regional Science and
Urban Economics,42(3), 389–395.
Gonzalez, Libertad, & Ortega, Francesc. 2013. Immigration and housing booms: Evidence from
Spain. Journal of Regional Science,53(1), 37–59.
Harding, John P, Rosenblatt, Eric, & Yao, Vincent W. 2009. The contagion effect of foreclosed
properties. Journal of Urban Economics,66(3), 164–178.
He, Zhiguo, Hu, Maggie, Wang, Zhenping, & Yao, Vincent. 2020. Valuation of Long-Term Prop-
erty Rights under Political Uncertainty. NBER Working Paper.
24
Electronic copy available at: https://ssrn.com/abstract=3477421

Holmes, Thomas J. 1989. The effects of third-degree price discrimination in oligopoly. American
Economic Review,79(1), 244–250.
Hood, Matthew, & Malik, Farooq. 2013. Is gold the best hedge and a safe haven under changing
stock market volatility? Review of Financial Economics,22(2), 47–52.
Hu, Maggie, Xu, Ke, & Abdullah, Yavas. 2020. Demystifying the mini-flats frenzy in Hong Kong.
Working paper.
Ihlanfeldt, Keith, & Mayock, Tom. 2012. Information, search, and house prices: Revisited. The
Journal of Real Estate Finance and Economics,44(1-2), 90–115.
Klingler, Sven, & Lando, David. 2018. Safe haven CDS premiums. The Review of Financial
Studies,31(5), 1856–1895.
Kwan, Shawna. 2020. Hong Kong Homes Remain World’s Least Affordable for 10th Year.
Bloomberg.
Lam, Kit-Chun, & Liu, Pak Wai. 1998. Immigration and the economy of Hong Kong. Vol. 18. City
University of Hong Kong Press.
Lambson, Val E, McQueen, Grant R, & Slade, Barrett A. 2004. Do out-of-state buyers pay more for
real estate? An examination of anchoring-induced bias and search costs. Real Estate Economics,
32(1), 85–126.
Li, Zhimin, Shen, Leslie Sheng, & Zhang, Calvin. 2020. Capital flows, asset prices, and the real
economy: A “China shock” in the US real estate market. Working paper.
Ling, David C, Naranjo, Andy, & Petrova, Milena T. 2018. Search costs, behavioral biases, and
information intermediary effects. The Journal of Real Estate Finance and Economics,57(1),
114–151.
Liu, Pearl. 2018. Meet the ‘new wave’ of buyers who could keep Hong Kong’s property market
hot. South China Morning Post, Jul.
25
Electronic copy available at: https://ssrn.com/abstract=3477421

M¨
oller, Marc, & Watanabe, Makoto. 2010. Advance purchase discounts versus clearance sales.
The Economic Journal,120(547), 1125–1148.
Ng, Naomi, & Ng, Kang-chung. 2018. Number of mainland Chinese migrants coming to Hong
Kong drops by almost 15,000 from last year. South China Morning Post, Aug.
Ottaviano, Gianmarco IP, & Peri, Giovanni. 2006. The economic value of cultural diversity: Evi-
dence from US cities. Journal of Economic Geography,6(1), 9–44.
Ottaviano, Gianmarco IP, & Peri, Giovanni. 2012. The effects of immigration on US wages and
rents: A general equilibrium approach. In: Migration Impact Assessment. Edward Elgar Pub-
lishing.
Pavlov, Andrey, & Somerville, Tsur. 2020. Immigration, capital flows and housing prices. Real
Estate Economics,48(3), 915–949.
Piazzesi, Monika, & Schneider, Martin. 2009. Momentum traders in the housing market: survey
evidence and a search model. American Economic Review,99(2), 406–11.
Ranaldo, Angelo, & S¨
oderlind, Paul. 2010. Safe haven currencies. Review of Finance,14(3),
385–407.
Rosen, Sherwin. 1974. Hedonic Prices and Implicit Markets: Product Differentiation in Pure
Competition. The Journal of Political Economy,82(1), 34–55.
S´
a, Filipa. 2015. Immigration and House Prices in the UK. The Economic Journal,125(587),
1393–1424.
Saiz, Albert. 2007. Immigration and housing rents in American cities. Journal of Urban Eco-
nomics,61(2), 345–371.
Saiz, Albert, & Wachter, Susan. 2011. Immigration and the neighborhood. American Economic
Journal: Economic Policy,3(2), 169–88.
26
Electronic copy available at: https://ssrn.com/abstract=3477421

Stavins, Joanna. 2001. Price discrimination in the airline market: The effect of market concentra-
tion. Review of Economics and Statistics,83(1), 200–202.
Stole, Lars. 2003. Price discrimination and imperfect competition. Handbook of Industrial Orga-
nization,3, 34–47.
Tu, Yong, Li, Pei, & Qiu, Leiju. 2017. Housing search and housing choice in urban China. Urban
Studies,54(8), 1851–1866.
Turnbull, Geoffrey K, & Sirmans, Casey F. 1993. Information, search, and house prices. Regional
Science and Urban Economics,23(4), 545–557.
Wilhelmsson, Mats. 2008. Evidence of buyer bargaining power in the Stockholm residential real
estate market. Journal of Real Estate Research,30(4), 475–500.
Zhang, Junfu, & Zheng, Liang. 2015. Are people willing to pay for less segregation? Evidence
from US internal migration. Regional Science and Urban Economics,53, 97–112.
Zhou, Xiaorong, Gibler, Karen, & Zahirovic-Herbert, Velma. 2015. Asymmetric buyer information
influence on price in a homogeneous housing market. Urban Studies,52(5), 891–905.
Zunz, Olivier. 1982. The changing face of inequality: Urbanization, industrial development, and
immigrants in Detroit, 1880-1920. University of Chicago Press.
27
Electronic copy available at: https://ssrn.com/abstract=3477421

Figure 1: Socioeconomic Events and Housing Price in Hong Kong
28
Electronic copy available at: https://ssrn.com/abstract=3477421

(a) Total Price
(b) Unit Price Per Square Foot
Figure 2: Housing Price by Origin of Buyers
29
Electronic copy available at: https://ssrn.com/abstract=3477421

(a) Total Price
(b) Transaction Number
Figure 3: Hong Kong Housing Market and Percentage of Mainland Buyers by Year
Note: Price adjusted by CPI of the month.
30
Electronic copy available at: https://ssrn.com/abstract=3477421

Figure 4: Percentage of Mainland Buyer and HKD/CNY Exchange Rate
31
Electronic copy available at: https://ssrn.com/abstract=3477421

Figure 5: By-year Estimates of Mainland Buyers’ Impact on Housing Price
Note: Standard errors clustered by district and year. 95% confidence intervals are plotted with error bars.
32
Electronic copy available at: https://ssrn.com/abstract=3477421

Table 1: Summary Statistics
(1) (2) (3) (4) (5) (6) (7) (8)
Local Buyer Migrant Buyer t-test
Obs Mean Std. Dev. Obs Mean Std. Dev. Diff. (2)-(5) Std. Err.
total price (million HKD) 662,334 3.126 2.486 25,264 4.421 3.508 -1.295*** 0.016
unit price (thousand HKD p.s.f) 662,334 5.824 3.368 25,264 7.527 4.020 -1.703*** 0.004
net unit area (hundred sq. ft.) 662,334 5.208 1.854 25,264 5.567 2.160 -0.359*** 0.012
number of bedrooms 597,317 1.925 1.011 23,730 1.970 1.033 -0.045*** 0.007
number of living rooms 601,592 1.530 0.778 23,789 1.579 0.764 -0.049*** 0.005
floor 662,334 16.465 11.715 25,264 17.906 13.259 -1.441*** 0.076
building age 662,334 18.186 10.133 25,264 17.012 10.889 1.174*** 0.065
bay-window size (sq. ft.) 662,334 16.192 16.013 25,264 18.633 16.021 -2.441*** 0.103
building completion year 662,334 1991 10.018 25,264 1993 11.151 -2.512*** 0.065
to train station (km) 662,334 0.705 0.894 25,264 0.670 0.890 0.035*** 0.006
to bus stop (km) 662,334 0.349 0.337 25,264 0.352 0.346 -0.003 0.002
to hospital (km) 662,334 1.602 1.331 25,264 1.601 1.454 0.001 0.009
to school (km) 662,334 0.144 0.202 25,264 0.137 0.182 0.007*** 0.001
to university (km) 662,334 3.430 2.743 25,264 3.518 3.215 -0.088*** 0.018
to coastal line (km) 662,334 1.360 1.583 25,264 1.263 1.675 0.097*** 0.010
Note: This table presents the summary statistics of our regression sample of resale housing transactions in Hong Kong from 2001 to 2017. Columns (1) to (3)
summarize the transactions made by local Hong Kong buyers. Columns (4) to (6) summarize the transactions made by mainland Chinese buyers. Columns (7) and
(8) present the t-test results of Columns (2) and (5). *** p<0.01, ** p<0.05, * p<0.1
33
Electronic copy available at: https://ssrn.com/abstract=3477421

Table 2: Effect of Mainland Buyers on Housing Price
(1) (2) (3)
2001 - 2017
log(price) log(price) log(price)
Mainland Buyer (Yes =1) 0.0139*** 0.0156*** 0.0200***
(0.0015) (0.0014) (0.0021)
Unit-level Features Y Y Y
Estate Fixed Effects Y N N
Year*Quarter Fixed Effects Y N N
Estate*Year Quarter Fixed Effects N Y N
Building-level Features N N Y
District*Year Quarter Fixed Effects N N Y
Observations 687,598 687,598 687,598
R-squared 0.957 0.978 0.932
Note: This table presents the estimated effects of mainland Chinese buyers on housing price in Hong Kong.
Unreported control variables for unit-level features include number of bedrooms (in categories, with the missing
values in an extra category), number of living rooms (in categories, with the missing values in an extra category),
bay-window indicator (whether or not included in housing price), bay-window size (in 10 equally sized categories),
net unit size (in 10 equally sized categories), direction facing dummies, and floor group dummies (each floor group is
formed by firstly classifying buildings to four groups, which are VeryLowRise (on or less than 10 floors), LowRise
(11 to 30 floors), MidRise (31 to 60 floors), and HighRise (61 floors or higher). Secondly, within each building group,
we form the floor groups for every 5 floors. For example, group VeryLowRise1 includes units on or below 5 floors in
buildings belong to VeryLowRise category.) Unreported control variables for building-level features include age of
buildings (in 10 equally sized categories), building completion years (in 10 equally sized categories), swimming pool
indicator, club house indicator, and distance to train station (MTR)/bus stop/hospital/school/university/coastal line
(each in 10 equally sized categories). In Column (1), we include the controls for unit-level features, the estate fixed
effects and year-quarter fixed effects. Private single buildings are considered as individual estates. In Column (2), we
include the controls for unit-level features and the estate times year quarter fixed effects. In Column (3), we include
the controls for unit-level and building-level features, as well as the district times year quarter fixed effects. Standard
errors clustered by district and year. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
34
Electronic copy available at: https://ssrn.com/abstract=3477421

Table 3: Effect of Mainland Buyers on Housing Price: Heterogeneity Analysis
Panel A. Heterogeneity Across Regions
(1) (2) (3)
HK Island Kowloon New Territories
log(price) log(price) log(price)
Mainland Buyer (Yes =1) 0.0155*** 0.0137*** 0.0107***
(0.0028) (0.0030) (0.0018)
Unit-level Features Y Y Y
Estate Fixed Effects Y Y Y
Year*Quarter Fixed Effects Y Y Y
Observations 139,412 165,431 382,755
R-squared 0.962 0.963 0.956
Panel B. Heterogeneity Across Net Unit Sizes
(1) (2) (3)
<430 sq. ft. 430-861 sq. ft. >861 sq. ft.
log(price) log(price) log(price)
Mainland Buyer (Yes =1) 0.0026 0.0153*** 0.0352***
(0.0017) (0.0018) (0.0044)
Unit-level Features Y Y Y
Estate Fixed Effects Y Y Y
Year*Quarter Fixed Effects Y Y Y
Observations 233,165 415,111 39,322
R-squared 0.953 0.954 0.902
Panel C. Heterogeneity Across Periods
(1) (2) (3)
2001-2006 2007-2012 2013-2017
log(price) log(price) log(price)
Mainland Buyer (Yes =1) -0.0048 0.0196*** 0.0100***
(0.0032) (0.0016) (0.0016)
Unit-level Features Y Y Y
Estate Fixed Effects Y Y Y
Year*Quarter Fixed Effects Y Y Y
Observations 201,881 350,118 135,599
R-squared 0.948 0.960 0.948
Note: This table presents the estimated heterogeneous effects of mainland Chinese buyers on housing price across
regions (Panel A), net unit sizes (Panel B), and periods (Panel C). Unreported control variables for unit-level features
are the same as in Table 2. Standard errors clustered by district and year. Robust standard errors in parentheses. ***
p<0.01, ** p<0.05, * p<0.1
35
Electronic copy available at: https://ssrn.com/abstract=3477421

Table 4: Effect of Lagged Mainland Buyers on Housing Price: First-stage Results of IV Estimation
(1) (2)
2011-2017
Proportion of Mainland Buyers log(Number of Mainland Buyers)
Predicted Proportion of Mainland Buyers 0.0738***
(0.0071)
Predicted Number of Mainland Buyers 0.1637***
(0.0057)
Unit-level Features Y Y
Estate Fixed Effects Y Y
Year*Quarter Fixed Effects Y Y
First-stage F-stats 107.24 819.53
Observations 221,854 226,925
Note: This table presents the first-stage IV estimation results for the effects of lagged mainland Chinese buyers on subsequent housing price in the same building
in Hong Kong. We use the Bartik shift-share predictions of mainland buyers as the instrument. In Column (1), the independent variable is the predicted proportion
of mainland buyers in the building in the previous one year. In Column (2), the independent variable is the predicted number of mainland buyers in the building in
the previous one year. Subsamples from 2011 to 2017 are included. Unreported control variables for unit-level features are the same as in Table 2. Standard errors
clustered by district and year. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
36
Electronic copy available at: https://ssrn.com/abstract=3477421

Table 5: Effect of Lagged Mainland Buyers on Housing Price: OLS and Second-stage IV
Estimation Results
(1) (2) (3) (4)
2011-2017
OLS IV OLS IV
log(price) log(price) log(price) log(price)
Lagged Proportion of Mainland
Buyers
0.0752*** 0.1706
(0.0088) (0.1091)
log(Lagged Number of Mainland
Buyers)
0.0155*** 0.0542***
(0.0024) (0.0087)
Mainland Buyer 0.0240*** 0.0240*** 0.0233*** 0.0221***
(0.0019) (0.0019) (0.0019) (0.0017)
Unit-level Features Y Y Y Y
Estate Fixed Effects Y Y Y Y
Year*Quarter Fixed Effects Y Y Y Y
First-stage F-stats 107.24 819.53
Observations 221,854 221,854 226,925 226,925
R-squared 0.948 0.947 0.948 0.947
Note: This table presents the estimated effects of lagged mainland Chinese buyers on subsequent housing price in the
same building in Hong Kong. In Column (1), the independent variable is the standardized proportion of mainland
buyers in the building in the previous one year. In Column (2), the independent variable is the logarithmic form of the
number of mainland buyers in the building in the previous one year. Columns (1) and (3) present the OLS estimation
results. Columns (2) and (4) report the second-stage IV estimation results, using the Bartik shift-share predictions of
mainland buyers as the instrument. Subsamples from 2011 to 2017 are included. Unreported control variables for
unit-level features are the same as in Table 2. Standard errors clustered by district and year. Robust standard errors in
parentheses. *** p<0.01, ** p<0.05, * p<0.1
37
Electronic copy available at: https://ssrn.com/abstract=3477421

Table 6: Channel of Residential Sorting
(1) (2)
Logit Logit
Mainland Buyer (Yes =1)
Lagged SD Proportion of Mainland Buyers 0.0030***
(0.0002)
Lagged Log(Number of Mainland Buyers) 0.0090***
(0.0006)
Unit-level Features Y Y
Building-level Features Y Y
District Fixed Effects Y Y
Year*Quarter Fixed Effects Y Y
Observations 673,757 687,473
Pseudo R-squared 0.062 0.063
Note: This table presents the estimated effects of lagged mainland Chinese buyers on attracting subsequent mainland
Chinese buyers in the same building in Hong Kong. The dependent variable is a dummy variable indicating the
mainland buyers. In Column (1), the independent variable is the proportion of mainland buyers in the building in the
previous one year. In Column (2), the independent variable is the logarithmic form of the number of mainland buyers
in the building in the previous one year. Unreported control variables for unit-level and building-level features are the
same as in Table 2. Standard errors clustered by district and year. Robust standard errors in parentheses. *** p<0.01,
** p<0.05, * p<0.1
38
Electronic copy available at: https://ssrn.com/abstract=3477421

Table 7: Horse Racing Analysis of Channels
(1) (2) (3) (4) (5) (6)
2010-2017
log(price) log(price) log(price) log(price) log(price) log(price)
Mainland Buyer 0.0250*** 0.0411*** 0.0172*** 0.0252*** -0.0036 0.0024
(0.0018) (0.0087) (0.0012) (0.0018) (0.0059) (0.0105)
Mainland Buyer*Lagged SD HKD/CNY Exchange Rate 0.0188** 0.0186**
(0.0093) (0.0082)
Mainland Buyer*Lagged SD Proportion of Mainland
Buyers
0.0136*** 0.0130***
(0.0017) (0.0018)
Mainland Buyer*SD Buyer’s Prior Transaction Times -0.0027*** -0.0017*
(0.0010) (0.0009)
Mainland Buyer*Local Seller 0.0279*** 0.0320***
(0.0056) (0.0057)
Mainland Seller -0.0045*** -0.0042***
(0.0016) (0.0016)
Unit-level Features Y Y Y Y Y Y
Estate Fixed Effects Y Y Y Y Y Y
Year*Quarter Fixed Effects Y Y Y Y Y Y
Observations 304,227 304,227 298,681 304,227 279,633 274,866
R-squared 0.953 0.953 0.953 0.953 0.953 0.953
Note: This table presents the horse racing analysis results for the impacts of different channels on housing price premium paid by mainland Chinese buyers. The
measurements for each channel include the standardized lagged HKD/CNY exchange rate, standardized lagged proportion of mainland buyers, standardized buyers’
prior transaction times, and dummy variables for local (mainland) sellers. The lag period for the HKD/CNY exchange rate is one month. The lag period for the
proportion of mainland buyers is one year, and the proportion is calculated within the same building. Unreported control variables for unit-level features are the
same as in Table 2. Standard errors clustered by district and year. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
39
Electronic copy available at: https://ssrn.com/abstract=3477421

Appendix
Figure A1: Number of Mainland Chinese One-way Permit (OWP) Holders Entering Hong Kong
Note: The time period used is from the middle of one year to the next. Statistics are obtained from the Census and
Statistics Department (CSD) of Hong Kong.
40
Electronic copy available at: https://ssrn.com/abstract=3477421

Figure A2: Housing Price and Housing Market Policies in China
41
Electronic copy available at: https://ssrn.com/abstract=3477421

Table A1: Definition of Variables
Variable Name Definition
Region 1 =Hong Kong Island; 2 =Kowloon; 3 =New Territories
District District codes assigned by EPRC, including:
1=Aberdeen/Ap Lei Chau; 2 =Causeway Bay; 3 =Central; 4 =Chai
Wan; 5 =Happy Valley; 6 =Kennedy Town; 7 =Mid-level West; 8 =
Mid-level Central; 9 =Mid-level East; 10 =North Point; 11 =North
Point Hill; 12 =Peak; 13 =Pokfulam; 14 =Quarry Bay; 15 =Repulse
Bay; 16 =Sai Ying Pun; 17 =Shau Kei Wan; 18 =Sheung Wan; 19
=Siu Sai Wan; 20 =Stanley; 21 =Tai Tam; 22 =Wan Chai; 23 =
Wong Chuk Hang; 24 =Cheung Sha Wan; 25=Diamond Hill; 26 =
Ho Man Tin; 27 =Hung Hom; 28 =Kai Tak; 29 =Kowloon Bay; 30
=Kowloon City; 31 =Kowloon Tong; 32 =Kwun Tong; 33 =Lai
Chi Kok; 34 =Lam Tin; 35 =Mong Kok; 36 =Ngau Chi Wan; 37 =
Ngau Tau Kok; 38 =San Po Kong; 39 =Sham Shui Po; 40 =Shek
Kip Mei; 41 =Tai Kok Tsui; 42 =Tsim Sha Tsui; 43 =Tsz Wan Shan;
44 =Wang Tau Hom; 45 =Wong Tai Sin; 46 =Yau Ma Tei; 47 =Yau
Tong; 48 =Fan Ling; 49 =Islands; 50 =Kwai Chung; 51 =Ma On
Shan; 52 =Sai Kung; 53 =Sha Tin; 54 =Sheung Shui; 55 =Tai Po;
56 =Tseung Kwan O; 57 =Tsing Yi; 58 =Tsuen Wan; 59 =Tuen
Mun; 60 =Yuen Long.
Building Type 1 =Private Estate; 2 =Private Single; 3=Government
Private estate building refers to an apartment building from a estate
(project) that has multiple buildings. The buildings in the same estate
usually share some common facilities, such as swimming pool or club
house. Private single building refers to an estate that only have one
single apartment building, which normally has less facilities than the
multi-block estates. A government building refers to the public hous-
ing initially constructed by the Hong Kong government, of which the
units are permitted to be resold in the private market at market price
if the sellers pay back the full subsidies and the land premiums to the
Housing Authority. We exclude the landed houses (i.e., the Village
Houses) in Hong Kong, which are mostly located in the suburban areas
of the New Territories region and are subject to restrictions in resales.
Price Total pretax contract price.
Unit Price Total price divided by the net area of the unit.
42
Electronic copy available at: https://ssrn.com/abstract=3477421

Table A2: Distribution of Residential Building Types in Hong Kong
(1) (2) (3) (4) (5) (6) (7) (8)
All Hong Kong Island Kowloon New Territories
Freq. Pct. Freq. Pct. Freq. Pct. Freq. Pct.
Private Estate 468,176 68.09 71,936 51.60 98,520 59.55 297,720 77.78
Private Single 119,695 17.41 51,669 37.06 46,024 27.82 22,002 5.75
Government 99,727 14.50 15,807 11.34 20,887 12.63 63,033 16.47
Total 687,598 100 139,412 100 165,431 100 382,755 100
Note: This table summarizes the distribution of residential property transactions in Hong Kong by regions and
building types. A private estate contains multiple buildings that share the common property management and facilities
in the estate. A private single building is a stand-alone residential building which does not belong to a residential
estate. A government building refers to the public housing initially constructed by the government. In Hong Kong,
public housing units are permitted to be resold to any buyers in the private market at market price if the sellers pay
back the full subsidies and the land premiums to the Housing Authority.
43
Electronic copy available at: https://ssrn.com/abstract=3477421