The E¤ect of Foreign Investors on Local Housing Markets:
Evidence from the UK
y
Filipa Sá
June 30, 2017
Abstract
I use newly-released administrative data on properties owned by overseas companies to study
the e¤ect of foreign investment on the housing market in England and Wales. To estimate the
causal e¤ect, I construct an instrument for foreign investment based on economic shocks abroad.
Foreign investment is found to have a positive e¤ect on house price growth. This e¤ect is present
at di¤erent percentiles of the distribution of house prices and is stronger in local authorities
where housing supply is less e lastic. Foreign investment is also found to reduce the rate of home
ownership. There is no evidence of an e¤ect on the housing stock or the share of vacant homes.
Key words: foreign investors, house prices
JEL Classi…cation: R 21, F21
Filipa Sá, School of Management & Business, King’s College London, Franklin-Wilkins Building, 150 Stamford
Street, London SE1 9NH, UK. Email: …lipa.sa@kcl.ac.uk.
y
I would like to thank Beatrice Faleri and Xinyu Li for excellent research assistance. I am very grateful to Philippe
Bracke and Alan Manning for their insightful comments and sug gestions and to Stuart Gray at the Land Registry
for help with the data . I have also received useful comments from seminar participants at the European M eeting of
the Urban Economics Association in C openh agen and the Bank of England.
1
1 Introduction
House prices in the UK have increased signi…cantly since the late 1990s. Figure 1 reports average
house prices in England and Wales using data from the Land Registry house price index database.
Average house prices almost tripled during the period shown, from just over £ 70,000 in 1999 to
about £ 215,000 in 2014. Apart from a reduction in 2009, at the height of the global …nancial crisis,
house prices increased every year during this period. What factors may be behind this upward
trend in house prices?
On the supply side, lack of available land for construction and regulatory constraints — such as
planning delays and restrictions — may be contributing to house price appreciation. In 2004, the
UK government commissioned a review on housing supply (Barker 2004). This review concluded
that a much higher rate of house building would be necessary to reduce the trend rate of house
price growth to a level comparable with the EU average. A related review on land use planning
(Barker 2006) argued for a need to make the planning application process more e¢ cient and to
incentivise the use of vacant previously developed land. A recent study by Hilber and Vermeulen
(2015) looks at how supply constraints a¤ect the transmission of income shocks to house prices.
They …nd that an increase in earnings raises house prices by more in areas with tighter regulatory
constraints (measured by the refusal rate of major residential projects).
On the demand side, Sá (2015) uses an IV approach to examine the causal e¤ect of immigration
on house prices. The …ndings suggest that an in‡ow of immigrants into a local area generates native
outmigration and has a negative e¤ect on local house prices. Badarinza and Ramadorai (2016) look
at the e¤ect of foreign investment on house prices in London. They construct a proxy for foreign
investment based on two ideas: …rst, foreign investors are more likely to invest in the UK property
market when their home countries face negative economic con ditions; second, foreign investors
exhibit "home bias abroad", i.e., they tend to choose areas in the UK where people from their home
country live. The focus of their paper is di¤erent from Sá (2015) because foreign buyers are not
necessarily residents in the UK, but may be purchasing properties purely for investment purposes.
The authors …nd a signi…cant "safe haven" e ¤ect, with increases in foreign risk bein g associated
with higher house prices in parts of London with a high share of foreign-origin residents. A related
strand of literature uses macro data to examine the e¤ect of foreign capital in‡ows on the housing
2
market. Aizenman and Jinjarak (2009) use panel regressions to study the association between the
current account and real estate prices across countries. They conclude that there is a positive
association between current account de…cits and appreciation of real estate prices. Sá, Towbin and
Wieladek (2014) estimate a panel VAR for OECD countries and look at the e¤ect of capital-in‡ow
shocks on the housing market. They …nd that capital-in‡ow shocks have a signi…cant and positive
e¤ect on real house prices, real credit to the private sector, and real residential investment.
In this paper, I look at the e¤ect of foreign investment on the housing market in England and
Wales. Unlike Badarinza and Ramadorai (2016), I do not use economic shocks abroad as a proxy
for foreign investment, but measure foreign investment directly using a new dataset released by
the Land Re gistry. This dataset contains information on all property transactions in England and
Wales registered to overseas companies. My sample is broader than the one in Badarinza and
Ramadorai (2016) and includes all local authorities in England and Wales and not just in London.
Also, while Badarinza and Ramadorai (2016) focus on the average e¤ect on hous e prices, I also
look at heterogeneous e¤ects across the distribution of house prices and across local authorities
with di¤erent levels of supply constraints. In addition, I look at other outcome variables, such as
the housing stock, vacant homes and home ownership.
The newly-released Land Registry dataset is used to calculate the share of total residential
transactions registered to overseas companies in each year and local authority for the period from
1999 to 2014. This share is calculated in volume and in value (using information on transaction
prices). To see how these shares correlate with house prices, Figure 1 plots the evolution of house
prices and foreign transactions calculated as averages across all local authorities in England and
Wales. All three series display an upward trend. The value share is more volatile than the volume
share, probably due to measurement error in reported transaction prices. This issue is discussed in
more detail in the data section.
Behind these averages, there is signi…cant regional variation, as illustrated in Figures 2 and 3
and in Table 1. Average house prices in the S outh East, particularly in London, are signi…cantly
higher than in other regions, reaching a value of almost £ 1.3 million in Kensington and Chelsea
in 2014. Foreign investment also tends to be concentrated in the South East — Westminster
and Kensington and Chelsea have by far the largest shares of transactions registered to overseas
companies. Some of the major cities in the North — such as Liverpool, Leeds and Manchester —
3
also attract large shares of foreign investment. Interestingly, the city of Salford (part of Greater
Manchester) had a signi…cant share of foreign investment in 2014. After looking at the addresses
of properties in Salford sold to overseas investors in that year, it appears that many of them are in
an apartment block partly funded by Chinese investors
1
.
This paper makes use of this regional variation and identi…es the e¤ect of foreign investment
on house prices from spatial correlations between the share of foreign transactions and changes
in house prices across local authorities. A potential problem in interpreting these correlations as
causal e¤ects is that the direction of causality between foreign investment and hou se prices is not
clear, because foreign investors are not randomly allocated across geographic areas. To overcome
this issue, I borrow from Badarinza and Ramadorai (2016) and use measures of economic shocks
abroad and the idea of "home bias" to construct an instrument for foreign investment.
To preview the results, I …nd that foreign investment has a positive and signi…cant e¤ect on
house prices. An increase of one percentage point in the volume share of residential transactions
registered to overseas companies leads to an increase of about 2:1 percent in house prices. To have a
better idea of the magnitude of this e¤ect, I use the model to construct the counterfactual evolution
of house prices when the share of foreign investment is set to zero. I …nd that average house prices in
England and Wales in 2014 would have been about 19% lower in the absence of foreign investment
(at approximately $174; 000, compared with an actual average of about $215; 000). Looking at
the e¤ect at di¤erent points of the distribution of house prices, I …nd that foreign investment
does not just raise prices of expensive homes, but has a positive e¤ect at di¤erent percentiles of the
distribution. To examine how supply restrictions a¤ect the propagation of foreign investment shocks
to house prices, I use the data in Hilber and Vermeulen (2015) to calculate the house price-earnings
elasticity for each local authority in England. I then divide local authorities into di¤erent quartiles
of this elasticity and estimate the model separately for each quartile. As expected, I …nd that
foreign investment increases prices by more in local authorities with a larger house price-earnings
elasticity, which are the ones with a less elastic housing supply.
I also examine th e e¤ect of foreign investment on other outcome variables. Looking at the
e¤ect on the housing stock, I do not …nd evidence that an increase in foreign investment leads
1
A report by the BBC on Chinese investment in major cities in the North, particularly Salford, can be found at:
http://www.bbc.co.uk/new s/business-36086 012
4
to an increase in housing construction, contrary to the suggestions of some estate agents in prime
London areas
2
. I also do not …nd evidence in favour of "buy-to-leave " — the hypothesis that foreign
buyers purchase properties purely for capital appreciation and do not occupy them or rent them
out. However, I do …nd evidence that foreign inve stment reduces home ownership rates, suggesting
that some residents may be priced out of the market in areas where foreign investors are more
active and have to rent rather than own their homes.
The …ndings in this paper are useful to inform the policy debate on the impact of foreign invest-
ment on the housing market. This topic has attracted the attention of the Mayor of London (Sadiq
Kahn), who has recently launched an inquiry into the consequences of foreign property ownership
in the capital. Other countries have also been debating this issue and have introduced policies to
control foreign investment in the housing market in an attempt to reduce house price appreciation.
For example, Australia has a legislative framework which encourages foreign investment in new
residential projects, but imposes tighter controls on the purchase of existing dwellings (Gauder,
Houssard and Orsmond (2014)); Switzerland has quotas on the number of residential properties
that can be s old to foreigners; and the Canadian city of Vancouver h as recently introduced a 15%
property tax on foreign home buyers.
The rest of the paper is organised as follows. The next section describes the key data sources,
particularly the new Land Registry Overseas Companies Dataset (OCD). Section 3 discusses the
empirical methodology use d to estimate the e¤ect of foreign investment on house prices and presents
the results. Section 4 discusses some robustness checks. Section 5 looks at the e¤ec t of foreign
investment on the h ousing stock, vacant homes and home ownership. Section 6 concludes.
2 Data
2.1 Investment by overseas companies
Data on all land, commercial and residential properties in England and Wales registered to overseas
companies are obtained from the Land Registry Overseas Companies Dataset (OCD), published in
March 2016. The dataset contains around 100,000 title records and collects information on tenure
2
Fo r examp le, research by Savills (2014) suggests that foreign investors play a major role in sti mulating the supply
of new housi ng in L ondon.
5
(freehold or leasehold), address, price paid (where available), name and country of incorporation
of the legal owner and date of registration. While the data go as far back as the mid-1960s,
the majority of records were registered from 1999 onwards. The public version of the datase t
contains data up to October 2015. The dataset includes overseas companies only and does not
cover properties owned by private individuals, UK companies or charities.
Prior to the release of this dataset, similar d ata were published on the website of the Private
Eye magazine, together with an informative map showing all land and properties in England and
Wales registered to overseas companies
3
. The Private Eye obtained these data from the Land
Registry under a Freedom of Information request. Their dataset covers the period from 1999 to
2014 and contains more complete information on price paid than the Land Registry dataset. I use
this dataset to …ll in missing information on prices in the Land Registry OCD. Therefore, the data
used for the empirical analysis cover the period from 1999 to 2014.
To assess the relative importance of foreign investment in a given local authority and year, it
is important to scale foreign transactions by the total number of transactions. To do this, I use
information on all property sales in En gland and Wales from the Land Registry Price Paid Data
(PPD). This dataset only covers residential properties sold at full market value. Until October
2013, only sales to private individuals were included. Since then, the dataset also includes buy-
to-let sales and sales to companies. To ensure consistency across time, I focus on transfers of
residential property to private individuals for all years.
I use two measu res of foreign investment by local authority and year. The …rst measure is
the volume share of foreign transactions and is calculated by dividing the number of reside ntial
properties registered to overseas companies (from the Land Registry OCD) by the total number of
residential transfers to private individuals (from the Land Registry PPD). The second measure is
the value share of foreign transactions and is obtained by dividing the total value of all residential
properties registered to overseas companies by the total value of all residential properties bought
by private individuals. The append ix contains more details on the construction of the volume and
value shares.
There are some potential measurement problems with these shares of foreign investment. On e
problem is that the Land Registry OCD only includes purchases by overseas companies, but not
3
http://www.private-eye.co.u k/registry
6
by non-resident private individuals. This implies th at the shares of foreign transactions may un-
derestimate the true importance of foreign investment in the local housing market. In practice,
however, most foreign investment is likely to be directed through a company, because tax rates on
rental income are generally lower for non-resident companies than for individuals
4
.
Another issue is that the Land Registry OCD only contains information on the country of
incorporation, bu t does not reveal the country of ultimate ownership of the companies that invest
in UK property. Table 2 lists the countries of incorporation with the largest shares of investment
in UK property in 2014 (in volume). About 34% of p urchases of property in England and Wales by
overseas companies in 2014 were done by companies incorporated in the British Virgin Islands. The
channel islands of Guernsey and Jersey also had large shares, as well as the Isle of Man. All these
territories have low tax rates. It is possible that some UK investors may register a company overseas
in order to pay less taxes. If this is the case, the volume and value shares would overestimate the
true importance of foreign investment in the local h ousin g market because the numerator would
include some properties bought by UK investors via a company registered overseas. To get a sense of
the extent of this problem, the robustness section reports results excluding properties registered to
companies incorporated in Guernsey, Jersey and the Isle of M an. Given their geographic proximity,
these are the most likely countries of incorporation of UK companies seeking to reduce their tax
liabilities.
An additional issue with the calculation of the foreign shares is that the transaction prices
reported in the Private Eye dataset are likely to su¤e r from measurement error. Some of the
information on prices is reported by visitors to the Private Eye website, who can click on a link on
the map to report additional information about a record. Occasionally, price paid …gures contain
errors and may not refer to an individual property, but to all properties registered to the same
company. Because of this issue with the price data, it is useful to consider the results for the value
share alongside those for the volume share, which is calculated solely from administrative data and
does not su¤er from measureme nt problems. In the empirical analysis, I use an IV approach to
deal with endogeneity in the shares of foreign investment, which should help address measurement
error in the value share.
4
Fo reign companies pay corporation tax on rental income at 20%, whereas tax rates for non-resident and resident
individuals vary between 20% and 45%, depending on i ncome.
7
2.2 House prices and other controls
Data on house prices are obtained from the Land Registry house price index dataset. The index is
based on repeated sales and measured at monthly frequency (it is converted to annual by taking
averages). Using repeated sales to measure house price changes has the advantage of holding
constant the quality of the housing stock.
In the regression analysis, I control for local economic conditions by including lags of the unem-
ployment rate and the bene…ts rate (the proportion of the population receiving any state bene…ts).
All data sources and de…nitions are listed in Table A1 in the appendix. Table 3 reports descriptive
statistics for the key variables.
3 Foreign buyers and house prices
3.1 Speci…cation
The following model is used to estimate the e¤ect of foreign investment on house prices:
ln(P
it
) =
F T
it
T T
it
+ X
it1
+
t
+
i
+ "
it
(1)
where ln(P
it
) is the change in the log of the house price index in local authority i between
years t 1 and t. The main independent variable is the share of foreign transactions to total
transactions (
F T
it
T T
it
). As described in the data section, this is obtained by dividing the volume or
value of residential properties registered to overseas companies by the total volume or value of
residential transfers to private individuals. The coe¢ cient can be interpreted as the percentage
change in hou se prices corresponding to an annual increase of one percentage point in the share of
foreign transactions.
X
it1
is a set of controls and includes a lagged dependent variable and one year lags of the
local unemployment rate and the share of the local population claiming state bene…ts. The un-
employment rate and the bene…ts rate capture local macroeconomic conditions, which may a¤ect
housing demand. The model includes a lagged dependent variable to allow for inertia in house price
growth. Case and Shiller (1989) …nd evidence that an increase in house prices in one year tends
8
to be followed by an increase in the subsequent year. Other studies examining the determinants
of house price growth — for example, Favara and Imbs (2015) and Jordà, Schularick and Taylor
(2015) — also include a lagged dependent variable in the model.
Year dum mies (
t
) capture national trends in in‡ation and other economic variables. Since the
model is written in …rst-di¤erences, time-invariant factors that are speci…c to each local authority
and that a¤ect the level of house prices have been di¤erenced out. However, local authority …xed
e¤ects (
i
) are still included to capture di¤erent trends in house prices at the local level. The
model is es timated in …rst di¤erences to account for heterogeneous trends across local authorities
and because house prices are measured as an index, whose level has no economic interpretation.
Following the recommendation in Bertrand, Du‡o and Mullainathan (2004) and Angrist and Pischke
(2009), standard errors are heteroskedasticity-robust and are clustered by local authority to account
for correlation within groups.
The e¤ect of foreign investment on house prices is identi…ed from spatial correlations between
the share of foreign transactions and changes in house prices across local authorities. Identi…cation
relies on variation in the s hare of foreign buyers across local authorities and time.
There are two potential problems in interpreting these correlations as causal e¤ects. First,
foreign investment and hous e prices may b e spatially correlated because of common …xed in‡uences,
for example, the climate or local amenities. This would lead to a correlation between the two
variables, even in the absence of any genuine e¤e cts of foreign investment. The second problem
is that the direction of causality between foreign investment and house prices is not clear because
foreign investors are not randomly allo cated across geographic areas.
To address the …rst problem, the model is estimated with the dependent variable in …rst-
di¤erences. This eliminates time-invariant, area-speci…c factors that a¤ect foreign investment and
house prices. To address the second problem, I construct an instrument for the value and volume
shares of foreign investment. The instrument is based on the analysis in Badarinza and Ramadorai
(2016), who look at the e¤ect of foreign investment on house prices in London. The authors do not
look directly at measures of foreign investment, since this information was not available until the
recent release of the Land Registry OCD. Instead, they construct a proxy for foreign investment
based on two ideas: …rst, foreign investors are more likely to invest in the UK property market when
their home countries face negative economic conditions; second, foreign investors exhibit "home bias
9
abroad", i.e., they tend to choose areas in the UK where people from their home country live.
"Home bias abroad" may arise if foreign investors …nd it easier to rent property to residents from
their home country, because there are no language or cultural barriers. Also, in areas where some
nationalities are more highly represented, local estate agents specialise in dealing with investors
from those countries, reducing transaction costs for foreign investors who buy property in those
areas
5
. The notion of "home bias abroad" is closely related to an instrument that is typically used
in the literature on immigration, which relies on the historical settlement pattern of immigrants by
country of origin to predict the current geographic distribution of the immigrant population — see,
for example, Cortes (2008) and Sá (2015). This instrument is based on the notion that immigrant
networks are an important determinant of the locational choices of new immigrants, because they
facilitate the job search process and assimilation into a new culture (Munshi 2003).
I follow Badarinza and Ramadorai (2016) and construct the following instrument for the share
of foreign investment in local authority i in year t:
P
c
f
c
i
z
c
t1
where f
c
i
is the share of residents in local authority i that were born in foreign country c from the
2001 Census and z
c
t1
is a measure of economic conditions in country c in year t 1, speci…cally the
economic risk index from the International Country Risk Guide (ICRG). The index is constructed
by awarding risk points for 5 components: GDP per capita, real annual GDP growth, annual
in‡ation rate, budget balance as a percentage of GDP and current account as a percentage of GDP.
Countries with lower risk are awarded a higher value for the index. The index is available for 140
countries for the period 1984 to 2015. Because population data from the Census is available for
only 60 countries and regions, it is necessary to c ombine the economic risk indices to match the
regions in the Census. I do this by calculating weighted indices us ing population shares from the
IMF World Economic Outlook as weights. In the robustness section, I estimate the model using
an alternative measure of economic conditions abroad to construct the instrument.
The validity of this instrument relies on two identi…cation assumptions. First, I assume that the
geographic distribution of the foreign-born population in 2001 is uncorrelated with recent changes
5
In prime central London locations, there are seve ral property consultancy …rms that specialise in helping Russian
and Chinese investors buy property in those areas.
10
in the economic performance of di¤erent UK local authorities. In that case, f
c
i
is correlated with
changes in house prices on ly through its relation with current foreign investment in those areas.
The second identifying assumption is the exogeneity of economic con ditions abroad to economic
conditions of UK local authorities. This is a plausible assumption because economic conditions
abroad should be largely determine d by country-speci…c political and institutional factors and year
…xed e¤ects should capture global macroeconomic shocks.
3.2 Results
Table 4 reports the results of estimating model (1) by OLS and IV. The OLS results suggest a
positive correlation between the share of foreign investment in volume and in value and house price
growth. The coe¢ cients are very similar with and without controlling for changes in the local
unemployment rate and the bene…ts rate.
The IV coe¢ cients are signi…cantly larger, pointing to considerable attenuation bias in the OLS
estimates, likely due to measu rement error in the shares of foreign investment. The IV results imply
that, on impact, house prices increase by about 2% when the volume share of total transactions
registered to overseas companies increases by one percentage point. For the value share, the increase
in house prices is about 1:4%. The instrument is highly signi…cant in the …rst stage an d has the
expected sign: when economic conditions abroad improve — corresponding to an increase in the
index of economic conditions abroad (weighted by local foreign population) — foreign buyers invest
less in the UK housing market. The F-statistic on the excluded instruments is around 8. This is
below the benchmark value of 10 suggested by Stock, Wright and Yogo (2002), but is above the
20% maximal bias threshold (6:66) in Stock and Yogo (2005). The table also reports the Anderson-
Rubin Wald test of the signi…cance of the foreign shares in the structural equation, which is robust
to the presence of weak instruments. The test indicates that the foreign shares have a signi…cant
e¤ect on house price growth.
To better un derstand the magnitude of these results, I use the IV coe¢ cients to predict the
evolution of average hou se prices when the volume share of overseas transactions is set to zero.
Figure 4 reports the evolution of actual average house prices across local authorities in England
and Wales (solid line), as well as the evolution of average house prices predicted by the model
with the volume share of overseas transactions observed in the data (long dashed line) and with a
11
volume share of overseas transactions equal to zero (short dashed line). Predicted house prices are
close to observed house prices, indicating that the model does well in explaining the evolution of
house prices. Without any foreign investment in the housing market, average house prices would
be lower. For e xample, in 2014, average house prices without foreign investment would be about
19% lower (at approximately $174; 000, compared with an actual average of about $215; 000).
To get a sense for how the response of house prices changes over time, I follow the method
introduced in Jordà (2005) and estimate impulse responses at di¤erent points in time using local
projections. This method estimates impulse responses directly, without the need to specify the
unknown multivariate dynamic process, as would be the case with a vector autoregression (VAR).
Local projections are obtaine d by estimating sequential regressions of the endogenous variable
shifted forward:
ln(P
it+h
) ln(P
it+h1
) =
h
F T
it
T T
it
+ X
it1
+
t
+
i
+ "
it
(2)
The vector of estimates f
h
g
h=0;1;:::
measures the e¤ect of foreign investment on house prices at
horizon h. To acc ount for endogeneity in the share of foreign investment, I estimate the regressions
using the ins trument based on economic shocks abroad. This approach is described in Jordà,
Schularick and Taylor (2015) as local projection instrumental variables (LP-IV) and has also been
used in Favara and Imbs (2015) to study the e¤ect of sho cks to credit supply on house price growth.
Figure 5 reports the impulse responses over a period of four years. The e¤ect of an inc rease in
foreign investment on house price growth is quite persistent and only becomes insigni…cant four
years after the shock.
3.3 E¤ect along the distribution of house prices
Foreign buyers are more active at the top end of the housing market. For example, in 2012, 13%
of all residential property transactions in England and Wales above £ 1 million were registered to
overseas companies, compared with only 2% of properties under £ 1 million. Therefore, it is possible
that most of the impact of foreign investment on house prices is felt at the top end of the market.
To test this hypothesis, I estimate the following model to study the e¤ect of foreign investment
at di¤erent percentiles of the distribution of house prices:
12
ln(P
pit
) =
p
F T
it
T T
it
+
p
X
it1
+
pt
+
pi
+ "
pit
(3)
The d ependent variable is the log of the p th percentile of house prices in local authority i
in year t. As before, the vector X
it1
includes a lagged dependent variable and two controls for
local macroeconomic conditions — one-year lags of the local unemployment rate and of the share
of the local population claiming state bene…ts. The coe ¢ cient
p
captures the e¤ect of overseas
investment on each percentile of the distribution of house prices. This model is similar to the one
adopted in Dustmann, Frattini and Preston (2013) to study the e¤ect of immigration along the
distribution of wages.
Table 5 reports the results of estimating the model for selected percentiles of the distribution
of house prices with the share of foreign transactions measured in volume and in value. Panel A
reports OLS results and panel B rep orts IV results using the instrument based on economic shocks
abroad. Results are reported with and without controlling for the lagged unemployment rate and
bene…ts rate. The OLS coe¢ cients suggest a slightly larger e¤ect of foreign investment at the top
end of the distribution of house prices. However, the IV coe¢ cients point to a similar positive e¤ect
of foreign investment on house prices at all points of th e distribution. For example, an increase in
the volume share of residential properties registered to overseas companies of one percentage point
increases house prices at the 95th percentile by about 2:3% and increase s house prices at the 5th
percentile by about 2:1%.
To obtain a more detailed picture, I estimate the model using a …ner grid of house price per-
centiles. Figure 6 reports the IV coe¢ cients in percentile intervals of …ve percentage points. These
coe¢ cients correspond to the results in columns (2) and (4) of Table 5. The …gure suggests that
foreign investment does not increase prices only for expensive homes, but has a positive e¤ect at
all points of the house price distribution.
3.4 Supply constraints and the e¤ect of foreign investment on house prices
The response of house prices to foreign investment should depend on supply conditions. In local
authorities where housing supply is more constrained by regulation or geography, house prices
should increase by more in response to an increase in f oreign investment. To test this hypothesis, I
13
use the estimates of the house price-earnings elasticity for lo cal authorities in England constructed
by Hilber an d Vermeulen (2015). The h ouse price-earnings elasticity is a proxy for the elasticity of
housing supply — in areas where supply is less elastic, a positive shock to demand (for example,
an increase in earnings) would lead to a larger increase in house prices. Therefore, a higher house
price-earnings elasticity re‡ects a less elastic supply of housing.
Hilber and Vermeulen (2015) estimate the house price-earnings elasticity by running a regression
of the log of house prices on earnings and interactions of earnings with the share of planning
applications for major residential projects that are refused permission, the share of land already
developed and the elevation range. The refusal rate measures regulatory constraints to house
building, while the share of land already developed and the elevation range measure land scarcity.
To address endogeneity in the refusal rate, the authors use the share of votes for the Labour Party
in the 1983 general election as an instrument. This is motivated by the fact that Labour voters
tend to be less protective of housing value s than Conservative voters. As an additional instrument,
they use the change in the delay rate of major planning applications after a reform introduced by
the Labour government in 2002, which set a target to speed up the planning process. To address
endogeneity in the share of land already developed, the authors instrument it with population
density in 1911. The model is estimated using annual data for the period from 1974 to 2008.
I use the estimated coe¢ cients in Hilber and Vermeulen (2015) and their data to construct
estimates of the house price-earnings elasticity for each local authority in England
6
. Figure 7
reports these elasticities and shows that local authorities in the South East are considerably more
elastic, especially in Greater London. Local authorities in the North and East of England have
much lower house price-earnings elasticities, re‡ecting lower regulatory restrictions and more space
available for construction. I then separate local authorities in England into four quartiles, according
to the house price-earnings elasticity, and estimate model (1) separately for each of these quartiles.
The results, reported in Table 6, suggest that foreign investment only has a positive e¤ect on house
prices for local authorities with a higher hous e price-earnings elasticity (i.e., a lower elasticity of
housing supply). The e¤ect of foreign investment is insigni…cant for local authorities in the bottom
two quartiles.
This analysis is related to some recent studies using US data which look at how the elasticity
6
More details on the constructi on of the elasti citie s c an be found in the append ix.
14
of housing supply a¤ects the propagation of shocks to house prices. These studies make use of
the elasticities of housing supply by metropolitan area constructed by Saiz (2010), which take into
account geographic and regulatory constraints to house building. Favara and Imbs (2015) analyse
whether branching deregulations across US s tates have a di¤erential e¤ect in counties with di¤erent
elasticities of housing supply. They …nd that the response of house prices is more muted in counties
where supply is more elastic, whereas the response of the housing stock is more muted in counties
where supply is less elastic. Adelino, Schoar and Severino (2012) look at a di¤erent sh ock to credit
supply — changes to the conforming loan limit, which determines the maximum size of a mortgage
that can be purchased or securitised by Fannie Mae or Freddie Mac. They …nd that cheaper credit
increases house prices by more in regions where housing supply is less elastic. Mian and Su… (2009)
compare house price growth between 1997 and 2007 in US metropolitan areas with high and low
supply elasticities. They …nd a strong increase in house price growth in inelastic areas until 2005,
followed by a large collapse in 2006 and beyond. By contrast, house price growth in elastic areas
remains low and ‡at during this period. All these …ndings are consistent with the results in Table
6.
4 Extensions and Robustness Checks
4.1 Alternative Samples
Many local authorities with large shares of foreign investment are located in London, as shown in
Figure 3 and Table 1. To test whether the e¤ect of foreign investment on house prices is di¤erent
in London and other parts of England and Wales, I estimate model (1) for the 32 London local
authorities. The results, reported in the …rst two columns of Table 7, suggest that foreign investment
has a positive e¤ect on house prices in London. However, the e¤ect is smaller than the average
e¤ect across all local authorities reported in Table 4. An increase of one percentage point in the
volume share of foreign investment increases house prices in London by 0:6%, compared with 2:1%
across all local authorities in England and Wales. For the value share, the e¤ect in London is 0:5%,
compared with 1:4% across all local authorities.
Another fact illustrated by Figure 3 is that many local authorities have a very small share of
foreign investment. To test how this a¤ects the results, I estimate model (1) for local authorities
15
with a volume share of foreign transactions above 0:5%. This leaves only 26 out of 172 local
authorities. The results, reported in the last two columns of Table 7, point again to a positive e¤ect
of foreign investment on house price growth. Comparing with the results for all local authorities in
Table 4, the IV coe¢ cients are somewhat smaller for local authorities with a high share of foreign
transactions.
A possible limitation with the de…nition of foreign transactions used in this paper is that the
Land Registry OCD includes all properties bought by overseas companies, but does not provide
information on the country of ultimate ownership. It is possible that some UK investors register
companies overseas to bene…t from lower taxes and use those companies to invest in property at
home. To assess how this may a¤ect the results, I recalculate the foreign shares excluding properties
registered to companies incorporated in Guernsey, Jersey and the Isle of Man. Because of their
geographic proximity, these are the most likely countries of incorporation of UK companies seeking
to reduce their tax liabilities. Table 8 reports the results of estimating model (1) f or this sample.
The coe¢ cients are similar to the ones reported in Table 4, suggesting that the results are not
being driven by transactions registered to companies in these territories.
4.2 Alternative Instrument
The IV results rep orted so far are based on an instrument that captures economic conditions
abroad, measured by the ec onomic risk index of the ICRG. To check robustness of the results to
the choice of instrument, I construct an alternative instrument based on the index of economic
policy uncertainty of Baker, Bloom and Davis (2016). The index is constructed by searching key
words in newspaper articles and is available for 15 countries (in addition to the UK). High values
of the index denote a higher degree of economic policy uncertainty. Using these data, I construct
the following instrument f or the volume and value shares of foreign investment in local authority i
in year t:
P
c
f
c
i
z
c
t1
where f
c
i
is the share of residents from foreign country c living in local authority i in the 2001
Census (as before) and z
c
t1
is the one-year lagged value of the index of economic policy uncertainty
16
for country c from Baker, Bloom and Davis (2016).
The IV estimates of the coe¢ cients in model (1) with this alternative instrument are reported
in Table 9. Compared with the results in Table 4, the coe¢ cients with this alternative instrument
are smaller. For the model with the full set of controls, the coe¢ cients imply that, on impact, house
prices increase by about 1:4% when the volume share of total transactions registered to overseas
companies increases by one percentage point. For the value share, the increase in house prices
is about 0:7%, which is half the size of the e¤ect reported in Table 4. The instrument is highly
signi…cant in the …rst stage and has the expected sign: an increase in economic policy uncertainty
abroad increases foreign investment in the UK housing market. The F-statistic on the excluded
instruments is somewhat lower for the volume share than the one reported in Table 4. However,
for the value share, the new instrument is stronger, with an F-statistic above the benchmark value
of 10 suggested by Stock, Wright and Yogo (2002). The Anderson-Rubin test con…rms that foreign
investment has a signi…cant e¤ect on house price growth.
5 Foreign buyers and other housing market outcomes
5.1 Housing stock
To test whether foreign investment encourages the construction of new housing, I estimate model (1)
with the change in the log stock of dwellings as the dependent variable (instead of the change in the
log of the house price index). Annual data on the dwelling stock are obtained from the Department
for Communities and Local Government (DCLG). Data are only available for local authorities in
England from 2001. The results of estimating the model by OLS and IV are reported in Table 10
and suggest there is no signi…cant e¤ect of foreign investment on the change in the housing stock.
It appears that foreign investment does not signi…cantly increase housing construction, resulting
instead in a signi…cant increase in prices. I have also estimated the model separately for local
authorities with a high and low house price-earnings elasticity (split at the median). I …nd that
this insigni…cant e¤ect is present in both groups of local authorities (the results are rep orted in
Table A2 in the Appendix).
17
5.2 Vacant Homes
One of the concerns expressed by commentators and policy makers when discussing the e¤ect of
foreign investment on the housing market is that homes bought by foreign investors are likely
to be left empty, with a negative e¤ect on local communities. This is sometimes described as
"buy-to-leave" (Financial Times (2016)).
Annual data on vacant dwellings are available from the DCLG for English local authorities from
2004 onwards. These data are obtained f rom council tax returns. Dwellings reported to the local
authority as vacant may be eligible for a council tax exemption or discount. Vacant dwellings are
classi…ed into long-term or short-term vacant, depending on whether they have been unoccupied or
substantially unfurnished for more or less than six months. To study the e¤ect of foreign investment
on vacant dwellings, I use thes e data to estimate model (1) with the change in the log number of
short-term and long-term vacant dwellings as the dependent variable. The results — reported
in Table A3 in the Appendix — point to an insigni…cant e¤ect of foreign investment on vacant
dwellings. The OLS coe¢ cients point to a positive correlation between the volume share of foreign
transactions and the number of short-term vacant dwellings, but this relation becomes insigni…cant
in the IV regress ions.
Council tax data on vacant dwellings have two potential limitations. First, local authorities have
discretion over the level of council tax discount th ey apply to vacant dwellings and m ay decide not
to apply any discount. In local authorities that do not award any discount, there is less incentive
for home owners to report their properties as vacant, which could lead to underreporting of vacant
homes. Another issue is that foreign investors may not be aware that they are eligible for a council
tax exemption or may view the exemption as too small to warrant reporting the home as vacant,
which may be particularly true at the high-end of the market.
To overcome these limitations, I use an alternative source of data on vacant homes. The 2011
Census reports the number of household spaces with no us ual residents. These are hous ehold
spaces that are vacant or used as second addresses, but may still be used by short-term residents
and visitors. Table 11 lists the 10 local authorities in England and Wales with the largest share of
household spaces with no usual residents. Many of these local authorities (for example, in Wales,
Cornwall and the Isle of Wight) are holiday destinations. However, the large share of homes with
18
no usual residents in the City of London, Westminster and Kensington and Chelsea is probably a
result of foreign investors buying in prime areas for long-term capital appreciation.
To examine the e¤ec t of foreign investment on vacant homes using this alternative data source,
I regress the share of household space s with no usual residents in each local authority on the share
of foreign transactions (in volume and in value):
vacant
i
=
F T
i
T T
i
+ "
it
(4)
The model is estimated by OLS and IV (using the alternative instrument based on econ omic
policy uncertainty abroad) for the cross-section of local authorities in England and Wales
7
. The
results are reported in the …rst two columns of Table 12. The OLS results sugge st a positive and
signi…cant correlation between foreign investment and vacant homes. However, the IV coe¢ cients
are insigni…cant. These results, together with the …ndings based on data from the DCLG, indicate
that there is no clear evidence that foreign investment in the housing market increases the number
of vacant homes.
5.3 Home Ownership
The results on the e¤ect of foreign investment on house prices suggest that an increase in foreign
investment leads to a signi…cant increase in prices at all points of the house price distribution. A
potential consequence of this is that residents may not b e able to a¤ord to buy a home in areas
where foreign investors are more active and may be forced to rent th eir homes instead.
To test this hypothesis, I collect data from the 2011 Census on the share of households in
owner-occupied accommodation for local authorities in England and Wales. I then estimate model
(4) with this share as the dependent variable. The results, reported in the last two columns of
Table 12, suggest that an increase in foreign investment in the hou sing market leads to a reduction
in the share of households who own their homes. The IV coe¢ cients imply that an increase of
one percentage point in the volume share of foreign transactions redu ce s the share of households
who own their homes by 5:6 percentage points. For the value share, the e¤ect is also negative and
signi…cant at 3 percentage points. There is evidence that residents are priced out of the market in
7
The usual instrument (based on t he economic risk index) produces weak …rst-s tage results for this c ross-sect ional
sample.
19
areas where foreign investors are more active.
6 Conclusions
This paper identi…es the causal e¤ect of foreign investment on the housing market in England
and Wales. It u ses a dataset recently released by the Land Registry on property transactions
registered to overseas companies. The paper uncovers a number of interesting results. Foreign
investment is found to have a positive e¤ect on house prices at di¤erent percentiles of the house
price distribution. This suggests that foreign investment in the housing market does not only drive
up prices of expensive homes, but has a "trickle down" e¤ect to less expensive properties. I also
highlight an important interaction between housing demand shocks and housing supply — increases
in foreign investment only appear to drive up prices in areas wh ere housing supply is particularly
constrained, either because there is less land available for construction or because of regulatory
constraints.
Looking beyond the impact on prices, I …nd no evidence that foreign investment has encouraged
construction of new housing. I also do not …nd evidence that more homes are left vacant in local
areas where foreign buyers are more active. Howe ver, I do …nd evidence that the rate of home
ownership declines as a result of foreign investment. These results should help inform the debate
on the impact of foreign investment on the housing market.
King’s College London, IZA, CEPR and Centre for Macroeconomics
20
References
[1] Adelino, M, Schoar, A., and Severino, F. (2012). ‘Credit Supply and House Prices: Evidence
from Mortgage Market Segmentation’, NBER Working Paper No. 17832.
[2] Aizenman, J. and Jinjarak, Y. (2009). ‘Current Account Patterns and National Real Estate
Markets’, Journal of Urban Economics, vol. 66, pp. 75–89.
[3] Angrist, J. D. and Pischke, J.-S. (2009). Mostly Harmless Econometrics: An Empiricist’s
Companion, Princeton: Princeton University Press.
[4] Badarinza. C. and Ramadorai, T. (2016). ‘Home Away from Home? Foreign Demand and
London House Prices’, Journal of Financial Economics, forthcoming, available at SSRN:
http://ssrn.com/abstract=2353124
[5] Baker, S. R., Bloom, N. and Davis, S. J. (2016). ‘Measuring Economic Policy Uncertainty’,
The Quarterly Journal of Economics, forthcoming.
[6] Barker, K. (2004). ‘Review of Housing Supply: Final Report — Recommendations’, London:
HMSO.
[7] Barker, K. (2006). ‘Barker Review of Land Use Planning: Final Report –Recommendations’,
London: HMSO.
[8] Bertrand, M., Du‡o, E. and Mullainathan, S. (2004). ‘How Much Should We Trust Di¤erence-
in-Di¤erences Estimates?’, Quarterly Journal of Economics, vol. 119(1), pp. 249-275.
[9] Case, K. E. and Shiller, R. J. (1989). ‘The E¢ ciency of the Market for Single-Family Homes’,
The American Economic Review, vol. 79(1), pp. 125-137.
[10] Cortes, P. (2008). ‘The E¤ect of Low-Skilled Immigration on U.S. Prices: Evidence from CPI
Data’, Journal of Political Economy, vol. 116(3), pp. 381-422.
[11] Dustmann, C., Frattini. T. and Preston, I. P. (2013). ‘The E¤ect of Immigration along the
Distribution of Wages’, Review of Economic Studies, vol. 80(1), pp. 145-173.
21
[12] Favara, G. and Imbs, J. (2015). ‘Credit Supply and the Price of Housing’, The American
Economic Review, vol. 105(3), pp. 958-992.
[13] Financial Times (2016). ‘The impact of ‘buy to leave’ on prime London’s housing market’,
February 12, 2016.
[14] Gauder, M., Houssard, C. and Orsmond, D. (2014). ‘Foreign Investment in Residential Real
Estate’, Reserve Bank of Australia Bulletin, June quarter.
[15] Hilber, C. A. L. and Vermeulen, W. (2015). ‘The Impact of Supply Constraints on House
Prices in England’, The Economic Journal, vol. 126(591), pp. 358–405.
[16] Jordà, O. (2005). ‘Estimation and Inference of Impulse Responses by Local Projections’, The
American Economic Review, vol. 95 (1), pp. 161–82.
[17] Jordà, O., Schularick, M. an d Taylor, A. M. (2015). ‘Betting the House’, Journal of Interna-
tional Economics, vol. 96(S1), pp. S1-S140.
[18] Mian, A. and Su…, A. (2009). ‘The Consequences of Mortgage Credit Expansion: Evidence
from the U.S. Mortgage Default Crisis’, The Quarterly Journal of Economics, vol. 124(4), pp.
1449-1496.
[19] Munshi, K. (2003). ‘Networks in the Mod ern Economy: Mexican Migrants in the US Labor
Market’, The Quarterly Journal of Economics, vol. 118(2), pp. 549–599.
[20] Sá, F., Towbin, P. and Wieladek, T. (2014). ‘Capital In‡ows, Financial Structure and Housing
Booms’, Journal of the European Economic Association, vol. 12(2), pp. 522–546.
[21] Sá, F. (2015). ‘Immigration and House Prices in the UK’, The Economic Journal, vol. 125, pp.
1393–1424.
[22] Saiz, A. (2010). ‘The Geographic Determinants of Housing Supply’, The Quarterly Journal of
Economics, vol.125(3), pp. 1253-1296.
[23] Savills (2014). ‘The World in London: Dynamics of a Global City — Overseas investors help
fund housing for the cosmopolitan city’, July 2014.
22
[24] Stock, J. H., Wright, J. and Yogo, M. (2002). ‘A Survey of Weak Instruments and Weak
Identi…cation in Generalised Method of Moments’, Journal of Business & Economic Statistics,
vol. 20, pp. 518-529.
[25] Stock, J. H., and Yogo, M. (2005). ‘Testing for Weak Instruments in Linear IV Regression’, in
Identi…cation and Inference for Econometric Models: Essays in Honor of Thomas Rothenberg,
D. W. K. Andrews and J. H. Stock, eds., Cambridge, UK: Cambridge University Press.
23
A Appendix. Construction of foreign shares
To construct the shares of foreign investment, I divide the number (or value) of residential properties
registered to overseas companies (from the Land Registry OCD) by the total number (or value)
of residential transfers to private individuals (from the Land Registry PPD) in a given year and
local authority. Because the Land Registry PPD only contains residential properties, I classify
transactions in the Land Registry OCD into residential and commercial and retain only transactions
of residential properties.
This classi…cation is done in three stages. First, I use certain key words in the address …eld
to classify properties. For example, if the address …eld contains the words "‡at" or "apartment",
the property is classi…ed as residential; if the address …eld contains the words "land", "garage",
"industrial estate", "store", "farm", etc., the property is classi…e d as non-residential. In a second
stage, I merge properties that remain unclassi…ed after the …rst stage with data from the Ordnance
Survey AddressBase dataset, which contains information on whether an address is commercial or
residential. The merge is done by property number or name, street and postcode. Finally, any
remaining unclassi…ed properties are searched manually in the Royal Mail Address Finder software
or on Google maps and classi…ed as residential or commercial. At the end of this process, 98; 271
records were classi…ed out of a total of 99; 345 transactions in the Land Registry OCD. About 50%
of the classi…ed records are residential properties.
B Appendix. Calculation of house price-earnings elasticities
The calculation of house price-earnings elasticities is based on equation (8) in Hilber and Vermeulen
(2015):
log(house price
j;t
) =
0
+
1
log(earnings
j;t
) +
2
log(earnings
j;t
) refusal rate
j
+
3
log(earnings
j;t
) %developed
j
+
4
log(earnings
j;t
)
elevation
j
+
34
P
i=1
4+i
D
t
+
352
P
i=1
38+i
D
j
+ "
j;t
where j denotes local planning authority (LPA) and t denotes year. The dependent variable
24
is the log of the mix-adjusted house price index and the main regressor is the log of male weekly
earnings. The model is estimated on 34 years of data (1974-2008). Regulatory constraints are
captured by the refusal rate of planning applications for major projects (consisting of 10 or more
dwellings), averaged over the period from 1979 to 2008. Land scarcity is captured by the share of
land already developed in 1990 and the elevation range. The model includes year …xed e¤ects (D
t
)
and LPA …xed e¤ects (D
j
). The refusal rate is instrumented with the share of votes for the Labour
Party in the 1983 general election and the change in the delay rate of major planning applications
following a reform introduced by the Labour government in 2002, which set a target to speed up
the planning process. The share of land already developed is instrumented with population density
in 1911.
The authors standardise the three measures of supply constraints by subtracting the mean and
dividing by the standard deviation. With this standardisation, the coe¢ cient
1
can be interpreted
as the house price-earnings elasticity for an LPA with average levels of the supply constraints.
I use the coe¢ cients and data in Hilber and Vermeulen (2015) to calculate the house-price
earnings elasticity in each LPA as follows:
\
elasticity
j
=
b
1
+
b
2
refusal rate
j
+
b
3
%developed
j
+
b
4
elevation
j
The local authorities in my sample do not exactly match the LPAs in Hilber and Vermeulen
(2015), because the Land Registry house price index is available at a lower level of disaggregation
than the mix-adjusted house price index used in their p aper. In cases where there is no match, I take
the average of the house price-earnings elasticities for all LPAs in a given local authority. Because
the measures of supply constraints are standardised, it is possible to obtain negative elasticities for
some local authorities, as shown in Figure 7.
25
Figure 1. Evolution of average house prices and share of foreign transactions
Source: Land Registry house price index, Land Registry Overseas Companies Dataset and Private Eye
offshore companies dataset (for values).
Figure 2. Average house prices in England and Wales, 2014
Source: Land Registry house price index.
0 .5 1
1.5 2
%
50000 100000
150000 200000 250000
£
1999
2002
2005
2008 2011
2014
year
average house prices (£)
foreign transactions - volume (%)
foreign transactions - value (%)
Figure 3. Average share of residential transactions in England and Wales registered to a foreign-owned
company, 2014
Volume
Value
Source: Land Registry Overseas Companies Dataset and Private Eye offshore companies dataset (for
values).
Table 1. Local authorities with the largest share of foreign transactions, 2014
Local authority
Foreign
transactions -
volume (%)
House prices (£)
Local authority
Foreign
transactions -
value (%)
House prices (£)
Westminster
13.1
922,702
Westminster
27.2
922,702
Kensington and Chelsea
12.1
1,288,406
Kensington and Chelsea
23.5
1,288,406
Salford
3.9
116,588
Greenwich
14.4
298,352
Camden
3.6
756,487
Salford
13.1
116,588
Liverpool
2.6
111,859
Bournemouth
7.6
198,537
Hammersmith
2.6
721,100
Manchester
6.4
130,355
Tower Hamlets
2.4
382,242
Camden
6.1
756,487
Lambeth
1.8
433,625
Bexley
5.7
244,459
Leeds
1.7
146,745
Kingston Upon Thames
5.2
406,106
Barnet
1.4
430,363
Leicester
4.7
129,463
Source: Land Registry house price index, Land Registry Overseas Companies Dataset and Private Eye offshore
companies dataset (for values).
Table 2. Countries of incorporation with the largest shares of investment, 2014
Country
Share of overseas investment in 2014
(in volume)
British Virgin Islands
33.5
Guernsey
19.4
Jersey
11.5
Isle of Man
10.1
Seychelles
2.9
Hong Kong
2.4
Luxembourg
2.1
Cyprus
1.7
Singapore
1.4
Panama
1.4
Source: Land Registry Overseas Companies Dataset.
Table 3. Descriptive statistics (1999 – 2014)
Variable
Observations
Mean
SD
Min
Max
Δ log house price index
2,580
0.069
0.091
-0.180
0.406
Δ total dwellings
1,963
0.007
0.010
-0.221
0.334
Share dwellings with no usual residents
(a)
173
0.045
0.024
0.019
0.207
Share households in owner-occupied
173
0.621
0.112
0.238
0.797
accommodation
(a)
Share foreign transactions - volume
2,752
0.003
0.011
0.000
0.214
Share foreign transactions - value
2,752
0.007
0.028
0.000
0.499
Unemployment rate
2,386
0.069
0.028
0.011
0.180
Benefits rate
2,752
0.149
0.049
0.044
0.331
House price - earnings elasticity
(b)
2,411
0.175
0.347
-0.492
1.151
(a) Cross-section from the 2011 Census.
(b) Author’s calculations, based on coefficients and data in Hilber and Vermeulen (2015).
Table 4. House prices and foreign transactions
Δ log house prices
(1)
(2)
(3)
(4)
Panel A. OLS
Share foreign transactions - volume
0.647***
0.659***
(0.123)
(0.132)
Share foreign transactions - value
0.161***
0.160**
(0.060)
(0.066)
Lagged Δ log house price
0.510***
0.446***
0.520***
0.460***
(0.011)
(0.017)
(0.011)
(0.017)
Lagged Δ unemployment rate
-0.085*
-0.101**
(0.048)
(0.049)
Lagged Δ benefits rate
-1.139***
-1.071***
(0.289)
(0.291)
R
2
within
0.858
0.862
0.857
0.860
Panel B. IV
Share foreign transactions - volume
2.166***
2.116***
(0.828)
(0.787)
Share foreign transactions - value
1.394***
1.427***
(0.535)
(0.537)
Lagged Δ log house price
0.467***
0.388***
0.456***
0.361***
(0.014)
(0.022)
(0.019)
(0.028)
Lagged Δ unemployment rate
-0.050
-0.099
(0.048)
(0.067)
Lagged Δ benefits rate
-1.493***
-1.795***
(0.294)
(0.343)
First-stage coefficients
-0.018***
-0.017***
-0.028***
-0.025***
(0.006)
(0.006)
(0.010)
(0.008)
Kleinbergen-Paap Wald rk F-statistic
7.797
8.065
8.364
8.695
Anderson-Rubin Wald test
49.826***
48.912***
49.826***
48.912***
Observations
2,408
2,005
2,408
2,005
Number of clusters
172
172
172
172
Notes: Panel A reports OLS coefficients from regressions of the log change in house prices on the share of
residential transactions registered to foreign-owned companies in volume (columns 1 and 2) and in value
(columns 3 and 4). Panel B reports coefficients of an IV specification in which the share of foreign transactions
is instrumented with a variable based on economic shocks abroad (see text for details on the construction of
the instrument). All variables are defined in appendix Table A1. The sample includes 172 local authorities in
England and Wales for the period 1999-2014. Regressions include local authority and year fixed effects and a
lagged dependent variable. Standard errors are clustered by local authority.
Stock-Yogo weak identification critical values: 16.38 (10%), 8.96 (15%), 6.66 (20%) and 5.53 (25%).
***Significant at the 1 percent level; ** significant at the 5 percent level; * significant at the 10 percent level.
Figure 4. Counterfactual analysis
Notes: The figure reports the evolution of average house prices for 172 local authorities in England and
Wales (solid line), the evolution of average house prices predicted by the model with the actual observed
value share of foreign transactions (dashed line) and the evolution of average house prices predicted by
the model when the value share of foreign transactions is set to zero in all local authorities (short dashed
line). The coefficients used in the predictions are from an IV regression of the log change in house prices
on the share of foreign transactions in volume. The share of foreign transactions is instrumented with a
variable based on economic shocks abroad (see text for details on the construction of the instrument).
The sample includes 172 local authorities in England and Wales for the period 1999-2014. The regression
includes local authority and year fixed effects and a lagged dependent variable.
50000 100000 150000 200000 250000
£
2000
2002
2004
2006
2008
2010
2012
2014
year
average house prices
predicted house prices
predicted house prices with no foreign buyers
Notes: The figure reports estimated coefficients and 90 percent confidence interval from local projection
instrumental variables (LP-IV) equations, which look at the effect of an increase in the share of foreign
investment (in volume and in value) on house price growth for four years after the shock. The share of
foreign transactions is instrumented with a variable based on economic shocks abroad (see text for
details on the construction of the instrument). The sample includes 172 local authorities in England and
Wales for the period 1999-2014. The regression includes local authority and year fixed effects, a lagged
dependent variable and lagged changes in the unemployment rate and in the benefits rate.
0 2 4 6 8
0 1 2
3 4
Years
Panel A. House price response to instrumented volume share of foreign transactions
0 2
4 6 8 10
0 1 2
3 4
Years
Panel B. House price response to instrumented value share of foreign transactions
(dashed lines are 90 percent confidence bands)
Figure 5. House prices: impulse responses to instrumented shock to share of foreign transactions
Table 5. House prices and foreign transactions – impact on different percentiles of the distribution of
house prices
Share foreign transactions
volume
value
Dependent variable – percentile of
distribution of house prices
(1)
(2)
(3)
(4)
Panel A. OLS
5
th
percentile
0.814***
0.784***
0.175**
0.185**
(0.198)
(0.163)
(0.078)
(0.084)
10
th
percentile
0.837***
0.834***
0.202**
0.210**
(0.192)
(0.184)
(0.094)
(0.104)
25
th
percentile
0.772***
0.808***
0.160*
0.167*
(0.164)
(0.189)
(0.082)
(0.093)
50
th
percentile
0.818***
0.864***
0.200***
0.211***
(0.165)
(0.164)
(0.064)
(0.073)
75
th
percentile
1.020***
1.064***
0.254***
0.264***
(0.147)
(0.155)
(0.078)
(0.088)
90
th
percentile
1.279***
1.270***
0.328***
0.317***
(0.196)
(0.195)
(0.108)
(0.117)
95
th
percentile
1.597***
1.556***
0.401***
0.378***
(0.223)
(0.199)
(0.128)
(0.137)
Panel B. IV
5
th
percentile
2.053**
2.139***
1.314***
1.423***
(0.817)
(0.822)
(0.491)
(0.508)
10
th
percentile
1.712***
1.852***
1.092***
1.229***
(0.662)
(0.688)
(0.400)
(0.429)
25
th
percentile
1.669***
1.840***
1.069***
1.230***
(0.606)
(0.639)
(0.388)
(0.432)
50
th
percentile
2.087***
2.186***
1.331***
1.444***
(0.776)
(0.803)
(0.492)
(0.527)
75
th
percentile
1.945***
1.867***
1.246***
1.239***
(0.687)
(0.638)
(0.432)
(0.420)
90
th
percentile
2.465***
2.156***
1.572***
1.422***
(0.850)
(0.733)
(0.524)
(0.467)
95
th
percentile
2.754***
2.316***
1.757***
1.527***
(0.865)
(0.695)
(0.529)
(0.439)
Controls
No
Yes
No
Yes
Observations
2,408
2,005
2,408
2,005
Number of clusters
172
172
172
172
Notes: Panel A reports OLS coefficients from regressions of the log change of different percentiles of the
distribution of house prices on the share of residential transactions registered to foreign-owned companies
in volume (columns 1 and 2) and in value (columns 3 and 4). Panel B reports coefficients of an IV
specification in which the share of foreign transactions is instrumented with a variable based on economic
shocks abroad (see text for details on the construction of the instrument). The sample includes 172 local
authorities in England and Wales for the period 1999-2014. Regressions include local authority and year
fixed effects and a lagged dependent variable. Controls include lagged changes in the unemployment rate
and in the benefits rate. Standard errors are clustered by local authority.
***Significant at the 1 percent level; ** significant at the 5 percent level; * significant at the 10 percent
level.
Notes: The figures report the estimated IV regression coefficients and 90 percent confidence interval
from a regression of the log change in different percentile of the distribution of house prices on the share
of foreign transactions in volume and in value. The share of foreign transactions is instrumented with a
variable based on economic shocks abroad (see text for details on the construction of the instrument).
The sample includes 172 local authorities in England and Wales for the period 1999-2014. Regressions
include local authority and year fixed effects, lagged changes in the unemployment rate and in the
benefits rate and a lagged dependent variable.
0 1 2 3 4
0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
Year of observation
Panel A. Impact of volume share of foreign transactions
.5 1 1.5 2 2.5
0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
Year of observation
Panel B. Impact of value share of foreign transactions
(dashed lines are 90 percent confidence bands)
Figure 6. Effect of foreign transactions across the distribution of house prices
Figure 7. House price-earnings elasticity
Source: Author’s calculations, based on coefficients and data in Hilber and Vermeulen (2015). See the
appendix for more details on the calculation of the elasticities. Local authorities in England only.
Table 6. House prices and foreign transactions – effect across different quartiles of the distribution of
house price-earnings elasticity
Δ log house prices
volume share
value share
(1)
(2)
(3)
(4)
Panel A. OLS
Quartile 1
-0.405
-0.166
-0.035
-0.025
(0.626)
(0.705)
(0.022)
(0.030)
Quartile 2
-0.171
-0.136
-0.017
-0.007
(0.162)
(0.135)
(0.032)
(0.031)
Quartile 3
1.033*
1.150**
0.380**
0.411**
(0.602)
(0.559)
(0.164)
(0.159)
Quartile 4
0.298***
0.299***
0.156***
0.155***
(0.078)
(0.098)
(0.034)
(0.038)
Panel B. IV
Quartile 1
49.926
118.252
0.718
0.642
(103.028)
(489.064)
(0.912)
(0.728)
Quartile 2
-5.942
-0.210
-0.500
-0.018
(11.022)
(8.408)
(0.913)
(0.737)
Quartile 3
5.940***
6.370***
4.073***
3.812***
(2.052)
(2.060)
(1.443)
(1.064)
Quartile 4
0.712**
0.652**
0.496**
0.483*
(0.325)
(0.330)
(0.251)
(0.263)
Controls
No
Yes
No
Yes
Notes: Panel A reports OLS coefficients from regressions of the log change in house prices on the share of
residential transactions registered to foreign-owned companies in volume (columns 1 and 2) and in value
(columns 3 and 4). Panel B reports second-stage coefficients of an IV specification in which the share of
foreign transactions is instrumented with a variable based on economic shocks abroad (see text for
details on the construction of the instrument). All variables are defined in appendix Table A1. Local
authorities are classified into four groups, corresponding to each quartile of the distribution of the house
price-earnings elasticity. Regressions are run separately for each of these four groups of local authorities.
The sample includes 150 local authorities in England for the period 1999-2014. Regressions include local
authority and year fixed effects and a lagged dependent variable. Controls include lagged changes in the
unemployment rate and in the benefits rate. Standard errors are clustered by local authority.
***Significant at the 1 percent level; ** significant at the 5 percent level; * significant at the 10 percent
level.
Table 7. House prices and foreign transactions – alternative samples
Δ log house prices
London LAs
High transaction LAs
(1)
(2)
(3)
(4)
Panel A. OLS
Share foreign transactions - volume
0.281***
0.387***
(0.087)
(0.122)
Share foreign transactions - value
0.160***
0.199***
(0.035)
(0.045)
Lagged Δ log house price
0.321***
0.314***
0.413***
0.408***
(0.075)
(0.072)
(0.036)
(0.034)
Lagged Δ unemployment rate
-0.026
-0.037
-0.060
-0.084
(0.065)
(0.064)
(0.141)
(0.144)
Lagged Δ benefits rate
-0.755
-0.801
-1.832**
-1.817**
(0.501)
(0.501)
(0.850)
(0.835)
R
2
within
0.896
0.897
0.781
0.781
Panel B. IV
Share foreign transactions - volume
0.644**
1.458**
(0.301)
(0.623)
Share foreign transactions - value
0.481*
1.057**
(0.250)
(0.473)
Lagged Δ log house price
0.264***
0.210***
0.328***
0.251***
(0.067)
(0.066)
(0.040)
(0.049)
Lagged Δ unemployment rate
-0.005
-0.027
0.033
-0.041
(0.058)
(0.063)
(0.141)
(0.173)
Lagged Δ benefits rate
-1.011**
-1.294**
-2.641***
-3.016***
(0.516)
(0.545)
(0.928)
(1.058)
First-stage coefficients
-0.032***
-0.043***
-0.027***
-0.037***
(0.011)
(0.015)
(0.009)
(0.013)
Kleinbergen-Paap Wald rk F-statistic
8.627
8.129
8.040
8.021
Anderson-Rubin Wald test
6.003**
6.003**
16.798***
16.798***
Observations
400
400
338
338
Number of clusters
32
32
26
26
Notes: Panel A reports OLS coefficients from regressions of the log change in house prices on the share of
residential transactions registered to foreign-owned companies in volume (columns 1 and 3) and in value
(columns 2 and 4). Panel B reports coefficients of an IV specification in which the share of foreign
transactions is instrumented with a variable based on economic shocks abroad (see text for details on the
construction of the instrument). All variables are defined in appendix Table A1. The sample covers the
period 1999-2014 and is restricted to 32 local authorities in London (columns 1 and 2) and to 26 local
authorities where the value share of foreign transactions in 2014 was higher than 0.5% (columns 3 and 4).
Regressions include local authority and year fixed effects and a lagged dependent variable. Standard
errors are clustered by local authority.
Stock-Yogo weak identification critical values: 16.38 (10%), 8.96 (15%), 6.66 (20%) and 5.53 (25%).
***Significant at the 1 percent level; ** significant at the 5 percent level; * significant at the 10 percent
level.
Table 8. House prices and foreign transactions – excluding transactions registered to companies
incorporated in Guernsey, Jersey and the Isle of Man
Δ log house prices
(1)
(2)
(3)
(4)
Panel A. OLS
Share foreign transactions - volume
0.808***
0.813***
(0.201)
(0.219)
Share foreign transactions - value
0.172*
0.164*
(0.090)
(0.093)
Lagged Δ log house price
0.511***
0.447***
0.522***
0.464***
(0.012)
(0.017)
(0.012)
(0.018)
Lagged Δ unemployment rate
-0.083*
-0.095*
(0.048)
(0.049)
Lagged Δ benefits rate
-1.131***
-1.055***
(0.288)
(0.293)
R
2
within
0.858
0.862
0.856
0.860
Panel B. IV
Share foreign transactions - volume
2.690***
2.591***
(1.000)
(0.929)
Share foreign transactions - value
1.860**
1.874**
(0.749)
(0.736)
Lagged Δ log house price
0.468***
0.391***
0.460***
0.370***
(0.015)
(0.021)
(0.021)
(0.029)
Lagged Δ unemployment rate
-0.045
-0.040
(0.048)
(0.072)
Lagged Δ benefits rate
-1.463***
-1.847***
(0.279)
(0.366)
First-stage coefficients
-0.014***
-0.014***
-0.021***
-0.019***
(0.005)
(0.005)
(0.007)
(0.007)
Kleinbergen-Paap Wald rk F-statistic
8.464
8.948
7.782
8.252
Anderson-Rubin Wald test
49.826***
48.912***
49.826***
48.912***
Observations
2,408
2,005
2,408
2,005
Number of clusters
172
172
172
172
Notes: Panel A reports OLS coefficients from regressions of the log change in house prices on the share of
residential transactions registered to foreign-owned companies in volume (columns 1 and 2) and in value
(columns 3 and 4), excluding transactions registered to companies incorporated in Guernsey, Jersey and the
Isle of Man. Panel B reports coefficients of an IV specification in which the share of foreign transactions is
instrumented with a variable based on economic shocks abroad (see text for details on the construction of the
instrument). All variables are defined in appendix Table A1. The sample includes 172 local authorities in
England and Wales for the period 1999-2014. Regressions include local authority and year fixed effects and a
lagged dependent variable. Standard errors are clustered by local authority.
Stock-Yogo weak identification critical values: 16.38 (10%), 8.96 (15%), 6.66 (20%) and 5.53 (25%).
***Significant at the 1 percent level; ** significant at the 5 percent level; * significant at the 10 percent level.
Table 9. House prices and foreign transactions – alternative instrument
Δ log house prices
(1)
(2)
(3)
(4)
IV – instrument based on economic policy uncertainty abroad
Share foreign transactions - volume
2.008**
1.391**
(0.859)
(0.651)
Share foreign transactions - value
0.973***
0.680**
(0.374)
(0.303)
Lagged Δ log house price
0.471***
0.417***
0.478***
0.419***
(0.013)
(0.025)
(0.015)
(0.025)
Lagged Δ unemployment rate
-0.068
-0.100*
(0.048)
(0.053)
Lagged Δ benefits rate
-1.317***
-1.368***
(0.321)
(0.338)
First-stage coefficients
0.002***
0.002***
0.003***
0.003***
(0.001)
(0.001)
(0.001)
(0.001)
Kleinbergen-Paap Wald rk F-statistic
6.771
6.774
10.208
11.085
Anderson-Rubin Wald test
35.250***
14.038***
35.250***
14.038***
Observations
2,408
2,005
2,408
2,005
Number of clusters
172
172
172
172
Notes: The table reports IV coefficients from regressions of the log change in house prices on the share of
residential transactions registered to foreign-owned companies in volume (columns 1 and 2) and in value
(columns 3 and 4). The share of foreign transactions is instrumented with a variable based on economic
policy uncertainty abroad (see text for details on the construction of the instrument). All variables are
defined in appendix Table A1. The sample includes 172 local authorities in England and Wales for the
period 1999-2014. Regressions include local authority and year fixed effects and a lagged dependent
variable. Standard errors are clustered by local authority.
Stock-Yogo weak identification critical values: 16.38 (10%), 8.96 (15%), 6.66 (20%) and 5.53 (25%).
***Significant at the 1 percent level; ** significant at the 5 percent level; * significant at the 10 percent
level.
Table 10. Housing stock and foreign transactions
Δ log dwelling stock
(1)
(2)
(3)
(4)
Panel A. OLS
Share foreign transactions - volume
-0.005
-0.006
(0.015)
(0.017)
Share foreign transactions - value
-0.009
-0.010
(0.008)
(0.008)
Lagged Δ log dwelling stock
-0.061***
-0.075***
-0.061***
-0.075***
(0.020)
(0.022)
(0.020)
(0.022)
Lagged Δ unemployment rate
-0.006
-0.006
(0.009)
(0.009)
Lagged Δ benefits rate
-0.044
-0.043
(0.126)
(0.127)
R
2
within
0.025
0.027
0.026
0.028
Panel B. IV
Share foreign transactions - volume
-0.072
-0.071
(0.107)
(0.107)
Share foreign transactions - value
-0.049
-0.049
(0.071)
(0.074)
Lagged Δ log dwelling stock
-0.061***
-0.075***
-0.062***
-0.076***
(0.021)
(0.023)
(0.021)
(0.023)
Lagged Δ unemployment rate
-0.008
-0.005
(0.010)
(0.009)
Lagged Δ benefits rate
-0.045
-0.040
(0.126)
(0.123)
First-stage coefficients
-0.017***
-0.016***
-0.025***
-0.023***
(0.006)
(0.006)
(0.009)
(0.009)
Kleinbergen-Paap Wald rk F-statistic
7.616
7.960
7.832
7.535
Anderson-Rubin Wald test
0.516
0.478
0.516
0.478
Observations
1,800
1,613
1,800
1,613
Number of clusters
150
150
150
150
Notes: Panel A reports OLS coefficients from regressions of the log change in the stock of dwellings on the
share of residential transactions registered to foreign-owned companies in volume (columns 1 and 2) and
in value (columns 3 and 4). Panel B reports coefficients of an IV specification in which the share of foreign
transactions is instrumented with a variable based on economic shocks abroad (see text for details on the
construction of the instrument). All variables are defined in appendix Table A1. The sample includes 150
local authorities in England for the period 2002-2014. Regressions include local authority and year fixed
effects and a lagged dependent variable. Standard errors are clustered by local authority.
Stock-Yogo weak identification critical values: 16.38 (10%), 8.96 (15%), 6.66 (20%) and 5.53 (25%).
***Significant at the 1 percent level; ** significant at the 5 percent level; * significant at the 10 percent
level.
Table 11. Local authorities with the largest share of vacant homes, 2011
Local authority
Vacant homes (%)
City of London
20.7
Gwynedd (Wales)
14.1
Pembrokeshire (Wales)
12.5
Westminster
11.9
Cornwall
11.3
Anglesey (Wales)
10.5
Kensington and Chelsea
10.5
Isle of Wight
9.7
Ceredigion (Wales)
9.0
Conwy (Wales)
8.9
Notes: The share of vacant homes measures the percentage of household spaces
with no usual residents (from the 2011 Census).
Table 12. Effect on share of household spaces with no usual residents and share of households in owner-
occupied accommodation
Share of household spaces
with no usual residents
Share of households in owner-
occupied accommodation
(1)
(2)
(3)
(4)
Panel A. OLS
Share foreign transactions - volume
0.245**
-2.511***
(0.107)
(0.506)
Share foreign transactions - value
0.184***
-1.069***
(0.036)
(0.264)
R
2
within
0.074
0.180
0.254
0.200
Panel B. IV
Share foreign transactions - volume
-0.124
-5.592***
(0.248)
(1.834)
Share foreign transactions - value
-0.065
-2.957***
(0.135)
(1.120)
First-stage coefficients
0.003***
0.006**
0.003***
0.006**
(0.001)
(0.002)
(0.001)
(0.002)
Kleinbergen-Paap Wald rk F-statistic
8.591
6.514
8.591
6.514
Anderson-Rubin Wald test
0.340
0.340
94.259***
94.259***
Observations
172
172
172
172
Number of clusters
172
172
172
172
Notes: Panel A reports OLS coefficients from regressions of the share of household spaces with no usual
residents and the share of households in owner-occupied accommodation on the share of residential
transactions registered to foreign-owned companies in volume (columns 1 and 3) and in value (columns 2
and 4). Panel B reports coefficients of an IV specification in which the share of foreign transactions is
instrumented with a variable based on economic policy uncertainty abroad (see text for details on the
construction of the instrument). All variables are defined in appendix Table A1. The sample includes 172
local authorities in England and Wales in 2011 (Census data). Standard errors are clustered by local
authority.
Stock-Yogo weak identification critical values: 16.38 (10%), 8.96 (15%), 6.66 (20%) and 5.53 (25%).
***Significant at the 1 percent level; ** significant at the 5 percent level; * significant at the 10 percent
level.
Table A1. Data sources
Variable
House price index
Land Registry house price index
Average house prices
Land Registry house price index
Percentiles of the distribution of house prices
Land Registry Price Paid Data (PPD)
Dwelling stock
Department for Communities and Local Government
(DCLG) Housing statistics Table 125, from 2001, local
authorities in England only
Number of vacant dwellings
DCLG Housing statistics Table 615, from 2004, local
authorities in England only
Share of household spaces with no usual
2011 Census, from ONS Neighbourhood Statistics
residents
Share of households in owner-occupied
2011 Census, from ONS Neighbourhood Statistics
accommodation
Share of foreign transactions - volume
Number of residential transactions registered to foreign-
owned companies divided by total number of residential
transactions
Number of foreign transactions: Land Registry Overseas
Companies Dataset (OCD)
Total number of transactions: Land Registry PPD
Share of foreign transactions - value
Value of residential transactions registered to foreign-
owned companies divided by total value of residential
transactions
Value of foreign transactions: Land Registry OCD
and Private Eye offshore companies dataset
Total value of transactions: Land Registry PPD
Unemployment rate
Labour Force Survey (until 2003) and Annual Population
Survey (from 2004), from Nomis.
Benefits rate
Proportion of working age population receiving any state
benefits, from Nomis
House price - earnings elasticity
Hilber and Vermeulen (2015), local authorities in
England only
Economic risk index
International Country Risk Guide (ICRG). The index
includes: inflation, GDP per head, GDP growth, budget
balance and current account as % of GDP. High values of
the index denote lower risk.
Economic policy uncertainty index
Baker, Bloom and Davis (2016). The index is based on
frequency counts of some key words in newspaper
articles. It is available for: Australia, Brazil, Canada,
China, France, Germany, India, Ireland, Italy, Japan,
Korea, Netherlands, Russia, Spain and the US. High
values of the index denote a higher degree of economic
policy uncertainty.
Table A2. Housing stock and foreign transactions – effect by house price-earnings elasticity
Δ log dwelling stock
volume share
value share
(1)
(2)
(3)
(4)
Panel A. OLS
Low elasticity
-0.021
-0.030
-0.018
-0.018
(0.122)
(0.118)
(0.019)
(0.021)
High elasticity
-0.003
-0.004
-0.002
-0.003
(0.013)
(0.014)
(0.005)
(0.005)
Panel B. IV
Low elasticity
-15.008
-23.399
-0.462
-0.506
(14.505)
(27.800)
(0.388)
(0.414)
High elasticity
0.000
0.011
0.000
0.008
(0.032)
(0.033)
(0.023)
(0.026)
Controls
No
Yes
No
Yes
Notes: Panel A reports OLS coefficients from regressions of the log change in the stock of dwellings on the
share of residential transactions registered to foreign-owned companies in volume (columns 1 and 2) and
in value (columns 3 and 4). Panel B reports second-stage coefficients of an IV specification in which the
share of foreign transactions is instrumented with a variable based on economic shocks abroad (see text
for details on the construction of the instrument). All variables are defined in appendix Table A1. Local
authorities are split by the median value of the house price-earnings elasticity. Regressions are run
separately for each of these two groups of local authorities. The sample includes 150 local authorities in
England for the period 2002-2014. Regressions include local authority and year fixed effects and a lagged
dependent variable. Controls include lagged changes in the unemployment rate and in the benefits rate.
Standard errors are clustered by local authority.
***Significant at the 1 percent level; ** significant at the 5 percent level; * significant at the 10 percent
level.
Table A3. Vacant dwellings and foreign transactions
Δ log short-term vacant
dwellings
Δ log long-term vacant
dwellings
(1)
(2)
(3)
(4)
Panel A. OLS
Share foreign transactions - volume
1.093**
1.076**
-0.492
-0.455
(0.427)
(0.432)
(0.520)
(0.527)
Share foreign transactions - value
0.150
0.156
-0.154
-0.158
(0.169)
(0.168)
(0.269)
(0.266)
Panel B. IV
Share foreign transactions - volume
-0.431
-0.340
-1.141
-1.001
(2.105)
(2.078)
(2.242)
(2.230)
Share foreign transactions – value
-0.340
-0.271
-0.884
-0.785
(1.691)
(1.679)
(1.698)
(1.717)
Controls
No
Yes
No
Yes
Observations
1,348
1,348
1,346
1,346
Number of clusters
150
150
150
150
Notes: Panel A reports OLS coefficients from regressions of the log change in the stock of short-term and
long-term vacant dwellings on the share of residential transactions registered to foreign-owned
companies in volume (columns 1 and 2) and in value (columns 3 and 4). Panel B reports second-stage
coefficients of an IV specification in which the share of foreign transactions is instrumented with a
variable based on economic shocks abroad (see text for details on the construction of the instrument). All
variables are defined in appendix Table A1. The sample includes 150 local authorities in England for the
period 2005-2014. Regressions include local authority and year fixed effects and a lagged dependent
variable. Controls include lagged changes in the unemployment rate and in the benefits rate. Standard
errors are clustered by local authority.
***Significant at the 1 percent level; ** significant at the 5 percent level; * significant at the 10 percent
level.
