Do Foreign Buyers Pay More Than Domestic Buyers? Evidence from International Transaction-Level Data

Abstract

In this paper, we examine the role of international capital flows in real estate prices by quantifying the extent foreign buyers overpay for their realty investments as well as the spillover effect of such behavior on property prices domestic buyers pay. Using a unique dataset accounting for about 30,000 realty investment transactions in Australia, Canada, France, Hong Kong, Japan, the Netherlands, the United Kingdom, and the United States, we find the following. First, foreign investors pay significantly higher prices than domestic investors, even after taking a wide variety of controls into account. Second, this paying over the odds becomes smaller the larger the buyers’ exposure to realty investments in the host countries. These results indicate that foreign investors are overcharged when they are less informed about the property market and that the extent to which they are overcharged decreases the more investment experience they have. Third, we did not find any significant spillover effects from overpaying by foreign investors to real estate prices in host countries. This finding is consistent with a group of extant studies employing aggregate-level data to examine the link between international capital flows and real estate prices.

Introduction

Given that real estate is one of the major investment objects in international financial market, a large body of literature has attempted to examine the impact of international capital flows on realty prices. The interaction between international capital flows and real estate markets has become even more relevant in the age of a “global saving glut,” in which a large influx of capital from emerging countries has lowered long-term interest rates and contributed to a run-up in asset prices (Bernanke, 2005). In recent years, many pieces of anecdotal evidence suggest that foreign investors are the central cause of local property booms.¹ Also, a considerable number of studies have examined the argument that global imbalances in financial flows have resulted in massive fluctuations in asset prices, above all in real estate prices. On the one hand, Jordà et al. (2014) argue that a change in monetary policy in one country can potentially generate large fluctuations in realty prices in other countries through changes in international capital flows. On the other hand, Favilukis et al. (2013) have expressed doubt that changes in international capital flows lead to large fluctuations in realty prices.

All of these pieces of evidence point to the importance of geography when we examine the real estate market. As demonstrated by Coval and Moskowitz (2001), who focus on the distance between fund managers and portfolio companies in the stock market, geography matters in the asset market in general since the physical distance between investors and assets is one of the major determinants of asset returns. It matters even more when investors and assets are geographically separated by national borders.

However, there has been a paucity of studies that employ disaggregated data and examine the role of geographical distances between investors and real estate properties, especially when they are separated by national borders.² This is presumably due to a lack of international transaction-level data on realty investment, which are needed for a precise analysis of international financial flows and their impact on property prices. As a result, it is not sufficiently clear how foreign investors pay prices for their property purchase and how these investors interact with domestic buyers given that there is an increase in demand for real estate properties in a host country.

Against this background, the present study employs a unique transaction-level dataset on real estate properties and seeks to investigate how foreign investors play a role in the real estate market. It does so by focusing on the case when investors and properties are separated by national borders. Specifically, in this study, we conduct the following three analyses. First, we identify the country where each of the investors in the market is located and examine how realty prices paid for by foreign investors differ from those paid by domestic investors. The estimations control for a comprehensive list of property characteristics (such as location, type, size, and age) and transaction characteristics (such as the geographical location of the buyer and the seller, and the type of buyer and seller), so that we can exclude potential confounding factors affecting the price paid by investors.

Second, we consider the possibility that the informational disadvantage associated with investing over a longer distance and across borders diminishes as international investors learn about local real estate markets through so-called “learning by investing” (see, e.g., Sorensen, 2008; Gompers et al., 2008). Such learning by investing may be particularly important in the case of real estate investment, since the extent of information asymmetry between buyers and sellers is substantial.

Third, after examining whether foreign investors indeed pay over the odds and, finding that this is the case, we investigate whether this has any effect on property prices in the neighborhood. For the analysis, we employ information on the timing and geographical location of each property sale to test if the prices paid by domestic investors for properties in the vicinity of a property purchased by a foreign investor are significantly higher after the purchase by the foreign investor. Regarding prices paid by local investors before the purchase by the foreign investor as control observations and prices paid by local investors after the purchase by the foreign investor as treatment observations, we compare the two sets of prices to explicitly test whether any causal impact from foreign property purchases on local real estate market can be observed.

The novelty of this study is at least threefold. First, this study is the first to explicitly examine differences in the behavior of foreign and domestic investors in the realty market. Second, this study is also the first to focus on international investors’ real estate experience and examine how this experience alleviates the information asymmetry over time. Third, this study is the first to explicitly study the causal effect of foreign real estate investment on local markets. Our dataset providing information on both the precise geographical location and timing of each transaction allows us to address this identification challenge.

The main findings of our analysis can be summarized as follows. First, we find that foreign investors pay substantially higher prices than domestic investors, even after taking a wide variety of controls into account. Second, this price difference becomes smaller the greater buyers’ investment exposure in a host country. These results are fairly robust even when we limit the sample to repeated sales observations allowing us to control for unobserved property characteristics such as quality. Taken together, these results indicate that foreign investors tend to pay inflated prices when they are less informed about the local realty market. Third, however, we did not find any significant spillover effects from foreign investors paying over the odds on local real estate prices. This finding is consistent with a group of extant studies employing aggregate data suggesting that the impact of international capital flows on real estate markets is limited (e.g., Favilukis et al., 2013), while it is not in line with a strand of studies that emphasize the role of capital flows (e.g. Aizenman & Jinjarak, 2009).

The remainder of this study is organized as follows. In Sect. "Related Literature and Theoretical Underpinnings", we briefly survey the related literature that provides the theoretical underpinnings of our empirical analysis. Next, in Sect. "Data and Empirical Approach", we explain the data and our empirical framework. In Sect. "Empirical Analysis", we then present and discuss the empirical results on realty prices paid by foreign and domestic investors and on spillover effects. Sect. "Conclusion" concludes and considers questions for future research.

Related Literature and Theoretical Underpinnings

In this section, we first provide a brief survey of studies on the impact of international capital flows on local realty prices. We then survey the literature examining how the information asymmetry resulting from the geographical distribution of investors affects realty prices of various types. Finally, we present the related literature on price spillovers in the real estate market.

A considerable number of studies quantitatively examine the determinants of real estate prices, including the role of international capital flows. For instance, using aggregate-level data on 43 countries from 1978 to 2008, Aizenman and Jinjarak (2009) show that there is a positive association between countries’ current account deficits, which reflect capital inflows, and increases in real estate prices. Similarly, Justiniano et al. (2014), using a quantitative equilibrium model, suggest that international capital flows accounted for a sizable portion of the increase in U.S. house prices before the 2008 financial crisis. On the other hand, Favilukis et al. (2013), also focusing on U.S. house prices, argue that international capital flows at most played a minor role. Ferrero (2015) argues that domestic factors such as lower collateral requirements rather than foreign factors including current account deficits facilitate access to external funds and drive up housing prices.

These studies have yet to reach a decisive conclusion regarding the extent of the impact of a country’s current account deficit and resulting capital inflows on the local realty market. Also, a country’s current account deficit means that capital can flow into the country in a variety of forms, such as the purchase of riskless sovereign bonds, securities issued by private companies, and real estate properties. Therefore, simply establishing causality between capital inflows and realty prices does not necessarily indicate that foreign buyers purchase a substantial amount of local real estate properties in the market. This issue needs to be addressed using more disaggregated data.

As for studies examining the impact of information asymmetry caused by the geographical distribution of buyers and properties in the realty market, these mostly focus on buyer-property distances within a country.³ Based on the theoretical discussion in Kurlat (2016), Kurlat and Stroebel (2015), for example, use data on realty transactions for Los Angeles County in the United States and analyze the determinants of changes in realty prices. They find that the physical characteristics of both the property itself and nearby properties as well as information advantage insiders (i.e. residents in the area) have over outsiders about these characteristics determine realty prices. Specifically, they find that price increases are smaller the larger the share of informed sellers and the less informed buyers are. Based on this finding, they conclude that information asymmetry is an important determinant of realty prices. In a similar vein, using realty transaction data for the United States, Garmaise and Moskowitz (2004) find that the geographical distances between buyers and properties are shorter the greater the information asymmetry faced by buyers. They further find that the median distance between buyers and properties is short (i.e., 47 km) and the distance between buyers and properties is shorter the greater the dispersion of realty prices, although the latter finding is less pronounced the older properties are. In sum, they show that information asymmetry associated with the geographical distance between buyers and properties is an important determinant of changes in property prices.

We will employ this strand of papers, which focus on the information advantage insiders have over outsiders, as a benchmark for our theoretical underpinnings. In the framework, we expect that informed buyers are less likely to suffer from information asymmetry between investors and sellers than are uninformed buyers.⁴ One important difference in our approach from the previous studies such as Kurlat and Stroebel (2015) is that we examine the impact of distance-induced information asymmetry by comparing foreign buyers and domestic buyers, while they do so by comparing neighboring and distant buyers located in the same country.

Note that there exist several types of mechanisms which reduce the extent of information asymmetry in the market. One of them emphasizes the role of experience. Gompers et al. (2008) focus on the behavior of venture capital investors and find that it is experienced rather than inexperienced investors that respond quickly to positive information in the market. Similarly, Sorensen (2008), also focusing on venture capitalists, posits a dynamic model of “learning by investing” and arrives at empirical findings consistent with his theory. We will employ this strand of papers, which focus on the role of experience to reduce the extent of information asymmetry, as another benchmark for our theoretical underpinnings. We expect that uninformed foreign buyers are less likely to suffer from information asymmetry as they accumulate investment experience.

One of the few studies examining the role of distance when buyers and properties are separated by borders is that by Badarinza and Ramadorai (2018) and Badarinza et al. (2021), who focus on the specific circumstances when shocks in foreign countries are transmitted to the local realty market. Employing realty transaction information for London, they show that an exogenous shock in a home country (i.e., outside of UK) affects the realty prices in spatially limited areas in London where many people born in the home country reside. This result, which they label the “safe haven effect,” indicates that certain types of shocks in foreign countries can be transmitted to the local realty market. Based on this study, we directly identify transactions made by foreign investors and contrast them with transactions by domestic investors using data on multiple host countries rather than one.

However, none of the above studies employing transaction level data and examining the impact of distance-induced information asymmetry on realty prices analyze the spillovers of price shocks on other properties or the comovement of prices across real estate properties. An examination of such spillovers or the comovement of prices is important for evaluating the aggregate impact of investments by outside buyers, especially those from foreign countries. While there are no such study that directly address this issue (to the best knowledge of the authors), Autor et al. (2014) focus on the abolishment of rent control in Cambridge, Boston, and examine spillovers of realty prices from previously-rent controlled properties on properties that were not subject to rent controls.

Given the abovementioned studies, we will focus on the behavior of realty investors that are located in foreign countries in order to disentangle the mechanism of an influx of foreign capital affecting domestic real estate market. These foreign buyers are faced with information asymmetry caused by geographical distance from local real estate properties, but their investment experience in the local market may substantially reduce such asymmetry. Further, we will examine the existence of spillover effects of such foreign realty investment on the prices paid in other realty transactions in the neighborhood. Combining the results of these analyses, we are able to evaluate the role of foreign investors in the realty market during the period that includes the years of a “global saving glut.”

Data and Empirical Approach

Data Overview

The data used for this study are transaction-level data for the period from 2005 to 2015. We obtain the data from Real Capital Analytics Inc. (RCA), one of the largest data vendors specializing in real estate investments. The data provided by RCA reflect institutional investment activities and cover relatively large investment transactions, involving real estate properties worth in excess of about US$ 1 million. The original data cover 71,000 realty transactions in Australia, Canada, France, Hong Kong, Japan, the Netherlands, the United Kingdom, and the United States. Although the data cover properties in a large number of cities (namely, 1,223 cities), a large part of the observations are concentrated in the major cities of eight countries, namely, Amsterdam (Netherlands); Chicago, Los Angeles, New York, and San Francisco (United States); Tokyo and Osaka (Japan); Paris (France); London (United Kingdom); Sydney (Australia); Toronto and Vancouver (Canada); and Hong Kong (Hong Kong). This means that the data we use are mainly for large investments in major cities.

The data contain various types of information regarding the investment transactions. Information included concerns the property involved in the transaction, namely, the price in US$, the floor area of the building in square feet, and the land area in acres. For the analysis, we use the natural logarithm of these variables (LN_PriceUSD, LN_Floor, and LN_Land). The data also contain the age of each structure (Age) as well as its type. Properties are classified into the following types: apartment, development site, hotel, industrial, office, other, retail, and seniors housing and care facility. Given this classification, we use eight dummy variables (Property type) to represent the property type, with seniors housing and care facilities serving as the reference group.

In addition, the dataset contains a wide range of transaction-related information. This includes the country in which the property is located (Property country), the country in which the buyer is located (Buyer country), and the country in which the seller is located (Seller country). In our empirical analysis, we control for these characteristics by including eight dummy variables for Property country and up to 102 dummy variables for Buyer country and Seller country.⁵

The data further contain information on the type of buyer investors and the type of seller investors of a property. Buyer and seller investors are classified into various categories (see Table 1d and e) and we construct dummy variables to represent them. The Buyer/Seller investor type variables capture the detailed characteristics of investment funds (e.g., corporate, developer/owner/operator, investment manager, or REIT). We construct dummy variables for these investor types in order to represent differences in the funding environment they face.

Table 1 Tabulation of transaction-level data

Table 1 provides an overview of our data, which consists of observations for 28,525 transactions. Note that the number of observations reduces from the original 71,000 to fewer than 30,000 due to the lack of information on some variables. Panel (a) shows the distribution of properties by type. As can be seen, apartments make up the largest share of properties in the database, followed by offices, industrial sites, and retail properties. Panel (b) shows the distribution by year of transaction. The figures indicate that there was a substantial reduction around 2008 and 2009 due to the global financial crisis. Panel (c) shows the distribution by country where properties are located. The United States makes up the largest shares, followed by Japan and Australia. The numbers for several countries – such as France and the UK – are much smaller than one would expect given the size of their countries. This is presumably due to missing information for several variables used in the analysis. Panels (d) and (e) show the distributions by buyer and seller investor type. In both panels, Developer/Owner/Operator make up the largest share, followed by REITs, equity funds, corporates, and investment managers.

We use Property country and Buyer country to construct a dummy variable, ForeignBuyer, that equals one if the two countries differ and zero otherwise. We hypothesize that information asymmetry is likely to be greater in the case of ForeignBuyer = 1, which, in turn, is likely to lead to higher transaction prices (or possibly lower returns) than in the case of ForeignBuyer = 0 (i.e., domestic buyer). Further, in order to take buyers’ investment experience in a particular country into account, we calculate the cumulative amount of investment of all buyers located in country A in real estate in country B. Note here that we also calculate the cumulative amount of domestic investment, in which case we regard countries A and B the same. This pairwise variable is measured for the period up to month t-1. Since this variable monotonically increases over the observation period, following Gompers et al. (2008), we standardize it to construct a variable, CUMINV, by dividing it by the cumulative total sum of investment of all buyers from country A in all host countries, including country A itself, measured up to month t-1.

There are three notes on this investment experience variable. First, buyers’ investment can be directed to their home country, in which case the variable measures the extent of domestic investment experience in the buyer’s own country. Second, although we could compute this variable for each buyer, we choose to construct the variable at the country level. This choice reflects our assumption based on Badarinza and Ramadorai’s (2018) finding that there is information sharing among buyers in one country. Third, we interpret this ratio variable as representing a probability that a country’s prospective buyer considering to invest in a host country may encounter experienced buyers who have made property investment in the host country before.⁶

Table 2 provides summary statistics for our variables. We have excluded several observations with extreme values for LN_PriceUSD, LN_Floor, LN_Land, and Age to have the sample size of 28,525 transactions.

Table 2 Summary statistics

Empirical Framework

Using our transaction-level data, we examine how buyers’ characteristics (in particular ForeignBuyer, CUMINV, and the interaction term of these two variables) as well as other transaction-specific factors affect the transaction price employing the following linear regression model:

$$\begin{array}{c}{LN\_PriceUSD}_{i,p,b,s,t}=\alpha +{\beta }_{1}{ForeignBuyer}_{i,p,b}+{\beta }_{2}{CUMINV}_{p,b,t}\\ +{\beta }_{3}{ForeignBuyer}_{i,p,b}\times {CUMINV}_{p,b,t}+{{\varvec{X}}}_{it}{\varvec{\gamma}}+{\eta }_{p}^{1}+{\eta }_{b}^{2}+{\eta }_{s}^{3}+{\eta }_{t}^{4}+{\varepsilon }_{i,t}\end{array}$$

(1)

The variable on the left-hand side of the equation is the natural logarithm of the transaction price of property i in country p sold by a seller in country s to a buyer in country b at time t (measured monthly). On the right-hand side of the equation, ${{\varvec{X}}}_{it}$ represents property characteristics such as the size, age (time-variant) and type of the property, which are important determinants of the price of a property.⁷${ForeignBuyer}_{i,p,b}$ is a dummy variable that equals one if country p and country b are different. ${CUMINV}_{p,b,t}$ is the standardized cumulative investment amount from country b in country p for the period up to the month preceding t.

We include the interaction term ${ForeignBuyer}_{i,p,b}\times {CUMINV}_{p,b,t}$ to test for the possibility that the impact associated with ${ForeignBuyer}_{i,p,b}$ varies with ${CUMINV}_{p,b,t}$. The four variables $\left\{{\eta }_{p}^{1}, {\eta }_{b}^{2},{\eta }_{s}^{3},{\eta }_{t}^{4}\right\}$ respectively represent country fixed-effects for the country in which the property is located, country fixed-effects for the country in which the buyer is located, country fixed-effects for the country in which the seller located, and time fixed effects. We also include other fixed-effects for each of the property types, buyer investor types, and seller investor types.

As another main specification, we also estimate the following equation:

$$\begin{array}{c}{LN\_PriceUSD}_{i,p,b,s,t}=\alpha +{\beta }_{1}{ForeignBuyer}_{i,p,b}\\ +{\beta }_{3}{ForeignBuyer}_{i,p,b}\times {CUMINV}_{p,b,t}+{{\varvec{X}}}_{it}{\varvec{\gamma}}+{\eta }_{t,p}^{1}+{\eta }_{t,b}^{2}+{\eta }_{t,s}^{3}+{\varepsilon }_{i,t}\end{array}$$

(2)

In this specification, the time-invariant country effects are replaced with time-variant country effects, i.e., we replace $\left\{{\eta }_{p}^{1}, {\eta }_{b}^{2},{\eta }_{s}^{3}\right\}$ with ${\{\eta }_{t,p}^{1},{\eta }_{t,b}^{2},{\eta }_{t,s}^{3}$}. We also include other time-variant fixed-effects for each of the property types, buyer investor types, and seller investor types. In this model, some variables reflect changes in macroeconomic conditions in host/buyer/seller countries including their output, employment, and exchange rates, while other variables such as buyer/seller investor types show how different types of investors react to macroeconomic shocks during the observation period. Note that we exclude the variable ${CUMINV}_{p,b,t}$, which is also time-variant, from the equation in order to avoid possible collinearity with other variables.

While we include a fair number of property characteristics that likely affect transaction prices, it is still possible that there may be omitted variables. If, for example, we have failed to include an important property characteristic that affects ${LN\_PriceUSD}_{i,p,b,s,t}$ and is correlated with ${ForeignBuyer}_{i,p,b}$, the coefficient ${\beta }_{1}$ will suffer from endogeneity bias. One characteristic that might affect property prices and that we have not controlled for is detailed information on the location of a property (such as the street). To examine this possibility, the six panels in Fig. 1 depict the location of properties bought by foreign investors (marked by a star) and domestic investors (marked by a dot) in Los Angeles, Paris, Toronto, London, Tokyo, and Sydney as illustrative examples. As can be seen, there does not appear to be a substantial systemic difference in terms of the areas where foreign and domestic investors bought properties. We therefore do not include further location-related variables in the baseline estimations. However, we will implement several additional analyses that control for the potential heterogeneity of property locations in a more appropriate manner than in the baseline. We employ information on the latitude and longitude of each transacted property in order to calculate the geographical distances between properties purchased by domestic buyers and those purchased by foreign buyers. The first way in which we try to control for the potential heterogeneity of property locations is to match each property purchased by domestic buyers with a property purchased by foreign buyers and located closest to the property purchased by domestic buyers. We limit the sample to properties purchased by domestic buyers whose matched properties are located within a radius of certain distances. The second way in which we try to control for the potential heterogeneity of property locations is to employ the repeat-sales approach. We focus on properties that have been purchased by both domestic and foreign buyers at different times and compare the difference in transacted prices.⁸

Fig. 1

Location of properties (purchased by foreign and domestic investors. Note: The figures show the locations of properties bought by foreign investors (stars) and domestic investors (dots). (a) Los Angeles and Paris. (b) Toronto and London. (c)Tokyo and Sydney

Empirical Analysis

Baseline Estimation

In this section, we present the results of the estimation of Eqs. (1) and (2). The results are displayed in Table 3, where columns (1) and (2) show the coefficient estimates for Eq. (1) and columns (3) and (4) show the coefficient estimates for Eq. (2). In each estimation, the most important coefficients are those on the foreign investor variable $({ForeignBuyer}_{i,p,b})$, the cumulative investment amount of the buyer’s country ${(CUMINV}_{p,b,t})$, and the interaction term of these two variables $({ForeignBuyer}_{i,p,b}\times {CUMINV}_{p,b,t})$.

Table 3 Baseline estimation

We start by considering the results in columns (1) and (2). In column (1), the coefficient on ${ForeignBuyer}_{i,p,b}$ is positive and significant, indicating that purchase prices are higher when the buyer is foreign (i.e., ${\beta }_{1}>0$). In Column (2), when we add ${CUMINV}_{p,b,t}$ and ${ForeignBuyer}_{i,p,b}\times {CUMINV}_{p,b,t}$ to the estimation, the coefficients on ${ForeignBuyer}_{i,p,b}$ and ${CUMINV}_{p,b,t}$ are both positive and significant, while the coefficient on ${ForeignBuyer}_{i,p,b}\times {CUMINV}_{p,b,t}$ is negative and significant. These coefficient estimates indicate that foreign investors pay higher prices (i.e., ${\beta }_{1}>0$), but that the extent to which foreign investors pay higher prices diminishes the larger the cumulative investment from country b in country p (i.e., ${\beta }_{3}<0$).

The estimation results further imply that a country’s cumulative investment has a positive impact on purchase prices when the investment is domestic (i.e., ${\beta }_{2}>0$) but the opposite impact when the investment is in another country (i.e., ${\beta }_{3}<0$ and $\left|{\beta }_{3}\right|>\left|{\beta }_{2}\right|$). This differential impact of ${CUMINV}_{p,b,t}$ depending on whether the investment is at home or abroad suggests that the variable may be capturing different things in the case of domestic and foreign investment. While in our analysis we assume that the variable represents foreign buyers’ investment experience in a country, in the case of domestic investment it might be a proxy for strong demand for property. That is, domestic buyers are already well informed about the domestic property market, so that the variable does not capture investment knowledge and experience; instead, the link between greater cumulative investment and property prices means that the property market is heating up.

The coefficients for the remaining variables employed in the estimations in columns (1) and (2) are generally as expected. Purchase prices are higher the larger the floor area of the property structure (LN_Floor), the younger the structure (Age), and the larger investment from countries other than the country of the buyer is (INV_OTHER). Note, however, that the size of the land of the property (LN_Land) has a negative impact on price once the size of the floor area is controlled for.

Next, let us turn to the estimation results based on Eq. (2), which are shown in columns (3) and (4). In column (3), the coefficient on ${ForeignBuyer}_{i,p,b}$ is positive and significant. Meanwhile, in column (4), where we add the ${ForeignBuyer}_{i,p,b}\times {CUMINV}_{p,b,t}$ variable, we find that the coefficient on ${ForeignBuyer}_{i,p,b}$ remains positive and significant, while the coefficient on ${ForeignBuyer}_{i,p,b}\times {CUMINV}_{p,b,t}$ is negative and significant. These coefficients indicate that foreign investors pay higher purchase prices (i.e., ${\beta }_{1}>0$) and that the extent to which foreign investors pay higher prices diminishes the larger the cumulative investment from country b in country p (i.e., ${\beta }_{3}<0$). The coefficients of the remaining variables are qualitatively similar to those in columns (1) and (2).

We evaluate the varying impact of buyers’ being from foreign countries on transaction prices according to the different values of ${CUMINV}_{p,b,t}$ and show the results in Fig. 2. The figure shows the level of ${CUMINV}_{p,b,t}$ along the horizontal axis, the marginal impact of ${ForeignBuyer}_{i,p,b}$ conditional on the level of ${CUMINV}_{p,b,t}$ on the purchase price along the vertical axis, and the 95% confidence interval (dashed lines). The estimated coefficients for ${ForeignBuyer}_{i,p,b}$ (0.435) and ${ForeignBuyer}_{i,p,b}\times {CUMINV}_{p,b,t}$ (-0.844) indicate that the conditional marginal impact of ${ForeignBuyer}_{i,p,b}$ slopes downward as ${CUMINV}_{p,b,t}$ increases. The figure shows that the point estimate of the conditional marginal impact of ${ForeignBuyer}_{i,p,b}$ is statistically significant and positive for ${CUMINV}_{p,b,t}$ of 0.27 or less and not statistically different from zero for the variable being larger than that. Since the summary statistics in Table 2 shows that the mean and standard deviation of ${CUMINV}_{p,b,t}$ are 0.78 and 0.18, respectively, it appears that only in a relatively minor share of purchases by foreign investors do they pay a higher price than do domestic counterparts.

Fig. 2

Marginal effect of ForeignBuyer (baseline). Note: The figure shows the marginal effect of ForeignBuyer conditional on the level of CUMINV. The estimated coefficients are taken from Table 3, column (2)

Note, however, that the mean of ${CUMINV}_{p,b,t}$ for foreign buyers is 0.126 and substantially smaller than the mean in case buyers are domestic (0.808). This is because buyers usually invest their money into domestic real estate properties rather than into foreign properties. Further, the median and 75 percentile point value of the variable ${CUMINV}_{p,b,t}$ for foreign buyers is 0.061 and 0.154, respectively, which are still smaller than its threshold value of 0.27, below which the conditional impact of ${ForeignBuyer}_{i,p,b}$ is statistically significant and positive. In sum, most of foreign realty buyers, even when they are from relatively experienced countries in realty investment abroad, pay more than domestic buyers do.

In addition to the link between ${ForeignBuyer}_{i,p,b}$ and property prices, the results in columns (3) and (4) allow us to examine the time-varying investment performance of each investor category. We focus on the results in column (4), since they are quite similar to those in column (3). We use the coefficients on Buyer investor type*Year and Seller investor type*Year dummies in the column in order to construct Figs. 3 and 4. First, Fig. 3 shows the difference between (i) coefficients on time-variant seller investor type dummies and (ii) coefficients on time-variant buyer investor type dummies. Specifically, the figure shows the differences for equity funds and pension funds. The estimated coefficients are taken from Table 3, column (4). We interpret these differences as the profit margin for each type of investor for each year. The figure shows a sharp decline in these margins right after the financial crisis. Interestingly, the extent of the drop is larger for equity funds than for pension funds, possibly reflecting the fact that debt turnover is more frequent among equity funds than pension funds and that they are more likely to fire-sell assets in their portfolio. Second, Fig. 4 plots the means and standard deviations of these margins for all types of investors. We see that investors such as equity funds and “finance” exhibit higher risk and return profiles, while pension funds show lower risk and return profiles.

Fig. 3

Difference between time-variant seller effects and buyer effects: The cases of equity funds and pension funds. Note: The figure shows the difference between (i) coefficients on time-variant seller investor type dummies and (ii) coefficients on time-variant buyer investor type dummies. Specifically, the figure shows the differences for equity funds and pension funds. The estimated coefficients are taken from Table 3, column (4)

Fig.4

Average and standard deviation of the difference between time-variant seller effects and buyer effects. Note: The figure shows the averages and standard deviations of the differences between (i) coefficients on time-variant seller investor type dummies and (ii) coefficients on time-variant buyer investor type dummies. We calculate the differences for all the types of investors. The estimated coefficients are taken from Table 3, column (4)

Estimations Controlling for Geographical Proximity and Property Fixed Effects

In the baseline estimation in the previous subsection, we assumed that the heterogeneity in property prices across observations is explained by a number of factors, including the host, buyer, and seller countries, property characteristics, and the buyer type. However, it is still possible that our estimation did not adequately take geographical proximity between real estate properties into account, since the baseline estimation includes only country dummies for geographical information on real estate properties. In order to deal with this issue, we implement two additional estimations in this subsection: (1) We employ information on geographical proximity between properties and pair properties purchased by foreign and domestic investors. And (2) we focus on properties that were sold multiple times and apply the repeat-sales methodology.

We begin by controlling for geographical proximity by collecting information on real estate properties that are located close to each other. We take the following steps in order to pair properties purchased by domestic investors with properties in close proximity purchased by foreign investors. We start by measuring the geographical distances between all pairs of real estate properties using the latitude and longitude information on each real estate property. Then, for each property purchased by domestic investors, we match the properties that are purchased by foreign investors and located closest to the property purchased by domestic investors. In the process of this matching, we set a threshold for the maximum radius within which the matching is done. We employ a radius of 100 m and of 500 m. Finally, using these matched observations, we regress realty price on the set of explanatory variables employed in the baseline estimation.

As a result of this procedure, we drop from our observations a certain number of properties that were purchased by foreign (domestic) investors but located far from any of the properties purchased by domestic (foreign) investors. When we construct the dataset by limiting the sample to observation that satisfy the constraint of a 500 m radius, the sample size drops to slightly above 20,000 observations. On the other hand, when we limit the sample using the constraint of a 100 m radius, the sample size falls to about 5,300 observations.

The results of the matched sample estimation are presented in columns (1) and (2) of Table 4. The most important coefficients are again those on ${ForeignBuyer}_{i,p,b}$, ${CUMINV}_{p,b,t}$, and ${ForeignBuyer}_{i,p,b}\times {CUMINV}_{p,b,t}$. The coefficients on these variables are qualitatively the same as in the baseline estimation; that is, they are positive and significant for ${ForeignBuyer}_{i,p,b}$ and ${CUMINV}_{p,b,t}$ and negative and significant for the interaction term of the two. Somewhat surprisingly, the coefficient on ${ForeignBuyer}_{i,p,b}$ is substantially larger in column (2) than in column (1) even though the matched samples of properties purchased by domestic and foreign buyers are located close to each other.

Table 4 Robustness check

Next, taking one step further, we employ the repeat-sales approach in order to more rigorously control for property fixed effects. Specifically, we identify real estate properties that are sold multiple times using the latitude and longitude information of each property. We further limit the sample of such repeat-sales properties to those that were purchased both by a domestic and a foreign investor at least once during the observation period. We end up with a sample of slightly less than 4,500 properties, which we use for the estimation controlling for property fixed effects. Note that some of the property characteristics variables, such as LN_Floor and LN_Land, were dropped, since these variables are time-invariant within the repeat-sales sample.

The results are shown in columns (3) and (4) of Table 4. The coefficients in these estimations on ${ForeignBuyer}_{i,p,b}$, ${CUMINV}_{p,b,t}$, and ${ForeignBuyer}_{i,p,b}\times {CUMINV}_{p,b,t}$ are qualitatively the same as those in the baseline estimation. They are positive and significant for ${ForeignBuyer}_{i,p,b}$ and ${CUMINV}_{p,b,t}$ and negative and significant for the interaction terms. The only exception is the coefficient on ${ForeignBuyer}_{i,p,b}$ in column (3), which is positive but statistically insignificant. This result implies that, at least in this sample, the unconditional marginal impact associated with ${ForeignBuyer}_{i,p,b}$ is not statistically different from zero.

To summarize, we control for locational proximity and property-level fixed-effects in a more rigorous manner than in the baseline so as to correctly identify the impact of purchases by foreign investors. In both approaches, i.e., when limiting the sample to matched properties in close proximity and when employing the repeat-sales sample, we find that foreign investors pay significantly higher prices than domestic investors and that the extent to which foreign investors overpay is smaller the larger is the cumulative investment from the country of foreign buyers in the host country.

Subsample Estimations

In this subsection, we examine how the estimation results in Table 3 are affected when we use different subsamples. First, we split the observation period into two subperiods. The results are presented in Table 5. The first two columns show the results when we split the period into the period through 2010 (column (1)) and after 2010 (column (2)), while the next two columns show the results for the subperiod through 2008 (column (3)) and after 2008 (column (4)). Overall, the coefficients are qualitatively the same as those in Table 3 with the exception of those in column (3), where all the coefficients on ${ForeignBuyer}_{i,p,b}$, ${CUMINV}_{p,b,t}$, and ${Foreignbuyer}_{i,p,b}\times {CUMINV}_{p,b,t}$ are statistically insignificant.⁹ A notable feature in the results is that the impact associated with ${ForeignBuyer}_{i,p,b}$ is substantially larger in columns (2) and (4) than in columns (1) and (3). Given that the later periods were periods when real estate markets were recovering from the global financial crisis, this result shows that the importance of information asymmetry increased after the financial crisis. This result somewhat contradicts the widely held view that prices are more likely to be above fundamentals during boom periods.

Table 5 Estimation for different observation periods

Second, we split the sample based on the type of property purchased and present the results in Table 6. Specifically, we focus on the following five categories: apartments (column (1)), hotels (column (2)), industrial (column (3)), offices (column (4)), and retail (column (5)). Looking at the coefficients on ${ForeignBuyer}_{i,p,b}$, ${CUMINV}_{p,b,t}$, and ${ForeignBuyer}_{i,p,b}\times {CUMINV}_{p,b,t}$, we find that only for industrial and office properties do we obtain qualitatively similar results as in the baseline. In contrast, for the other types of properties including apartments, the signs on the coefficients of interest are consistent with our baseline results, but the coefficients are insignificant. The above results suggest that information asymmetry is more likely to be present in the case of types of property whose value is more difficult to measure (e.g., properties for business use) than in the case types of properties whose structures are standardized and whose values are relatively easy to measure (e.g., residential properties).

Table 6 Estimation for different property types

Estimations with Additional Explanatory Variables

In this subsection, we examine how the baseline results in Table 3 change when we introduce additional variables. In the baseline analysis, we employed dummies for the type of investor purchasing a property as well as buyer country-year dummies and property host country-year dummies without specifying the mechanisms how these dummies affect real estate prices. To be more specific about the mechanisms, we add two sets of variables in Eq. (1). The first set concerns the buyer’s views and objectives regarding the investment: whether the buyer regards the investment as low-risk and low-return (Core = 1) or otherwise (Core = 0), the buyer intends to increase the value of the property and regards the investment as medium-to-high risk and medium-to-high return (ValueAdded = 1) or otherwise (ValueAdded = 0), or the buyer intends to occupy the property him/herself (Occ = 1) or otherwise (Occ = 0). In our estimation, we drop Occ from the independent variables and include only Core and ValueAdded to measure the impact associated with Core and ValueAdded, using Occ as the baseline case.¹⁰ The second set of variables measures the buyer’s investment opportunities both in his/her home country (Buyer_YoY_Return) and in the host country (Host_YoY_Return). Note that the sample size decreases substantially, since these additional explanatory variables are available only for a limited number of observations. For example, we employ Buyer_YoY_Return constructed from the housing price index for a limited number of countries by the Federal Reserve Bank of Dallas.

Table 7 shows the results, which reveal a number of notable findings. First, the signs of the coefficients on ${ForeignBuyer}_{i,p,b}$, ${CUMINV}_{p,b,t}$, and ${ForeignBuyer}_{i,p,b}\times {CUMINV}_{p,b,t}$ are the same as those in the baseline results, although the coefficient on ${CUMINV}_{p,b,t}$ is statistically insignificant. Second, buyers’ investment objective has a significant impact on the price they pay. It turns out that buyers that regard their investment as medium-to-high risk and expect medium-to-high returns end up paying higher prices. It also shows that buyers that regard their investment as low-risk and low-return tend to pay higher prices as well. The fact that the coefficient on ValueAdded (0.147) is larger than that on Core (0.075) indicates that the extent of overpayment by a buyer is more sizable in case of middle-to-high risk investment than in the case of low-risk investment. Third, investment opportunities in buyers’ home country, but not those in the host country, have a significant positive impact on the purchase prices. Buyers whose home country real estate market is booming may be less likely to be financially constrained and end up in paying over the odds for real estate overseas.

Table 7 Additional independent variables

Testing for Spillover Effects

The results presented in Tables 3, 4, 5, 6 and 7 have shown that foreign real estate buyers tend to pay higher prices than domestic buyers. Given that this result is obtained controlling for a wide range of property and investor characteristics, it seems possible that property purchases by foreign investors might lead to demand pressure in host country real estate markets. However, whether property purchases by foreign investors at inflated prices create sufficient pressure to have a wider impact on the local real estate market is an empirical matter. Specifically, it will depend on the number of investable properties in the local market and hence the elasticity of supply. In addition, if the markets in which foreign and domestic buyers invest are segmented, the higher prices paid by foreign buyers will not necessarily generate spillover effects on the wider property market.

Based on these considerations, we focus only on properties bought by domestic investors and how the prices paid by them are related to the prices of properties nearby purchased by foreign investors. Properties nearby are those within a 100 m radius. In order to identify the causal impact of property purchases by foreign buyers, we distinguish properties bought by domestic investors into those bought before and after property purchases by foreign investors nearby. The former properties are then used as the control group, while the latter are the treatment group. If there are any causal spillover effects from the higher prices paid by foreign buyers on the prices paid by domestic buyers, we should see a statistically significant difference in the prices paid by the two different groups of domestic buyers.

Table 8 outlines the sample we use for this spillover test. Specifically, the table shows that among the 27,166 properties bought by domestic buyers, 23,190 do not have any properties bought by foreign investors nearby (i.e., within a radius of 100 m).¹¹ The number of properties bought by domestic investors that do have a property or properties bought by foreign investors nearby is a total of 3,976. In 1,470 of these case, the domestic investor bought the property before the foreign investor(s), while in 2,053 cases, the domestic investor bought the property after the foreign investor(s). Meanwhile, in 453 cases, foreign buyers bought properties nearby both before and after the domestic buyer.

Table 8 Sample for spillover test

Table 9 Spillover test

Using these 27,166 observations, we run the following estimation, which, in addition to various control variables, includes the following two dummy variables:${Placebo}_{i,p,b}$, which takes a value of one if property i has a property nearby purchased by a foreign buyer after the property was bought, and${Spillover}_{i,p,b}$, which takes a value of one if property i has a property nearby purchased by a foreign buyer before the property was bought. Note that as we focus only on properties bought by domestic investors, we cannot include ${ForeignBuyer}_{i,p,b}$ and ${CUMINV}_{p,b,t}$ in the equation. If there are any spillover effects as a result of purchases by foreign buyers, we expect ${\delta }_{1}>0$ but ${\delta }_{2}$=0, since purchases by foreign buyers after purchases by domestic buyers should not affect the price paid by domestic buyers. Thus, we estimate the following equation¹²:

$$\begin{array}{c}{LN\_PriceUSD}_{i,p,b,s,t}=\alpha +{\delta }_{1}{1\_spillover}_{i,p,b}+{\delta }_{2}{1\_placebo}_{i,p,b}\\ +{\delta }_{3}{1\_spillover}_{i,p,b}\times {1\_placebo}_{i,p,b}+{{\varvec{X}}}_{it}{\varvec{\gamma}}+{\eta }_{p}^{1}+{\eta }_{b}^{2}+{\eta }_{s}^{3}+{\eta }_{t}^{4}+{\varepsilon }_{i,t}\end{array}$$

(3)

Table 9 presents the estimation results. We find that both ${\delta }_{1}$ and ${\delta }_{2}$ are greater than zero, indicating that domestic buyers tend to pay more for properties that are located close to properties purchased by foreign buyers than for properties that are not close to properties purchased by foreign buyers. We also find that these coefficients are almost identical. This indicates that there exists no substantial difference in the premium paid by domestic buyers regardless of whether they purchased their property before or after the foreign buyer(s) purchased theirs. These results provide evidence for a selection effect in that prices are higher for properties purchased by foreign buyers, while we find no evidence for the causal impact of foreign realty investment on prices domestic buyers pay for real estate properties.

We summarize our findings in this entire section as follows. First, foreign investors pay a positive premium for the purchase of real estate properties over domestic buyers, while the size of the premium becomes smaller as the buyer country accumulates its investment experience in the host country (Tables 3, 4, and 7). The results are overall robust throughout the subsample analysis, whilst there are some exceptions when we employ the subsample before the global financial crisis and the subsample of non-business property transactions (Tables 5 and 6). Second, we fail to find a causal relationship that the positive premium foreign investors pay induces domestic buyers to raise their purchase prices (Table 9).

Conclusion

In this study, we examined whether foreign real estate buyers pay higher prices than their domestic counterparts and, finding that this is the case, investigate whether this has any spillover effects in terms of driving up property prices. Using about 30,000 observations covering realty investment transactions in eight countries and controlling for a comprehensive list of property and transaction characteristics, we find the following. First, foreign investors pay substantially higher prices than domestic investors even after taking various controls into account. Second, this price difference is smaller the larger the cumulative aggregate real estate investment from the buyers’ country in the host country. These results suggest that foreign investors tend to pay over the odds when they are less informed about local property markets and that the extent to which they overpay decreases with buyers’ investment experience, which, we assume, is represented by the cumulative investment. Third, despite the finding that foreign investors tend to pay over the odds, we did not find any evidence of significant spillover effects of foreign real estate investment on local real estate markets.

Our last result sheds new light on the previous literature that examined the impact of international capital flows on real estate prices. We disentangle the mechanism of how much prices foreign investors overpay for local properties and how these investors interact with local investors. The evidence against spillovers suggests that we need to provide a reason why. One of the potential explanations includes the supply of properties in host countries being fairly elastic and/or there being a certain degree of market segmentation between foreign and domestic property buyers.

Finally, we highlight potential avenues for future research. First, an important extension of this study is to examine if really no spillover effects can be observed under all circumstances and, if this is the case, to consider why. Second, another important research direction would be to examine investors’ choice among multiple investment locations. We believe these potential extensions could provide further insights for a better understanding of the price implications of international real estate transactions.