What types of capital in ows are associated with booms in real estate prices?

Transcription

1 What types of capital in ows are associated with booms in real estate prices? Alejandro Jara y Central Bank of Chile Eduardo Olaberría z OECD November 16, 2012 Abstract The importance of the real estate price bubbles as leading indicators of nancial crises is hardly disputable (Kaminsky and Reinhart (1999)). However, the empirical evidence about its causes is still weak. In this paper, we provide additional evidence about the link between capital in ows and booms in real estate prices, by building on the recent literature on this subject and focusing on the role of the composition of capital ows. We nd that the robust and strong positive association between large capital in ows and booms in real estate prices are mostly explained by debt related ows (bank and portfolio debt ows), rather that other types of capital ows. In addtion, the association between capital in ows and booms in real estate prices can be reduce by the use of capital controls, the adoption of a more exible exchange rate, and the quality of institutions. Financial development, on the other hand, increases this association between capital in ows and booms in real estate prices. Keywords: Capital in ows, asset prices, real estate, capital controls, exchange rate regimes, bubbles JEL Classi cation Number: E32; F32; F41; G10; G12; G15 Preliminary draft. Please do not cite. y ajara@bcentral.cl z Eduardo.OLABERRIA@oecd.org 1

2 1 Introduction and motivation The importance of asset price bubbles in general, and real estate price bubbles in particular, as leading indicators of nancial crises is hardly disputable. 1 Therefore, understanding what are the main determinants of such bubbles is highly relevant from the perspective of a policy maker. While there are several studies that have stressed the link between housing prices bubbles and the surge of capital ows, the empirical evidence remains weak. In fact, as Reinhart and Reinhart (2008) have argued, there has been only discussion and some anecdotal evidence that suggest that asset prices tend to increase during periods of surges of capital ow. 2 The starting point for this type of analysis considers that, when capital in ows are abundant, the demand for domestic nancial assets increases, including those related to housing, putting addition pressure on property prices, assuming that the supply of those assets remains constant. In this paper, we study the degree of association between the dynamics of capital in ows and booms in the domestic real estate prices, which in turn may increase the risk to nancial stability. In particular, we address the following questions: (i) Is there a signi cant relationship between capital in ows and real estate prices?, if so (ii) What kind of capital in ows generates the stronger impact?, (iii) What role does it play some structural country s characteristics in this relationship?. For this purpose, we build a panel of quarterly data for the period of 1990q1 2010q4. In particular, we construct a real housing price index for a selected number of advanced and emerging economies, and combine it with a dataset of di erent types of net capital in ows as percentage of GDP (FDI, portfolio equity, portfolio debt, and others). In addition, we construct a set of control variables, including GDP growth, and in ation; as well as a set of structural variables that account for the degree of nancial development, nancial openness, the exibility of the exchange rate, and a proxy for the quality of institutions. We believe that this is particularly important because the understanding of macroeconomic and nancial e ects of capital in ows is of key importance to policy makers, particularly when surges in capital in ows increase the risk of nancial distress as Ostry et al. (2010) and Ostry et al. (2011) have emphasized, since some countries may have incentives to establish controls on certain types of capital in- ows. In this context, identifying the mechanisms through which capital in ows increase nancial vulnerabilities and the role that policies can play in moderating 1 Reinhart and Reinhart (2008) present evidence that real house prices boomed on the eve of the worst post-world War II banking crises in emerging market economies. 2 Calvo et al. (2003), for example, provides some of these anecdotal evidence. 2

3 these vulnerabilities, would help countries in choosing the appropriate policy mix. Although the literature that studies the relationship between capital in ows and asset prices is rather limited, recent works by Jinjarak and coauthors have made an important e ort in this direction. Aizenman and Jinjarak (2008), for example, look at the association between the current account and real estate valuation across countries. Controlling for lagged GDP per capita, in ation, institutions and interest rates, they nd a robust and positive association between current account de cits and the appreciation of real estate prices. In a similar vein, Jinjarak and She rin (2011) explore the issue of causality between real estate prices and the current account. Using recently developed econometric methods to perform a number of case studies, they nd that current account de cits directly drove real estate prices in Ireland, Spain and United States, but that the e ect in England was only transitory. Nevertheless, none of these works study the association between di erent categories of capital in ows and real estate prices; which we believe is important because it helps to account for some key features that characterizes these in ows (maturity, currency composition, and sectorial relevance). Furthermore, the current literature, as far as we are concern, does not analyze the channels through which capital in ows are more likely to a ect booms in real estate prices. Therefore, our main contribution to the literature we believe is to provide evidence along these lines. 2 A conceptual framework In this section we presents a simple two period model to illustrate the macroeconomic consequences of large capital in ows. In particular, the model will illustrate how a sudden in ow of capital will give rise to, among other things, an increase in the real price of real state. Formally, consider a two period, real model of a small open economy. Consumers derive utility from consuming both tradable goods (c T ) and house services or non-tradable goods (c H ). The economy is endowed with a constant stream of tradable (y T ) and houses (y H ). The consumer s preferences are given by: u c T 1 + v c H 1 + u c T 2 + v c H 2 where is a preference parameter that capture a demand shocks in period 1, > 0 is the discount factor, u(:) and v(:) are strictly increasing and strictly concave functions. Assuming, for simplicity that initial net foreign assets are zero, period 1 budget constraint takes the following form: (1) 3

4 b 1 = y T + p 1 y H c T 1 p 1 c H 1 (2) where p denotes the relative price of houses in terms of the numeraire (tradable goods) and b 1 denotes net foreign assets at the end of period 1. Therefore, the consumer solves the following Lagrangean: L = u c T 1 + v c H 1 + u c T 2 + v c H 2 + y T + p 1 y H c T 1 p 1 c H 1 + yt + p 2 y H c T 2 p 2 c H r (3) The F.O.C. are given by: u 0 c T 1 = ; (4) v 0 c H 1 = p1 ; (5) u 0 c T 2 = 1 + r (6) u 0 c H p 2 2 = (7) 1 + r Combining these rst-order conditions, we get u 0 c T 1 = (1 + r) (8) u 0 (c T 2 ) u 0 c H 1 u 0 (c H 2 ) = (1 + r) p 1 (9) p 2 and imposing the equilibrium condition in the non-tradable real estate market (c H 1 = c H 2 = y H ), Finally, note that in equilibrium: = (1 + r) p 1 p 2 (10) 2 + r 1 + r yt = c T 1 + ct r (11) 4

5 Now we use this simple model to illustrate the macroeconomic consequences of a sudden increase in capital in ows. 3 Assuming that = 1 and (1 + r) = 1, we start at a balanced trade de cit. Now, lets consider a decrease in world interest rates: i.e (1 + r) < 1: In other words, the real interest rate is now lower than the rate of time preference. Then, from (8), the consumption of tradable goods in period one must be higher than in period two (c T 1 > c T 2 ). Hence, from (11), consumption of tradable goods is higher than production of tradable and the economy runs a trade de cit in period 1 (c T 1 > y T ). Additionally, from (10), the relative price of non-tradable houses in terms of tradable goods increases in period one relative to period two (p 1 > p 2 ), i.e. in period 1 the economy experiences high tradable consumption, high real price of houses, and a trade de cit, all relative to period 2. Intuitively, the higher demand for consumption in period 1 relative to period 2 translates itself into a trade de cit and, given that the supply of houses is completely inelastic, to a higher real price of houses. Note that the results would be identical if instead of a decrease in the international interest rate, the economy faces a demand shock. In sum, this very simple two period model shows that an exogenous capital in- ows leads to higher consumption, trade de cits, and an increase in the real price of houses. Hence, while the origin of capital in ows (push or pull) may vary from case to case, the macroeconomic consequences are likely to be very much the same, increasing the probability of a real state bubble. Note that in this model the association between housing prices and capital ows arises from the need to increase period 1 consumption, which in turn increases the current account de cit and capital in ows. Alternatively, we can stress the role played by the interest rate channel; as capital in ows help to sustain low levels of domestic interest rates, thereby strengthening domestic demand. While the previous model gives us a preliminary approach to understanding the relationship between capital ows and house prices, this model is limited in several aspects. First, it is not possible to distinguish which type of capital ows may be associated more strongly to a booms in housing prices. Moreover, it is not clear if nancial development, capital account openness, and exchange rate exibility could play a role in this relationship. Nevertheless, Calvo (2011), provides an alternative way to think about the rela- 3 Note that in this analysis it does not matter what generates the increase in capital in ow, because the e ects on the model are the same. In other words, capital in ows can be the result of pull factors (e.g. an exogenous change in the demand for foreign assets) or push factors (e.g. a decrease in the internatinal real interest rate). 5

6 tionship between capital ows and housing prices. According to his model, there is a direct relationship between capital ows and housing prices, because capital in ows increase the liquidity of real estate assets, thereby increasing its value. Moreover, when capital in ows are intermediated by a local banking system with loose credit standards, this relationship gets strengthened. Consequently, Calvo (2011) provides a conceptual framework for understanding the link between housing prices and some speci c type of capital in ow (i.e. banking capital ows). A somewhat di erent perspective is one that emphasizes the relationship between an increase in housing price and the value of the collateral. In this context, the increase in house prices increases consumers wealth by the so called "housing wealth e ect", which in turn increases consumption and decreases domestic savings. Therefore, the relationship between housing prices and capital in ows is indirect, as capital in ows increases as a consequence of the lower domestic savings. Note that when capital in ows are directly related to housing prices, it is posible to visualize a role for the exchange rate regime. In particular, if the exchange rate is more exible, international investors will have less incentives to take advantage of the higher value/liquidity in the real estate sector generated by capital in ows; because the exchange rate exibility may reduces the expected return on this transaction. Moreover, it is also possible to see a role for both the quality of institutions and the depth of the nancial sector. In particular, if the economy has institutions of high quality, it is less likely to come across with problems of asymmetric information in banking sector (adverse selection and moral hazard), reducing the risk of overborrowing (McKinnon and Pill (1996), Krugman (1998) and Kiyotaki and Moore (1997)). Moreover, in an economy with a more developed nancial system, it is more likely that the credit channel will operates, even if the bank has conservative lending standards. 3 Empirical approach 3.1 Data issues For the regression analysis, we construct a quarterly panel data for 35 advanced and emerging countries, from 1990q1 2010q4 (see table 1 for the list of countries considered). This panel data includes information on real housing prices, capital in ows, and a set of control variables. Nominal housing prices come from the dataset gathered by the Bank for International Settlements (BIS), which we then de ated by the consumer price index (CPI) to obtain our measure of real housing price index. 6

7 Net capital in ows comes directly from the capital account information of the Balance of Payments Statistics of the IMF (BoP). In this sense our approach di ers from Aizenman and Jinjarak (2008), where net capital in ows are approximated by the current account de cits. 4 Regarding the set of control variables, annual real GDP growth and quarterly in ation rate comes from the IFS from the IMF. Financial depth is proxy by the stock of credit issued by the banking sector, and the nominal GDP. These variables comes from the IFS and the WDI from the World Bank respectively. We also use the index created by Chinn and Ito (2008) as a proxy of capital account openness. Regarding the exchange rate regime, we use the coarse measure de ned by Reinhart and Rogo (2004). Finally, we use the rating of sovereign debt as a proxy of quality of institutions. See the Appendix B for more details about the sources and de nitions of all the variables included in our empirical approach. 3.2 Regression analysis This section studies the relationship between booms of real estate prices and di erent types of capital in ows. In particular, we estimate booms in real estates prices as a function of international capital in ows and domestic conditions; and look if the composition of capital in ows matter for this association. In addition, we estimate a set regressions where we analyses whether the association between capital in ows and real estate prices can be in uenced by the degree of nancial openness, nancial development, the exchange rate regime, and the quality of institutions. Methodologically we follow closely Olaberria (2011), which perform similar regression analysis, although its focus is on the relationship between capital in ows and stock prices. Our variable booms in real housing prices represents a measure of both, the occurrence and the intensity of a signi cant increase in real housing prices. In order to identify the occurrence of such increase, we follow Mendoza and Terrones (2008) de nition of credit booms. First, we obtain the cyclical component of the real housing price index as the di erence between the index and its long-run trend using Hodrick- Prescott (HP) lter, with the standard smoothing parameter set at 1600 for quarterly 4 Although the IMF data is one of the most comprehensive, there are several issues associated with the compilation of the BoP that need to be taken into account. For example, data is missing for many countries, in particular for the early nineties, which makes the data coverage to vary substantially from country to country. Additionaly, as Lane and Milesi-Ferretti (2001) suggests, there are a number of measurement problems with debt data related to di erent methodologies for recording non-payments, rescheduling, debt forgiveness, and reductions. We recongnize these are important limitations of the database that may bias the estimates, but is something we have to live with. 7

8 data. Second, we compute the standard deviation of the cyclical component. Then, we de ne the event of a signi cant increase in real housing prices as the period when the cyclical component is above one standard deviation. 5 Thus, our de nition of booms is equal zero during normal times, and equal to the deviation from the trend during periods of extreme increase in the real housing prices. Figures 1a and 1b in the Appendix shows the cyclical component of the real housing price index, its standard deviation, and the de nition on booms used in the regression analysis for a selected number of advanced and emerging economies. Note that the cyclical component of the real housing price index is stationary for most countries, as shown by the results of the unit root tests (see table 2 in the Appendix). We start by estimating a standard regression equation using a cross-country timeseries unbalanced panel data of 35 developed and emerging countries over the period 1990q1 2010q4: Boom it = X it + KF it + t + i + it (12) where the subscript i and t represent the country and time period respectively. Boom it is our measure of the booms in real housing price as explained above. X it represents the set of control variables, and KF it represents the di erent categories of net capital in ows as percentage of GDP, including foreign direct investment, portfolio investment (equity and debt), and other in ows. Finally, t and i denote unobserved time and country speci c e ects, respectively, and it is the error term. In order to avoid an omitted variable bias, we control for variables that can potentially be associated with housing prices. First, in order to control for the state of the economy we use the annualized real GDP growth rate, because if the economy is booming it is more likely that housing prices are booming too. Second, we include the annualized in ation rate to control for the nominal and monetary conditions, because if the economy is facing an expansionary environment, it can create conditions for increasing housing prices. Also, we use as control a set of variables that shown low variability across time, and that represent structural characteristics of each economy, such as the degree of nancial development, the openness of the nancial account, the degree of exchange rate exibility, and a proxy for the quality of institutions. We also use instrumental variables to deal with potential problems of endogeneity, as some variables that a ect housing princes can also a ect capital in ows. In particular, we use instruments that are associated with housing prices mainly through their relationship with capital in ows, such as external factors like: (1) VIX, (2) 5 Note that this de nition is quite subjective. Alternatively, we can consider making this de - nition more strict by increasing the number of standard devitions required to identify an extrem event. 8

9 GDP growth of the three main world economies (G3), (3) nominal short term interest rate of G3, and (4) the cyclical component of a commodity priced index. We include the VIX because as Forbes and Warnock (2011) claims, this variable is one of the most important drivers of both, net and gross capital ows. Table 3 reports our rst set of results of the determinants of booms in real estate prices. Column 1 presents a simple panel regression between booms in housing prices and di erent types of net capital in ows assuming xed e ects. Column 2 follows a similar methodology than in column 1, but now adds the control variable. Finally, column 3 uses instrumental variables. The results on column 1 shows that the simple regression between di erent types of net capital in ows and booms in housing prices is strong and signi cant at 1%. Moreover, "other in ows", which includes mainly cross-border bank ows, presents the stronger association with booms in housing prices. However, the results in column 1 may be biased by the fact that we are not controlling for other factors that can a ect housing prices. Column 2 presents the results after incorporating the annual GDP growth, and the quarterly in ation rate as controls; in addition to the structural variables described above. The association between GDP growth and booms in real estate prices is positive and statistically signi cant, similarly than with in ation, suggesting that booms in real estate prices are more likely to occur when the economy is growing fast and in ation is increasing. Moving to our variable of interest, column 2 indicates that all net capital in ows are positively associated with booms in real estate prices. However, the magnitude of this association is not the same for all types of capital in ows. In particular, the coe cients is larger for "other in ows", suggesting that bank related in ows are the most likely to increase housing prices during booms periods. Column 3 presents the results of endonenizing net capital in ows through instrumental variables. Comparing with column 1 and 2, column 3 stresses the point that not all types of capital in ows can be associated with booms in real estate prices. In particular, we nd that "other in ows" and "portfolio debt" in ows are positively and statistically signi cantly associated with booms in real estate prices. The coe cient of "other in ows" shows again the stronger association with booms in real estate prices. To summarize, table 3 provides the rst two statements of the paper: (1) net capital in ows can be associated with booms in housing prices, but (2) not all types of capital ows are created equal. Furthermore, the association is only relevant for debt related ows, both bank loans and portfolio debt. These results complement the ndings of Aizenman and Jinjarak (2008) and Sá et al. (2011) who nd that in OECD countries the current account balance (a close measure of total net capital 9

10 in ows) is signi cantly associated with booms in real estate prices. Theory suggest that some factors, such as the exibility of the exchange rate, the level of nancial development, the degree of nancial openness, and the quality of institutions can a ect the association between capital in ows and housing prices. We now assess these issues by allowing the association of each measure of capital in ows and booms in housing prices to vary with the level of nancial development, nancial openness, the exchange rate regime, and a proxy for the quality of institutions. We do this by interacting each capital in ows measure with linear measures of the variable of interest in each country and each period. Therefore, we estimate in this case the following regression: Boom it = X it + 2 KF it + 3 (KF it Interactions it ) + t + i + it (13) Table 4 reports in column 1 to 4 the estimations including the individual interactions using a panel regression methodology assuming xed e ects; while column 5 reports the results considering all interactions together. Column 1 interact net capital in ows with a measure of nancial depth, as represented by the domestic bank credit as percentage of GDP, measured in logs. Column 2 interact the di erent categories of net capital in ows with the measure of nancial openness proxy by the de jure index of Chinn and Ito (2008). Column 3 uses the coarse exchange rate classi cation de ned by Reinhart and Rogo (2004). Finally, column 4 considers the interaction of net capital in ows with the rating of the sovereign debt measured in logs, as a proxy of the quality of institutions. We also control for the growth rate of the economy, and the in ation rate, similarly than in previous estimations. In general, the results presented in table 4 are in harmony with our expectations. First, we nd that real GDP growth and in ation rate are signi cant in all regressions, except in column 5 when all interactions are included. We interpret this result as indicating that booms in housing prices are more likely to occur when the economy as a whole is booming. When we look at the di erent categories of capital in ows, the rst insight is that "other in ows" and "portfolio debt" continue to be positively and signi cantly associated with booms in real estate prices in most regressions, reinforcing our ndings of table 3. However, the coe cients are not always stable, showing that interactions play a signi cant role in understanding the association between capital in ows and booms in housing prices. Column 1, for example, shows that the association between booms in real estate prices and net capital in ows is stronger the more developed are nancial markets, as re ected by the coe cient accompanying the measure of nancial depth. This result is reasonable given that our measure of nancial development rather than measuring 10

11 the strength of the nancial system is a measure of the size of the banking sector in the whole economy. Furthermore, this is consistent with the conjecture described in Calvo (2011), which emphasizes that a more developed nancial system can help to increase the liquidity of some xed, and formerly not very liquid, assets, like houses; and that these liquidity considerations create a perfect environment for the creation of bubbles in housing prices. In the case of nancial openness (column 2), the interaction is positive and signi cant, suggests that stricter capital control help reduce the association between capital in ows and booms in housing prices, a result consistent with Korinek (2010). The negative coe cients for the interactions of capital in ows and the measure of exchange rate regime (column 3) suggest that a more exible exchange rate regime reduces the association between capital in ows and booms in housing prices, consistent with Mendoza and Terrones (2008). Column 4 shows that, when considered by itself, our proxy for the quality of institutions have not always the same sign and statistical signi cance. However, the positive coe cient accompanying "other in ows", which is positive and signi cant in column 4 and 5, implies that improving the rating/quality of institutions, helps to reduce the association between "other in ows" and housing price booms. 6 Column 5 reinforce the ideas describe above, as nancial depth, nancial openness, a less exible exchange rate, and poor quality of institutions increase the association between capital in ows and booms in housing prices, in particular, but not exclusively, for debt related ows. Finally, in table 5 we control for potential endogeneity problems using instrumental variables. We use the same set of instruments described above. Once again, we nd that, in general, booms in real estate prices are positively and strongly associated with capital in ows, but composition matters. Moreover, this association seems to be a ected by nancial depth, nancial openness, the exibility of the exchange rate and the quality of institutions, in particular for "other in ows", as shown in column 5. Most of our ndings are consistent with theory. Speci cally, these results con rm previous theoretical ndings that an increase in net capital in ows can potentially be linked to booms in real stock prices. The results are also consistent with the empirical literature. In particular, the nding that debt related ows, represented by both, "other in ows" and "portfolio debt" ows, are the more dangerous type of capital. For example, Tong and Wei (2011) shows that the volume of total capital in ows has no signi cant e ect on the severity of stock market declines, but that large pre-crisis exposure to debt related in ows tends to be associated with a more 6 Note that by de nition our "rating" variables improves when it falls. 11

12 severe decline during the crisis. Or results complement these ndings of Tong and Wei (2011) by showing that debt related ows are more likely to have been involved in the development of a bubble that creates the crisis. A result that may look surprising at rst is that a more developed nancial market can increase the association of "other in ows" and booms in real estate prices. This result is not always robust, but it holds in a number of estimations. However, is somewhat consistent with the model described in Calvo (2011), which shows that nancial development can increase the liquidity considerations of goods that are per se not liquid, and create the rationale for the creation and destruction of bubbles, and the related disturbances in credit markets. 4 Conclusions As a consequence of the global nancial crisis, policy makers are reassessing regulatory policies to reduce systemic vulnerabilities and costly nancial crisis. A key lesson we learned from the crisis is that wrong macroeconomic policies, weak regulation, and market failures pose a great risk to nancial stability. In this context, this paper makes a signi cant contribution for policy analysis. It contributes to our understanding of the role that institutions and policies can play in moderating the vulnerabilities associated with an increase in capital in ows. The paper s ndings can potentially help policymakers choose the appropriate policy options to handle an increase in capital in ows. A major implication of this paper is that capital in ows, in particular debt related in ows, are associated with booms in real estate prices; and, therefore, can potentially increase the risk of nancial crisis. Nonetheless, there are some factors that can help reduce this association. Consider as an example the prescription given in a recent IMF paper by Ostry et al. (2011). Ostry et al. (2011) claims that countries "may have incentives to establish administrative controls to capital in ows if they increase the risk of nancial distress." Since nancial distress is one of the (undesired) consequences of booms in asset prices (Kaminsky and Reinhart (1999)), one could reinterpret this claim as: if capital in ows were found to be associated with booms in asset prices, policymakers may have incentives to establish capital controls. This inquiry can, in part, be answered by the ndings of this paper. The ndings in this paper suggest that countries that adopt a more exible exchange rate and use, at least temporally, capital controls, can reduce the association between capital in ows are booms in real estate prices. On the other hand, nancial development should be handled with care because it can potentially increase the 12

13 association between capital in ows an bubbles. References Aizenman, J. and Y. Jinjarak (2008), Current account patterns and national real estate markets. NBER Working Paper. Calvo, G. (2011), On capital in ows, liquidity and bubbles. Columbia University, May. Available at www. columbia. edu/ gc2286. Calvo, G.A., A. Izquierdo, and E. Talvi (2003), Sudden stops, the real exchange rate, and scal sustainability: Argentina s lessons. NBER Working Paper. Chinn, M.D. and H. Ito (2008), A new measure of nancial openness. Journal of Comparative Policy Analysis, 10, Forbes, K.J. and F.E. Warnock (2011), Capital ow waves: surges, stops, ight, and retrenchment. NBER Working Paper. Jinjarak, Y. and S.M. She rin (2011), Causality, real estate prices, and the current account. Journal of Macroeconomics, 33, Kaminsky, G.L. and C.M. Reinhart (1999), The twin crises: the causes of banking and balance-of-payments problems. American economic review, Kiyotaki, N. and J. Moore (1997), Credit cycles. The Journal of Political Economy, 105, Korinek, A. (2010), Regulating capital ows to emerging markets: An externality view. Unpublished manuscript. Krugman, P.R. (1998), What happened to asia? Lane, P.R. and G.M. Milesi-Ferretti (2001), The external wealth of nations: measures of foreign assets and liabilities for industrial and developing countries. Journal of international Economics, 55, McKinnon, R.I. and H. Pill (1996), Credible liberalizations and international capital ows: the Şoverborrowing syndrome µt. In Financial Deregulation and Integration in East Asia, NBER-EASE Volume 5, 7 50, University of Chicago Press. 13

14 Mendoza, E.G. and M.E. Terrones (2008), An anatomy of credit booms: evidence from macro aggregates and micro data. NBER Working Paper. Olaberria, E. (2011), Capital ows and boom in asset prices, evidence from a panel of countries. Central Bank of Chile, Working Paper. Ostry, J.D., A.R. Ghosh, M. Chamon, and M.S. Qureshi (2011), Capital controls: When and why&quest. IMF Economic Review, 59, Ostry, J.D., A.R. Ghosh, K. Habermeier, M. Chamon, M.S. Qureshi, and D.B.S. Reinhardt (2010), Capital In ows: The Role of Controls. Reinhart, C.M. and V.R. Reinhart (2008), Capital ow bonanzas: An encompassing view of the past and present. NBER Working Paper. Reinhart, C.M. and K.S. Rogo (2004), The modern history of exchange rate arrangements: A reinterpretation*. The Quarterly journal of economics, 119, Sá, F., P. Towbin, and T. Wieladek (2011), Low interest rates and housing booms: the role of capital in ows, monetary policy and nancial innovation. Tong, H. and S.J. Wei (2011), The composition matters: Capital in ows and liquidity crunch during a global economic crisis. Review of Financial Studies, 24, A Appendix 14

15 SE ES IE GB FR US Note: solid line: boom dash line: cycle dot: std cycle Source: Author's calculation based on BIS data Figure 1a: Booms in real estate prices (advanced economies) 15

16 KR ID SK RU MX ZA Note: solid line: boom dash line: cycle dot: std cycle Source: Author's calculation based on BIS data Figure 1b: Booms in real estate prices (emerging economies) 16

17 B Tables 17

18 Tabla 1: List of countries Emerging economies Emerging Asia Emerging Europe Latin America Middle East and Africa MY Malaysia EE Estonia MX Mexico IL Israel ID Indonesia SK Slovak Rep ZA South Africa KR Korea LT Lithuania RU Russia CZ Czech Rep BG Bulgaria HU Hungary LV Latvia PL Poland Advanced economies Europe Other developed O -shore SE Sweden NZ New Zealand HK Hong Kong CH Switzerland CA Canada BE Belgium AU Australia ES Spain US US AT Austria JP Japan IE Ireland GR Greece GB UK NL Netherlands FR France SI Slovenia DK Denmark PT Portugal NO Norway 18

19 Tabla 2: Unit root tests for real property prices (cyclical HP) Country Obs. dfuller p-value AT AU BE BG CA CH CZ DK EE ES FR GB GR HK HU ID IE IL Country Obs. dfuller p-value JP KR LT LV MX MY NL NO NZ PL PT RU SE SI SK US ZA

20 Tabla 3: Determinants of real property prices overvaluation All countries (1) (2) (3) VARIABLES Fixed e ect Fixed e ect IV Net FDI % GDP *** *** (0.0349) (0.0433) (0.0965) Net Equity % GDP *** *** (0.0311) (0.0400) (0.1353) Net Debt % GDP *** *** ** (0.0293) (0.0379) (0.1037) Net OI % GDP *** *** *** (0.0226) (0.0342) (0.1047) GDP growth *** *** (0.0432) (0.0513) In ation *** ** (0.0019) (0.0022) Fin Depth (log) *** (0.0053) (0.0039) Openness *** ** (0.0028) (0.0020) ER regime ** (0.0035) (0.0013) Rating (log) *** (0.0029) (0.0018) Constant *** *** (0.0011) (0.0247) (0.0190) Observations 1,866 1,538 1,538 R-squared Number of IFS codes Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 20

21 Tabla 4: Determinants of real property prices overvaluation with interactions (Panel) All countries VARIABLES (1) Fin depth (2) Openness (3) ER regime (4) Rating (5) All Net FDI % GDP *** (0.3430) *** (0.1068) ** (0.0872) *** (0.0623) *** (0.5059) Net Equity % GDP * (0.3533) (0.1194) *** (0.0831) *** (0.0513) (0.7457) Net Debt % GDP *** (0.2773) (0.0972) *** (0.0793) *** (0.0513) (0.3981) Net OI % GDP (0.2348) (0.0712) *** (0.0708) ** (0.0528) ** (0.3402) GDP growth *** (0.0445) *** (0.0434) *** (0.0428) *** (0.0448) *** (0.0446) In ation ** (0.0019) ** (0.0019) *** (0.0019) *** (0.0019) (0.0019) Fin Depth (log) *** (0.0053) *** (0.0052) *** (0.0052) *** (0.0053) *** (0.0055) Openness *** (0.0028) *** (0.0031) *** (0.0029) *** (0.0028) *** (0.0032) ER regime (0.0035) (0.0035) ** (0.0034) * (0.0035) (0.0034) Rating (log) *** (0.0029) *** (0.0029) *** (0.0029) * (0.0029) (0.0030) INTERACTIONS Net FDI % GDP *** (0.0742) *** (0.0964) Net Equity % GDP *** (0.0715) (0.1136) Net Debt % GDP *** (0.0598) (0.0791) Net OI % GDP (0.0519) ** (0.0697) Net FDI % GDP *** (0.0406) *** (0.0432) Net Equity % GDP * (0.0501) (0.0760) Net Debt % GDP *** (0.0409) (0.0528) Net OI % GDP *** (0.0314) *** (0.0383) Net FDI % GDP (0.0437) (0.0474) Net Equity % GDP *** (0.0419) ** (0.0543) Net Debt % GDP *** (0.0362) *** (0.0380) Net OI % GDP *** (0.0319) *** (0.0323) Net FDI % GDP ** (0.0300) (0.0410) Net Equity % GDP (0.0375) (0.0689) Net Debt % GDP *** (0.0299) (0.0421) Net OI % GDP *** (0.0250) *** (0.0323) Constant *** (0.0253) *** (0.0244) *** (0.0245) *** (0.0253) *** (0.0264) Observations 1,538 1,538 1,538 1,538 1,538 R-squared Number of IFS codes Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 21

22 Tabla 5: Determinants of real property prices overvaluation with interactions (IV) All countries VARIABLES (1) Fin depth (2) Openness (3) ER regime (4) Rating (5) All Net FDI % GDP *** (0.2791) (0.0920) *** (0.0683) *** (0.0568) (0.4335) Net Equity % GDP (0.3451) (0.1194) *** (0.0719) *** (0.0443) (0.7067) Net Debt % GDP ** (0.2691) (0.0975) *** (0.0716) *** (0.0457) (0.3783) Net OI % GDP ** (0.2225) ** (0.0683) *** (0.0614) ** (0.0490) (0.3182) GDP growth *** (0.0432) *** (0.0421) *** (0.0421) *** (0.0425) *** (0.0417) In ation *** (0.0019) ** (0.0019) *** (0.0019) *** (0.0019) ** (0.0019) Fin Depth (log) *** (0.0039) *** (0.0037) *** (0.0033) *** (0.0037) *** (0.0034) Openness * (0.0020) *** (0.0021) (0.0018) *** (0.0020) *** (0.0020) ER regime (0.0019) (0.0018) ** (0.0016) (0.0018) (0.0015) Rating (log) ** (0.0020) (0.0020) (0.0018) (0.0020) ** (0.0018) INTERACTIONS Net FDI % GDP *** (0.0614) * (0.0866) Net Equity % GDP (0.0697) (0.1091) Net Debt % GDP *** (0.0577) (0.0729) Net OI % GDP (0.0492) (0.0634) Net FDI % GDP *** (0.0374) (0.0394) Net Equity % GDP (0.0501) (0.0737) Net Debt % GDP *** (0.0416) (0.0516) Net OI % GDP *** (0.0307) *** (0.0347) Net FDI % GDP (0.0356) (0.0378) Net Equity % GDP * (0.0381) ** (0.0502) Net Debt % GDP ** (0.0322) ** (0.0341) Net OI % GDP *** (0.0282) *** (0.0280) Net FDI % GDP * (0.0256) (0.0385) Net Equity % GDP (0.0351) (0.0628) Net Debt % GDP *** (0.0283) (0.0392) Net OI % GDP *** (0.0232) *** (0.0292) Constant *** (0.0198) *** (0.0191) *** (0.0171) *** (0.0190) (0.0180) Observations 1,538 1,538 1,538 1,538 1,538 Number of IFS codes Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 22

23 C Dataset description 1. Real estate prices (PP and Booms). Index of housing prices, de acted by the CPI. Source BIS and IMF. 2. Capital in ows (KF). Di erent de nitions of net capital in ows as percentage of GDO. Source IFS from the IMF. 3. Financial depth (Fin depth). Stock of bank domestic credit as percentage of GDO. Source IFS and WDI. 4. Capital account openness (Openness). Financial openness as de ned by Chinn and Ito (2008). 5. Exchange rate regime (ER regimes). Coarse de nition of exchange rate regime de ned by Reinhart and Rogo (2006). 6. Credit rating (Ratings): it is calculated considering the credit rating reported by the rating agency Moody s. The credit rating structure of this agency considers 21 ratings (notches), from Aaa (prime) to Ca (in default with little prospect to recovery), additionally, the agency assign a outlook status to the institution or economy evaluated: positive, stable or negative. We transform these string variables to numbers from 0 to 59, where 0 is Aaa positive and 59 is Ca Negative. i.e. if one economy reduce its credit rating one notch, maintaining the stable status, increase the credit rating variable in 3 units. While more risky is the country higher is the credit rating indicator we use. The countries with credit rating information are 55 in our data base. 7. GDP growth rate (GDP Growth). annual percentage variation of the annual gross domestic product (GDP) at constant prices, valued in national currency millions of USD, provided by the WEO database of de IMF. 8. Volatility Index (VIX). The Chicago Board Options Exchange Volatility Index (VIX). Source Bloomberg. 9. In ation rate. Annualized percentage change of the Consumer Price Index (CPI) reported by the IFS. 10. Short term interest rate. Source Bloomberg. 23