Journal Issue
Share
Article

House Price Synchronicity, Banking Integration, and Global Financial Conditions

Author(s):
Adrian Alter, Jane Dokko, and Dulani Seneviratne
Published Date:
November 2018
Share
  • ShareShare
Show Summary Details

I. Introduction

As global liquidity surged owing to accommodative financial conditions, house prices across advanced and emerging market economies have experienced greater synchronicity. IMF (2018a) finds that nearly 80 percent of countries and cities within a broad set of developed economies have experienced positive house price growth rates in the past decade, while this figure is over 60 percent for emerging market economies and cities. Moreover, over time, median synchronicity in house price gaps—measured by extracting the cyclical component of real house prices—has steadily increased over time across countries and cities (Figure 1). 2

Figure 1.House Price Gap Synchronicity Across Countries and Cities

(Closer to zero denotes higher synchronicity)

Source: Authors’ estimates.

Note: The Synch1 measure capturing the negative of the absolute difference between the house price gaps between two countries is used (see Annex II for further details). Upper and lower bounds are the 75th and 25th percentiles of the samples respectively. Solid median lines in city-pair panels denote the time span with significantly higher city-sample coverage. Shaded areas correspond to U.S. recession periods.

House price synchronicity is of particular interest given that greater comovement in house prices could amplify the propagation of external shocks. These shocks could be directly transmitted to the domestic economy through channels such as portfolio, balance sheet, and liquidity, or indirectly through risk premium and confidence channels (Allen and Gale 2000; Longstaff 2010). Simultaneous changes in mortgage rates due to global financial conditions could lead to greater house price synchronicity, thus propagating shocks to aggregate demand when financial conditions tighten sharply. At the same time, an increase in global demand for safe assets may compress sovereign spreads where risk is perceived to be low, thereby pushing down mortgage rates and supporting house price booms in those countries (Bernanke et al. 2011). For instance, foreign capital may be a driver of residential property markets in global cities such as London, New York, or Tokyo, especially during “flight to safety” episodes (Badarinza and Ramadorai 2018). In addition, as illustrated in Figure 2, asset managers may rebalance their portfolios to mitigate their losses, thus resulting in dwindling equity price returns (i.e., portfolio channel); this impact could be further amplified due to asset classes such as REITS. In addition, an exogenous shock to house prices may lead to asset fire sales and deleveraging that would result in declining collateral values and hindering the availability of credit in the economy (i.e., bank balance sheet channel). An exogenous shock could also heighten the rollover risk as investors suffering losses may find it difficult to obtain further financing opportunities, thereby affecting the aggregate demand (i.e., liquidity channel). A shock to the financial system in one country could also result in elevated risk premia in other countries, therefore affecting the aggregate demand through indirect channels (i.e., risk premium/confidence channel).

Figure 2.House Price Synchronicity and Transmission of External Shocks

Source: Authors’ illustration.

Even though housing is a non-tradable asset, Claessens et al. (2011a)—echoing past research such as Terrones (2004)—points out the presence of high synchronicity in their sample of countries, partially reflecting the importance of global factors such as global interest rates, U.S. business cycles, and global commodity prices. In the same spirit, Hirata et al. (2012) allude to the role of global integration of housing markets across advanced and emerging market economies as a determinant of house price synchronicity.3

Nevertheless, house price synchronicity may also be reflective of the co-movement in economic cycles (in other words, due to business cycle synchronicity). Claessens et al (2011b) notes that business cycles are highly synchronized with house price cycles. Indeed, past research has identified bilateral financial and trade linkages as two possible determinants of business cycle synchronicity between countries (IMF 2013; Kalemli-Ozcan et al. 2013a, 2013b; Duval et al 2016).

In this paper, building upon the literature on global financial conditions and house prices, we analyze the role of bilateral financial linkages and global financial conditions above-and-beyond that of business cycle synchronicity as a driver of house price synchronicity. We perform bilateral panel data analyses at country-pair level with nearly 50,000 observations and at major city-pair level with nearly 70,000 observations for a broader set of advanced and emerging economies (over 40 economies) and cities (over 70 cities) than previously analyzed. In particular, we aim to address the following questions: (1) Do global financial conditions amplify the house price synchronicity controlling for bilateral macro-financial linkages? (2) Is there an association between bilateral bank linkages and house price synchronicity above-and-beyond that of business cycle synchronicity? (3) What is the role of various institutional factors in either mitigating or amplifying the impact of global financial conditions on house price synchronicity? (4) Do policy tools such as macroprudential policies still turn out to be effective in addressing domestic vulnerabilities in the presence of heightened house price synchronicity?

Our main findings are fourfold. First, the importance of global factors in house price synchronicity as documented in past research still holds when a broader sample of countries and cities with coverage spanning through end-2016 is used. Notably, we find that abundant global liquidity as well as loose financial conditions (in addition to other global factors such as global interest rates) are positively associated with house price synchronicity across country-pairs as well as across major city-pairs. Thus, this paper sheds light on the important role played by mounting financial integration on housing markets across the globe. Second, we find that greater exchange rate flexibility attenuates the positive impact of global factors on house price synchronicity. Third, bilateral relationships such as past co-movement in business cycles and bilateral bank linkages are also positively associated with house price synchronicity. Finally, we find that the macroprudential policies aimed at tackling domestic vulnerabilities may have the additional impact of reducing countries’ house price synchronicity with the rest of the region and the world.

The rest of the paper is structured as follows. Section II describes the data and the construction of the main indicators used in the empirical analyses. Section III presents the main country-level empirical analysis and additional robustness checks. Section IV presents the city-level analysis where we first provide a network analysis on city-level interconnectedness dynamics followed by the empirical analysis. Section V extends the analysis further, looking at the impact of macroprudential policies on house price synchronicity. Section VI concludes.

II. Data and Measurement

This section presents a brief description of the construction of the main variables used in our regression analyses. Further information on underlying data sources, descriptions, and the economies and cities covered in this paper are presented in Annex I.

A. House Price Gap Synchronicity

We employ a measure of house price synchronicity that can be computed at any point in time (in other words at time-series level) rather than as period-wise computations; this measure also provides the additional advantage of not being bound between -1 and 1.

Synchronicity is calculated using the instantaneous quasi-correlation, originally presented by Morgan, Rime, and Strahan (2004) and used in recent business cycle literature (such as Duval et al. 2016; IMF 2013; Kalemli-Ozcan et al. 2013a, 2013b). 4 House price synchronicity (HPsynchijt) between country i and j at time t is measured as follows:

where HPgapit and HPgapjt stand for house price gap of country i and j respectively at quarter t and the gaps are measured as explained above. HPgapι¯andHPgapj¯ are the average house price gaps of countries i and j respectively, while σigap,σjgap the standard deviations of house piece gaps of countries i and j respectively.

House price gaps are measured by extracting the cyclical component of real house prices using the band-pass filter of Christiano and Fitzgerald (2003), with the maximum length of 30 years to capture medium-term financial cycles5. The above cyclical components of house prices are then taken as a ratio of the house price levels to obtain house price gaps6.

B. Business Cycle Synchronicity

Business cycle synchronicity (BCS) is analogous to the house price synchronicity measure presented above.

where Ygapit and Ygapjt represent output gaps of countries i and j respectively at quarter t and the gaps and measured using Christiano and Fitzgerald band-pass filter (2003), with the maximum length adjusted for business cycles instead of financial cycles. Ygapι¯andYgapj¯ are the average output gaps of countries i and j respectively, while σiqap,σjqap are the standard deviations of output gaps of countries i and j respectively.

C. Bilateral Banking Integration7

Banking integration is measured using bilateral locational banking statistics on residency basis obtained from BIS IBS restricted databases, to be conceptually consistent with balance of payments, national accounts, and external debt statistics. Bilateral banking integration is measured as the logarithm of the sum of bilateral claims of country i vis-à-vis country j and bilateral claims of country j vis-à-vis country i as a ratio of the sum of GDPs of country i and j8:

where Aijt is the bilateral claims of country i vis-à-vis country j at quarter t, Ajit is bilateral claims of country j vis-à-vis country i, GDPit is the nominal GDP of country i at time t, and GDPjt is the nominal GDP of country j at time t.

D. Global Financial Conditions

We control for the effect of global financial conditions on house price gap synchronicity as common shocks could propagate through global financial stability-related risks. In our main analyses, we focus on changes in Bank of International Settlements’ (BIS) global liquidity to capture global financial conditions. This measure captures the changes in banks’ cross-border claims denominated in all currencies plus local claims in foreign currency in percent of global GDP. In addition to global liquidity, as robustness checks, we also use global financial conditions index (FCI) and the U.S. FCI estimated in line with IMF (2017). 9 We also use Chicago Board Options Exchange volatility index (VIX), as well as Wu and Xia (2016) and Krippner (2013) U.S. shadow interest rates to capture global financial conditions in robustness specifications.

E. Other Controls

To further assess the impact of global financial conditions and bilateral bank linkages when countries have stronger institutions or when they are at different stages of economic development, we use several institutional characteristics and advanced/emerging market economy dummy variables. In particular, we use indicators for high capital account openness (measured using the Chinn-Ito index which is a de jure measure of financial openness), high exchange rate regime (measured using de facto exchange rate regime indices by Ilzetzki, Reinhart, and Rogoff 2017), and high financial openness (measured using the index developed by Lane and Milesi-Ferretti (2007), which is a de facto measure of financial openness) separately in specifications, where high is defined as a dummy variable that equals 1 when both countries in the country-pair are in the top fifth of the institutional characteristic during a given quarter. Dummy variables for advanced economies, emerging market economies, and advanced-emerging market economies take the value of 1 if both countries in the country-pair are either advanced economies, emerging market economies, or advanced-emerging market economies.

III. Country-Level Analysis

A. Empirical Strategy

This paper employs bilateral country-pair panel data analysis to estimate the impact of business cycle synchronicity, bilateral financial linkages, and global financial conditions on house price synchronicity at country-level10. Our baseline econometric specification presented below is estimated at quarterly frequency from 1990 to 2016, for 40 countries: 11

where HPsynchijt is the synchronicity of house price gaps between country-pair i and j at quarter t. BSCij denotes business cycle synchronicity between country i and j. FININTij refers to bilateral financial integration between country i and j.12GLOBALt is the global factor proxied by the changes in global liquidity. INSTij denote dummies which equal 1 if both countries have a high level of an institutional characteristic (i.e., economic development level, de jure capital account openness, exchange rate flexibility, or de facto financial account openness).13 All regressors are lagged by one quarter. In addition, linear and quadratic time trends (tr) are included. αij is the country-pair fixed effects capturing unobservable time-invariant idiosyncratic factors common to country-pair i and j such as geographic proximity. εijt is the error term.14 Importantly, country-pair fixed effects capture time-invariant supply-side and regulatory considerations that influence house price synchronicity between two countries.

B. Results

Impact of Global Financial Conditions on House Price Gap Synchronicity

In our main analyses, we estimate the impact of global financial conditions (also referred to as the global factor) on house price gap synchronicity using the changes in BIS’ global liquidity variable mentioned in the preceding section as the proxy for the global factor15 and instantaneous quasi-correlation (also mentioned in the previous section)16 as the synchronicity measure for house price and business cycle synchronicity. The results presented in Table 1 show that the global financial conditions are positively associated with house price synchronicity even when controlling for bilateral macro financial conditions including business cycle synchronicity and banking integration (column 4). This impact is also robust across various specifications, including where we control for different institutional characteristics and various error clustering methods are considered (see Tables 24 for robustness checks). This result could provide preliminary evidence for the positive association between the abundance of global liquidity and short-term co-movements in house price gaps.

Table 1.House Price Gap Synchronicity at Country Level and Global Factors
Dependent Variable: House Price Gap Synchronization of Country Pair i and j (quasi correlation)(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)
Business Cycle Synchronization of ij0.025*

(0.013)
0.030**

(0.014)
0.022

(0.014)
0.026*

(0.013)
0.026*

(0.013)
0.025*

(0.015)
0.026*

(0.014)
0.026*

(0.014)
0.026**

(0.013)
0.042

(0.033)
Bilateral Bank Integration of ij-0.011

(0.033)
0.012

(0.031)
0.012

(0.031)
0.011

(0.036)
0.022

(0.036)
0.022

(0.035)
0.012

(0.032)
-0.016

(0.034)
Global Factor (global liquidity)0.016**

(0.006)
0.016**

(0.008)
0.020**

(0.008)
0.019***

(0.007)
0.019**

(0.007)
0.018**

(0.007)
0.022*

(0.013)
Global Factor Interacted with:
x EMEs-EMEs Dummy-0.001

(0.009)
x EMEs-AEs Dummy0.000

(0.006)
x High Capital Account Openness with the World-0.002

(0.005)
x High Exchange Rate Regime (ij) (15 categories; high = more flexible)-0.023***

(0.008)
x High Exchange Rate Regime (ij) (6 categories; high = more flexible)-0.009

(0.007)
x High Financial Openness with the World (ij)0.003

(0.006)
GFC Period Dummy Interacted with:
x Business Cycle Synchronization of ij-0.032

(0.038)
x Bilateral Bank Integration of ij-0.022

(0.035)
x Global Factor-0.025*

(0.012)
Post-GFC Period Dummy Interacted with:
x Business Cycle Synchronization of ij-0.039

(0.035)
x Bilateral Bank Integration of ij0.010

(0.033)
x Global Factor-0.029

(0.018)
GFC Dummy-0.137**

(0.060)
Post-GFC Dummy-0.044

(0.052)
Observations65,45065,34349,38449,38449,38443,87146,70846,70847,35349,384
R-squared0.2270.3540.2510.2300.2300.2330.2240.2230.2410.232
Multiway ClusteringYesYesYesYesYesYesYesYesYesTwo-way
GroupAllAllAllAllAllAllAllAllAllAll
Time FE and Country-Pair FEYesYes
Time FE, Country-Pair FE, and country*time FEYes
Quadratic Trend and Country-Pair FEYesYesYesYesYesYes
Country-Pair FEYes
Source: Authors’ estimates.Note: GFC Dummy = a dummy variable that equals 1 during 2008–09, and zero otherwise. Post-GFC Dummy = a dummy variable that equals 1 during 2010–16, and zero otherwise. All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications 5 through 9, but are not shown above (specifically, dummy variables for EMEs-EMEs, EMEs-AEs, high capital account openness, high exchange rate regime, high financial openness are included in specifications 5 through 9, but not shown). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are three-way clustered (at country i, country j, and date), with the exception of regression (10), in which errors are two-way clustered (at country i, country j). AEs = advanced economies; EMEs = emerging market economies; FE = fixed effects; GFC = global financial crisis. *** p < 0.01; ** p < 0.05; * p < 0.1.
Source: Authors’ estimates.Note: GFC Dummy = a dummy variable that equals 1 during 2008–09, and zero otherwise. Post-GFC Dummy = a dummy variable that equals 1 during 2010–16, and zero otherwise. All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications 5 through 9, but are not shown above (specifically, dummy variables for EMEs-EMEs, EMEs-AEs, high capital account openness, high exchange rate regime, high financial openness are included in specifications 5 through 9, but not shown). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are three-way clustered (at country i, country j, and date), with the exception of regression (10), in which errors are two-way clustered (at country i, country j). AEs = advanced economies; EMEs = emerging market economies; FE = fixed effects; GFC = global financial crisis. *** p < 0.01; ** p < 0.05; * p < 0.1.
Table 2.House Price Gap Synchronicity at Country Level and Global Factors─Robustness Checks: Global Factors
Dependent Variable: House Price Gap Synchronization of Country Pair i and j (quasi-correlation)(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)
Global liquidityUS FCI (↑=loosening)Global FCI (↑=loosening)VIX (Inverse)US Shadow rate (Wu Xia)US Shadow rate (Krippner)
Business Cycle Synchronization of ij0.026*

(0.013)
0.026*

(0.014)
0.039***

(0.011)
0.043***

(0.012)
0.022*

(0.012)
0.023*

(0.013)
0.022***

(0.005)
0.024***

(0.005)
0.014***

(0.005)
0.014***

(0.005)
0.016***

(0.005)
0.016***

(0.005)
Bilateral Bank Integration of ij0.012

(0.031)
0.022

(0.036)
0.016

(0.033)
0.031

(0.033)
0.004

(0.032)
0.018

(0.032)
0.015

(0.029)
0.031

(0.030)
-0.003

(0.029)
0.002

(0.030)
0.003

(0.029)
0.007

(0.030)
Global Factor0.016**

(0.006)
0.019***

(0.007)
0.060***

(0.018)
0.076***

(0.020)
0.037**

(0.015)
0.049***

(0.018)
0.003**

(0.001)
0.005***

(0.001)
0.049***

(0.006)
0.057***

(0.006)
0.026***

(0.005)
0.032***

(0.005)
Global Factor Interacted with: x High Exchange Rate Regime (ij) (15 categories; high = more flexible)-0.023***

(0.008)
-0.069**

(0.034)
-0.056*

(0.033)
-0.011**

(0.004)
-0.025*

(0.014)
-0.026**

(0.011)
Observations49,38446,70849,38446,70848,89246,21649,38446,70849,38446,70849,38446,708
R-squared0.2300.2240.2280.2230.2290.2230.2250.2190.2340.2290.2290.224
ClusteringMulti-wayMulti-wayTwo-wayTwo-wayTwo-wayTwo-waycountry-paircountry-paircountry-paircountry-paircountry-paircountry-pair
Quadratic Trend and Country-Pair FEYesYesYesYesYesYesYesYesYesYesYesYes
Source: Authors’ estimates.Note: All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications, but are not shown above (specifically, dummy variables for high exchange rate regime). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are clustered as described above; FE = fixed effects.*** p < 0.01; ** p < 0.05; * p < 0.1.
Source: Authors’ estimates.Note: All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications, but are not shown above (specifically, dummy variables for high exchange rate regime). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are clustered as described above; FE = fixed effects.*** p < 0.01; ** p < 0.05; * p < 0.1.
Table 3.House Price Gap Synchronicity at Country Level and Global Factors─Robustness Checks: Additional Controls
Dependent Variable: House Price Gap Synchronization of Country Pair i and j (quasi-correlation)(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)
BaselineInterest rate synchronizationControlling for TradeControlling for Equity Synch
Business Cycle Synchronization of ij0.026*

(0.013)
0.026*

(0.014)
0.026*

(0.013)
0.026*

(0.014)
0.025*

(0.012)
0.024*

(0.014)
Interest Rate Synchronization of ij0.016

(0.040)
0.013

(0.041)
0.016**

(0.008)
0.013*

(0.008)
Bilateral Bank Integration of ij0.012

(0.031)
0.022

(0.036)
0.011

(0.033)
0.021

(0.036)
0.011

(0.011)
0.021*

(0.011)
0.009

(0.031)
0.019

(0.036)
0.015

(0.032)
0.024

(0.037)
Global Factor0.016**

(0.006)
0.019***

(0.007)
0.015**

(0.006)
0.018***

(0.006)
0.015***

(0.001)
0.018***

(0.001)
0.016**

(0.006)
0.019***

(0.007)
0.017**

(0.006)
0.019***

(0.007)
Global Factor Interacted with: x High Exchange Rate Regime (ij) (15 categories; high = more flexible)-0.023***

(0.008)
-0.024***

(0.008)
-0.024***

(0.004)
‐0.023***

(0.008)
-0.023***

(0.008)
Bilateral Trade Integration of ij0.008-0.001
Equity Return Synchronization of ij(0.038)(0.040)0.010

(0.009)
0.010

(0.009)
Observations49,38446,70847,83045,18847,83045,18848,89046,21548,97346,308
R-squared0.2300.2240.2280.2220.2280.2220.2320.2260.2310.225
ClusteringMulti-wayMulti-wayMulti-wayMulti-wayVCE robustVCE robustMulti-wayMulti-wayMulti-wayMulti-way
Quadratic Trend and Country-Pair FEYesYesYesYesYesYesYesYesYesYes
Source: Authors’ estimates.Note: All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications, but are not shown above (specifically, dummy variables for high exchange rate regime). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are clustered as described above; FE = fixed effects.*** p < 0.01; ** p < 0.05; * p < 0.1.
Source: Authors’ estimates.Note: All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications, but are not shown above (specifically, dummy variables for high exchange rate regime). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are clustered as described above; FE = fixed effects.*** p < 0.01; ** p < 0.05; * p < 0.1.
Table 4.House Price Gap Synchronicity at Country Level and Global Factors─Robustness Checks: Clustering of Standard Errors
Dependent Variable: House Price Gap Synchronization of Country Pair i and j (quasi-correlation)(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)
Business Cycle Synchronization of ij0.026*

(0.013)
0.026*

(0.014)
0.026**

(0.011)
0.026**

(0.012)
0.026**

(0.012)
0.026*

(0.013)
0.026**

(0.011)
0.026**

(0.012)
0.026***

(0.005)
0.026***

(0.005)
0.026***

(0.009)
0.026**

(0.010)
0.026***

(0.003)
0.026***

(0.003)
0.026***

(0.003)
0.026***

(0.003)
Bilateral Bank Integration of ij0.012

(0.031)
0.022

(0.036)
0.012

(0.030)
0.022

(0.034)
0.012

(0.027)
0.022

(0.030)
0.012

(0.036)
0.022

(0.038)
0.012

(0.029)
0.022

(0.030)
0.012

(0.017)
0.022

(0.018)
0.012

(0.011)
0.022**

(0.011)
0.012

(0.011)
0.022**

(0.011)
Global Factor (global liquidity)0.016**

(0.006)
0.019***

(0.007)
0.016***

(0.005)
0.019***

(0.006)
0.016**

(0.006)
0.019***

(0.006)
0.016***

(0.005)
0.019***

(0.005)
0.016***

(0.002)
0.019***

(0.002)
0.016***

(0.004)
0.019***

(0.004)
0.016***

(0.001)
0.019***

(0.001)
0.016***

(0.001)
0.019***

(0.001)
Global Factor Interacted with: x High Exchange Rate Regime (ij) (15 categories; high = more flexible)-0.023***

(0.008)
-0.023**

(0.009)
-0.023***

(0.007)
-0.023***

(0.008)
-0.023***

(0.007)
-0.023***

(0.003)
-0.023***

(0.004)
-0.023***

(0.004)
Observations49,38446,70849,38446,70849,38446,70849,38446,70849,38446,70849,38446,70849,38446,70849,38446,708
R-squared0.2300.2240.2300.2240.2300.2240.2300.2240.2300.2240.2300.2240.2300.2240.2300.224
ClusteringCtr1 ctr2 timeCtr1 ctr2 timeCtr1 ctr2Ctr1 ctr2ctr1 timectr1 timectr2 timectr2 timecountry-paircountry-pairtimetimeVCE robustVCE robustno clusteringno clustering
Quadratic Trend and Country-Pair FEYesYesYesYesYesYesYesYesYesYesYesYesYesYesYesYes
Source: Authors’ estimates.Note: All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications, but are not shown above (specifically, dummy variables for high exchange rate regime). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are clustered as described above; FE = fixed effects.*** p < 0.01; ** p < 0.05; * p < 0.1.
Source: Authors’ estimates.Note: All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications, but are not shown above (specifically, dummy variables for high exchange rate regime). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are clustered as described above; FE = fixed effects.*** p < 0.01; ** p < 0.05; * p < 0.1.

Moreover, the impact of the global financial conditions on house price synchronicity appears to be higher between advanced economies than in country-pairs that are emerging market economies (column 5). While the impact of the global financial conditions in advanced economies is statistically significant and positive, neither emerging market economies’ nor advanced-emerging market economy-pairs’ impact is statistically significant at conventional levels when standard errors are clustered in the most stringent manner.

Institutional characteristics such as higher exchange rate flexibility appear to be attenuating the positive association between global financial conditions and house price synchronicity (column 7). This impact is statistically significant at 1 percent confidence interval. Moreover, it is robust to various controls, as presented in Tables 25. We also find an attenuating effect of de jure financial openness (i.e., Chinn-Ito index of capital account openness) on the global financial conditions’ impact on house price synchronicity, but the impact of this interaction term is not statistically significant at conventional levels (column 6). Results in columns 4 to 6 are also presented in figure 3, where we have standardized the coefficients for comparability across specifications.

Table 5.House Price Gap Synchronicity at City Level and Global Factors─Two-Way Clustering
Dependent Variable: House Price Gap Synchronization of City Pair i and j (quasi-correlation)(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)
Business Cycle Synchronization of ij0.011

(0.010)
0.021*

(0.012)
0.011

(0.010)
0.019*

(0.010)
0.019*

(0.010)
0.016

(0.010)
0.018

(0.011)
0.017

(0.011)
0.016

(0.010)
0.079***

(0.027)
Bilateral Bank Integration of ij0.008

(0.045)
0.016

(0.045)
0.019

(0.045)
0.020

(0.050)
0.021

(0.046)
0.020

(0.046)
0.021

(0.047)
0.066

(0.057)
Global Factor (global liquidity )0.018***

(0.005)
0.030***

(0.008)
0.012**

(0.005)
0.021***

(0.006)
0.023***

(0.006)
0.018***

(0.005)
0.024**

(0.009)
Global Factor Interacted with:
x EMEs-EMEs Dummy-0.019

(0.011)
x EMEs-AEs Dummy-0.018**

(0.008)
x High Capital Account Openness with the World0.018**

(0.008)
x High Exchange Rate Regime (ij) (15 categories; high = more flexible)-0.017**

(0.008)
x High Exchange Rate Regime (ij) (6 categories; high = more flexible)-0.012*

(0.007)
x High Financial Openness with the World (ij)0.001

(0.016)
GFC Period Dummy Interacted with:
x Business Cycle Synchronization of ij-0.081***

(0.029)
x Bilateral Bank Integration of ij-0.058

(0.049)
x Global Factor-0.025**

(0.010)
Post-GFC Period Dummy Interacted with:
x Business Cycle Synchronization of ij-0.078***

(0.029)
x Bilateral Bank Integration of ij-0.071*

(0.040)
x Global Factor-0.026**

(0.011)
GFC Dummy0.010

(0.038)
Post- GFC Dummy0.014

(0.049)
Observations66,57566,57266,57566,57566,57559,35363,69163,69162,58866,575
R-squared0.2600.3430.2600.2540.2560.2650.2510.2520.2680.260
ClusteringTwo-wayTwo-wayTwo-wayTwo-wayTwo-wayTwo-wayTwo-wayTwo-wayTwo-wayTwo-way
Time FE and Country-Pair FEYesYes
Time FE, Country-Pair FE, and country*time FEYes
Quadratic Trend and Country-Pair FEYesYesYesYesYesYes
Country-Pair FEYes
Source: Authors’ estimates.Note: GFC Dummy = a dummy variable that equals 1 during 2008–09, and zero otherwise. Post-GFC Dummy = a dummy variable that equals 1 curing 2010–16, and zero otherwise. All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications 5 through 9, but are not shown above (specifically, dummy variables for EMEs-EMEs, EMEs-AEs, high capital account openness, high exchange rate regime, high financial openness are included in specifications 5 through 9, but not shown). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are two-way clustered (at country ij, and date). AEs = advanced economies; EMEs = emerging market economies; FE = fixed effects; GFC = global financial crisis.*** p < 0.01; ** p < 0.05; * p < 0.1.
Source: Authors’ estimates.Note: GFC Dummy = a dummy variable that equals 1 during 2008–09, and zero otherwise. Post-GFC Dummy = a dummy variable that equals 1 curing 2010–16, and zero otherwise. All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications 5 through 9, but are not shown above (specifically, dummy variables for EMEs-EMEs, EMEs-AEs, high capital account openness, high exchange rate regime, high financial openness are included in specifications 5 through 9, but not shown). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are two-way clustered (at country ij, and date). AEs = advanced economies; EMEs = emerging market economies; FE = fixed effects; GFC = global financial crisis.*** p < 0.01; ** p < 0.05; * p < 0.1.

Furthermore, the positive impact of global liquidity on house price synchronicity was substantially higher prior to the global financial crisis (GFC). This may provide evidence to the association between the global house price boom that occurred preceding the GFC and the abundance of global liquidity accumulated during that period.

Figure 3.Impact of Global Financial Conditions on House Price Synchronization

Source: Authors’ estimates.

Note: Global financial conditions are proxied by the BIS global liquidity variable mentioned in the previous section. Synchronicity is measured by the quasi correlation of gaps. Shaded bars denote joint significance of the F-test at or above 90 percent. Patterned bars denote interaction terms that are statistically significant. Coefficients are standardized. Standard deviation of the country-level dependent variable is approximately 0.85 (see Annex Table 1.3). AEs = advanced economies; EMEs = emerging market economies; FX = exchange rate.

The analysis concerning the impact of bilateral linkages on house price gap synchronicity is presented in Annex II. Using an alternative measure of house price synchronicity, which captures the medium-term dynamics through differences in house price gaps, we find evidence that both business cycle synchronicity and bilateral banking integration are positively and robustly associated with house price synchronicity.

C. Robustness Checks

In addition to the results presented above, various robustness checks were performed, with the main findings broadly unchanged. For instance, alternative proxies for global financial conditions including the U.S. financial conditions index (FCI), Global FCI, CBOE volatility index (VIX), U.S. shadow interest rates (Wu and Xia 2016; Krippner 2013) are used, where the global financial conditions and the high exchange rate regime interaction terms are still found to be statistically significant with the coefficient sign and the size broadly unchanged (Table 2).17 Specifications above were also estimated by replacing BCS with interest rate synchronicity to investigate the role of synchronized monetary policies in contributing to house price synchronicity. We find interest rate synchronicity to be a statistically significant driver of house price synchronicity on its own when either synchronicity measure is used (either synch1 or quasi correlation). However, the statistical significance of interest rate synchronicity above and beyond other financial factors such as the global liquidity and bilateral banking linkages is only robust to less stringent manners of standard error clustering (Table 3, columns 3–6). At the same time, trade integration was included as an additional control, but found not to be statistically significant (Table 3, columns 7–8). When equity price synchronicity is included as an additional control, the main results presented in the previous section remain broadly unchanged (Table 3, columns 9–10). However, equity price synchronicity itself does not consistently have a statistically significant relationship with house price synchronicity.

Various clustering alternatives were employed (clustering at country-pair level, two-way at country i and country j, two-way at country-pair and time level, and without clustering, Huber/White/sandwich estimator), and as expected, the level of significance improves under less restrictive clustering options (Table 4). Additional time controls, such as year fixed effects and linear time trends, were also considered with little changes to the main conclusions. Finally, further robustness checks were employed by dropping one country-pair at a time as well.

IV. City-Level Analysis

While house prices synchronicity may vary among country-pairs owing to their degree of exposure to bilateral linkages and global financial conditions as identified in the preceding section, house prices in major cities18 may move in tandem due to increasing global presence even if their country-level house prices may not portray such dynamics. To dig deeper into city-level house price synchronicity, we first explore house price interconnectedness dynamics through a network analysis, and then move on to analyzing the drivers of city-level house price synchronicity empirically.

A. Network Analysis: House Price Interconnectedness at City Level

Our network analysis uses the spillovers approach developed by Diebold and Yilmaz (2014) (see Annex III for detailed methodology) controlling for global financial conditions (proxied by the U.S. FCI). In fact, comparing the network analysis at country-level and city-level confirms that cities that are attractive to global investors may be at the core of the network and closer to other cities such as financial centers even if the respective countries are at the periphery (Figure 4). For instance, Tokyo and Rome are centrally located in the vicinity of global financial centers such as New York and London in the city-level network map below (Figure 4, Panel 2), while Japan and Italy are located at the periphery of the country-level network map (Figure 4, Panel 1).

Figure 4.House Price Interconnectedness Among Countries vs. Cities

Source: Authors’ estimates.

Note: The figure is based on a vector autoregression of country-level/city-level house price growth rates (quarter over quarter) controlling for global factors, spanning 1990:Q1 to 2016:Q4 for country-level and 2004:Q1 to 2017:Q2 for city-level. For methodology details, see Annex III. See the footnote 18 for city selection criteria, conditional on data availability. Node size is based on the city’s total outward spillovers. Pink nodes represent advanced economies and gray nodes represent emerging market economies. Arrows’ thickness is based on link distribution. Only links above the 50th percentile for country-level and 66th percentile for city-level are considered. The figure layout is based on the algorithm by Fruchterman and Reingold (1991), and plotted using the “qgraph” R package. Ack = Auckland; Ams = Amsterdam; Bgt = Bogotá; Brl = Berlin; Brs = Brussels; Dbl = Dublin; Dub = Dubai; HKG = Hong Kong SAR; Hls = Helsinki; Jkr = Jakarta; Lim = Lima; Lnd = London; Mdr = Madrid; Mmb = Mumbai; Mnl = Manila; Msc = Moscow; Mxc = Mexico City; NYC = New York City; Osl = Oslo; Prs = Paris; Rom = Rome; Sel = Seoul; SGP = Singapore; Shn = Shanghai; Snt = Santiago; Stc = Stockholm; Syd = Sydney; Tky = Tokyo; Trn = Toronto; Vnn = Vienna. Following Morgan Stanley Capital International markets classification criteria, Korea (and thus Seoul) is classified as an emerging market economy; moreover, Korea (and thus Seoul) was not classified as an advanced economy in the IMF’s World Economic Outlook country classification at the beginning of our sample period, which starts in 1990.

B. Empirical Strategy

The determinants of city-level house price synchronicity are analyzed using a bilateral panel data analysis, where we specifically estimate the impact of country-level measures such as business cycle synchronicity and bilateral financial linkages, and global financial conditions on house price synchronicity within major cities. The analysis is estimated at quarterly frequency from 2004 to 2016 for over 70 major cities19. The econometric specification for the city-level analysis takes the following form:

where HPsynchijt is the synchronicity of house price gaps between city-pair i and j at quarter t. αij stands for city-pair fixed effects and tr stands for quadratic and linear time trends. GLOBALt-1stands for global financial conditions proxied by changes in the BIS’ global liquidity in percent of global GDP. All other regressors are country-level variables that are defined in the section on the country-level analysis.

C. Results

Similar to our country-level analysis, we use the changes in BIS global liquidity to proxy for the global financial conditions and instantaneous quasi-correlation to measure city-level house price gap synchronicity. We find that global financial conditions are positively associated with city-level house price gap synchronicity; this impact is statistically significant even if standard errors are clustered using a more stringent form of multi-way clustering, while the significance level improves from a 10 percent confidence level to a 1 percent confidence level if two-way clustering is employed instead (column 4 in Tables 5 and 6).

Table 6.House Price Gap Synchronicity at City Level and Global Factors─Multi-Way Clustering
Dependent Variable: House Price Gap Synchronization of City Pair i and j (quasi-correlation)(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)
Business Cycle Synchronization of ij0.011

(0.013)
0.021

(0.013)
0.011

(0.013)
0.019

(0.012)
0.019

(0.012)
0.016

(0.014)
0.018

(0.014)
0.017

(0.013)
0.016

(0.013)
0.079

(0.060)
Bilateral Bank Integration of ij0.008

(0.029)
0.016

(0.031)
0.019

(0.030)
0.020

(0.046)
0.021

(0.037)
0.020

(0.037)
0.021

(0.038)
0.066

(0.051)
Global Factor (global liquidity)0.018*

(0.009)
0.030*

(0.016)
0.012

(0.008)
0.021*

(0.011)
0.023*

(0.011)
0.018*

(0.009)
0.024*

(0.014)
Global Factor Interacted with:
x EMEs-EMEs Dummy-0.019

(0.016)
x EMEs-AEs Dummy-0.018

(0.014)
x High Capital Account Openness with the World0.018

(0.014)
x High Exchange Rate Regime (ij) (15 categories; high = more flexible)-0.017**

(0.008)
x High Exchange Rate Regime (ij) (6 categories; high = more flexible)-0.012

(0.008)
x High Financial Openness with the World (ij)0.001

(0.024)
GFC Period Dummy Interacted with:
x Business Cycle Synchronization of ij-0.081

(0.062)
x Bilateral Bank Integration of ij-0.058

(0.073)
x Global Factor-0.025*

(0.014)
Post-GFC Period Dummy Interacted with:
x Business Cycle Synchronization of ij-0.078

(0.062)
x Bilateral Bank Integration of ij-0.071

(0.058)
x Global Factor-0.026

(0.017)
GFC Dummy0.010

(0.050)
Post- GFC Dummy0.014

(0.045)
Observations66,57566,57266,57566,57566,57559,35363,69163,69162,58866,575
R-squared0.2600.3430.2600.2540.2560.2650.2510.2520.2680.260
ClusteringMulti- wayMulti-wayMulti-wayMulti-wayMulti-wayMulti-wayMulti-wayMulti-wayMulti-wayMulti-way
Time FE and Country-Pair FEYesYes
Time FE, Country-Pair FE, and country*time FEYes
Quadratic Trend and Country-Pair FEYesYesYesYesYesYes
Country-Pair FEYes
Source: Authors’ estimates.Note: GFC Dummy = a dummy variable that equals 1 during 2008–09, and zero otherwise. Post-GFC Dummy = a dummy variable that equals 1 curing 2010–16, and zero otherwise. All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications 5 through 9, but are not shown above (specifically, dummy variables for EMEs-EMEs, EMEs-AEs, high capital account openness, high exchange rate regime, high financial openness are included in specifications 5 through 9, but not shown). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are three-way clustered (at country i, country j, and date). AEs = advanced economies; EMEs = emerging market economies; FE = fixed effects; GFC = global financial crisis.*** p < 0.01; ** p < 0.05; * p < 0.1.
Source: Authors’ estimates.Note: GFC Dummy = a dummy variable that equals 1 during 2008–09, and zero otherwise. Post-GFC Dummy = a dummy variable that equals 1 curing 2010–16, and zero otherwise. All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications 5 through 9, but are not shown above (specifically, dummy variables for EMEs-EMEs, EMEs-AEs, high capital account openness, high exchange rate regime, high financial openness are included in specifications 5 through 9, but not shown). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are three-way clustered (at country i, country j, and date). AEs = advanced economies; EMEs = emerging market economies; FE = fixed effects; GFC = global financial crisis.*** p < 0.01; ** p < 0.05; * p < 0.1.

In line with our country-level findings, city-level analysis also confirms that higher exchange rate flexibility tends to be attenuating the positive association between the global factor and the city-level house price synchronicity; this impact is statistically significant at a 5 percent confidence level even when more stringent form of standard error clustering is used (column 7 in Tables 5 and 6).

Furthermore, the impact of global financial conditions on city-level house price synchronicity is higher among city-pairs residing within advanced economies than that of city-pairs residing either within emerging economies or advance-emerging economy pairs (column 5 in Tables 5 and 6). While the impact for advanced economies is statistically significant even when more stringent forms of standard errors are used, the interaction term for advance-emerging pairs is significant only when a less stringent form of clustering is used (such as two-way clustering at country-pair and time level; column 5 in Table 5). The interaction term for emerging economies is not statistically significant when two-way clustering is used.

In contrast to our country-level analysis, the city-level empirical findings suggest that greater financial openness at country-level tends to amplify the positive association between global financial conditions and city-level house price synchronicity. In other words when a de jure measure of financial openness is used (i.e., Chinn-Ito index of capital account openness). However, we find that this impact is not statistically significant if standard errors are clustered in a more stringent manner (column 6 in Table 6). We fail to find statistically significant results when a de facto measure of financial openness is used (i.e., Lane and Milesi-Ferretti (2007) measure of financial openness).

The city-level analysis also confirms that the global financial conditions were positively associated with city-level house price synchronicity prior to the global financial crisis (column 10 in Tables 5 and 6).

V. Extensions: The Impact of Macroprudential Policies

In this section, we focus on the relationship between macroprudential policies (MPPs) and house price synchronicity with regional and global cycles.20 MPPs targeted at dampening the accumulation of domestic vulnerabilities in the financial and housing sectors may have indirect effects of weakening the correlation of house price cycles, thereby leaving room for policymakers to regain control over local house price dynamics.

Macroprudential tools, which have been used more actively since the global financial crisis (Alam et al. 2018; Cerutti, Claessens, and Laeven 2015), aim at curbing leverage and reducing financial vulnerabilities in order to decrease the likelihood of domestic asset bubbles and financial crises. MPPs are usually domestically targeted, with a large share of measures focused on domestic credit and housing market conditions. However, in countries experiencing deeper financial integration and where business cycles are more intertwined at the regional and global levels, house prices are, in part, driven by other factors, such as capital flows from global investors and by global financial conditions.21 Thus, the relationship between macroprudential tools and house price synchronicity might be ambiguous because it may be offset by other factors.

Recent empirical literature (Vandenbussche, Vogel, and Detragiache 2015; Cerutti, Dagher, and Dell’Ariccia 2015) suggests that the role of macroprudential policies in mitigating house prices is less clear and may vary according to policy type. For instance, measures targeting housing finance (Akinci and Olmstead-Rumsey 2017) and those that complement monetary policy (Bruno, Shim, and Shin 2017) seem to be most effective in mitigating house price growth. In contrast, there is no robust evidence for policies such as risk-weighting and provisioning requirements (Kuttner and Shim 2016).

A. Empirical Strategy

The analysis gauges the effectiveness of macroprudential tools in reducing house price synchronicity across 41 countries from the second quarter of 1990 through the last quarter of 2016. More specifically, the following panel regression specification is estimated, with i denoting country and t representing quarter:

where αi denotes country fixed effects. The dependent variable HPS refers to house price cycle synchronicity (instantaneous quasi-correlation) with either the regional or the global cycle. PCS is business cycle synchronicity with the region or the rest of the world. X is a vector of controls (including global financial conditions, financial integration with the region or the world, and institutional characteristics). MPP is a macroprudential tool (such as limits to loan-to-value ratios or debt-to-income ratios, or fiscal-based measures that include sellers’ and buyers’ stamp duty taxes) or a macroprudential group index (such as loan-targeted, supply-side [capital, general, loans], or demand-side tools).22

B. Results

House price growth evolved differently after the adoption of demand-side MPPs such as loan-to-value (LTV) limits, depending on the level of synchronicity (Figure 5). Before the adoption of these policies, house prices grew similarly in countries with high or low house price synchronicity. Following the adoption of MPPs, house price growth declined in both groups of countries, but the decline was stronger and more sustained in low-synchronicity countries. These simple patterns suggest that policymakers may have more control over the dynamics of the housing markets in low-synchronicity countries. At the same time, it suggests that a high degree of synchronicity does not render MPPs ineffective. This could be the case if the financial factors behind house price synchronicity operate, at least partially, through local financial intermediaries.

Figure 5.Average House Price Growth and Demand-side Macroprudential Policies

Source: Authors’ estimates.

Note: The figure depicts the average year-over-year house price growth for high-synchronicity and low-synchronicity countries within a period of plus or minus five quarters around the implementation of demand-side macroprudential policies (MPPs). Demand-side MPPs include limits to debt-service-to-income and loan-to-value (LTV) ratios. Total number of demand-side events is 47, and t = 0 is identified as the first quarter in which demand-side MPPs were implementated within the plus-or-minus-five-quarter window. Synchronicity is based on the quasi-correlation of house price gaps with the global cycle. A country is classified in the high-synchronicity group when its average syncronicity (over the sample period) with the global cycle is above the 50th percentile in the sample, and vice versa.

MPPs are also associated with a reduction in house price synchronicity (Figure 6, Panel 1 and Annex Table 4.1); in fact, tighter macroprudential tools targeting bank capital and credit conditions are found to be associated with lower house price synchronicity. Since these tools mostly affect local financial intermediaries and domestic demand, this finding also suggests that factors driving house price co-movement operate, to some degree, through these channels. The relationship between capital-based measures, which include countercyclical capital buffers, and house price synchronicity seems the most highly negative. Likewise, loan-targeted measures, including LTV limits, and supply-side loan-targeted tools, such as limits on foreign currency, are found to lessen correlations with the global and regional house price cycles. The adoption of fiscal-based measures, such as ad valorem and buyer’s stamp duty taxes that could potentially deter global investors from engaging in speculative real estate purchases is also associated with a decline in synchronicity, but to a lesser extent than other MPPs.23 When looking only at periods with credit booms, the results are both qualitatively and quantitatively similar, although the relationships are slightly less significant (Figure 6, panel 2 and Annex Table 4.2).

Figure 6.Impact of Macroprudential Measures on House Price Synchronicity

Source: Authors’ estimates.

Note: Figure depicts estimated average effects of macroprudential tools on house price synchronicity with the regional cycle (green) and global cycle (red). Shaded bars show statistically significant standardized coefficients, at the 10 percent confidence level. Estimated panel regressions use data for 41 countries (panel 1) spanning over 1990:Q2 – 2016:Q4 period. Regressions control for business cycle synchronicity, financial integration, and global financial conditions. All regressors are lagged one quarter. Supply side (loans) consists of limits on credit growth, loan loss provisions, loan restrictions, and limits on foreign currency loans. Supply side (capital) consists of capital requirements, conservation buffers, the leverage ratio, and the countercyclical capital buffer. Supply side (general) consists of reserve requirements, liquidity requirements, and limits on foreign exchange positions. Demand-side includes limits to debt-service-to-income and LTV ratios. All loans measures include demand side and supply side (loans). Fiscal-based measures include taxes such as ad valorem, seller’s and buyer’s stamp duty, or other taxes.

VI. Conclusions

Using various proxies for global financial conditions, this paper confirms that the abundance of liquidity owing to accommodative financial conditions is positively associated with house price synchronicity at country and city levels. While higher house price synchronicity may benefit countries in some cases, positive association with global financial conditions could also suggest a stronger transmission of external shocks into the domestic economy or to major cities within an economy. Moreover, house price synchronicity dynamics among major cities may vary from that of their respective countries’ owing to the attractiveness of these cities to global investors. Our analysis also finds that the positive association of global financial conditions with house price synchronicity was stronger preceding the global financial crisis.

Countries with more flexible exchange rate regimes, on average, may possess the ability to attenuate the positive impact of global financial conditions on house price synchronicity. Moreover, our empirical analysis suggests that major cities located in countries with more flexible exchange rate regimes possess the ability of attenuating the impact of global financial conditions on city-level house price synchronicity as well.

Finally, we find that house price growth in countries that experience lower house price synchronicity with the rest of the world, on average, are more sensitive to macroprudential policies that are aimed at reducing domestic vulnerabilities, compared to high synchronicity countries. However, our empirical analysis suggests that macroprudential policies intended at addressing domestic vulnerabilities also possess the unintended effect of reducing house price synchronicities, thereby allowing policymakers to regain partially control over local house price dynamics.

References

    AkinciOzge andJaneOlmstead-Rumsey. 2017. “How Effective Are Macroprudential Policies? An Empirical Investigation.” Journal of Financial Intermediation33 (January): 3357.

    • Search Google Scholar
    • Export Citation

    AlamZohairAdrianAlterJesseEisemanGastonGelosHeedonKangMachikoNaritaErlendNier andNaixiWang. 2018. “Digging Deeper—Evidence on the Effects of Macroprudential Policies from a New Database.” IMF Working PaperInternational Monetary FundWashington, DC.

    • Search Google Scholar
    • Export Citation

    AllenFranklin andDouglasGale. 2000. “Financial Contagion.” Journal of Political Economy108 (1): 133.

    BadarinzaCristian andTarunRamadorai. 2018. “Home away from Home? Foreign Demand and London House Prices.” Journal of Financial EconomicsJuly (26): 2134.

    • Search Google Scholar
    • Export Citation

    BernankeBenCarol C.BertautLaurieDemarco andSteven B.Kamin. 2011. “International Capital Flows and the Return to Safe Assets in the United States, 2003–2007.” FRB International Finance Discussion Paper 1014Federal Reserve BoardWashington, DC.

    • Search Google Scholar
    • Export Citation

    BrunoValentinaIlhyockShim andHyunSong Shin. 2017. “Comparative Assessment of Macroprudential Policies.” Journal of Financial Stability28: 183202.

    • Search Google Scholar
    • Export Citation

    CameronA.ColinJonah B.Gelbach andDouglas L.Miller. 2011. “Robust Inference with Multi-Way Clustering.” Journal of Business Economics29: 23849.

    • Search Google Scholar
    • Export Citation

    CeruttiEugenioStijnClaessens andLucLaeven. 2015. “The Use and Effectiveness of Macroprudential Policies: New Evidence.” Journal of Financial Stability28 (C): 20324.

    • Search Google Scholar
    • Export Citation

    CeruttiEugenioJihadDagher andGiovanniDell’Ariccia. 2017. “Housing Finance and Real-Estate Booms: A Cross-Country Perspective.” Journal of Housing Economics38 (C): 113.

    • Search Google Scholar
    • Export Citation

    ChinnMenzieD. andHiroIto. 2006. “What Matters for Financial Development? Capital Controls, Institutions, and Interactions.” Journal of Development Economics81 (1): 16392.

    • Search Google Scholar
    • Export Citation

    ChristianoLawrenceJ. andTerry J.Fitzgerald. 2003. “The Band Pass Filter.” International Economic Review44 (2): 43565.

    ClaessensStijnM.Ayhan Kose andMarco E.Terrones. 2008. “What Happens during Recessions, Crunches and Busts?IMF Working Paper 08/274International Monetary FundWashington, DC.

    • Search Google Scholar
    • Export Citation

    ClaessensStijnM.Ayhan Kose andMarco E.Terrones. 2011a. “Financial Cycles: What? How? When?IMF Working Paper 11/76International Monetary FundWashington, DC.

    • Search Google Scholar
    • Export Citation

    ClaessensStijnM.Ayhan Kose andMarco E.Terrones. 2011b. “How Do Business and Financial Cycles Interact?IMF Working Paper 11/88International Monetary FundWashington, DC.

    • Search Google Scholar
    • Export Citation

    Cushman & Wakefield. 2017. “Winning in Growth Cities.” December 19. https://cushmanwakefield.turtl.co/story/59d48cf5d1c76a6148cabb4f.

    • Search Google Scholar
    • Export Citation

    DemirerMertFrancis X.DieboldLauraLiu andKamilYilmaz. 2018. “Estimating global bank network connectedness.” Journal of Applied Econometrics33 (1): 115.

    • Search Google Scholar
    • Export Citation

    DieboldFrancisX. andKamilYilmaz. 2014. “On the Network Topology of Variance Decompositions: Measuring the Connectedness of Financial Firms.” Journal of Econometrics182 (1): 11934.

    • Search Google Scholar
    • Export Citation

    DuvalRomainNanLiRichaSaraf andDulaniSeneviratne. 2016. “Value-Added Trade and Business Cycle Synchronization.” Journal of International Economics99 (March): 25162.

    • Search Google Scholar
    • Export Citation

    FruchtermanThomasM. J. andEdward M.Reingold. 1991. “Graph Drawing by Force-Directed Placement.” Software: Practice and Experience21 (11): 112964.

    • Search Google Scholar
    • Export Citation

    HirataHideakiM.Ayhan KoseChristopherOtrok andMarco E.Terrones. 2012. “Global house price fluctuations: Synchronization and determinants.” Working Paper No. w18362National Bureau of Economic Research.

    • Search Google Scholar
    • Export Citation

    HodrickRobertJ. andEdwardPrescott. 1997. “Postwar U.S. Business Cycles: An Empirical Investigation.” Journal of Money Credit and Banking29 (1): 116.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF). 2013. “Dancing Together? Spillovers, Common Shocks, and the Role of Financial and Trade Linkages.” World Economic OutlookOctoberWashington, DC.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF). 2016. “The Growing Importance of Financial Spillovers from Emerging Market Economies.” Global Financial Stability ReportAprilWashington, DC.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF). 2017. “Financial Conditions and Growth at Risk.” Global Financial Stability ReportOctoberWashington, DC.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF). 2018a. “House Price Synchronization: What Role for Financial Factors?Global Financial Stability ReportAprilWashington, DC.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF). 2018b. “People’s Republic of China––Hong Kong Special Administrative Region: Selected Issues.” IMF Country Report 18/17International Monetary FundWashington, DC.

    • Search Google Scholar
    • Export Citation

    IlzetzkiEthanCarmen M.Reinhart andKenneth S.Rogoff. 2017. “Exchange Rate Arrangements Entering the 21st Century: Which Anchor Will Hold?NBER Working Paper 23134National Bureau of Economic ResearchCambridge, MA.

    • Search Google Scholar
    • Export Citation

    JordàÒscaKatharinaKnollDmitryKuvshinovMoritzSchularick andAlanTaylor. 2017. “The Rate of Return on Everything, 1870–2015.” NBER Working Paper 24112National Bureau of Economic ResearchCambridge, MA.

    • Search Google Scholar
    • Export Citation

    Kalemli-ÖzcanSebnemEliasPapaioannou andFabrizioPerri. 2013a. “Global Banks and Crisis Transmission.” Journal of International Economics89 (2): 495510.

    • Search Google Scholar
    • Export Citation

    Kalemli-ÖzcanSebnemEliasPapaioannou andJosé-LuisPeydró. 2013b. “Financial Regulation, Financial Globalization, and the Synchronization of Economic Activity.” Journal of Finance68 (3): 1179228.

    • Search Google Scholar
    • Export Citation

    KatagiriMitsuru. 2018. “House Price Synchronization and Financial Openness: A Dynamic Factor Model Approach.” IMF Working Paper 18/209International Monetary FundWashington, DC.

    • Search Google Scholar
    • Export Citation

    KrippnerLeo. 2013. “Measuring the Stance of Monetary Policy in Zero Lower Bound Environments.” Economics Letters118 (1): 13538.

    • Search Google Scholar
    • Export Citation

    KuttnerKennethN. andIlhyockShim. 2016. “Can Non-Interest Rate Policies Stabilize Housing Markets? Evidence from a Panel of 57 Economies.” Journal of Financial Stability26 (October): 3144.

    • Search Google Scholar
    • Export Citation

    LandierAugustinDavidSraer andDavidThesmar. 2017. “Banking Integration and House Price Co-movement.” Journal of Financial Economics125 (1): 125.

    • Search Google Scholar
    • Export Citation

    LanePhilip andGian-MariaMilesi-Ferretti. 2007. “The External Wealth of Nations Mark II: Revised and Extended Estimates of Foreign Assets and Liabilities, 1970–2004.” Journal of International Economics73 (2): 22350.

    • Search Google Scholar
    • Export Citation

    LongstaffFrancis. 2010. “The Subprime Credit Crisis and Contagion in Financial Markets.” Journal of Financial Economics97: 436450.

    • Search Google Scholar
    • Export Citation

    MorganDonaldBertrandRime andPhilipStrahan. 2004. “Bank Integration and State Business Cycles.” Quarterly Journal of Economics119 (4): 155585.

    • Search Google Scholar
    • Export Citation

    SongSong andPeter J.Bickel. 2011. “Large Vector Auto Regressions.” Working Paper SFB649DP2011–048 Humboldt University Collaborative Research Center 649.

    • Search Google Scholar
    • Export Citation

    TerronesMarco. 2004. “The Global House Price BoomIMF World Economic OutlookSeptember pp.71136.

    WuJing Cynthia andFanDora Xia. 2016. “Measuring the Macroeconomic Impact of Monetary Policy at the Zero Lower Bound.” Journal of Money Credit and Banking48 (2–3): 25391.

    • Search Google Scholar
    • Export Citation

    VandenbusscheJeromeUrsulaVogel andEnricaDetragiache. 2015. “Macroprudential Policies and Housing Prices: A New Database and Empirical Evidence for Central, Eastern, and Southeastern Europe.” Journal of Money Credit and Banking47 (S1): 34377.

    • Search Google Scholar
    • Export Citation
Annex I: Data Sources, Coverage, and Summary Statistics
Annex Table 1.1.Data Sources
VariableDescriptionSource
Country-Level Variables
Real House Price IndicesResidential property prices (seasonally adjusted) at country level (also at city level)Bank for International Settlements; CEIC Data Co. Ltd; Emerging Markets Econom Data Ltd; Global Financial Data Solutions; Global Property Guide; Haver Analytics IMF, Research Department house price dataset; Organisation for Economic Co-operation and Development; Thomson Reuters Datastream; IMF staff calculations
Real House Price Indices (long historical) Real GDPAnnual nominal house prices starting 1870 for 17 advanced economies (adjusted for inflation) GDP at constant prices, seasonally adjustedJordà-Schularick-Taylor Macrohistory database; IMF staff calculations

Haver Analytics; Organisation for Economic Co-operation and Development; IMF, Global Data Source database; IMF, World Economic Outlook database
Real GDP (long historical) Nominal GDPAnnual real GDP starting 1870 for 17 advanced economies

GDP at current prices, seasonally adjusted (both in national currency and US dollars)
Jordà-Schularick-Taylor Macrohistory database

Haver Analytics; Organisation for Economic Co-operation and Development; IMF, Global Data Source database; IMF, World Economic Outlook database
InflationPercent change in the consumer price indexHaver Analytics; IMF, Global Data Source database; IMF staff calculations
Inflation (long historical)Percent change in the consumer price index for 17 advanced economies starting 1870Jordà-Schularick-Taylor Macrohistory database
Total Bank Claims and LiabilitiesTotal locational assets and liabilities vis-à-vis th e world in percent of GDPBank for International Settlements; IMF staff calculations
Financial OpennessForeign assets plus foreign liabilities in percent of GDPLane Milesi-Ferretti dataset (2007; updated)
Capital Account OpennessChin-Ito index, measuring a country’s degree of capital account opennessChinn and Ito (2006) dataset (updated)
Exchange Rate RegimeDe facto exchange rate regime of a country (variables based on 15 categories and 6 categories are used)Ilzetzki, Reinhart, and Rogoff (2017) dataset
Macroprudential PoliciesMacroprudential policy tools at quarterly frequencyAlam and others (forthcoming)
Bilateral-Level Variables
Bilateral Bank Claims vis-à-vis Counterparty EconomiesBilateral locational cross-border claims on residency basisBank for International Settlements International Banking Statistics confidential databa
Bilateral Gross Trade vis-à-vis Counterparty EconomiesGross exports vis-à-vis counterparty economiesIMF, Direction of Trade database; IMF staff calculations
Global-Level Variables
Global Liquidity

US Financial Conditions Index
Total claims of all Bank for International Settlements reporters vis-à-vis the world, in percent of world GDP Positive values of the FCI indicate tighter-than-average financial conditions. For methodology and variables included in the FCI, refer to Annex 3.2 of the October 2017 Global Financial Stability Report.Bank for International Settlements; Haver Analytics

IMF, October 2017 Global Financial Stability Report (Chapter 3)
Global Financial Conditions IndexBased on a PCA of all FCIs estimated; Positive values of the FCI indicate tighter-than-average financial conditions. For methodology and variables included in the FCI, refer to Annex 3.2 of the October 2017

Global Financial Stability Report.
IMF, October 2017 Global Financial Stability Report (Chapter 3)
VIXChicago Board Options Exchange Volatility IndexHaver Analytics
US Shadow Interest RatesWu-Xia and Krippner Shadow Federal Funds RatesBloomberg Finance L.P.; Haver Analytics
Source: Authors.Note: FCI = financial conditions index; PCA = principal component analysis; VIX = Chicago Board Options Exchange Volatility Index.
Source: Authors.Note: FCI = financial conditions index; PCA = principal component analysis; VIX = Chicago Board Options Exchange Volatility Index.

Annex Figure 1.1.Sample Coverage

Source: Authors’ calculations.

Note: Cities selected are the largest cities based on population, and overlap with the top 50 cities for global investors identified by Cushman & Wakefield (2017). The sample comprises over 70 cities based on the top 30 cities for global investors in Cushman & Wakefield’s (2017) Global Capital Markets 2017 report’s economic scale, financial center, technology hub, and innovation pillars are also used in robustness checks. If none of the cities in a country (where data are available) are chosen based on the four pillars stated above, the largest city by population in the country is included. Moreover, an additional sample with 44 major cities is also constructed.

Annex Table 1.2.List of Economies and Cities in the Analysis
Economies
AustraliaEuro areaItalySingapore
AustriaFinlandJapanSlovenia
BelgiumFranceKoreaSouth Africa
CanadaGermanyMalaysiaSpain
ChileGreeceMexicoSweden
ChinaHong Kong SARNetherlandsSwitzerland
ColombiaHungaryNew ZealandTaiwan Province of China
CyprusIndiaNorwayThailand
Czech RepublicIndonesiaPortugalTurkey
DenmarkIrelandRussiaUnited Kingdom
EstoniaIsraelSerbiaUnited States
Cities 1
Amsterdam*Dublin*Manila*Seattle
Athens*DusseldorfMelbourneShanghai*
AtlantaFrankfurtMexico City*Shenzhen
Auckland*GuangzhouMiamiSingapore (core central region)*
AustinGreater Stockholm*MilanSuzhou
Bangkok*HamburgMontrealSydney*
BarcelonaFinland metro area*Moscow*Taipei*
BeijingHong Kong SAR (urban areas)*Mumbai*Tallinn*
Belgrade*HoustonMunichTianjin
Berlin*Inner Paris*NagoyaTokyo*
Bogotá*Istanbul*New York*Toronto*
BostonJakarta*OsakaVancouver
Brussels*Kuala Lumpur*Oslo*Vienna*
Budapest*Lake Geneva AreaPhiladelphiaWashington DC
Buenos Aires*Lima*Prague*Zurich*
ChicagoLisbon*Rome*
Copenhagen*Ljubljana*San Francisco
DallasLondon*South Santiago*
DelhiLos AngelesSouthern Seoul*
Dubai*Madrid*São Paulo*
Source: Authors’ calculations.

See the Annex Figure 1.1 note above for city selection criteria. Cities with asterics are included in the smaller sample.

Source: Authors’ calculations.

See the Annex Figure 1.1 note above for city selection criteria. Cities with asterics are included in the smaller sample.

Annex Table 1.3.Standard Deviations of the Variables Used in Empirical Analyses
Country-levelCity-level
House price synchronization [Synch1]0.100.10
Business cycle synchronization [Synch1]0.010.02
House price synchronization [Quasi-correlation]0.840.99
Business cycle synchronization [Quasi-correlation]1.331.28
Bilateral bank integration of ij1.040.97
Global factor (global liquidity)3.904.48
Global liquidity : AE-AE pairs2.513.09
Global liquidity : EM-EM pairs0.950.95
Global liquidity : AE-EM pairs2.843.10
Global liquidity : Sample with high capital account openness2.843.25
Global liquidity : Rest of the sample2.773.22
Global liquidity : Sample with high FX regime1.011.93
Global liquidity : Rest of the sample3.764.04
Global liquidity : Sample with high financial openness1.110.71
Global liquidity : Rest of the sample3.814.52
Global liquidity : Pre-crisis sample2.131.99
Global liquidity : GFC sample2.245.84
Global liquidity : Post-GFC sample1.622.31
Analysis on the Impact of Macroprudential Policies:
House price synchronization with the region [Quasi-correlation]0.99
House price synchronization with the world [Quasi-correlation]0.97
LTV0.39
Fiscal-based measures0.17
All measures1.36
All loans0.73
Demand side0.51
Supply side1.02
Supply side: general0.79
Supply side: capital0.40
Supply side: loans0.39
Source: Authors’ calculations.
Source: Authors’ calculations.

Annex Figure 1.2.House Price Synchronicity with Global and Regional Cycles

Source: Authors’ estimates.

Note: Panel 1 depicts distributions of the house price synchronization with global cycle (green) and regional cycle (red) for the overall sample (41 countries) and for each region: Asia (13), Americas (5), and Europe and Other (23). Top and bottom horizontal lines show min and max; top/middle/bottom box lines show 75th/50th/25th percentile of the distribution. The global cycle is computed as the median house price cycle across all countries. The regional cycle is the median house price cycle in each region. House price cycles are extracted using the band pass filter developed by Christiano and Fitzgerald (2003).

Annex II: Alternative Measures of House Price Synchronicity

A. Alternative Measure 1: Inverse Absolute Gap Difference (Synch1)

Following Kalemli-Ozcan et al. (2013a; 2013b), HPsynchijt is calculated as the inverse of the absolute difference of house price gaps in country i and j at quarter t as below:

where HPgapit and HPgapjt stand for house price gap of countries i and j respectively at quarter t.24

Empirical Strategy

Similar to equation 4 in Section III of this paper, we estimate the impact of business cycle synchronicity, bilateral financial linkages, and global financial conditions on house price synchronicity using Synch1 measure of house price synchronicity.

where Synchlijt is the synchronicity of house price gaps between country-pair i and j at quarter t measured as presented in equation A2.1BSCij denotes business cycle synchronicity between country i and j measured as presented in footnote 24. All other independent variables are as same as in equation 4 introduced in Section III.

Results

Impact of Bilateral Linkages on House Price Gap Synchronicity

We estimate the impact of bilateral linkages—business cycle synchronicity (BCS) and bilateral banking integration—on house price synchronicity using Synch1ijt as an alternative measure of house price and business cycle synchronicity. Given Synch1ijt is perceived more as of a medium-term measure of synchronicity compared to the instantaneous quasi-correlation, while bilateral banking linkages are measured using stock of assets and liabilities from balance sheet side, we believe Synch1ijt is a better measure to capture the bilateral linkages that we are analyzing. The global factor that measures global financial conditions (more of a short-term indicator) is included in these specifications only as a control variable and to provide consistency across regression tables.

The results are presented in Table A2.1 below. We find that both BCS and bilateral banking integration have statistically significant positive association with house price synchronicity (columns 1 to 3). For comparability of the coefficients, we present the results of the baseline specification (column 4), further standardized in Figure 3. The figure suggests that the impact of both BCS and bilateral bank integration on house price gap synchronicity is comparable in magnitude.

Annex Figure 2.1.Impact of Bilateral Linkages on House Price Synchronicity

Source: Authors’ estimates.

Note: Synchronicity is measured by the Synch1 of gaps measure. Figure shows statistically significant standardized coefficients that are calculated using the coefficients in specification 4 in Table 1 and their respective standard deviations and presented in terms of standard deviations of the dependent variable; this specification also controls for the global financial conditions (proxied through the global liquidity) in addition to country-pair fixed effects, quadratic and linear time trends (standard errors are clustered at multi-way at time, country i and country j). Standard deviation of the country-level dependent variable (synch1) is approximately 0.10; standard deviation of the country-level BCS (measured using synch1) is 0.01; standard deviation of the bilateral bank integration is 1.04. i = country 1 and j = country 2 in the country pair.

We also find that the impact of bilateral banking integration on house price gap synchronicity is lower among emerging market economy country pairs, compared to that of advanced economy country-pairs (column 5). Moreover, when both countries in the country-pair are de facto more financially open, the positive impact of bilateral banking integration on house price gap synchronicity is muted; this result is statistically significant at a 5 percent confidence interval (column 9). While the impact of banking integration on house price gap synchronicity is positive in our baseline specification in Table A2.1, we fail to find statistically significant impact for the post-GFC period (column 10).

Annex Table 2.1.House Price Gap Synchronicity at Country-Level and Bilateral Linkages
Dependent Variable: House Price Gap Synchronization of Country Pair i and j (Synch1)(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)
Business Cycle Synchronization of ij0.766***

(0.254)
0.675**

(0.293)
0.733***

(0.243)
0.657**

(0.254)
0.658**

(0.253)
0.746***

(0.262)
0.725***

(0.261)
0.725***

(0.262)
0.675**

(0.253)
0.706**

(0.337)
Bilateral Bank Integration of ij0.006*

(0.003)
0.007**

(0.003)
0.012

(0.007)
0.009*

(0.004)
0.007**

(0.003)
0.007*

(0.004)
0.007**

(0.003)
0.004

(0.005)
Global Factor (global liquidity)-0.001

(0.001)
-0.001

(0.001)
-0.001

(0.001)
-0.001

(0.001)
-0.001

(0.001)
-0.001

(0.001)
0.001

(0.001)
Bilateral Bank Integration Interacted with:
x EMEs-EMEs Dummy-0.016*

(0.009)
x EMEs-AEs Dummy-0.009

(0.010)
x High Capital Account Openness with the World-0.005

(0.003)
x High Exchange Rate Regime (ij) (15 categories; high = more flexible)-0.005

(0.004)
x High Exchange Rate Regime (ij) (6 categories; high = more flexible)-0.001

(0.004)
x High Financial Openness with the World (ij)-0.019***

(0.004)
GFC Period Dummy Interacted with:
x Business Cycle Synchronization of ij-0.080

(0.516)
x Bilateral Bank Integration of ij0.008**

(0.004)
x Global Factor0.001

(0.001)
Post-GFC Period Dummy Interacted with:
x Business Cycle Synchronization of ij0.380

(0.456)
x Bilateral Bank Integration of ij0.007

(0.005)
x Global Factor0.004

(0.003)
GFC Dummy0.048***

(0.011)
Post-GFC Dummy0.042***

(0.009)
Observations65,45065,34349,38449,38449,38443,87146,70846,70847,35349,384
R-Squared0.3530.4980.3860.3560.3560.3610.3560.3560.3600.360
Multiway ClusteringYesYesYesYesYesYesYesYesYesTwo-way
GroupAllAllAllAllAllAllAllAllAllAll
Time FE and Country-Pair FEYesYes
Time FE, Country-Pair FE, and country*time FEYes
Quadratic Trend and Country-Pair FEYesYesYesYesYesYes
Country-Pair FEYes
Source: Authors’ estimates.Note: GFC Dummy = a dummy variable that equals 1 during 2008–09, and zero otherwise. Post-GFC Dummy = a dummy variable that equals 1 curing 2010–16, and zero otherwise. All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications 5 through 9, but are not shown above (specifically, dummy variables for EMEs-EMEs, EMEs-AEs, high capital account openness, high exchange rate regime, high financial openness are included in specifications 5 through 9, but not shown). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are three-way clustered (at country i, country j, and date), with the exception of regression (10), in which errors are two-way clustered (at country i, country j). AEs = advanced economies; EMEs = emerging market economies; FE = fixed effects; GFC = global financial crisis. *** p < 0.01; ** p < 0.05; * p < 0.1.
Source: Authors’ estimates.Note: GFC Dummy = a dummy variable that equals 1 during 2008–09, and zero otherwise. Post-GFC Dummy = a dummy variable that equals 1 curing 2010–16, and zero otherwise. All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications 5 through 9, but are not shown above (specifically, dummy variables for EMEs-EMEs, EMEs-AEs, high capital account openness, high exchange rate regime, high financial openness are included in specifications 5 through 9, but not shown). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are three-way clustered (at country i, country j, and date), with the exception of regression (10), in which errors are two-way clustered (at country i, country j). AEs = advanced economies; EMEs = emerging market economies; FE = fixed effects; GFC = global financial crisis. *** p < 0.01; ** p < 0.05; * p < 0.1.

B. Alternative Measure 2: Pearson Correlations

In a separate exercise, regressions were run using a panel of three non-overlapping seven-year periods (in other words, three non-overlapping 28 quarter periods), in which the house price and business cycle synchronicity is captured by the bilateral Pearson correlation coefficients for the period. All other explanatory variables are the average values for the period. Further robustness checks in this exercise were employed by collapsing the other explanatory variables using the last value of the previous period instead. The interaction term of the global factor and foreign exchange regime continues to be statistically significant, in addition to the global factor itself (Table A2.2).

Annex Table 2.2.House Price Gap Synchronicity at Country Level and Global Factors—Pearson Correlations with 3 Non-overlapping Periods
Dependent Variable: House Price Gap Synchronization of Country Pair i and j (Non-overlapping period-wise Pearson correlation)(1)(2)(3)(4)
Control variables collapsed by MeanControl variables collapsed by last obs of previos period
Business Cycle Synchronization of ij0.104*

(0.061)
0.058

(0.062)
0.089

(0.060)
0.058

(0.062)
Bilateral Bank Integration of ij0.019

(0.029)
0.008

(0.032)
0.029

(0.023)
0.037

(0.024)
Global Factor (global liquidity)0.013*

(0.007)
0.019**

(0.007)
0.043***

(0.013)
0.051***

(0.014)
Global Factor Interacted with: x High Exchange Rate Regime (ij) (15 categories; high = more flexible)-0.117***

(0.023)
-0.168***

(0.064)
Observations1,6601,5531,6601,553
R-squared0.3690.3800.3750.380
Clusteringcountry-paircountry-paircountry-paircountry-pair
Source: Authors’ estimates.Note: Pearson correlation coefficients of house price synchronicity are measured as HousePriceSynchijt=PCORRijt=cov(HPgapi,HPgapj)σigapσjgap, where Hpgapi and Hpgapj stand for house price gap of country i and j respectively, cov is the covariance, and σ is the standard deviation (business cycle synchronicity presented here is also measured similarly using pearson correlation coeficients of the output gaps). All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications, but are not shown above (specifically, dummy variables for high exchange rate regime). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are clustered as described above; FE = fixed effects. *** p < 0.01; ** p < 0.05; * p < 0.1.
Source: Authors’ estimates.Note: Pearson correlation coefficients of house price synchronicity are measured as HousePriceSynchijt=PCORRijt=cov(HPgapi,HPgapj)σigapσjgap, where Hpgapi and Hpgapj stand for house price gap of country i and j respectively, cov is the covariance, and σ is the standard deviation (business cycle synchronicity presented here is also measured similarly using pearson correlation coeficients of the output gaps). All regressors are lagged by one quarter. Institutional characteristics dummies are included in specifications, but are not shown above (specifically, dummy variables for high exchange rate regime). High = a dummy variable that equals 1 when both countries are in the top fifth of the institutional characteristic. Standard errors (in parentheses) are clustered as described above; FE = fixed effects. *** p < 0.01; ** p < 0.05; * p < 0.1.

C. Alternative Measure 3: Synchronicity with Longer time series

The relationship between house price gap synchronicity and BCS is found to be positive and statistically significant when Jordà-Schularick-Taylor (2017) dataset is considered. This analysis contains annual observations from 1870 to 2013 for 17 advanced economies is used (Table A2.3). Additional analysis was limited by data availability.

Annex Table 2.3.House Price Gap Synchronicity at Country Level and Business Cycle Synchronicity—Estimations using Jordà-Schularick-Taylor Dataset—1870–2013
Dependent Variable: House Price Gap Synchronization of Country Pair i and j(1)(2)(3)(4)(5)(6)(7)(8)
Synch 1Quasi-correlation
Business Cycle Synchronization of ij0.902**

(0.385)
0.902**

(0.311)
0.902***

(0.153)
0.902***

(0.089)
0.042

(0.032)
0.042*

(0.024)
0.042***

(0.015)
0.042***

(0.008)
Observations9,8189,8189,8189,8189,8189,8189,8189,818
R-squared0.1430.1430.1430.1430.0710.0710.0710.071
ClusteringMulti-wayTwo-wayVCE robustNoMulti-wayTwo-wayVCE robustNo
Quadratic Trend and Country-Pair FEYesYesYesYesYesYesYesYes
Source: Authors’ estimates.Note: See equations 1, 2, A2.1 and A2.2 for Synch1 and Quasi-correlation methodologies. Regressors are lagged by one year; FE = fixed effects.*** p < 0.01; ** p < 0.05; * p < 0.1.
Source: Authors’ estimates.Note: See equations 1, 2, A2.1 and A2.2 for Synch1 and Quasi-correlation methodologies. Regressors are lagged by one year; FE = fixed effects.*** p < 0.01; ** p < 0.05; * p < 0.1.
Annex III: Methodology-House Price Interconnectedness Analysis

Following the methodology proposed by Diebold and Yilmaz (2014), we measure house price interconnectedness based on a large-scale vector autoregressive (VAR) model. The econometric framework is estimated separately using quarter-on-quarter house price growth rates at the country- and city-level, while controlling for global financial conditions. Within the VAR model, the interconnectedness is defined as the fraction of H-quarter-ahead forecast error variance of country/city j’s house price growth that can be accounted for by country/city i’s house prices growth dynamics.

Quarterly house price growth rates are computed using seasonally adjusted real house prices either at country-level or at city-level. Global financial conditions in this analysis are proxied by the U.S. Financial Conditions Index (FCI) constructed in line with IMF 2017.25 The estimation period for the country-level analysis spans from 1990:Q1 to 2016:Q4, while for city-level interconnectedness analysis, owing to data limitations, is estimated for a period spanning from 2004:Q1 to 2017:Q2.

The number of countries and cities in our samples ─ that enters as the set of variables in the VAR setting ─ is large (n=30). Following Demirer et al. (2018), Song and Bickel (2011), the VAR is estimated using machine learning techniques such as lasso and elastic net which allow for the estimation of large-scale VARs.

The baseline house price interconnectedness specification we estimated can be described as follows:

, where Y stands for quarterly house price growth variavles for 32 countries or 30 major cities that enter as endogenous variables in the VAR setting. U.S. FCI is used to contol for global financial conditions, where robustness checks were performed with variables mentioned in the footnote 25.

Annex IV: Impact of Macroprudential Measures on House Price Synchronicity─Regression Results
Annex Table 4.1.Unconditional Estimation Sample
Dependent variable: house price gap synchronicity (quasi‐correlation) with:(1) Region(2) World(3) Region(4) World(5) Region(6) World(7) Region(8) World(9) Region(10) World(11) Region(12) World(13) Region(14) World(15) Region(16) World(17) Region(18) World
Global factor (FCI)-0.051

(0.037)
-0.084**

(0.040)
-0.050

(0.037)
-0.083**

(0.040)
-0.053

(0.037)
-0.086**

(0.040)
-0.052

(0.037)
-0.085**

(0.040)
-0.052

(0.037)
-0.084**

(0.040)
-0.052

(0.037)
-0.085**

(0.040)
-0.049

(0.037)
-0.083**

(0.040)
-0.058

(0.037)
-0.089**

(0.040)
-0.051

(0.037)
-0.084**

(0.040)
Business cycle synchronicity with the region0.028

(0.020)
0.030

(0.019)
0.029

(0.019)
0.028

(0.019)
0.029

(0.020)
0.029

(0.019)
0.030

(0.020)
0.027

(0.019)
0.029

(0.019)
Bank integration with the region0.013

(0.013)
0.012

(0.013)
0.013

(0.013)
0.013

(0.013)
0.013

(0.013)
0.012

(0.013)
0.012

(0.013)
0.013

(0.013)
0.013

(0.013)
Business cycle synchronicity with the world0.041**

(0.020)
0.043**

(0.019)
0.042**

(0.019)
0.042**

(0.019)
0.042**

(0.020)
0.043**

(0.019)
0.043**

(0.019)
0.041**

(0.019)
0.043**

(0.019)
Bank integration with the world0.027

(0.020)
0.026

(0.020)
0.027

(0.020)
0.027

(0.020)
0.027

(0.020)
0.027

(0.020)
0.027

(0.020)
0.028

(0.020)
0.027

(0.020)
Macroprudential measures
LTV-0.097**

(0.045)
-0.128***

(0.040)
Fiscal-based measures-0.132

(0.144)
-0.122*

(0.067)
All measures-0.027

(0.019)
-0.032**

(0.016)
All loan-targeted-0.064***

(0.022)
-0.064**

(0.025)
Demand side-0.077**

(0.033)
-0.059

(0.038)
Supply side: all-0.020

(0.029)
-0.026

(0.024)
Supply side: general0.042

(0.030)
0.012

(0.031)
Supply side: capital-0.230***

(0.074)
-0.174**

(0.071)
Supply side: loans-0.092**

(0.044)
-0.123**

(0.055)
Observations3,5203,5203,5203,5203,5203,5203,5203,5203,5203,5203,5203,5203,5203,5203,5203,5203,5203,520
R-squared0.0080.0170.0070.0150.0080.0170.0090.0170.0080.0160.0070.0150.0080.0150.0150.0200.0080.017
Source: Authors’ estimates.Note: All regressors are lagged by one quarter. Supply side (loans) consists of limits on credit growth, loan loss provisions, loan restrictions, and limits on foreign currency loans. Supply side (capital) consists of capital requirements, conservation buffers, the leverage ratio, and the countercyclical capital buffer. Supply side (general) consists of reserve requirements, liquidity requirements, and limits on foreign exchange positions. Demand-side includes limits to debt-service-to-income and LTV ratios. All loan-targeted measures include demand side and supply side (loans). Fiscal-based measures include taxes such as ad valorem, seller’s and buyer’s stamp duty, or other taxes. All regressions include country fixed effects. Robust standard errors are presented in parentheses.*** p < 0.01; ** p < 0.05; * p < 0.1.
Source: Authors’ estimates.Note: All regressors are lagged by one quarter. Supply side (loans) consists of limits on credit growth, loan loss provisions, loan restrictions, and limits on foreign currency loans. Supply side (capital) consists of capital requirements, conservation buffers, the leverage ratio, and the countercyclical capital buffer. Supply side (general) consists of reserve requirements, liquidity requirements, and limits on foreign exchange positions. Demand-side includes limits to debt-service-to-income and LTV ratios. All loan-targeted measures include demand side and supply side (loans). Fiscal-based measures include taxes such as ad valorem, seller’s and buyer’s stamp duty, or other taxes. All regressions include country fixed effects. Robust standard errors are presented in parentheses.*** p < 0.01; ** p < 0.05; * p < 0.1.
Annex Table 4.2.Conditional on Positive Credit Gaps
Dependent variable: house price gap synchronicity (quasi-correlation) with:(1) Region(2) World(3) Region(4) World(5) Region(6) World(7) Region(8) World(9) Region(10) World(11) Region(12) World(13) Region(14) World(15) Region(16) World(17) Region(18) World
Global factor (FCI)-0.127***

(0.040)
-0.159***

(0.042)
-0.128***

(0.040)
-0.159***

(0.042)
-0.127***

(0.040)
-0.162***

(0.042)
-0.127***

(0.040)
-0.161***

(0.042)
-0.127***

(0.040)
-0.160***

(0.042)
-0.127***

(0.040)
-0.160***

(0.042)
-0.125***

(0.041)
-0.158***

(0.042)
-0.133***

(0.040)
-0.165***

(0.041)
-0.127***

(0.040)
-0.160***

(0.041)
Business cycle synchronicity with the region0.043**

(0.019)
0.043**

(0.019)
0.043**

(0.019)
0.042**

(0.019)
0.043**

(0.019)
0.043**

(0.019)
0.044**

(0.019)
0.041**

(0.019)
0.042**

(0.019)
Bank integration with the region0.036***

(0.012)
0.036***

(0.012)
0.036***

(0.012)
0.036***

(0.012)
0.036***

(0.012)
0.036***

(0.012)
0.036***

(0.012)
0.037***

(0.012)
0.037***

(0.012)
Business cycle synchronicity with the world0.060***

(0.017)
0.062***

(0.017)
0.060***

(0.017)
0.060***

(0.017)
0.060***

(0.017)
0.061***

(0.017)
0.061***

(0.017)
0.059***

(0.017)
0.061***

(0.017)
Bank integration with the world0.056***

(0.013)
0.056***

(0.013)
0.058***

(0.013)
0.057***

(0.013)
0.057***

(0.014)
0.059***

(0.014)
0.058***

(0.014)
0.060***

(0.014)
0.058***

(0.014)
Macroprudential measures
LTV-0.028

(0.051)
-0.131*

(0.074)
Fiscal-based measures-0.257***

(0.074)
-0.178

(0.174)
All measures-0.011

(0.021)
-0.037

(0.027)
All loan-targeted-0.033

(0.025)
-0.082**

(0.037)
Demand side-0.023

(0.039)
-0.087

(0.054)
Supply side: all-0.004

(0.038)
-0.023

(0.039)
Supply side: general0.054

(0.048)
0.019

(0.047)
Supply side: capital-0.194**

(0.096)
-0.188*

(0.102)
Supply side: loans-0.075

(0.046)
-0.142*

(0.074)
Observations2,1392,1392,1392,1392,1392,1392,1392,1392,1392,1392,1392,1392,1392,1392,1392,1392,1392,139
R-squared0.0360.0520.0370.0490.0360.0510.0370.0530.0360.0510.0360.0490.0380.0490.0420.0540.0370.052
Source: Authors’ estimates.Note: All regressors are lagged by one quarter. Supply side (loans) consists of limits on credit growth, loan loss provisions, loan restrictions, and limits on foreign currency loans. Supply side (capital) consists of capital requirements, conservation buffers, the leverage ratio, and the countercyclical capital buffer. Supply side (general) consists of reserve requirements, liquidity requirements, and limits on foreign exchange positions. Demand-side includes limits to debt-service-to-income and LTV ratios. All loan-targeted measures include demand side and supply side (loans). Fiscal-based measures include taxes such as ad valorem, seller’s and buyer’s stamp duty, or other taxes. All regressions include country fixed effects. Robust standard errors are presented in parentheses.*** p < 0.01; ** p < 0.05; * p < 0.1.
Source: Authors’ estimates.Note: All regressors are lagged by one quarter. Supply side (loans) consists of limits on credit growth, loan loss provisions, loan restrictions, and limits on foreign currency loans. Supply side (capital) consists of capital requirements, conservation buffers, the leverage ratio, and the countercyclical capital buffer. Supply side (general) consists of reserve requirements, liquidity requirements, and limits on foreign exchange positions. Demand-side includes limits to debt-service-to-income and LTV ratios. All loan-targeted measures include demand side and supply side (loans). Fiscal-based measures include taxes such as ad valorem, seller’s and buyer’s stamp duty, or other taxes. All regressions include country fixed effects. Robust standard errors are presented in parentheses.*** p < 0.01; ** p < 0.05; * p < 0.1.
1This paper is background to a subsection of analyses presented in Chapter 3 of the April 2018 GFSR, “House Price Synchronization: What Role for Financial Factors?” (IMF 2018a). We would like to thank Tobias Adrian, Alan Feng, Mitsuru Katagiri, Romain Lafarguette, and Claudio Raddatz, for helpful comments, suggestions, and guidance. All remaining errors are our own. The views expressed in this paper do not necessarily represent the views of the IMF, its Executive Board, IMF management, the Federal Reserve Bank of Chicago, or the Federal Reserve Board.
2See IMF (2018a) for a detailed discussion on trends in house prices across countries and cities. House price synchronicity measures are presented in detail in section II and Annex II.
3Several papers focus on house price co-movement within a country. For instance, Landier, Sraer, and Thesmar (2017) find evidence for increased correlation of U.S. housing market across states owing to the rise of large banks.
4For robustness purposes, alternative measures of house price synchronicity are considered in Annex II.
5For emerging market economies, we use 20 years as the maximum length instead.
6As a robustness check, we also constructed house price gaps using Hodrick and Prescott (1997) filter with a lambda of 400,000 which is commonly used as the lambda relevant for financial cycles. We obtain house price gaps broadly consistent to that of the Christiano and Fitzgerald (CF) filter. CF filter is chosen for our analysis as it computes the cyclical component for all observations without being prone to tail bias.
7Additional forms of bilateral financial integration measures such as bilateral portfolio linkages and bilateral direct investment linkages are not used in our analysis due to their lower frequency and shorter time span.
8To address the issue of mirror data asymmetry, following Kalemli-Ozcan et al. (2013a; 2013b), we take the average of country i’s assets vis-à-vis country j and county j's liabilities vis-à-vis country i as the assets of country i vis-à-vis country j and vice versa.
9See Annex 3.2 of the IMF’s October 2017 GFSR (Chapter 3) for FCI construction methodology.
10See Annex II for the analysis on the impact of Bilateral Linkages on House Price Gap Synchronicity.
11Although our house price time series, particularly for advanced economies, start several decades prior to 1990, we restrict our econometric analysis to begin in 1990 as the availability of data on bilateral banking linkages significantly improves starting from 1990. We exclude four EMs out of our original sample of 44 countries in the econometric analysis due to the short length of their house price time series.
12Financial integration is measured using bilateral locational banking statistics on residency basis obtained from BIS IBS restricted databases. Bilateral banking integration is measured as the logarithm of the sum of bilateral claims of country i vis-à-vis country j and bilateral claims of country j vis-à-vis country i as a ratio of the sum of GDPs of country i and j. Additional forms of bilateral financial integration measures such as bilateral portfolio linkages and bilateral direct investment linkages are not used in our analysis due to their lower frequency and much shorter time span.
13High level is defined based on the top 1/5 of the distribution of institutional characteristics, at any point in time. In addition, robustness checks were performed by defining the institutional factors as high using 75th or 66th percentile instead of the 80th percentile as cutoff rates.
14To account for serial correlation, following Cameron et al. (2011), standard errors are multi-way clustered (at country i, country j, and time level, where appropriate).
15See Table 2 in the robustness checks section for results using additional proxies for global factors.
16Given global financial conditions index is more of a short-term indicator, we believe instantaneous quasi-correlation (that purges the mean house price gap) as the synchronicity measure is better suited for this analysis.
17While results are robust to these alternative proxies for the global factor, the level of statistical significance declines, especially when the most stringent manner of standard error clustering is considered.
18The selection of cities is based on population and overlaps with the top 50 cities for global investors identified by Cushman & Wakefield (2017). The sample comprises over 70 cities (see Annex I) combining the Top 30 cities in global investors’ ranking by Cushman & Wakefield’s (2017) Global Capital Markets 2017 report, where economic scale, financial center, technology hub, and innovation criteria are considered. If none of the cities in a country (where data are available) are chosen based on the four pillars stated above, the largest city by population in the country is included. Moreover, an additional sample with 44 major cities off the above sample is also constructed.
19The coefficients in the regression analysis are weighted by the number of major cities in each country.
20Synchronicity with regional cycles may pose further financial stability concerns, as macro-financial shocks could transmit more easily from one country to another through interconnected bank balance sheets and collateral values. In some regions, house price synchronicity with the regional cycle is stronger than with the global cycle, reflecting deeper intra-regional financial and trade integration (see also Katagiri 2018). As depicted in Annex Figure 1.2., we find that the median correlation with the global cycle is roughly 0.4, while the one with the regional cycle is about 0.5 across all countries and time.
21House price synchronicity with the global cycle is heterogeneous across regions, potentially reflecting deeper intraregional financial and trade integration.
22For more details regarding the macroprudential tools database, see Alam et al. (2018).
23In some instances, fiscal-based measures target speculative investments, including by foreign buyers (see IMF 2018b).
24Business cycle synchronicity (BCS) measure is similar to the house price synchronicity measure presented above. BCSijt=|YgapitYgapjt|, where Ygapit and Ygapjt stand for output gap of country i and j respectively at quarter t and measured using Christiano and Fitzgerald band-pass filter (2003), with the maximum length adjusted for business cycles instead of financial cycles.
25Robustness checks were performed using the global FCI and the VIX index, and results are found to be very similar.

Other Resources Citing This Publication