## 6. Impact of Macroprudential Measures on House Price Synchronicity

## Source details

**Canonical URL:** [6. Impact of Macroprudential Measures on House Price Synchronicity](https://www.imf.org/-/media/files/publications/wp/2018/wp18250.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp18250.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp18250.pdf.json)

---

### Overview and research questions
- Paper analyzes role of bilateral financial linkages and global financial conditions above-and-beyond business cycle synchronicity as drivers of house price synchronicity.
- Empirical scope:
  - Bilateral panel data at country-pair level with nearly 50,000 observations.
  - Major city-pair level with nearly 70,000 observations.
  - Coverage: over 40 economies and over 70 cities, with coverage spanning through end-2016.
  - Baseline econometric analysis: quarterly frequency from 1990 to 2016, for 40 countries.
- Core research questions:
  1. Do global financial conditions amplify 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 business cycle synchronicity?
  3. What is the role of institutional factors in mitigating or amplifying the impact of global financial conditions on house price synchronicity?
  4. Do macroprudential policies remain effective in addressing domestic vulnerabilities amid heightened house price synchronicity?

### Data and measurement (main variables and construction)
- House price gap synchronicity:
  - Measured using instantaneous quasi-correlation (Morgan, Rime, and Strahan 2004).
  - Synchronicity computed at time-series level (not bounded between -1 and 1).
  - House price gaps = cyclical component of real house prices extracted with Christiano and Fitzgerald (2003) band-pass filter, with maximum length of 30 years to capture medium-term financial cycles (20 years max for emerging market economies).
  - Robustness: Hodrick and Prescott (1997) filter with lambda of 400,000 used as a check; CF filter chosen to avoid tail bias.
- Business cycle synchronicity:
  - Analogous instantaneous quasi-correlation measure using output gaps from CF band-pass filter (maximum length adjusted for business cycles).
- Bilateral banking integration:
  - Measured using BIS locational banking statistics on residency basis.
  - Bilateral banking integration = ln[(bilateral claims i→j + bilateral claims j→i) / (GDP_i + GDP_j)] * 100.
  - Mirror data asymmetry addressed by averaging assets and liabilities per Kalemli-Ozcan et al. (2013a; 2013b).
- Global financial conditions (global factor):
  - Main proxy: changes in BIS global liquidity (changes in banks’ cross-border claims denominated in all currencies plus local claims in foreign currency in percent of global GDP).
  - Robustness proxies: Global FCI and U.S. FCI (IMF 2017 methodology), CBOE VIX, U.S. shadow interest rates (Wu and Xia 2016; Krippner 2013).
- Institutional and other controls:
  - High capital account openness: Chinn-Ito index (de jure).
  - High exchange rate regime/flexibility: Ilzetzki, Reinhart, and Rogoff (2017) de facto index.
  - High financial openness: Lane and Milesi-Ferretti (2007) de facto index.
  - “High” defined as dummy = 1 when both countries in the pair are in the top fifth of the institutional characteristic during a given quarter (80th percentile), with robustness checks using 75th or 66th percentiles.
  - Dummy variables for both countries being advanced economies, emerging market economies, or advanced–emerging market pairs.

### Empirical strategy and specification
- Baseline country-pair regression (quarterly, 1990–2016, 40 countries):
  - Dependent variable: house price gap synchronicity between country-pair i and j at quarter t.
  - Regressors (all lagged by one quarter): business cycle synchronicity; bilateral financial integration; global factor (changes in global liquidity); interaction terms of global factor with dummies for high institutional characteristics; linear and quadratic time trends; country-pair fixed effects.
  - Country-pair fixed effects capture time-invariant idiosyncratic factors (e.g., geographic proximity, supply-side/regulatory considerations).
  - Standard errors: multi-way clustered (at country i, country j, and time level, where appropriate); various clustering alternatives used in robustness checks.

### Main empirical findings — country-pair and city-pair levels
- Headline findings:
  - Global factors matter:
    - Abundant global liquidity and loose global financial conditions are positively associated with house price synchronicity across country-pairs and major city-pairs.
    - Result robust when controlling for bilateral macro-financial linkages including business cycle synchronicity and banking integration.
  - Exchange rate flexibility dampens global influence:
    - Greater exchange rate flexibility attenuates the positive impact of global factors on house price synchronicity.
    - Interaction statistically significant at 1 percent confidence interval and robust across specifications.
  - Bilateral linkages matter:
    - Past co-movement in business cycles and bilateral bank linkages are positively associated with house price synchronicity.
  - Macroprudential policies can reduce synchronicity:
    - Macroprudential policies aimed at tackling domestic vulnerabilities may additionally reduce a country’s house price synchronicity with the rest of the region and the world.
- Heterogeneity by country group:
  - Impact of global financial conditions on house price synchronicity is higher between advanced economies than between emerging market economy pairs.
  - For advanced economies, the impact is statistically significant and positive; for emerging market economies and advanced–emerging market pairs, the impact is not statistically significant at conventional levels when using the most stringent standard error clustering.
- Time variation:
  - Positive impact of global liquidity on house price synchronicity was substantially higher prior to the global financial crisis (GFC).

### Representative coefficient patterns (country-pair regressions)
- Business Cycle Synchronization of ij: coefficients frequently reported around 0.025, 0.026, 0.039, 0.043 and with significance levels ranging from * to ***; representative standard errors include (0.013), (0.014), (0.011), (0.012), (0.005).
- Bilateral Bank Integration of ij: coefficients mixed in sign and magnitude (examples: -0.011, 0.012, 0.022, -0.016, 0.031) with representative standard errors (0.033), (0.031), (0.036), (0.035), (0.032), (0.034).
- Global Factor (global liquidity): many positive and significant coefficients reported (examples: 0.016**, 0.019***, 0.022*, 0.060***, 0.076***), representative standard errors include (0.006), (0.007), (0.008), (0.013), (0.001), (0.005).
- Interaction: Global Factor × High Exchange Rate Regime (ij): consistently negative and often statistically significant (e.g., -0.023*** with standard error (0.008) and other robustness values -0.024***, -0.069**, -0.056*, -0.011**, -0.025*, -0.026** with reported standard errors).
- GFC Dummy: -0.137** (0.060); Post-GFC Dummy: -0.044 (0.052).
- Model fit and sample:
  - Representative Observations counts: 65,450; 65,343; 49,384; 43,871; 46,708; 47,353; 48,890; 46,215; 48,973; 46,308.
  - Representative R-squared values: 0.227, 0.354, 0.251, 0.230, 0.233, 0.224, 0.223, 0.241, 0.232, 0.230.
  - Standard errors typically three-way clustered (country i, country j, date) except select regressions using two-way clustering.

### Robustness checks and sensitivity
- Alternative global-factor proxies:
  - U.S. FCI, Global FCI, CBOE VIX, U.S. shadow rates — main sign and magnitude broadly unchanged, though statistical significance may decline under the most stringent clustering.
- Alternative synchronicity measures:
  - Annex II alternative measures capturing medium-term dynamics; results on bilateral banking integration and business cycle synchronicity remain positive and robust.
- Clustering and estimators:
  - Explored clustering at country-pair level, two-way clustering (country i and country j), two-way clustering (country-pair and time), and Huber/White/sandwich estimator.
- Time controls:
  - Year fixed effects and linear time trends produce little change to main results.
- Cutoff percentiles:
  - Robustness to defining “high” institutional characteristics at the 75th and 66th percentiles checked.

### City-level analysis — network and regressions
- Network analysis (VAR-based interconnectedness):
  - Quarterly country-level span: 1990:Q1 to 2016:Q4.
  - Quarterly city-level span: 2004:Q1 to 2017:Q2.
  - Interconnectedness defined as fraction of H-quarter-ahead forecast error variance of j explained by i (Diebold and Yilmaz 2014); estimation uses lasso and elastic net for high-dimensional VAR.
  - Only links above 50th percentile (country-level) and 66th percentile (city-level) considered in network plots.
  - Key qualitative finding: global-investor-attractive cities can be central even when their countries are peripheral (example: Tokyo and Rome).
- City-level panel regression (quarterly 2004–2016, over 70 cities):
  - Dependent variable: synchronicity of house price gaps between city-pair i and j at quarter t (instantaneous quasi-correlation).
  - Controls include country-level business cycle synchronicity, bilateral financial linkages, global financial conditions (BIS global liquidity).
  - Fixed effects: city-pair FE, quadratic and linear time trends; regressors lagged one quarter.
- City-level main findings:
  - Global financial conditions positively associated with city-level house price gap synchronicity.
    - Significance: statistically significant under multi-way clustering; significance improves under two-way clustering (e.g., 10% → 1% in some specifications).
    - Representative coefficients: Global Factor (global liquidity) values reported such as 0.018*** (0.005), 0.030*** (0.008), 0.012** (0.005), 0.021*** (0.006).
  - Exchange rate flexibility attenuates the global factor’s effect on city-level synchronicity (e.g., interaction -0.017** (0.008) two-way clustering; -0.017** (0.008) multi-way clustering).
  - Advanced-economy city-pairs show higher impact of global financial conditions than emerging-economy city-pairs.
  - De jure financial openness tends to amplify global-factor impact under less stringent clustering; de facto openness yields no consistent significant results.
  - Pre-crisis relation: global financial conditions positively associated with city-level synchronicity prior to the GFC.
  - Representative city-level sample sizes and fit:
    - Observations examples: 66,575; 66,572; 66,575; 66,575; 66,575; 59,353; 63,691; 63,691; 62,588.
    - R-squared examples: 0.260; 0.343; 0.260; 0.254; 0.256; 0.265.

### Macroprudential policies (MPPs) — extensions and empirical results
- Motivation:
  - MPPs targeted at dampening domestic vulnerabilities may weaken correlation of domestic house price cycles with regional and global cycles, but effects may be ambiguous due to capital flow channels.
- Empirical strategy for MPPs:
  - Panel regression for 41 countries from 1990:Q2 through 2016:Q4.
  - Dependent variable HPS: house price cycle synchronicity (instantaneous quasi-correlation) with regional or global cycle.
  - Key regressors: business cycle synchronicity, global financial conditions, financial integration, institutional characteristics, and MPP indicators or macroprudential group indices (loan-targeted, supply-side capital/general/loans, demand-side, fiscal-based).
  - Specification includes country fixed effects; all regressors lagged one quarter.
- MPPs — empirical findings:
  - Demand-side MPPs (e.g., LTV, debt-service-to-income limits):
    - Total number of demand-side events: 47; event analysis window ±5 quarters around implementation.
    - Before implementation, house prices grew similarly in high- and low-synchronicity countries; after implementation, house price growth declined in both groups, with a stronger and more sustained decline in low-synchronicity countries.
    - Interpretation: policymakers may have more control in low-synchronicity countries; high synchronicity does not render MPPs ineffective.
  - Aggregate and tool-specific effects (Annex Table 4.1 and Figure 6):
    - Tighter macroprudential tools targeting bank capital and credit conditions are associated with lower house price synchronicity.
    - Most negative associations: supply-side (capital) measures (including countercyclical capital buffers).
    - Loan-targeted measures (including LTV limits) and supply-side loan-targeted tools (limits on foreign currency loans) lessen correlations with global and regional house price cycles.
    - Fiscal-based measures (ad valorem and buyer’s/seller’s stamp duty taxes) associated with declines in synchronicity but to a lesser extent.
    - Results conditional on positive credit gaps (Annex Table 4.2) broadly similar though some coefficients differ in significance.
- Representative MPP coefficient examples (Annex Table 4.1, unconditional sample; all regressions use 3,520 observations):
  - 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: capital: -0.230*** (0.074); -0.174** (0.071)
  - Supply side: loans: -0.092** (0.044); -0.123** (0.055)
- Representative MPP coefficient examples (Annex Table 4.2, conditional on positive credit gaps; 2,139 observations):
  - Global factor (FCI): -0.127*** (0.040); -0.159*** (0.042)
  - Business cycle synchronicity with the region: 0.043** (0.019)
  - Bank integration with the region: 0.036*** (0.012)
  - LTV: -0.028 (0.051); -0.131* (0.074)
  - Fiscal-based measures: -0.257*** (0.074)
  - Supply side: capital: -0.194** (0.096)
  - All loan-targeted: -0.033 (0.025); -0.082** (0.037)

### Additional empirical patterns and alternative measures
- Alternative synchronicity measure Synch1 (Inverse Absolute Gap Difference) at country-level (Annex Table 2.1):
  - Business Cycle Synchronization of ij: coefficients such as 0.766*** (0.254), 0.675** (0.293), 0.733*** (0.243).
  - Bilateral Bank Integration of ij: small positive coefficients (e.g., 0.006* (0.003); 0.007** (0.003)).
  - Global Factor: coefficients near zero in some specifications (e.g., -0.001 (0.001)).
  - Interaction: Bilateral Bank Integration × High Financial Openness with the World (ij): -0.019*** (0.004).
- Alternative aggregation and periods:
  - Pearson correlations over three non-overlapping seven-year periods (Annex Table 2.2) find global factor coefficients like 0.013* (0.007), 0.019** (0.007), 0.043*** (0.013), 0.051*** (0.014), and strong negative interactions with High Exchange Rate Regime (e.g., -0.117*** (0.023)).
  - Long-run annual series (Jordà-Schularick-Taylor dataset, 1870–2013) show large positive coefficients for Business Cycle Synchronization (e.g., 0.902** (0.385) for Synch1; and 0.042*** (0.015) for quasi-correlation).

### Key summary statistics (selected; preserve values as reported)
- Country-level house price synchronization [Synch1]: 0.10
- City-level house price synchronization [Synch1]: 0.10
- Business cycle synchronization [Synch1]: country-level 0.01; city-level 0.02
- House price synchronization [Quasi-correlation]: country-level 0.84; city-level 0.99
- Business cycle synchronization [Quasi-correlation]: country-level 1.33; city-level 1.28
- Bilateral bank integration of ij: country-level 1.04; city-level 0.97
- Global factor (global liquidity): country-level 3.90; city-level 4.48
- Global liquidity : AE-AE pairs: 2.51; 3.09
- Global liquidity : EM-EM pairs: 0.95; 0.95
- Global liquidity : AE-EM pairs: 2.84; 3.10
- Global liquidity : Sample with high capital account openness: 2.84; 3.25
- Global liquidity : Rest of the sample: 2.77; 3.22
- Global liquidity : Sample with high FX regime: 1.01; 1.93
- Global liquidity : Rest of the sample: 3.76; 4.04
- Global liquidity : Sample with high financial openness: 1.11; 0.71
- Global liquidity : Rest of the sample: 3.81; 4.52
- Global liquidity : Pre-crisis sample: 2.13; 1.99
- Global liquidity : GFC sample: 2.24; 5.84
- Global liquidity : Post-GFC sample: 1.62; 2.31

### Policy implications (drawn from empirical results)
- Global liquidity and loose global financial conditions can amplify cross-border co-movement in house prices, implying that global financial cycles can complicate domestic housing stability.
- Exchange rate flexibility can attenuate the transmission of global financial conditions to domestic house price synchronicity; exchange rate policy may play a role in dampening external amplification of domestic housing cycles.
- Bilateral banking integration and business cycle synchronicity increase vulnerability to cross-border propagation of housing shocks; monitoring and managing bilateral financial exposures matters for housing stability.
- Macroprudential policies targeted at domestic vulnerabilities can also reduce house price synchronicity with regional and global markets, suggesting macroprudential tools retain effectiveness even amid heightened global synchronization.

*Source: wp18250 — “6. Impact of Macroprudential Measures on House Price Synchronicity” (IMF Working Paper content as provided).*

### REFERENCES ___________________________________________________________27

### REFERENCES ___________________________________________________________27

### Figures

- 1. House Price Gap Synchronicity Across Countries and Cities  _______________________6
- 2. House Price Synchronicity and Transmission of External Shocks ____________________7
- 3. Impact of Global Financial Conditions on House Price Synchronization _____________13
- 4. House Price Interconnectedness Among Countries vs. Cities ______________________18
- 5. Average House Price Growth and Demand-side Macroprudential Policies ____________24

*Source: wp18250 - REFERENCES ___________________________________________________________27*

### 6. Impact of Macroprudential Measures on House Price Synchronicity  ________________25

### 6. Impact of Macroprudential Measures on House Price Synchronicity

### Overview and research questions
- Paper analyzes role of bilateral financial linkages and global financial conditions above-and-beyond business cycle synchronicity as drivers of house price synchronicity.
- Empirical scope:
  - Bilateral panel data at country-pair level with nearly 50,000 observations.
  - Major city-pair level with nearly 70,000 observations.
  - Coverage: over 40 economies and over 70 cities, with coverage spanning through end-2016.
  - Baseline econometric analysis: quarterly frequency from 1990 to 2016, for 40 countries.
- Core research questions:
  1. Do global financial conditions amplify 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 business cycle synchronicity?
  3. What is the role of institutional factors in mitigating or amplifying the impact of global financial conditions on house price synchronicity?
  4. Do macroprudential policies remain effective in addressing domestic vulnerabilities amid heightened house price synchronicity?

### Data and measurement (main variables and construction)
- House price gap synchronicity:
  - Measured using instantaneous quasi-correlation (Morgan, Rime, and Strahan 2004).
  - Synchronicity computed at time-series level (not bounded between -1 and 1).
  - House price gaps = cyclical component of real house prices extracted with Christiano and Fitzgerald (2003) band-pass filter, with maximum length of 30 years to capture medium-term financial cycles (20 years max for emerging market economies).
  - Robustness: Hodrick and Prescott (1997) filter with lambda of 400,000 used as a check; CF filter chosen to avoid tail bias.
- Business cycle synchronicity:
  - Analogous instantaneous quasi-correlation measure using output gaps from CF band-pass filter (maximum length adjusted for business cycles).
- Bilateral banking integration:
  - Measured using BIS locational banking statistics on residency basis.
  - Bilateral banking integration = ln[(bilateral claims i→j + bilateral claims j→i) / (GDP_i + GDP_j)] * 100.
  - Mirror data asymmetry addressed by averaging assets and liabilities per Kalemli-Ozcan et al. (2013a; 2013b).
- Global financial conditions (global factor):
  - Main proxy: changes in BIS global liquidity (changes in banks’ cross-border claims denominated in all currencies plus local claims in foreign currency in percent of global GDP).
  - Robustness proxies: Global FCI and U.S. FCI (IMF 2017 methodology), CBOE VIX, U.S. shadow interest rates (Wu and Xia 2016; Krippner 2013).
- Institutional and other controls:
  - High capital account openness: Chinn-Ito index (de jure).
  - High exchange rate regime/flexibility: Ilzetzki, Reinhart, and Rogoff (2017) de facto index.
  - High financial openness: Lane and Milesi-Ferretti (2007) de facto index.
  - “High” defined as dummy = 1 when both countries in the pair are in the top fifth of the institutional characteristic during a given quarter (80th percentile), with robustness checks using 75th or 66th percentiles.
  - Dummy variables for both countries being advanced economies, emerging market economies, or advanced–emerging market pairs.

### Empirical strategy and specification
- Baseline country-pair regression (quarterly, 1990–2016, 40 countries):
  - Dependent variable: house price gap synchronicity between country-pair i and j at quarter t.
  - Regressors (all lagged by one quarter): business cycle synchronicity; bilateral financial integration; global factor (changes in global liquidity); interaction terms of global factor with dummies for high institutional characteristics; linear and quadratic time trends; country-pair fixed effects.
  - Country-pair fixed effects capture time-invariant idiosyncratic factors (e.g., geographic proximity, supply-side/regulatory considerations).
  - Standard errors: multi-way clustered (at country i, country j, and time level, where appropriate); various clustering alternatives used in robustness checks.

### Main empirical findings
- Four headline findings:
  1. Global factors matter: Abundant global liquidity and loose global financial conditions (in addition to other global factors such as global interest rates) are positively associated with house price synchronicity across country-pairs and major city-pairs. Result robust when controlling for bilateral macro-financial linkages including business cycle synchronicity and banking integration.
  2. Exchange rate flexibility dampens global influence: Greater exchange rate flexibility attenuates the positive impact of global factors on house price synchronicity. This interaction is statistically significant at 1 percent confidence interval and robust across specifications.
  3. Bilateral linkages matter: Past co-movement in business cycles and bilateral bank linkages are positively associated with house price synchronicity (evidence documented using alternative synchronicity measures in Annex II).
  4. Macroprudential policies can reduce synchronicity: Macroprudential policies aimed at tackling domestic vulnerabilities may additionally reduce a country’s house price synchronicity with the rest of the region and the world.
- Heterogeneity by country group:
  - Impact of global financial conditions on house price synchronicity is higher between advanced economies than between emerging market economy pairs.
  - For advanced economies, the impact is statistically significant and positive; for emerging market economies and advanced–emerging market pairs, the impact is not statistically significant at conventional levels when using the most stringent standard error clustering.
- Time variation:
  - Positive impact of global liquidity on house price synchronicity was substantially higher prior to the global financial crisis (GFC), consistent with association between the global pre-GFC house price boom and abundant global liquidity in that period.
- Additional findings on synchrony drivers:
  - Interest rate synchronicity is a statistically significant driver of house price synchronicity on its own. However, its significance above and beyond other financial factors (global liquidity and bilateral banking linkages) is robust only under less stringent clustering of standard errors.
  - Trade integration included as a control is not statistically significant.
  - Equity price synchronicity included as a control does not consistently have a statistically significant relationship with house price synchronicity.
- Standardized-coefficient reporting:
  - Figure 3 presents standardized coefficients for comparability; standard deviation of the country-level dependent variable is approximately 0.85 (see Annex Table 1.3).

### Robustness checks and sensitivity
- Robustness across alternative global-factor proxies: U.S. FCI, Global FCI, CBOE VIX, U.S. shadow rates (Wu and Xia 2016; Krippner 2013) — main sign and magnitude broadly unchanged, though statistical significance may decline under the most stringent clustering.
- Alternative synchronicity measures: Annex II explores alternative measures capturing medium-term dynamics; results on bilateral banking integration and business cycle synchronicity remain positive and robust.
- Alternative estimators and clustering:
  - Explored clustering at country-pair level, two-way clustering (country i and country j), two-way clustering (country-pair and time), and Huber/White/sandwich estimator; less restrictive clustering raises statistical significance as expected.
- Additional time controls: year fixed effects and linear time trends produce little change to main results.
- Robustness to different cutoff percentiles for defining “high” institutional characteristics (75th, 66th) checked.

### Policy implications (drawn from empirical results)
- Global liquidity and loose global financial conditions can amplify cross-border co-movement in house prices, implying that global financial cycles can complicate domestic housing stability.
- Exchange rate flexibility can attenuate the transmission of global financial conditions to domestic house price synchronicity; exchange rate policy may play a role in dampening external amplification of domestic housing cycles.
- Bilateral banking integration and business cycle synchronicity increase vulnerability to cross-border propagation of housing shocks; monitoring and managing bilateral financial exposures matters for housing stability.
- Macroprudential policies targeted at domestic vulnerabilities can also reduce house price synchronicity with regional and global markets, suggesting macroprudential tools retain effectiveness even amid heightened global synchronization.

*Source: wp18250 - 6. Impact of Macroprudential Measures on House Price Synchronicity (IMF working paper content as provided).*

### conclusions. Finally, further robustness checks were employed by dropping one country-pair

### conclusions. Finally, further robustness checks were employed by dropping one country-pair at a time as well.

### Main regression findings (country-pair level)
- Dependent Variable: House Price Gap Synchronization of Country Pair i and j (quasi-correlation).
- Business Cycle Synchronization of ij:
  - Coefficients reported across tables: 0.025*, 0.026*, 0.026*, 0.026*, 0.026**, 0.026***, 0.026**, 0.026***, 0.039***, 0.043***, 0.022*, 0.023*, 0.022***, 0.024***, 0.014***, 0.014***, 0.016***, 0.016***, 0.026* (multiple specifications shown).
  - Standard errors shown in parentheses for representative specifications: (0.013), (0.014), (0.011), (0.012), (0.005), (0.005).
- Bilateral Bank Integration of ij:
  - Coefficients across specifications: -0.011, 0.012, 0.012, 0.011, 0.022, 0.022, 0.012, -0.016, 0.012, 0.022, 0.016, 0.031, 0.004, 0.018, 0.015, 0.031, -0.003, 0.002, 0.003, 0.007.
  - Representative standard errors: (0.033), (0.031), (0.036), (0.035), (0.032), (0.034).
- Global Factor (global liquidity):
  - Coefficients reported: 0.016**, 0.016**, 0.020**, 0.019***, 0.019**, 0.018**, 0.022*, 0.016**, 0.019***, 0.015**, 0.018***, 0.015***, 0.018***, 0.016**, 0.019***, 0.017**, 0.019***, 0.060***, 0.076***, 0.037**, 0.049***, 0.003**, 0.005***, 0.049***, 0.057***, 0.026***, 0.032***.
  - Representative standard errors: (0.006), (0.007), (0.008), (0.007), (0.013), (0.001), (0.005).
- GFC Dummy and Post-GFC Dummy:
  - GFC Dummy coefficient: -0.137** (standard error (0.060)).
  - Post-GFC Dummy coefficient: -0.044 (standard error (0.052)).
- Interactions and other reported coefficients:
  - Global Factor interacted with x High Exchange Rate Regime (ij) (15 categories; high = more flexible): -0.023*** (standard error (0.008)) and further robustness values -0.024***, -0.024***, -0.023***, -0.023*** in various specifications (standard errors (0.008), (0.008), (0.004), (0.008), (0.008)) and additional interactions reported as -0.069**, -0.056*, -0.011**, -0.025*, -0.026** (standard errors (0.034), (0.033), (0.004), (0.014), (0.011)).
  - Global Factor interacted with x EMEs-EMEs Dummy: -0.001 (standard error (0.009)).
  - Global Factor interacted with x EMEs-AEs Dummy: 0.000 (standard error (0.006)).
  - Global Factor interacted with x High Capital Account Openness with the World: -0.002 (standard error (0.005)).
  - Global Factor interacted with x High Financial Openness with the World (ij): 0.003 (standard error (0.006)).
  - GFC Period Dummy interacted with Global Factor: -0.025* (standard error (0.012)).
  - Post-GFC Period Dummy interacted with Global Factor: -0.029 (standard error (0.018)).
  - GFC Period Dummy interacted with Business Cycle Synchronization of ij: -0.032 (standard error (0.038)).
  - GFC Period Dummy interacted with Bilateral Bank Integration of ij: -0.022 (standard error (0.035)).
  - Post-GFC Period Dummy interacted with Business Cycle Synchronization of ij: -0.039 (standard error (0.035)).
  - Post-GFC Period Dummy interacted with Bilateral Bank Integration of ij: 0.010 (standard error (0.033)).

### Robustness checks and alternative controls
- Tables report multiple robustness exercises:
  - Robustness Checks: Global Factors (Table 2) and Additional Controls (Table 3).
  - Controls include: Interest Rate Synchronization of ij; Bilateral Trade Integration of ij; Equity Return Synchronization of ij; US Shadow rate (Krippner); Global liquidity; US FCI (↑ = loosening); Global FCI (↑ = loosening); VIX (Inverse); US Shadow rate (Wu Xia).
- Interest Rate Synchronization of ij coefficients in some specifications: 0.016, 0.013, 0.016**, 0.013* (standard errors (0.040), (0.041), (0.008), (0.008)).
- Bilateral Trade Integration of ij: coefficients 0.008 and -0.001 (standard errors (0.038) and (0.040)).
- Equity Return Synchronization of ij: coefficients 0.010 and 0.010 (standard errors (0.009), (0.009)).
- Observations and R-squared reported across specifications:
  - Representative Observations counts: 65,450; 65,343; 49,384; 43,871; 46,708; 47,353; 48,890; 46,215; 48,973; 46,308; 49,384; 46,708; 49,384; 46,708.
  - Representative R-squared values: 0.227, 0.354, 0.251, 0.230, 0.233, 0.224, 0.223, 0.241, 0.232, 0.230, 0.224, 0.228, 0.222, 0.232.
- Clustering and fixed effects:
  - Standard errors are three-way clustered (at country i, country j, and date) except regression (10) where errors are two-way clustered (at country i, country j).
  - Specifications include combinations of Time FE, Country-Pair FE, country*time FE, Quadratic Trend, and Country-Pair FE as indicated per table.
  - Clustering variations explored extensively in Table 4 (Clustering of Standard Errors) with many specifications reporting multi-way, two-way, VCE robust, and no clustering options.

### Key empirical patterns highlighted
- A positive and statistically significant association of the Global Factor (global liquidity) with house price gap synchronization appears repeatedly (coefficients often around 0.016–0.019 with significance levels ** or *** in many specifications).
- Business Cycle Synchronization of ij is consistently positive and often statistically significant (coefficients commonly around 0.025–0.026, with significance at * or ** or *** depending on specification).
- The interaction of the Global Factor with a High Exchange Rate Regime (more flexible) is generally negative and statistically significant (e.g., -0.023***), indicating the global factor’s effect on house price gap synchronization is attenuated when both countries have more flexible exchange rate regimes in multiple specifications.
- Bilateral Bank Integration of ij shows mixed results, with coefficients small in magnitude and varying in sign across specifications; many coefficients are statistically insignificant in the reported specifications.

### Additional empirical scope: city-level analysis (transition)
- Motivation: house prices in major cities may move in tandem due to increasing global presence even if country-level house prices do not.
- Methodology overview:
  - First step: explore house price interconnectedness dynamics through a network analysis.
  - Second step: analyze drivers of city-level house price synchronicity empirically.
- Sample construction note:
  - 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 combining the Top 30 cities in global investors’ ranking by Cushman & Wakefield’s (2017) Global Capital Markets 2017 report.
  - If none of the cities in a country (where data are available) are chosen based on the four pillars, the largest city by population in the country is included.
  - An additional sample with 44 major cities off the above sample is also constructed.

*Source: Authors’ estimates (WP/18/250 conclusions and robustness tables).*

### 1. Country-Level House Price Interconnectedness 2. City-Level House Price Interconnectedness

### wp18250 - 1. Country-Level House Price Interconnectedness 2. City-Level House Price Interconnectedness

### Network analysis: city-level interconnectedness
- Data and scope:
  - Vector autoregression of country-level/city-level house price growth rates (quarter over quarter).
  - Country-level span: 1990:Q1 to 2016:Q4.
  - City-level span: 2004:Q1 to 2017:Q2.
  - Node size based on the city’s total outward spillovers.
  - Pink nodes represent advanced economies and gray nodes represent emerging market economies.
  - Only links above the 50th percentile for country-level and 66th percentile for city-level are considered.
  - Figure layout based on Fruchterman and Reingold (1991); plotted using the “qgraph” R package.
- Key qualitative finding:
  - Cities attractive to global investors can be central in the city-level network even when their countries lie at the periphery (example: Tokyo and Rome are centrally located near global financial centers such as New York and London in the city-level map while Japan and Italy are peripheral in the country-level map).

### City-level empirical strategy
- Estimation framework:
  - Bilateral panel data analysis at quarterly frequency from 2004 to 2016 for over 70 major cities.
  - Dependent variable: synchronicity of house price gaps between city-pair i and j at quarter t (instantaneous quasi-correlation).
  - Controls include: country-level measures (business cycle synchronicity, bilateral financial linkages), and global financial conditions.
  - Global financial conditions proxied by changes in the BIS’ global liquidity in percent of global GDP.
  - Econometric specification includes city-pair fixed effects, quadratic and linear time trends, and other country-level regressors (see Annex III for methodology).

### City-level results — summary of main findings
- Global financial conditions:
  - Global financial conditions (changes in BIS global liquidity) are positively associated with city-level house price gap synchronicity.
  - Statistical significance:
    - The impact is statistically significant under multi-way clustering; significance improves from a 10 percent confidence level to a 1 percent confidence level under two-way clustering in column 4 (Tables 5 and 6).
- Exchange rate flexibility:
  - Higher exchange rate flexibility attenuates the positive association between the global factor and city-level house price synchronicity.
  - Statistical significance: significant at a 5 percent confidence level even under more stringent multi-way clustering (column 7 in Tables 5 and 6).
- Advanced vs emerging economies:
  - The impact of global financial conditions on city-level house price synchronicity is higher among city-pairs within advanced economies than among city-pairs within emerging economies or advance-emerging economy pairs (column 5 in Tables 5 and 6).
  - For advanced economies the impact remains statistically significant under more stringent clustering; the interaction term for advance-emerging pairs is significant only under less stringent two-way clustering; the interaction term for emerging economies is not statistically significant under two-way clustering.
- Financial openness:
  - De jure financial openness (Chinn-Ito index) tends to amplify the positive association between global financial conditions and city-level synchronicity, but this impact is not statistically significant under more stringent standard error clustering (column 6 in Table 6).
  - De facto financial openness (Lane and Milesi-Ferretti (2007) measure) yields no statistically significant results.
- Pre-crisis relation:
  - 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).
- Weighting remark:
  - The coefficients in the regression analysis are weighted by the number of major cities in each country.

### Representative coefficient estimates and model diagnostics (Tables 5 and 6)
- Table 5. House Price Gap Synchronicity at City Level and Global Factors — Two-Way Clustering
  - Business Cycle Synchronization of ij: 0.011 (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 ij: 0.008 (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 terms shown selectively:
    - x EMEs-AEs Dummy: -0.018** (0.008)
    - x High Capital Account Openness with the World: 0.018** (0.008)
    - x High Exchange Rate Regime (15 categories; high = more flexible): -0.017** (0.008)
    - x High Exchange Rate Regime (6 categories; high = more flexible): -0.012* (0.007)
    - x High Financial Openness with the World (ij): 0.001 (0.016)
  - GFC Period Dummy Interactions:
    - 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 Interactions:
    - 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 Dummy: 0.010 (0.038)
  - Post-GFC Dummy: 0.014 (0.049)
  - Observations: 66,575 | 66,572 | 66,575 | 66,575 | 66,575 | 59,353 | 63,691 | 63,691 | 62,588 | 66,575
  - R-squared: 0.260 | 0.343 | 0.260 | 0.254 | 0.256 | 0.265 | 0.251 | 0.252 | 0.268 | 0.260
  - Clustering: Two-way (at country ij, and date) for all columns shown.
  - Fixed effects and trends: combinations of Time FE, Country-Pair FE, country*time FE, Quadratic Trend used as indicated in table.

- Table 6. House Price Gap Synchronicity at City Level and Global Factors — Multi-Way Clustering
  - Business Cycle Synchronization of ij: 0.011 (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 ij: 0.008 (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 terms shown selectively:
    - x EMEs-EMEs Dummy: -0.019 (0.016)
    - x EMEs-AEs Dummy: -0.018 (0.014)
    - x High Capital Account Openness with the World: 0.018 (0.014)
    - x High Exchange Rate Regime (15 categories; high = more flexible): -0.017** (0.008)
    - x High Exchange Rate Regime (6 categories; high = more flexible): -0.012 (0.008)
    - x High Financial Openness with the World (ij): 0.001 (0.024)
  - GFC Period Dummy Interactions:
    - 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 Interactions:
    - 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 Dummy: 0.010 (0.050)
  - Post-GFC Dummy: 0.014 (0.045)
  - Observations: 66,575 | 66,572 | 66,575 | 66,575 | 66,575 | 59,353 | 63,691 | 63,691 | 62,588 | 66,575
  - R-squared: 0.260 | 0.343 | 0.260 | 0.254 | 0.256 | 0.265 | 0.251 | 0.252 | 0.268 | 0.260
  - Clustering: Multi-way (three-way clustered at country i, country j, and date) for all columns shown.

### Extensions — macroprudential policies (MPPs) and house price synchronicity
- Motivation and interpretation:
  - MPPs targeted at dampening domestic vulnerabilities in the financial and housing sectors may indirectly weaken the correlation of house price cycles with regional and global cycles, potentially giving policymakers greater control over local house price dynamics.
  - Because house prices in financially integrated countries are also driven by capital flows from global investors and global financial conditions, the net effect of MPPs on synchronicity may be ambiguous.
  - Literature context:
    - Measures targeting housing finance (Akinci and Olmstead-Rumsey 2017) and those complementing monetary policy (Bruno, Shim, and Shin 2017) appear most effective in mitigating house price growth.
    - Risk-weighting and provisioning requirements show no robust evidence of effectiveness (Kuttner and Shim 2016).
- Empirical strategy for MPPs:
  - Panel regression for 41 countries from 1990:Q2 through 2016:Q4.
  - Dependent variable HPS: house price cycle synchronicity (instantaneous quasi-correlation) with regional or global cycle.
  - Key regressors: business cycle synchronicity (BCS) with region or rest of world, global financial conditions, financial integration, institutional characteristics, and MPP indicators or macroprudential group indices (loan-targeted, supply-side capital/general/loans, demand-side, fiscal-based measures).
  - Specification includes country fixed effects; all regressors lagged one quarter.
- MPPs — empirical results:
  - Demand-side MPPs (e.g., loan-to-value (LTV) limits, debt-service-to-income limits):
    - Figure 5 evidence: Average year-over-year house price growth for high-synchronicity and low-synchronicity countries around implementation of demand-side MPPs (total number of demand-side events is 47; t = 0 is first quarter of implementation within a ±5-quarter window).
    - Before implementation, house prices grew similarly in high- and low-synchronicity countries; after implementation, house price growth declined in both groups, with a stronger and more sustained decline in low-synchronicity countries.
    - Interpretation: policymakers may have more control in low-synchronicity countries; high synchronicity does not render MPPs ineffective.
  - Aggregate and tool-specific effects (Figure 6 and Annex Table 4.1):
    - Tighter macroprudential tools targeting bank capital and credit conditions are associated with lower house price synchronicity.
    - Most negative association: supply-side (capital) measures, which include countercyclical capital buffers.
    - Loan-targeted measures (including LTV limits) and supply-side loan-targeted tools (limits on foreign currency loans) lessen correlations with global and regional house price cycles.
    - Fiscal-based measures (ad valorem and buyer’s/seller’s stamp duty taxes) are associated with declines in synchronicity but to a lesser extent than other MPPs.
    - When focusing on credit boom periods, results are qualitatively and quantitatively similar though slightly less significant (Figure 6, panel 2 and Annex Table 4.2).
- Definitions used in MPP analysis (as reported in notes):
  - Supply side (loans): limits on credit growth, loan loss provisions, loan restrictions, limits on foreign currency loans.
  - Supply side (capital): capital requirements, conservation buffers, the leverage ratio, countercyclical capital buffer.
  - Supply side (general): reserve requirements, liquidity requirements, limits on foreign exchange positions.
  - Demand-side: limits to debt-service-to-income and LTV ratios.
  - All loans: includes demand side and supply side (loans).
  - Fiscal-based measures: taxes such as ad valorem, seller’s and buyer’s stamp duty, or other taxes.
- Statistical note:
  - Figure 6 depicts estimated average effects; shaded bars show statistically significant standardized coefficients at the 10 percent confidence level.
  - Regressions control for business cycle synchronicity, financial integration, global financial conditions; all regressors lagged one quarter; estimated using data for 41 countries spanning 1990:Q2–2016:Q4.

*Source: Authors’ estimates (wp18250).*

### 1. Unconditional

### 1. Unconditional

### Main conclusions
- Using various proxies for global financial conditions, the paper confirms that the abundance of liquidity owing to accommodative financial conditions is positively associated with house price synchronicity at country and city levels.
- Higher house price synchronicity can benefit countries in some cases, but positive association with global financial conditions could also suggest stronger transmission of external shocks into the domestic economy or to major cities within an economy.
- 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.
- The positive association of global financial conditions with house price synchronicity was stronger preceding the global financial crisis.

### Exchange rate regime and attenuation
- 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.
- Major cities located in countries with more flexible exchange rate regimes also possess the ability of attenuating the impact of global financial conditions on city-level house price synchronicity.

### Macroprudential policies and house price sensitivity
- 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.
- 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.

### Key empirical findings (selected)
- Country-level house price synchronization [Synch1]: 0.10
- City-level house price synchronization [Synch1]: 0.10
- Business cycle synchronization [Synch1]: country-level 0.01; city-level 0.02
- House price synchronization [Quasi-correlation]: country-level 0.84; city-level 0.99
- Business cycle synchronization [Quasi-correlation]: country-level 1.33; city-level 1.28
- Bilateral bank integration of ij: country-level 1.04; city-level 0.97
- Global factor (global liquidity): country-level 3.90; city-level 4.48
- Global liquidity : AE-AE pairs: 2.51; 3.09
- Global liquidity : EM-EM pairs: 0.95; 0.95
- Global liquidity : AE-EM pairs: 2.84; 3.10
- Global liquidity : Sample with high capital account openness: 2.84; 3.25
- Global liquidity : Rest of the sample: 2.77; 3.22
- Global liquidity : Sample with high FX regime: 1.01; 1.93
- Global liquidity : Rest of the sample: 3.76; 4.04
- Global liquidity : Sample with high financial openness: 1.11; 0.71
- Global liquidity : Rest of the sample: 3.81; 4.52
- Global liquidity : Pre-crisis sample: 2.13; 1.99
- Global liquidity : GFC sample: 2.24; 5.84
- Global liquidity : Post-GFC sample: 1.62; 2.31

### Findings on bilateral linkages and alternative synchronicity measures
- Using Synch1 (Inverse Absolute Gap Difference) as an alternative, both business cycle synchronicity (BCS) and bilateral banking integration have statistically significant positive association with house price synchronicity (columns 1 to 3 in Table A2.1).
- The impact of both BCS and bilateral bank integration on house price gap synchronicity is comparable in magnitude (baseline specification, column 4; further standardized in Figure 3).
- The impact of bilateral banking integration on house price gap synchronicity is lower among emerging market economy country pairs compared to advanced economy country-pairs (column 5).
- 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).
- The positive impact of banking integration on house price gap synchronicity observed in the baseline specification is not statistically significant for the post-GFC period (column 10).

### Data sources and sample notes (selected)
- Real House Price Indices: Bank for International Settlements; CEIC Data Co. Ltd; Emerging Markets Economy 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 GDP, Inflation, Nominal GDP: Haver Analytics; OECD; IMF Global Data Source database; IMF, World Economic Outlook database.
- Total Bank Claims and Liabilities: Bank for International Settlements; IMF staff calculations.
- Financial Openness: Lane-Milesi-Ferretti dataset (2007; updated); Chin-Ito index (Chinn and Ito 2006 dataset, updated).
- Exchange Rate Regime: Ilzetzki, Reinhart, and Rogoff (2017) dataset.
- Macroprudential Policies: Alam and others (forthcoming) — tools at quarterly frequency.
- Global liquidity measured as total claims of all Bank for International Settlements reporters vis-à-vis the world, in percent of world GDP.
- US Financial Conditions Index and Global Financial Conditions Index: IMF, October 2017 Global Financial Stability Report (Chapter 3).
- VIX: Chicago Board Options Exchange Volatility Index; US Shadow Interest Rates: Wu-Xia and Krippner Shadow Federal Funds Rates.

*Source: IMF Working Paper (wp18250), “1. Unconditional” section.*

### Annex Figure 2.1. Impact of Bilateral Linkages

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

### Measurement and Notes
- Synchronicity is measured by the Synch1 of gaps measure.
- Figure shows statistically significant standardized coefficients calculated using coefficients in specification 4 in Table 1 and respective standard deviations, presented in terms of standard deviations of the dependent variable.
- Specification controls for global financial conditions (proxied through the global liquidity), country-pair fixed effects, quadratic and linear time trends; standard errors are clustered 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.

### Standard-deviation scale shown in figure
- 0.00
- 0.01
- 0.02
- 0.03
- 0.04
- 0.05
- 0.06
- 0.07
- 0.08
- 0.09

### Variables displayed (as labeled in figure)
- Business cycle synchronization of ij
- Bilateral bank integration of ij
- Standard deviations

---

### Annex Table 2.1. House Price Gap Synchronicity at Country-Level and Bilateral Linkages

### Core regression coefficients (excerpt across specifications)
- Dependent Variable: House Price Gap Synchronization of Country Pair i and j (Synch1)
- Business Cycle Synchronization of ij:
  - (1) 0.766*** (0.254)
  - (2) 0.675** (0.293)
  - (3) 0.733*** (0.243)
  - (4) 0.657** (0.254)
  - (5) 0.658** (0.253)
  - (6) 0.746*** (0.262)
  - (7) 0.725*** (0.261)
  - (8) 0.725*** (0.262)
  - (9) 0.675** (0.253)
  - (10) 0.706** (0.337)
- Bilateral Bank Integration of ij:
  - (1) 0.006* (0.003)
  - (2) 0.007** (0.003)
  - (3) 0.012 (0.007)
  - (4) 0.009* (0.004)
  - (5) 0.007** (0.003)
  - (6) 0.007* (0.004)
  - (7) 0.007** (0.003)
  - (8) 0.004 (0.005)
- Global Factor (global liquidity):
  - (1) -0.001 (0.001)
  - (2) -0.001 (0.001)
  - (3) -0.001 (0.001)
  - (4) -0.001 (0.001)
  - (5) -0.001 (0.001)
  - (6) -0.001 (0.001)
  - (7) 0.001 (0.001)

### Bilateral Bank Integration interactions
- 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)

### Period interactions
- GFC Period Dummy Interacted with:
  - x Business Cycle Synchronization of ij: -0.080 (0.516)
  - x Bilateral Bank Integration of ij: 0.008** (0.004)
  - x Global Factor: 0.001 (0.001)
- Post-GFC Period Dummy Interacted with:
  - x Business Cycle Synchronization of ij: 0.380 (0.456)
  - x Bilateral Bank Integration of ij: 0.007 (0.005)
  - x Global Factor: 0.004 (0.003)
- GFC Dummy: 0.048*** (0.011)
- Post-GFC Dummy: 0.042*** (0.009)

### Sample, fit, and clustering
- Observations by specification: 65,450; 65,343; 49,384; 49,384; 49,384; 43,871; 46,708; 46,708; 47,353; 49,384
- R-Squared by specification: 0.353; 0.498; 0.386; 0.356; 0.356; 0.361; 0.356; 0.356; 0.360; 0.360
- Multiway Clustering: Yes for most; regression (10) uses two-way clustering (at country i, country j)
- Fixed effects: country-pair FE included; various specifications include time FE, country*time FE, quadratic trend

---

### Alternative Measure 2: Pearson Correlations (Three non-overlapping seven-year periods)

### Setup notes
- Panel of three non-overlapping seven-year periods (three non-overlapping 28 quarter periods).
- House price and business cycle synchronicity captured by bilateral Pearson correlation coefficients for the period.
- All other explanatory variables are the average values for the period; robustness checks collapse explanatory variables using last value of previous period.
- Interaction term of the global factor and foreign exchange regime remains statistically significant, in addition to the global factor itself.

### Selected coefficients (Annex Table 2.2)
- Dependent Variable: House Price Gap Synchronization of Country Pair i and j (Non-overlapping period-wise Pearson correlation)
- Business Cycle Synchronization of ij:
  - (1) 0.104* (0.061)
  - (2) 0.058 (0.062)
  - (3) 0.089 (0.060)
  - (4) 0.058 (0.062)
- Bilateral Bank Integration of ij:
  - (1) 0.019 (0.029)
  - (2) 0.008 (0.032)
  - (3) 0.029 (0.023)
  - (4) 0.037 (0.024)
- Global Factor (global liquidity):
  - (1) 0.013* (0.007)
  - (2) 0.019** (0.007)
  - (3) 0.043*** (0.013)
  - (4) 0.051*** (0.014)
- Global Factor Interacted with High Exchange Rate Regime (ij) (15 categories; high = more flexible):
  - (3) -0.117*** (0.023)
  - (4) -0.168*** (0.064)
- Observations: 1,660; 1,553; 1,660; 1,553
- R-squared: 0.369; 0.380; 0.375; 0.380
- Clustering: country-pair

---

### Alternative Measure 3: Synchronicity with Longer Time Series (Jordà-Schularick-Taylor dataset)

### Setup notes
- Annual observations from 1870 to 2013 for 17 advanced economies (Jordà-Schularick-Taylor dataset).
- Regressors are lagged by one year; FE = fixed effects.

### Selected coefficients (Annex Table 2.3)
- Dependent Variable: House Price Gap Synchronization of Country Pair i and j
- Synch1 (columns 1–4):
  - Business Cycle Synchronization of ij: 0.902** (0.385); 0.902** (0.311); 0.902*** (0.153); 0.902*** (0.089)
  - Observations: 9,818 (all columns)
  - R-squared: 0.143 (columns 1–4)
- Quasi-correlation (columns 5–8):
  - Business Cycle Synchronization of ij: 0.042 (0.032); 0.042* (0.024); 0.042*** (0.015); 0.042*** (0.008)
  - Observations: 9,818 (all columns)
  - R-squared: 0.071 (columns 5–8)
- Clustering: mixture of Multi-way, Two-way, VCE robust, No depending on column
- Quadratic Trend and Country-Pair FE: Yes in indicated specifications

---

### Annex III: Methodology — House Price Interconnectedness Analysis

### VAR-based interconnectedness measure
- Methodology follows Diebold and Yilmaz (2014).
- Interconnectedness defined as fraction of H-quarter-ahead forecast error variance of country/city j’s house price growth explained by country/city i’s house price growth dynamics.
- Quarterly house price growth rates computed using seasonally adjusted real house prices at country-level or city-level.
- Global financial conditions proxied by the U.S. Financial Conditions Index (FCI) constructed in line with IMF 2017.
- Estimation period:
  - Country-level analysis: 1990:Q1 to 2016:Q4
  - City-level interconnectedness analysis: 2004:Q1 to 2017:Q2
- VAR sample size noted as large (n=30); estimation uses lasso and elastic net (machine learning techniques) following Demirer et al. (2018), Song and Bickel (2011).
- Baseline VAR specification includes U.S. FCI to control for global financial conditions; robustness checks performed with alternative variables (footnote 25).

### Robustness note (footnote 25)
- Robustness checks were performed using the global FCI and the VIX index, and results are found to be very similar.

---

### Annex IV: Impact of Macroprudential Measures on House Price Synchronicity — Regression Results

### Annex Table 4.1. Unconditional Estimation Sample (selected coefficients)
- Dependent variable: house price gap synchronicity (quasi-correlation) with:
- Global factor (FCI):
  - Examples across columns: -0.051 (0.037); -0.084** (0.040); -0.050 (0.037); -0.083** (0.040); -0.058 (0.040); -0.089** (0.037)
- Business cycle synchronicity with the region:
  - Example values: 0.028 (0.020); 0.030 (0.019); 0.029 (0.019)
- Bank integration with the region:
  - Example values: 0.013 (0.013); 0.012 (0.013)
- Business cycle synchronicity with the world:
  - Example values: 0.041** (0.020); 0.043** (0.019); 0.042** (0.019)
- Bank integration with the world:
  - Example values: 0.027 (0.020); 0.026 (0.020)
- Macroprudential measures (representative coefficients):
  - 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: general: 0.042 (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)
- Observations: 3,520 (all columns)
- R-squared examples: 0.008; 0.017; 0.007; 0.015; 0.009; 0.017; 0.008; 0.015; 0.015; 0.020

### Annex Table 4.2. Conditional on Positive Credit Gaps (selected coefficients)
- Dependent variable: house price gap synchronicity (quasi-correlation) with:
- Global factor (FCI):
  - Example values: -0.127*** (0.040); -0.159*** (0.042); -0.128*** (0.040); -0.162*** (0.042)
- Business cycle synchronicity with the region:
  - Example values: 0.043** (0.019); 0.043** (0.019)
- Bank integration with the region:
  - Example values: 0.036*** (0.012); 0.037*** (0.012)
- Business cycle synchronicity with the world:
  - Example values: 0.060*** (0.017); 0.062*** (0.017)
- Bank integration with the world:
  - Example values: 0.056*** (0.013); 0.059*** (0.014)
- Macroprudential measures (representative coefficients):
  - 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: general: 0.054 (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)
- Observations: 2,139 (all columns)
- R-squared examples: 0.036; 0.052; 0.037; 0.049; 0.038; 0.049; 0.042; 0.054

*Source: Authors’ estimates (as presented in the supplied content).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18250.pdf_