## Annex 3.1–3.5. Macroprudential Regulation: Data, Dampening Effects, and Policy Responses

## Source details

**Canonical URL:** [Annex 3.1–3.5. Macroprudential Regulation: Data, Dampening Effects, and Policy Responses](https://www.imf.org/-/media/files/publications/weo/2020/april/english/annexch3.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/weo/2020/april/english/annexch3.pdf.md)
- [Structured JSON version](/-/media/files/publications/weo/2020/april/english/annexch3.pdf.json)

---

### Data sources, sample, and country coverage
- Sample: 38 emerging markets, based on the April 2020 World Economic Outlook classification, from the first quarter of 2000 to the last quarter of 2016.
- Country selection criteria:
  - a population larger than one million,
  - at least 10 years of GDP data,
  - at least 5 years of data for net capital inflows,
  - with data on macroprudential regulation from the dataset of Alam and others (2019).
- Exchange rate regime classification: coarse classification in Ilzetzki and others (2019). Flexible exchange rate regimes include bands, crawls, and managed floats (categories 2, 3, and 4). Fixed exchange rate regimes include hard pegs, currency board arrangements, horizontal bands, and de facto pegs (category 1). Freely falling exchange rate regimes (category 5) are excluded.
- Key data sources (selected indicators):
  - Capital controls: Fernandez and others (2016)
  - Chicago Board Options Exchange Volatility Index (VIX): Haver Analytics
  - Commodity terms of trade: Gruss and Kebhaj (2019)
  - Composite risk rating: International Country Risk Guide, The PRS Group, Inc, www.prsgroup.com
  - Cyclically adjusted balance: IMF, World Economic Outlook
  - Financial development: Sahay and others (2015)
  - Exchange rate regime: Ilzetzki and others (2019)
  - Expected inflation: Consensus Economics, Haver Analytics, IMF staff calculations
  - Gross capital inflows / outflows / Official Reserves / Net capital inflows: IMF, Balance of Payments and International Investment Position Statistics
  - Gross public debt / Gross public debt in foreign currency / Population: IMF, World Economic Outlook
  - Inflation expectation anchoring index: Bems and others (2018)
  - Macroprudential regulation: Integrated Macroprudential Policy (iMaPP) dataset, Alam and others (2019)
  - US monetary policy shock: Iacoviello and Navarro (2019)
  - Nominal GDP, Real GDP, other series: Haver Analytics; IMF staff calculations where noted
  - Policy rates, Real credit from banks: Bank for International Settlements; Haver Analytics; and IMF, International Financial Statistics
  - Institutional quality: Worldwide Governance Indicators
- Country coverage (selected listing): Albania; Argentina; Belarus; Bosnia and Herzegovina; Brazil; Bulgaria; Chile; China; Colombia; Costa Rica; Croatia; Dominican Republic; Ecuador; El Salvador; Georgia; Hungary; India; Indonesia; Jamaica; Jordan; Kazakhstan; Malaysia; Mexico; Morocco; Northern Macedonia; Pakistan; Paraguay; Peru; Philippines; Poland; Romania; Russia; Serbia; South Africa; Thailand; Turkey; Ukraine; Uruguay.

### Macroprudential measures (iMaPP): index and categories
- iMaPP provides information about 12 macroprudential tools, grouped into an overall index and five subcategories: Capital, Credit Demand, Credit Supply, FX Exposure, Liquidity.
- Mapping (category → measures listed as iMaPP variables, notes preserved):
  - Capital: Capital requirements (Capital) — Including risk weights; Leverage limits (LVR); Loan loss provision requirements (LLP) — Including dynamic and sector-specific provisioning (e.g. housing); Countercyclical capital buffer (CCB); Capital conservation buffer requirements (Conservation); Measures targeted at SIFIs (SIFI) — Including capital and liquidity surcharges.
  - Credit Demand: Limits to loan-to-value ratio (LTV) — Mostly targeted at housing loans; Limits to debt-service-to-income ratio (DSTI); Tax on transactions (Tax) — Including stamp duties and capital gain taxes.
  - Credit Supply: Limits on credit growth or volume (LCG) — Including penalties for exceeding limits; Loan restrictions (LoanR) — Tailored LCG conditional on loan and bank characteristics, or other factors.
  - FX Exposure: Limits on foreign currency lending (LFC); Limits on gross open FX positions (LFX) — Including currency mismatch regulations; Reserve requirements on FC assets (RR_FCD).
  - Liquidity: Reserve requirements (RR_dom) — On domestic currency assets; Liquidity measures (Liquidity) — Including liquidity coverage ratios, liquid asset ratios, net stable funding ratios, core funding ratios, and external debt restrictions; Limits to loan-to-deposit ratio (LTD) — Including penalties for exceeding limits.
- Capital and liquidity measures account for about two thirds of the overall tightening in macroprudential regulation between 1990 and 2016.

### Empirical framework and identification (Dampening effects of MPru)
- Baseline panel regression (notation preserved):
  - Y_{i,t} = α_i + β · S_{i,t} + γ · (S_{i,t} * MPru_{i,t}) + δ · (S_{i,t} * MPru_{i,t}^2) + ζ MPru_{i,t} + θ MPru_{i,t}^2 + κ · C_{i,t} + ε_{i,t}  (equation (3.1))
  - Y_{i,t}: quarterly real GDP growth for country i at time t.
  - S_{i,t}: vector of global financial shocks including US policy rate shocks (Iacoviello and Navarro (2019)), VIX (risk premium proxy), and net capital inflows.
  - Net capital inflows are instrumented to isolate global “push” variation: instrument net inflows to country i with gross inflows to the other emerging markets following Blanchard and others (2017).
  - Interaction terms with MPru_{i,t} and MPru_{i,t}^2 allow assessment of (nonlinear) dampening effects; coefficients γ reveal dampening.
  - Control vector C_{i,t} includes lagged GDP growth, lagged log real GDP per capita, institutional quality, linear trend, lagged output gap, and commodity terms of trade.
- Estimation approach:
  - Two-stage least squares (2SLS) where net capital inflows and their interactions with MPru and MPru^2 are instrumented with gross capital inflows to all other emerging markets and corresponding interactions.
  - Results presented in terms of net outflows (by inverting sign).
  - Standard errors: Driscoll and Kraay (1998) correction to account for cross-sectional dependence, autocorrelation, and heteroscedasticity.
- Derivative used to assess impact of global shocks as a function of MPru:
  - ∂Y_{i,t}/∂S_{j,t} = β + γ_j MPru_{i,t} + δ_j MPru_{i,t}^2  (equation (3.2)) — nonlinear in MPru.

### Baseline regression key coefficients (column (1) — full sample)
- Lag dependent variable: 0.071 (standard error 0.057)
- Lag output gap: -0.378*** (0.035)
- Lag ln real GDP per capita: -0.880* (0.505)
- Linear trend: 0.003 (0.005)
- Commodity terms of trade: 0.050 (0.035)
- US monetary policy shock: -0.158 (0.168)
- Ln VIX: -0.712*** (0.182)
- Net outflows: -0.186*** (0.046)
- Macroprudential regulation (MPru): -1.380*** (0.401)
- US monetary policy shock * MPru: -0.100 (0.109)
- Ln VIX * MPru: 0.631*** (0.143)
- Net outflows * MPru: 0.108*** (0.036)
- MPru^2: 0.237 (0.209)
- Observations: 2,260
- Countries: 38
- F-statistic: 73.1
- Note: MPru is divided by 10 to ease visualization of coefficients. Sample: 2000:Q1 to 2016:Q4. All specifications include country fixed effects. Driscoll-Kraay standard errors in parentheses. * p < .10; ** p < .05; *** p < .01.

### Baseline results — qualitative summary and quantitative magnitudes
- VIX increases and capital outflows negatively affect GDP growth in emerging markets; US monetary policy shocks are insignificant in baseline.
- Instrumentation for net capital flows is supported by a strong F-statistic (above conventional threshold).
- Interaction coefficients indicate macroprudential regulation dampens the effects of global shocks on GDP growth; inclusion of MPru^2 and its interactions shows decreasing marginal returns to MPru.
- Quantitative examples:
  - When the VIX doubles (an increase similar to the one during the global financial crisis), an emerging market at the lowest level of macroprudential regulation in the sample would experience a decline in quarterly GDP growth by 1.8pp.
  - Net outflows worth one percent of GDP would lead to a fall in GDP growth by 0.6pp.
  - An economy with the median level of macroprudential regulation would experience a GDP decline by only 0.5pp in the case of a doubling of the VIX and by 0.3pp in the case of a 1pp capital outflow.
- Conclusion: macroprudential policy can play an important role in dampening the effects of global financial shocks in emerging markets.

### Robustness checks and heterogeneity
- Reverse causality tests:
  - Excluding observations with negative GDP growth; replacing MPru with its first lag and fourth lag; replacing time-varying MPru with time-invariant country averages.
  - Coefficients on interaction terms between MPru and VIX or net outflows remain statistically significant in all specifications.
- Omitted-variable tests:
  - Controls and interactions added for institutional quality, financial development, gross public debt (percent of GDP), gross public debt in foreign currency (percent of gross public debt), cyclically adjusted balance (percent of trend GDP), monetary policy rate, inflation expectation anchoring index, exchange rate regime, capital controls, official reserves (percent of trend GDP).
  - Dampening effects remain statistically significant across these tests; instrumentation becomes more cumbersome and F-statistic falls below ten when certain controls are included.
  - Results robust to exclusion of any single country; excluding 2008:Q3–2009:Q4 confirms baseline results. Only in some specifications with squared MPru does significance on the VIX interaction occasionally drop below 10 percent.
- Symmetry of dampening effects:
  - Extended specification distinguishes positive and negative shocks; triple interaction coefficients γ^+ and γ^- are statistically significant and similar in magnitude.
  - Wald test rejects hypothesis that γ^+ and γ^- differ statistically: dampening effects operate symmetrically for positive and negative VIX and net outflow shocks.
- Disaggregated categories:
  - For VIX shocks: measures on credit demand, foreign exchange exposure, and liquidity exhibit significant interactions.
  - For capital outflows: interaction terms significant for measures targeting capital, credit demand, and credit supply (capital-based measures to be taken with caution due to low F-statistic).
  - Overall: dampening effects arise from a broad range of measures rather than a single category.

### Effects on credit and exchange rates
- Dependent variables: real bank credit growth, nominal effective exchange rate (NEER), real effective exchange rate (REER) (positive values denote appreciation).
- Findings:
  - VIX and net capital shocks lead to depreciation of the NEER and the REER.
  - Net outflows tend to slow down real credit growth.
  - Macroprudential regulation reduces sensitivity of NEER and REER to movements in the VIX.
  - MPru dampens effects of net capital outflows on both credit growth and the exchange rate.
- Selected coefficients (Annex Table 3.2.6):
  - Lag dependent variable: Real credit growth 0.226*** (0.051); NEER appreciation 0.206*** (0.041); REER appreciation 0.174*** (0.043).
  - Ln VIX: Real credit growth -1.232 (0.981); NEER appreciation -2.469** (1.031); REER appreciation -2.248** (1.000).
  - Net outflows: Real credit growth -0.563* (0.288); NEER appreciation -0.396** (0.173); REER appreciation -0.410*** (0.158).
  - MPru: Real credit growth -1.065 (1.125); NEER appreciation -2.552* (1.332); REER appreciation -2.412* (1.353).
  - Ln VIX * MPru: NEER appreciation 1.066** (0.521); REER appreciation 1.037** (0.520).
  - Net outflows * MPru: Real credit growth 0.173* (0.093); NEER appreciation 0.134** (0.065); REER appreciation 0.129** (0.060).
  - Observations: Real credit growth 2,049; NEER 2,164; REER 2,164. Countries: 38, 35, 35.

### Macroprudential regulation and monetary policy responses
- Panel specification (equation 3.4):
  - I_{i,t} = α_i + β∙S_{i,t} + γ∙(S_{i,t} * MPru_{i,t}) + ζ MPru_{i,t} + κ∙C_{i,t} + ε_{i,t}
  - Dependent variable I_{i,t} denotes policy rates. Global shocks S_{i,t}: (i) US policy rate, (ii) VIX, (iii) net capital outflows.
- Measurement notes:
  - US policy rate: federal funds rate except during the zero lower bound period where implied policy rate from Wu and Xia (2015) is used.
  - Regression includes expected inflation (next 12 months), output gap, real credit growth, commodity terms of trade; country fixed effects included.
  - Net capital inflows instrumented; estimation via 2SLS; Driscoll and Kraay standard errors.
  - Sample: quarterly data for up to 30 EMs (2000–2016); hard pegs excluded.

### Baseline results on policy rates (Annex Table 3.3.1)
- Selected coefficients:
  - US policy rate: (1) 0.597*** (0.068); (2) 0.359*** (0.034); (3) 0.295*** (0.037); (4) 0.366*** (0.066).
  - Ln VIX: (1) 1.396*** (0.482); (2) 0.587** (0.283); (3) 0.599** (0.302); (4) 2.098*** (0.451).
  - Net outflows: (1) 0.404*** (0.118); (2) 0.347*** (0.079); (3) 0.337*** (0.084); (4) 0.418*** (0.127).
  - Expected inflation, next 12 months: (2)/(3) 1.235*** / 1.224*** / 1.227*** (0.075–0.077).
  - Output gap: (2)/(3) 0.250*** / 0.148** / 0.208*** (0.082–0.071).
  - Macroprudential regulation (MPru) in full specification (4):
    - Level: 0.220*** (0.053)
    - US policy rate * MPru: -0.013*** (0.004)
    - Ln VIX * MPru: -0.116*** (0.020)
    - Net outflows * MPru: -0.004 (0.004)
- Sample and fit:
  - Observations: (1) 1,360; (2) 1,262; (3) 1,250; (4) 1,250.
  - Countries: (1) 30; (2) 25; (3) 25; (4) 25.
  - F-statistic: (1) 73.98; (2) 56.42; (3) 49.70; (4) 21.98.

### Interpretation: monetary policy responses
- Policy rates in emerging markets respond procyclically to tightening global financial conditions: they increase with rises in the US policy rate, the VIX, and with higher net capital outflows.
- Interaction results in full specification:
  - Significant and negative interaction terms between MPru and (i) the US policy rate and (ii) the VIX, implying higher MPru mutes the procyclical monetary response.
  - If MPru is sufficiently tight, the monetary response to increases in US policy rates and the VIX can become countercyclical (emerging markets lowering policy rates).
  - MPru does not appear to affect the monetary policy response to net capital outflows in column 4 (interaction insignificant).

### Robustness to reverse causality (policy rates)
- Lagged MPru specifications: one-quarter lag and one-year lag of MPru reported; interactions of MPru with US policy rate and Ln VIX remain negative and statistically significant in lagged specifications.
- Country-average MPru: interaction of MPru with the VIX remains negative and statistically significant; interaction with the US policy rate loses significance, but MPru supports a more countercyclical response to net capital outflows in country-average specification (Net outflows * MPru -0.014** (0.007)).
- Summary: lagging MPru and using country-average MPru alleviate reverse causality concerns; negative interactions between MPru and global shocks are robust.

### Cross-country spillovers (Annex Table 3.4.1)
- No evidence of negative spillovers from higher domestic macroprudential regulation to other countries.
- Spillovers tend to be positive with respect to net outflows: higher average MPru in other EMs dampens the adverse effect of net outflows on a country’s real GDP growth.
- Dampening effects from domestic and foreign macroprudential regulation are similar in magnitude, noting that others' MPru coefficient captures a one-unit increase averaged across all other EMs.

### Policy response functions of macroprudential measures (Annex Table 3.5.1)
- Main empirical result: EMs tend to loosen macroprudential policies when global financial conditions tighten, and tighten when global conditions ease.
- Results robust to controls for expected inflation, output gap, commodity terms of trade, and real credit growth.
- Selected coefficients:
  - Net outflows (presented as net outflows so shocks are negative for EMs): (1) -0.035** (0.015); (2) -0.044** (0.023); (3) -0.038** (0.018).
  - US monetary policy shocks: (1) -0.056* (0.030); (2) -0.066** (0.031); (3) -0.056** (0.026).
  - Ln VIX: (1) -0.170** (0.077); (2) -0.160* (0.086); (3) -0.128** (0.055).
- Interpretation: behavior is consistent with prior findings that the probability of tightening macroprudential measures is higher when capital flows surge.

*Source: IMF staff compilation (Annex 3.1 and Annex 3.2, World Economic Outlook, April 2020).*

### Annex 3.1. Data Sources and Country Coverage

### Annex 3.1. Data Sources and Country Coverage

### Data sources and sample
- Sample: 38 emerging markets, based on the April 2020 World Economic Outlook classification, from the first quarter of 2000 to the last quarter of 2016.
- Country selection criteria:
  - a population larger than one million,
  - at least 10 years of GDP data,
  - at least 5 years of data for net capital inflows,
  - with data on macroprudential regulation from the dataset of Alam and others (2019).
- Exchange rate regime classification: coarse classification in Ilzetzki and others (2019). Flexible exchange rate regimes include bands, crawls, and managed floats (categories 2, 3, and 4). Fixed exchange rate regimes include hard pegs, currency board arrangements, horizontal bands, and de facto pegs (category 1). Freely falling exchange rate regimes (category 5) are excluded.

### Key data sources (selected indicators)
- Capital controls: Fernandez and others (2016)
- Chicago Board Options Exchange Volatility Index (VIX): Haver Analytics
- Commodity terms of trade: Gruss and Kebhaj (2019)
- Composite risk rating: International Country Risk Guide, The PRS Group, Inc, www.prsgroup.com
- Cyclically adjusted balance: IMF, World Economic Outlook
- Financial development: Sahay and others (2015)
- Exchange rate regime: Ilzetzki and others (2019)
- Expected inflation: Consensus Economics, Haver Analytics, IMF staff calculations
- Gross capital inflows / outflows / Official Reserves / Net capital inflows: IMF, Balance of Payments and International Investment Position Statistics
- Gross public debt / Gross public debt in foreign currency / Population: IMF, World Economic Outlook
- Inflation expectation anchoring index: Bems and others (2018)
- Macroprudential regulation: Integrated Macroprudential Policy (iMaPP) dataset, Alam and others (2019)
- US monetary policy shock: Iacoviello and Navarro (2019)
- Nominal GDP, Real GDP, other series: Haver Analytics; IMF staff calculations where noted
- Policy rates, Real credit from banks: Bank for International Settlements; Haver Analytics; and IMF, International Financial Statistics
- Institutional quality: Worldwide Governance Indicators

_Source note: IMF staff compilation._

### Macroprudential measures (iMaPP): index and categories
- iMaPP provides information about 12 macroprudential tools, grouped into:
  - overall index and five subcategories: Capital, Credit Demand, Credit Supply, FX Exposure, Liquidity.
- Mapping (category → measures listed as iMaPP variables, notes preserved):
  - Capital: Capital requirements (Capital) — Including risk weights; Leverage limits (LVR); Loan loss provision requirements (LLP) — Including dynamic and sector-specific provisioning (e.g. housing); Countercyclical capital buffer (CCB); Capital conservation buffer requirements (Conservation); Measures targeted at SIFIs (SIFI) — Including capital and liquidity surcharges.
  - Credit Demand: Limits to loan-to-value ratio (LTV) — Mostly targeted at housing loans; Limits to debt-service-to-income ratio (DSTI); Tax on transactions (Tax) — Including stamp duties and capital gain taxes.
  - Credit Supply: Limits on credit growth or volume (LCG) — Including penalties for exceeding limits; Loan restrictions (LoanR) — Tailored LCG conditional on loan and bank characteristics, or other factors.
  - FX Exposure: Limits on foreign currency lending (LFC); Limits on gross open FX positions (LFX) — Including currency mismatch regulations; Reserve requirements on FC assets (RR_FCD).
  - Liquidity: Reserve requirements (RR_dom) — On domestic currency assets; Liquidity measures (Liquidity) — Including liquidity coverage ratios, liquid asset ratios, net stable funding ratios, core funding ratios, and external debt restrictions; Limits to loan-to-deposit ratio (LTD) — Including penalties for exceeding limits.
- Capital and liquidity measures account for about two thirds of the overall tightening in macroprudential regulation between 1990 and 2016.

### Country coverage (selected listing)
- Countries included (as reported): Albania; Argentina; Belarus; Bosnia and Herzegovina; Brazil; Bulgaria; Chile; China; Colombia; Costa Rica; Croatia; Dominican Republic; Ecuador; El Salvador; Georgia; Hungary; India; Indonesia; Jamaica; Jordan; Kazakhstan; Malaysia; Mexico; Morocco; Northern Macedonia; Pakistan; Paraguay; Peru; Philippines; Poland; Romania; Russia; Serbia; South Africa; Thailand; Turkey; Ukraine; Uruguay.

_Source note: IMF staff compilation._

---

### Annex 3.2. Dampening Effects of Macroprudential Regulation

### Empirical framework and identification
- Baseline panel regression (notation preserved):  
  Y_{i,t} = α_i + β · S_{i,t} + γ · (S_{i,t} * MPru_{i,t}) + δ · (S_{i,t} * MPru_{i,t}^2) + ζ MPru_{i,t} + θ MPru_{i,t}^2 + κ · C_{i,t} + ε_{i,t}  (equation (3.1))
  - Y_{i,t}: quarterly real GDP growth for country i at time t.
  - S_{i,t}: vector of global financial shocks including US policy rate shocks (Iacoviello and Navarro (2019)), VIX (risk premium proxy), and net capital inflows.
  - Net capital inflows are instrumented to isolate global “push” variation: instrument net inflows to country i with gross inflows to the other emerging markets following Blanchard and others (2017).
  - Interaction terms with MPru_{i,t} and MPru_{i,t}^2 allow assessment of (nonlinear) dampening effects of macroprudential regulation; coefficients γ reveal dampening.
  - Control vector C_{i,t} includes lagged GDP growth, lagged log real GDP per capita, institutional quality, linear trend, lagged output gap, and commodity terms of trade (following Obstfeld and others (2019)). The specification omits contemporaneous credit-to-GDP ratio to avoid blocking a potential transmission channel.
- Estimation approach:
  - Two-stage least squares (2SLS) where net capital inflows and their interactions with MPru and MPru^2 are instrumented with gross capital inflows to all other emerging markets and corresponding interactions.
  - Results presented in terms of net outflows (by inverting sign).
  - Standard errors: Driscoll and Kraay (1998) correction to account for cross-sectional dependence, autocorrelation, and heteroscedasticity.
- Derivative used to assess impact of global shocks as a function of MPru:
  - ∂Y_{i,t}/∂S_{j,t} = β + γ_j MPru_{i,t} + δ_j MPru_{i,t}^2  (equation (3.2)) — nonlinear in MPru.

### Baseline regression key coefficients (column (1) — full sample)
- Lag dependent variable: 0.071 (standard error 0.057)
- Lag output gap: -0.378*** (0.035)
- Lag ln real GDP per capita: -0.880* (0.505)
- Linear trend: 0.003 (0.005)
- Commodity terms of trade: 0.050 (0.035)
- US monetary policy shock: -0.158 (0.168)
- Ln VIX: -0.712*** (0.182)
- Net outflows: -0.186*** (0.046)
- Macroprudential regulation (MPru): -1.380*** (0.401)
- US monetary policy shock * MPru: -0.100 (0.109)
- Ln VIX * MPru: 0.631*** (0.143)
- Net outflows * MPru: 0.108*** (0.036)
- MPru^2: 0.237 (0.209)
- Observations: 2,260
- Countries: 38
- F-statistic: 73.1
- Note: MPru is divided by 10 to ease visualization of coefficients. Sample: 2000:Q1 to 2016:Q4. All specifications include country fixed effects. Driscoll-Kraay standard errors in parentheses. * p < .10; ** p < .05; *** p < .01.

### Baseline results — qualitative summary and key quantitative magnitudes
- VIX increases and capital outflows negatively affect GDP growth in emerging markets; US monetary policy shocks are insignificant in baseline.
- Instrumentation for net capital flows is supported by a strong F-statistic (above conventional threshold).
- Interaction coefficients indicate macroprudential regulation dampens the effects of global shocks on GDP growth; inclusion of MPru^2 and its interactions shows decreasing marginal returns to MPru.
- Quantitative examples:
  - When the VIX doubles (an increase similar to the one during the global financial crisis), an emerging market at the lowest level of macroprudential regulation in the sample would experience a decline in quarterly GDP growth by 1.8pp.
  - Net outflows worth one percent of GDP would lead to a fall in GDP growth by 0.6pp.
  - An economy with the median level of macroprudential regulation would experience a GDP decline by only 0.5pp in the case of a doubling of the VIX and by 0.3pp in the case of a 1pp capital outflow.
- Conclusion: macroprudential policy can play an important role in dampening the effects of global financial shocks in emerging markets.

### Robustness checks
- Reverse causality tests (Annex Table 3.2.2):
  - Excluding observations with negative GDP growth; replacing MPru with its first lag and fourth lag; replacing time-varying MPru with time-invariant country averages.
  - Coefficients on interaction terms between MPru and VIX or net outflows remain statistically significant in all specifications, supporting robustness to reverse causality concerns.
- Omitted-variable tests (Annex Table 3.2.3):
  - Controls and interactions added for institutional quality, financial development, gross public debt (percent of GDP), gross public debt in foreign currency (percent of gross public debt), cyclically adjusted balance (percent of trend GDP), monetary policy rate, inflation expectation anchoring index, exchange rate regime, capital controls, official reserves (percent of trend GDP).
  - Dampening effects of MPru against VIX and net outflows remain statistically significant across these tests.
  - Instrumentation becomes more cumbersome: F-statistic falls below ten when institutional quality, gross debt, gross debt in foreign currency, and the inflation expectation anchoring index are used as controls.
  - Results robust to exclusion of any single country; excluding the global financial crisis period (2008:Q3–2009:Q4) confirms baseline results. Only in some specifications with squared MPru does significance on the VIX interaction occasionally drop below 10 percent.
  - Results also robust to inclusion of time fixed effects (at cost of excluding global shocks).

### Symmetry of dampening effects
- Extended specification (equation (3.3)) distinguishes positive and negative shocks using dummies D^+ and D^- interacting with shocks and MPru.
- Results (Annex Table 3.2.4):
  - Triple interaction coefficients γ^+ and γ^- are statistically significant and similar in magnitude.
  - Wald test rejects the hypothesis that γ^+ and γ^- differ statistically: dampening effects of MPru operate symmetrically for positive and negative VIX and net outflow shocks.

### Disaggregated categories of macroprudential measures
- MPru index disaggregated into five categories: bank capital, credit demand, credit supply, foreign exchange positions, liquidity.
- Findings (Annex Table 3.2.5):
  - For VIX shocks: measures on credit demand, foreign exchange exposure, and liquidity exhibit significant interaction coefficients.
  - For capital outflows: interaction terms significant for measures targeting capital, credit demand, and credit supply (results using capital-based measures to be taken with caution due to low F-statistic).
  - Overall: dampening effects arise from a broad range of measures rather than a single category.

### Dampening effects on credit and exchange rates
- Dependent variables replaced with real bank credit growth, nominal effective exchange rate (NEER), and real effective exchange rate (REER) (positive values denote appreciation).
- Annex Table 3.2.6 findings:
  - VIX and net capital shocks lead to depreciation of the NEER (column 2) and the REER (column 3).
  - Net outflows tend to slow down real credit growth (column 1).
  - Macroprudential regulation reduces sensitivity of NEER and REER to movements in the VIX.
  - MPru dampens effects of net capital outflows on both credit growth and the exchange rate.

---

*Source: IMF staff compilation (Annex 3.1 and Annex 3.2, World Economic Outlook, April 2020).*

### Annex Table 3.2.1.  Dampening Effects of Macroprudential Regulation on GDP Growth

### Annex Table 3.2.1.  Dampening Effects of Macroprudential Regulation on GDP Growth

### Main regression results (specifications 1–4)
- Lag dependent variable:
  - (1) -0.054 (0.039)
  - (2) 0.076 (0.051)
  - (3) 0.076 (0.054)
  - (4) 0.079 (0.052)
- Lag output gap:
  - (1) -0.180*** (0.024)
  - (2) -0.367*** (0.035)
  - (3) -0.353*** (0.034)
  - (4) -0.368*** (0.035)
- Lag ln real GDP per capita:
  - (1) -0.791* (0.474)
  - (2) -1.608** (0.736)
  - (3) -1.833** (0.822)
  - (4) -1.675** (0.754)
- Institutional quality:
  - (1) -0.650 (0.501)
  - (2) -0.197 (0.813)
  - (3) -0.541 (0.811)
  - (4) -0.343 (0.766)
- Linear trend:
  - (1) -0.002 (0.005)
  - (2) 0.005 (0.007)
  - (3) 0.010 (0.008)
  - (4) 0.007 (0.007)
- Commodity terms of trade:
  - (1) 0.079* (0.048)
  - (2) 0.039 (0.050)
  - (3) 0.041 (0.055)
  - (4) 0.052 (0.039)
- US monetary policy shock:
  - (1) 0.207 (0.166)
  - (2) -0.013 (0.298)
  - (3) -0.008 (0.265)
  - (4) -0.046 (0.314)
- Ln VIX:
  - (1) -0.597*** (0.221)
  - (2) -1.504*** (0.305)
  - (3) -1.498*** (0.298)
  - (4) -1.356*** (0.402)
- Net outflows:
  - (1) -0.258*** (0.071)
  - (2) -0.396*** (0.111)
  - (3) -0.424*** (0.108)
  - (4) -0.444*** (0.151)
- Macroprudential regulation (MPru):
  - (2) -0.639** (0.323)
  - (3) -1.286*** (0.375)
  - (4) -1.534*** (0.388)
  - Note: MPru is divided by 10 to ease visualization of coefficients.
- Interaction terms:
  - US monetary policy shock * MPru:
    - (1) -0.120* (0.068)
    - (2) -0.101 (0.100)
    - (3) -0.133 (0.090)
    - (4) -0.070 (0.116)
  - Ln VIX * MPru:
    - (1) 0.259** (0.120)
    - (2) 0.598*** (0.133)
    - (3) 0.683*** (0.137)
    - (4) 0.444** (0.208)
  - Net outflows * MPru:
    - (1) 0.067*** (0.024)
    - (2) 0.111*** (0.035)
    - (3) 0.125*** (0.035)
    - (4) 0.149** (0.062)
- Sample and fit:
  - Observations: (1) 1,846; (2) 2,235; (3) 2,153; (4) 2,260
  - Countries: 38, 38, 38, 38
  - F-statistic: (1) 35.1; (2) 32.9; (3) 32.6; (4) 29.7

### Notes on specifications
- Net inflows (in percent of trend GDP) are instrumented using gross inflows to other EMs (in percent of trend GDP); results are presented in terms of net outflows.
- Estimations based on a sample of EMs from 2000:Q1 to 2016:Q4. All specifications include country fixed effects. Driscoll-Kraay standard errors reported in parentheses. * p < .10; ** p < .05; *** p < .01.

---

### Annex Table 3.2.2.  Dampening Effects on GDP Growth, Robustness to Reverse Causality

### Robustness specifications (columns 1–11)
- Lag dependent variable (selected):
  - Column (1) 0.071 (0.050)
  - Column (4) 0.090* (0.050)
  - Column (7) 0.221*** (0.071)
  - Column (11) 0.058 (0.059)
- Lag output gap (selected):
  - Column (1) -0.370*** (0.035)
  - Column (5) -0.316*** (0.024)
  - Column (7) -0.252*** (0.053)
  - Column (11) -0.401*** (0.034)
- Lag ln real GDP per capita (selected):
  - Column (1) -1.758** (0.774)
  - Column (6) -2.705** (1.086)
  - Column (7) -5.289 (3.605)
  - Column (11) -0.660 (0.467)
- Ln VIX:
  - Column (1) -1.623*** (0.326)
  - Column (2) -2.169*** (0.488)
  - Column (7) -1.555*** (0.484)
  - Column (10) -1.592*** (0.343)
- Net outflows:
  - Column (1) -0.412*** (0.120)
  - Column (3) -0.552*** (0.160)
  - Column (7) -0.358** (0.157)
  - Column (11) -0.108*** (0.032)
- Macroprudential regulation (MPru):
  - Column (1) -1.412*** (0.418)
  - Column (3) -1.439*** (0.464)
  - Column (7) -1.679** (0.808)
  - Column (11) -1.090*** (0.402)
- Interaction terms (selected):
  - Ln VIX * MPru:
    - Column (1) 0.644*** (0.152)
    - Column (3) 0.620*** (0.171)
    - Column (7) 0.816** (0.341)
    - Column (11) 0.449*** (0.156)
  - Net outflows * MPru:
    - Column (1) 0.117*** (0.042)
    - Column (3) 0.131*** (0.042)
    - Column (7) 0.097* (0.051)
    - Column (11) 0.025* (0.014)
- Additional interactions with variable X (various country characteristics) reported; coefficients vary in sign and significance across columns.
- Sample and fit (selected):
  - Observations range from 1,200 to 2,260 across columns.
  - Countries: mostly 38, with some specifications 35, 34, 31, 20 depending on X.
  - F-statistic examples: Column (1) 5.8; Column (2) 10.5; Column (5) 29.9; Column (11) 23.4

### Robustness interpretation
- Results presented in terms of net outflows; MPru divided by 10 to ease visualization.
- Robustness checks vary X (institutional quality, financial development, gross public debt, gross public debt in foreign currency, cyclically adjusted balance, monetary policy rate, inflation expectation anchoring, fixed exchange rate regime, capital controls, official reserves).
- Findings: core MPru effects and interactions with Ln VIX and net outflows remain robust across a range of controls and tests addressing reverse causality.

---

### Annex Table 3.2.3.–3.2.5.  Additional Robustness and Heterogeneity: Omitted Variables, Symmetric Dampening, Categories

### Robustness to omitted variables (selected coefficients)
- Including time fixed effects (columns for Ln VIX and Net outflows):
  - Lag dependent variable:
    - Ln VIX column 0.070 (0.047)
    - Net outflows column 0.100* (0.051)
  - Lag output gap:
    - Ln VIX column -0.396*** (0.036)
    - Net outflows column -0.388*** (0.037)
  - Lag ln real GDP per capita:
    - Ln VIX column -1.676** (0.738)
    - Net outflows column -1.968** (0.776)
  - Ln VIX: -1.555*** (0.329)
  - Net outflows: -0.420*** (0.106)
  - Interactions with D+ and D- and MPru reported with significance (see table for exact coefficients).

### Symmetric dampening effects (selected specifications: Capital, Credit demand, Credit supply, FX exposure, Liquidity)
- Lag dependent variable: 0.061 to 0.071 (standard errors 0.054–0.067)
- Lag output gap: approximately -0.374*** to -0.383*** (standard errors 0.035–0.039)
- Lag ln real GDP per capita: coefficients vary, e.g., -2.012** (0.972) in column (1) and -0.682 (0.531) in column (4)
- Macroprudential regulation (Mpru) and nonlinear (MPru^2) terms included in some specifications with varying significance; select examples:
  - Mpru: 8.040 (12.012) in column (1); -10.270*** (2.420) in column (4)
  - Net outflows * Mpru: 2.390** (1.129) in column (1); 0.638*** (0.223) in column (2); -0.115 (0.128) in column (4)
  - MPru^2 and associated interactions reported where nonlinear effects tested.

### Macroprudential categories, credit growth and exchange rates (Annex Table 3.2.6)
- Dependent variables: Real credit growth, NEER appreciation, REER appreciation (columns 1–3)
- Lag dependent variable:
  - Real credit growth 0.226*** (0.051)
  - NEER appreciation 0.206*** (0.041)
  - REER appreciation 0.174*** (0.043)
- Lag output gap:
  - Real credit growth 0.196** (0.083)
  - NEER appreciation -0.071 (0.058)
  - REER appreciation -0.050 (0.059)
- Lag ln real GDP per capita:
  - Real credit growth -3.606 (2.830)
  - NEER appreciation -2.225* (1.201)
  - REER appreciation -2.337* (1.233)
- Ln VIX:
  - Real credit growth -1.232 (0.981)
  - NEER appreciation -2.469** (1.031)
  - REER appreciation -2.248** (1.000)
- Net outflows:
  - Real credit growth -0.563* (0.288)
  - NEER appreciation -0.396** (0.173)
  - REER appreciation -0.410*** (0.158)
- Macroprudential regulation (MPru):
  - Real credit growth -1.065 (1.125)
  - NEER appreciation -2.552* (1.332)
  - REER appreciation -2.412* (1.353)
- Interactions:
  - Ln VIX * MPru:
    - Real credit growth 0.480 (0.413)
    - NEER appreciation 1.066** (0.521)
    - REER appreciation 1.037** (0.520)
  - Net outflows * MPru:
    - Real credit growth 0.173* (0.093)
    - NEER appreciation 0.134** (0.065)
    - REER appreciation 0.129** (0.060)
- Control: Credit to GDP coefficient -0.002** (0.001) reported in some specifications.
- Observations and fit:
  - Observations: Real credit growth 2,049; NEER appreciation 2,164; REER appreciation 2,164
  - Countries: 38, 35, 35
  - F-statistic examples: 25.5 (Real credit growth), 3.4 (NEER), 3.2 (REER)

---

### Annex 3.3. Macroprudential Regulation and Monetary Policy Responses

### Empirical framework (equation and identification)
- Panel specification (equation 3.4):
  - I_{i,t} = α_i + β∙S_{i,t} + γ∙(S_{i,t} * MPru_{i,t}) + ζ MPru_{i,t} + κ∙C_{i,t} + ε_{i,t}
  - Dependent variable I_{i,t} denotes policy rates for country i at time t.
  - Global shocks S_{i,t} include: (i) the US policy rate, (ii) the VIX, and (iii) net capital outflows.
- Notes on measurement and estimation:
  - US policy rate: federal funds rate except during the zero lower bound period where the implied policy rate from Wu and Xia (2015) is used.
  - Regression includes expected inflation over the next 12 months, output gap, real credit growth, commodity terms of trade; country fixed effects included.
  - Net capital inflows instrumented; estimation via two-stage least squares; standard errors use Driscoll and Kraay (1998) correction.
  - Sample: quarterly data for up to 30 EMs (2000–2016); hard pegs excluded to focus on countries with monetary autonomy.

### Baseline results (Annex Table 3.3.1)
- Column (1) to (4) summary (selected coefficients):
  - US policy rate:
    - (1) 0.597*** (0.068)
    - (2) 0.359*** (0.034)
    - (3) 0.295*** (0.037)
    - (4) 0.366*** (0.066)
  - Ln VIX:
    - (1) 1.396*** (0.482)
    - (2) 0.587** (0.283)
    - (3) 0.599** (0.302)
    - (4) 2.098*** (0.451)
  - Net outflows:
    - (1) 0.404*** (0.118)
    - (2) 0.347*** (0.079)
    - (3) 0.337*** (0.084)
    - (4) 0.418*** (0.127)
  - Expected inflation, next 12 months:
    - (2)/(3) 1.235*** / 1.224*** / 1.227*** (0.075–0.077)
  - Output gap:
    - (2)/(3) 0.250*** / 0.148** / 0.208*** (0.082–0.071)
  - Real credit growth:
    - (3) 0.050*** (0.014)
  - Commodity terms of trade:
    - (3) -0.085** (0.042)
  - Macroprudential regulation (MPru) in full specification (4):
    - Level: 0.220*** (0.053)
    - US policy rate * MPru: -0.013*** (0.004)
    - Ln VIX * MPru: -0.116*** (0.020)
    - Net outflows * MPru: -0.004 (0.004)
- Sample fit:
  - Observations: (1) 1,360; (2) 1,262; (3) 1,250; (4) 1,250
  - Countries: (1) 30; (2) 25; (3) 25; (4) 25
  - F-statistic: (1) 73.98; (2) 56.42; (3) 49.70; (4) 21.98

### Interpretation of baseline findings
- Policy rates in emerging markets respond procyclically to tightening global financial conditions: they increase with rises in the US policy rate, the VIX, and with higher net capital outflows.
- Interaction results in the full specification (column 4):
  - Significant and negative interaction terms between MPru and (i) the US policy rate and (ii) the VIX, implying that higher levels of macroprudential regulation mute the procyclical monetary response.
  - If MPru is sufficiently tight, the monetary response to increases in US policy rates and the VIX can become countercyclical (emerging markets lowering policy rates).
  - MPru does not appear to affect the monetary policy response to net capital outflows in column 4 (interaction insignificant).

### Robustness to reverse causality (Annex Table 3.3.2 summary)
- Lagged MPru specifications:
  - One-quarter lag and one-year lag of MPru reported; interactions of MPru with US policy rate and Ln VIX remain negative and statistically significant in lagged specifications.
- Country-average of MPru:
  - Interaction of MPru with the VIX remains negative and statistically significant.
  - Interaction with the US policy rate loses significance, but MPru now supports a more countercyclical response to net capital outflows (Net outflows * MPru -0.014** (0.007) in country-average specification).
- Selected coefficients from lagged and country-average specifications:
  - US policy rate in lagged specs: 0.366*** (0.065), 0.409*** (0.067), 0.351*** (0.063)
  - Ln VIX in lagged specs: 2.120*** (0.452), 2.175*** (0.480), 2.407*** (0.402)
  - Net outflows in lagged specs: 0.406*** (0.126), 0.391*** (0.130), 0.566*** (0.182)
  - MPru level in lagged specs: 0.217*** (0.055), 0.242*** (0.061)
  - US policy rate * MPru in lagged specs: -0.014*** (0.004), -0.018*** (0.004)
  - Ln VIX * MPru in lagged specs: -0.116*** (0.021), -0.129*** (0.025)
  - Net outflows * MPru in country-average: -0.014** (0.007)
- Sample and fit for robustness table:
  - Observations: 1,241 to 1,250 depending on column
  - Countries: 25
  - F-statistic examples: 22.97; 22.15; 20.53

### Robustness interpretation
- Lagging MPru and using country-average MPru alleviate concerns of reverse causality (i.e., that changes in MPru are driven by domestic policy rate changes).
- The negative interactions between MPru and global financial shocks (US policy rate and VIX) are robust to lagging MPru and to specifications relying on cross-country variation in MPru.

*Source: IMF staff calculations.*

### Annex Table 3.3.2.  Regressions of Domestic Policy Rates, Robustness to Reverse

### Annex Table 3.3.2.  Regressions of Domestic Policy Rates, Robustness to Reverse Causality

### Key findings from robustness to reverse causality tests
- Main empirical approach: Net inflows (in percent of trend GDP) for each country instrumented using gross inflows to other EMs (in percent of trend GDP). Results presented in terms of net outflows. Estimations use fixed effects on a panel of EMs, excluding countries with pegged and freely falling exchange rates, during 2000:Q1 to 2016:Q4. Driscoll-Kraay standard errors in parenthesis. * p < .10; ** p < .05; *** p < .01.
- Interaction of macroprudential regulation (MPru) and the VIX remains negative and significant across robustness tests.
- Interaction of MPru and the US policy rate remains negative and significant in most specifications; loses significance at the 10 percent level in the specification augmented with official reserves.

### Selected coefficient estimates (columns 1–9 correspond to alternative augmentations)
- US policy rate:
  - (1) 0.347*** (0.073)
  - (2) 0.053 (0.136)
  - (3) -0.061 (0.128)
  - (4) 0.583*** (0.090)
  - (5) 0.566*** (0.079)
  - (6) 0.235*** (0.070)
  - (7) 0.315*** (0.089)
  - (8) 0.324*** (0.097)
  - (9) 2.899*** (0.428)
- Ln VIX:
  - (1) 1.941*** (0.486)
  - (2) 2.757*** (0.699)
  - (3) 0.187 (0.963)
  - (4) 1.260** (0.540)
  - (5) 1.960*** (0.376)
  - (6) 1.884*** (0.401)
  - (7) 1.490*** (0.434)
  - (8) 1.637*** (0.435)
  - (9) 0.643** (0.312)
- Net outflows:
  - (1) 0.415*** (0.123)
  - (2) 0.178 (0.168)
  - (3) 0.347* (0.187)
  - (4) 0.591*** (0.142)
  - (5) 0.331*** (0.100)
  - (6) 0.146 (0.185)
  - (7) 0.221*** (0.081)
  - (8) 0.591*** (0.174)
  - (9) 0.076 (0.047)
- Macroprudential regulation (MPru):
  - (1) 0.201*** (0.055)
  - (2) 0.228*** (0.042)
  - (3) 0.172*** (0.062)
  - (4) 0.166*** (0.050)
  - (5) 0.148*** (0.055)
  - (6) 0.160 (0.101)
  - (7) 0.190*** (0.046)
  - (8) 0.164*** (0.053)
  - (9) 0.084** (0.040)
- US policy rate * MPru:
  - (1) -0.012*** (0.004)
  - (2) -0.012*** (0.003)
  - (3) -0.008* (0.004)
  - (4) -0.017*** (0.004)
  - (5) -0.021*** (0.004)
  - (6) -0.008** (0.004)
  - (7) -0.013*** (0.003)
  - (8) -0.005 (0.004)
  - (9) -0.016*** (0.002)
- Ln VIX * MPru:
  - (1) -0.111*** (0.021)
  - (2) -0.107*** (0.016)
  - (3) -0.097*** (0.024)
  - (4) -0.099*** (0.019)
  - (5) -0.088*** (0.020)
  - (6) -0.074** (0.037)
  - (7) -0.104*** (0.017)
  - (8) -0.091*** (0.020)
  - (9) -0.064*** (0.015)

### Sample and diagnostics
- Observations across specifications: 1250, 1250, 1239, 1157, 1212, 976, 1250, 1236, 1250.
- Countries: 25, 25, 25, 24, 24, 18, 25, 25, 25.
- F-statistic examples: (1) 3.952; (2) 6.900; (3) 11.871; (9) 5.46.

---

### Annex Table 3.4.1.  Regressions of Real GDP Growth, Spillovers

### Core findings on cross-country spillovers
- No evidence of negative spillovers from higher domestic macroprudential regulation to other countries.
- Spillovers tend to be positive with respect to net outflows: higher average MPru in other EMs dampens the adverse effect of net outflows on a country’s real GDP growth.
- Dampening effects from domestic and foreign macroprudential regulation are similar in magnitude, but the coefficient on others' MPru captures a one-unit increase averaged across all other EMs in the group (a larger aggregate tightening).

### Selected coefficient estimates (columns 1–6; different definitions of "others")
- Lag dependent variable:
  - (1) 0.024 (0.055)
  - (2) 0.070 (0.050)
  - (3) 0.071 (0.048)
  - (4) 0.055 (0.048)
  - (5) 0.067 (0.055)
  - (6) 0.052 (0.047)
- Lag output gap:
  - (1) -0.382*** (0.046)
  - (2) -0.381*** (0.037)
  - (3) -0.381*** (0.039)
  - (4) -0.360*** (0.034)
  - (5) -0.339*** (0.031)
  - (6) -0.359*** (0.032)
- Ln VIX:
  - (1) -1.469* (0.794)
  - (2) -1.748*** (0.472)
  - (3) -1.236** (0.579)
  - (4) -1.498*** (0.471)
  - (5) -1.817*** (0.484)
  - (6) -1.669*** (0.458)
- Net outflows:
  - (1) -1.367*** (0.479)
  - (2) -0.600*** (0.153)
  - (3) -0.683*** (0.226)
  - (4) -0.535*** (0.168)
  - (5) -0.828*** (0.225)
  - (6) -0.626*** (0.152)
- Macroprudential regulation (MPru):
  - (1) -2.864*** (0.932)
  - (2) -1.582*** (0.459)
  - (3) -1.855*** (0.556)
  - (4) -1.429*** (0.456)
  - (5) -1.479*** (0.463)
  - (6) -1.630*** (0.453)
- Ln VIX * MPru:
  - (1) 1.207*** (0.374)
  - (2) 0.709*** (0.168)
  - (3) 0.731*** (0.197)
  - (4) 0.635*** (0.165)
  - (5) 0.747*** (0.178)
  - (6) 0.692*** (0.165)
- Net outflows * MPru:
  - (1) 0.148*** (0.054)
  - (2) 0.120*** (0.040)
  - (3) 0.138*** (0.053)
  - (4) 0.121*** (0.044)
  - (5) 0.145*** (0.049)
  - (6) 0.105*** (0.034)
- Net outflows * others' MPru:
  - (1) 0.418** (0.175)
  - (2) 0.120*** (0.031)
  - (3) 0.113** (0.044)
  - (4) 0.068** (0.031)
  - (5) 0.170*** (0.050)
  - (6) 0.116*** (0.032)
- Others' MPru (average level of MPru in other EMs):
  - (1) 4.296* (2.388)
  - (2) -0.027 (0.367)
  - (3) 0.863 (0.682)
  - (4) 0.148 (0.433)
  - (5) 0.486 (0.480)
  - (6) 0.469 (0.336)

### Sample and diagnostics
- Observations: examples include 2,260; 2,192; 2,260; 2,260; 2,260; 1,082.
- Countries: 38, 37, 38, 35, 38, 35.
- F-statistic examples: (1) 4.11; (2) 9.6; (3) 9.4; (4) 10.4; (5) 10.8; (6) 2.7.
- Wald test (p-value) examples: (1) 0.063; (2) 0.993; (3) 0.384; (4) 0.078; (5) 0.511; (6) 0.739.

---

### Annex Table 3.5.1.  Policy Response Functions of Macroprudential Measures

### Main empirical result: Do EMs adjust MPru in response to global shocks?
- Evidence suggests policy makers in emerging markets tend to loosen macroprudential policies when global financial conditions tighten, and tighten when global conditions ease.
- Results robust to controlling for expected inflation, output gap, commodity terms of trade, and real credit growth.
- This behavior is consistent with prior findings that the probability of tightening macroprudential measures is higher when capital flows surge.

### Selected coefficient estimates (columns reported)
- Net outflows (presented as net outflows so shocks are negative for EMs):
  - (1) -0.035** (0.015)
  - (2) -0.044** (0.023)
  - (3) -0.038** (0.018)
- US monetary policy shocks:
  - (1) -0.056* (0.030)
  - (2) -0.066** (0.031)
  - (3) -0.056** (0.026)
- Ln VIX:
  - (1) -0.170** (0.077)
  - (2) -0.160* (0.086)
  - (3) -0.128** (0.055)
- Control variables (examples):
  - Expected inflation: (2) -0.011 (0.007); (3) -0.011* (0.006)
  - Output gap: (2) -0.017 (0.016); (3) -0.007 (0.013)
  - Commodity terms of trade: (2) 0.016** (0.006); (3) 0.010* (0.005)

### Sample and diagnostics
- Observations: (1) 2,286; (2) 1,798; (3) 1,798.
- Countries: (1) 38; (2) 32; (3) 32.
- F-statistic examples: (1) 69.64; (2) 42.61; (3) 42.61.

*Source: IMF staff calculations.*

---


_Source: https://www.imf.org/-/media/files/publications/weo/2020/april/english/annexch3.pdf_