## wpiea2020106-print-pdf

## Source details

**Canonical URL:** [wpiea2020106-print-pdf](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020106-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020106-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020106-print-pdf.pdf.json)

---

### 1. US policy rate and VIX — Data, measures, and macroprudential evolution
- Global financial shocks included:
  - shocks to US monetary policy identified by Iacoviello and Navarro (2019).
  - Chicago Board Options Exchange’s Volatility Index (VIX).
  - net capital inflows (normalized by the HP-trend of GDP and instrumented to isolate global push factors).
- Macroprudential regulation index:
  - Constructed by cumulating macroprudential tightening actions net of loosening actions reported in the iMaPP database from 1990 to 2016.
  - The index is rescaled upward across all countries so values are always positive because it enters quadratically in regressions.
- Sample timing and series:
  - Quarterly axes shown for 2000q1, 2005q1, 2010q1, 2015q1.
  - US policy rate is the federal funds rate except during the zero lower bound (ZLB) period where the implied rate from Wu and Xia (2015) is used.
- Data sources: Bank for International Settlements; IMF, International Financial Statistics database; Haver Analytics; Wu and Xia (2015); IMF, Balance of Payments; iMaPP database; authors’ calculations.
- Macroprudential policy evolution (iMaPP evidence):
  - Macroprudential regulation has been considerably tightened over the years, especially since 2005.
  - The global financial crisis led to a temporary loosening but emerging markets returned to tighten regulation during the subsequent recovery.
  - There is substantial cross-country dispersion in the level of macroprudential regulation.

### 1.1 Empirical strategy
- Regression specification:
  - Y_{i,t} = α_i + β S_{i,t} + γ(S_{i,t}·MPru_{i,t}) + δ(S_{i,t}·MPru^2_{i,t}) + ζ MPru_{i,t} + θ MPru^2_{i,t} + κ C_{i,t} + ε_{i,t}
  - Y_{i,t} = quarterly real GDP growth; S_{i,t} = vector of global financial shocks; MPru_{i,t} = level of macroprudential regulation; α_i = country fixed effect.
  - Quadratic interactions included to allow for non-linear effects.
- Controls C_{i,t}: lagged GDP growth, lagged log of real GDP per capita, institutional quality, a linear trend, lagged output gap, commodity terms of trade.
- Identification and estimation:
  - Net capital inflows instrumented with gross inflows to other emerging markets following Blanchard et al. (2017).
  - Two-stage least squares with Driscoll and Kraay (1998) correction to standard errors.
  - Time fixed effects excluded in baseline; included in robustness checks.

### 1.2 Main empirical findings (summary)
- Dampening effects on GDP growth:
  - A more stringent level of macroprudential regulation significantly reduces the sensitivity of GDP growth in emerging markets to global financial shocks.
  - Dampening effects display decreasing marginal returns: additional tightening is less effective at higher levels of MPru.
  - Dampening properties are not driven by a single tool; a broad range of measures contribute (tools that boost bank capital and liquidity, limit foreign exchange exposures, and avert risky credit).
  - Dampening effects are heterogeneous across types of global financial shock.
  - Macroprudential regulation leads to symmetric dampening effects: it supports GDP growth when global financial shocks are adverse but can lower economic activity when global financial conditions are favorable.
- Spillovers:
  - No evidence of negative spillovers from one country tightening macroprudential regulations; instead, higher average MPru in other emerging markets tends to enhance macroeconomic stability in a country.
- Interaction with monetary policy:
  - MPru allows for a more countercyclical monetary policy response to global financial shocks.
  - At low MPru, central banks tend to respond procyclically (increasing rates when global financial conditions tighten).
  - At more stringent MPru, monetary policy response becomes countercyclical (declining policy rates when global financial conditions tighten).
- Comparison with capital controls:
  - Analysis using capital control indicators (Chinn and Ito (2008); Fernández et al. (2015); Quinn and Toyoda (2008); Pasricha et al. (2018)) does not find evidence that tighter capital controls provide similar benefits to tighter macroprudential regulation.
  - Findings align with Forbes et al. (2015) and Frost et al. (2020) that macroprudential tools, especially FX-based measures, are more effective than capital controls in influencing macroeconomic outcomes.

### 1.3 Caveats and limitations
- Measurement limitations:
  - iMaPP action indicators record tighten/loosen but not intensity (except LTV limits); initial cross-country differences in 1990 exist.
  - Measurement imprecision likely biases results toward underestimating effects.
- Endogeneity concerns:
  - Robustness tests exploit time-series and cross-sectional identification, but residual omitted-variable bias cannot be entirely ruled out.
- Robustness:
  - Results robust to a broad set of endogeneity tests, inclusion of time fixed effects, and both time-series and cross-sectional approaches.

---

### 2.1 Baseline results — GDP and policy-rate regressions; nonlinearity and categories

- Baseline regression insights (Table 1):
  - Column (1): An increase in the VIX and an outflow of capital have negative and highly statistically significant effects on economic growth.
  - Column (2): US monetary shocks have detrimental effects on growth if not controlling for other global shocks, but lose significance when controlling for the VIX and capital flows.
  - Lag of the output gap coefficient is negative and significant.
  - Instrumentation: Kleibergen-Paap rk WaldF-statistic well above conventional threshold.
  - Diagnostics: Lagrange Multiplier tests indicate serial correlation; Modified Wald indicates heteroskedasticity; Pesaran, Frees, and Friedman tests reject cross-sectional independence.

- Dampening effects and nonlinearity:
  - Inclusion of MPru (column (3)): interactions between shocks and MPru are negative and highly statistically significant — more stringent MPru dampens shock effects on GDP growth.
  - Nonlinear specification (column (5)): squared MPru and its interactions included; interaction terms of VIX and capital flows with MPru^2 are negative and statistically significant — dampening subject to decreasing marginal returns.
  - Derivative: ∂Y_i,t/∂S_j,t = β + γ_j MPru_i,t + δ_j MPru^2_i,t (nonlinear in MPru).

- Quantitative illustrations (Figure 3):
  - At the lowest MPru in the sample:
    - A doubling of the VIX (100 percent increase) leads to a decline in quarterly GDP growth by 1.8 percentage points.
    - A net outflow worth three percent of GDP triggers similar effects.
  - At the median MPru in the sample:
    - Same shocks produce a GDP decline of 0.5 percentage points.
  - If MPru is sufficiently tight:
    - VIX and net capital outflows no longer have statistically significant effects on EM GDP.
  - Confidence intervals: shaded areas = 90 percent confidence intervals computed with Driscoll-Kraay standard errors.
  - Net capital outflows are scaled by the HP-trend of GDP.
  - Distribution of MPru shown for 2000–2016 and at end-2016.

- Robustness to reverse causality and omitted variables (Table 2 and Table 3):
  - Tests: excluding negative GDP observations; MPru lagged by one quarter; MPru lagged by one year; MPru replaced by time-invariant country averages (2000—2016).
  - In all specifications interaction coefficients between MPru and the VIX or net outflows remain positive and highly statistically significant.
  - Augmented regressions include interactions of shocks with institutional quality, financial development (IMF’s Financial Development Index), gross public debt (% of GDP), public foreign-currency debt (% of total public debt), cyclically-adjusted fiscal balance (% of GDP), monetary policy rate, inflation expectation anchoring, exchange rate regime, stringency of capital controls, and stock of official reserves (% of GDP).
  - Results: Dampening effects of MPru robust across these omitted-variable checks.
  - Instrumentation caveat: instrumenting three variables can reduce F-statistic below ten for institutional quality, public debt, and inflation expectation anchoring index.
  - Time fixed effects (column (11) Table 3): interactions with VIX and capital outflows remain positive and statistically significant.

- Symmetric dampening effects (equation (3), Table 4, Figure 4):
  - Separate γ^+ and γ^- estimated for positive and negative realizations of global shocks.
  - γ^+ and γ^- for both the VIX and capital outflows are statistically significant and similar in size; Wald tests cannot reject equality.
  - Interpretation: MPru dampens both positive and negative global financial shocks symmetrically.
  - Illustration: Growth differential between 75th and 25th percentile MPru over 2000–2016:
    - Example: Higher MPru would have increased quarterly GDP growth by about 0.6 percent between Q4 2008 and Q2 2009.
    - Conversely, MPru would have lowered economic growth considerably in years before the global financial crisis when global conditions were buoyant.
  - Policy implication: Maintaining high MPru is not costless — it forgoes growth when global conditions are favorable; however, constraining excessive risk-taking when conditions are loose is key to resilience.

- Categories of MPru measures (Table 5, Figure 5):
  - MPru disaggregated into: bank capital and liquidity; credit demand (LTV/DSTI); credit supply (limits on credit growth); foreign currency exposure; liquidity.
  - VIX (panel 1): measures targeted at credit demand, FX exposure, and liquidity dampen VIX impact on GDP.
  - Net capital outflows (panel 2): measures targeted at bank capital, credit demand, and credit supply strengthen resilience against net outflows.
  - Interpretation: Dampening effects are broad-based; adequate bank capital and liquidity and constraints on risky credit play crucial roles alongside FX measures.

---

### 2.5 Cross-country spillovers — specification and findings
- Objective: test whether tighter domestic MPru generates negative spillovers (via relocation/leakage) or positive spillovers (via greater resilience stabilizing trade/financial flows).
- Extended regression adds interaction of shocks with:
  - country i’s MPru_i,t, and
  - the average MPru in other EMs MPrū_i,t (weighted by average size of gross capital inflows).
- MPrū_i,t computed in alternative ways:
  - Column (1): average level in all other EMs.
  - Column (2): average in same geographical region.
  - Columns (3)-(4): differentiated by GDP per capita above/below median (quarterly GDP for time-varying grouping in column (3); average GDP over 2002/2016 for time-invariant grouping in column (4)).
  - Columns (5)-(6): differentiated by political/economic/financial/institutional risk class using a composite risk index from the Political Risk Service.
- Main findings:
  - No evidence of negative spillovers from macroprudential regulation.
  - Spillovers tend to be positive vis-à-vis net outflows: interaction between net outflows and average MPru in other EMs is positive and significant regardless of MPrū specification.
  - Magnitudes: coefficients on net outflows * MPrū_i,t across columns (2)–(6) are similar to those on net outflows * own MPru_i,t.
  - Caveat: coefficients on interactions with MPrū_i,t capture the effect of a one-unit increase in the average MPru across peer countries — a larger tightening than a one-unit increase of an economy’s own MPru.
- Implication:
  - No empirical support that tighter MPru in one EM increases susceptibility of others to global shocks; instead, higher average MPru across peers mitigates adverse impacts of net outflows on a given country’s GDP growth.

_Italic: Source: wpiea2020106-print-pdf - 2.5    Cross-country spillovers_

---

### 3.2 Robustness — reverse causality, omitted variables, instruments, and comparisons with capital controls

- Reverse causality tests for MPru use (Table 8):
  - MPru specifications: one-quarter lag; one-year lag; country average MPru (2000–2016).
  - Sample: EMs from 2000Q1 to 2016Q4.
  - Instrument: net inflows instrumented using gross inflows to other EMs (percent of trend GDP).
  - US policy rate: effective federal funds rate except during ZLB where Wu and Xia (2015) implied rate used.
  - Findings:
    - Across lag specifications MPru continues to support a more countercyclical monetary policy response.
    - Using country average MPru, regulation supports more countercyclical response to capital flow shocks rather than to US policy rate changes.
    - Table 8 reports observations = 1,241; 1,211; 1,250 and countries = 25; 25; 25 for columns (1)–(3) respectively.
    - F-statistics: 22.97; 22.15; 20.53.

- Omitted variable tests (Table 9):
  - Augmented regressions include X and interactions where X = institutional quality; financial development; gross public debt; public FX debt share; cyclically-adjusted balance; inflation expectation anchoring; capital flow measures; official reserves; time fixed effects.
  - Sample: EMs 2000Q1–2016Q4. MPru divided by 10 for visualization. Driscoll-Kraay standard errors reported.
  - Findings:
    - Ln VIX * MPru coefficients remain negative and significant across columns (examples): -0.111***; -0.107***; -0.097***; -0.099***; -0.088***; -0.074**; -0.104***; -0.091***; -0.064***.
    - US policy rate * MPru remains negative and significant in most specifications (examples): -0.012***; -0.012***; -0.008*; -0.017***; -0.021***; -0.008**; -0.013***; -0.005; -0.016***.
    - One exception: specification augmented with official reserves where US policy rate interaction loses significance at the 10 percent level.
    - Observations vary across columns (examples): 1,250; 1,239; 1,157; 1,129; 1,250; 1,236; 1,250; 1,236; 1,250. F-statistics reported (examples): 3.95; 26.90; 11.87; 12.45; 5.79; 2.74; 20.18; 2.77; 5.46.

- Robustness to reverse causality for GDP dampening (Table 2):
  - Specifications: exclude negative GDP growth (col 1); MPru lagged one quarter (col 2); MPru lagged one year (col 3); MPru = country average (col 4).
  - Findings: MPru continues to dampen effects of global shocks on GDP.
  - Example coefficients (MPru main effects): -0.639** (col 1); -1.286*** (col 2); -1.534*** (col 3).
  - Interaction terms (Ln VIX * MPru): 0.259**; 0.598***; 0.683***; 0.444** across columns (1)–(4).
  - Net outflows * MPru: 0.067***; 0.111***; 0.125***; 0.149**.
  - Observations: 1,846; 2,235; 2,153; 2,260. Countries = 38 across columns. F-statistics: 35.1; 32.9; 32.6; 29.7.

- Tests for omitted variables in GDP regressions (Table 3):
  - MPru main effect remains negative and significant in all specifications (examples): -1.412***; -1.285***; -1.439***; -1.428***; -1.402***; -1.399***; -1.679**; -1.348***; -1.357***; -1.128***; -1.090***.
  - Ln VIX main effect remains negative and significant (examples): -1.623***; -2.169***; -1.265*; -1.372***; -1.294***; -1.999***; -1.555***; -1.540***; -1.996***; -1.592***.
  - Net outflows main effect remains negative and significant (examples): -0.412***; -0.410***; -0.552***; -0.491***; -0.251***; -0.302**; -0.358**; -0.419***; -0.321***; -0.525***; -0.108***.
  - Interaction patterns between MPru and shocks (Ln VIX * MPru and Net outflows * MPru) remain positive and significant across specifications (examples): Ln VIX * MPru = 0.644***; 0.627***; 0.620***; 0.641***; 0.559***; 0.652***; 0.816**; 0.615***; 0.575***; 0.522***; 0.449***; Net outflows * MPru = 0.117***; 0.124***; 0.131***; 0.101***; 0.065**; 0.094**; 0.097*; 0.102***; 0.091***; 0.064**; 0.025*.

- Symmetry and category robustness:
  - Symmetric dampening (Table 4): MPru * D+ and MPru * D- coefficients indicate symmetric dampening; Wald tests p-values (examples) 0.421 and 0.943 do not reject equality.
  - Macroprudential categories (Table 5): heterogeneous effects across categories; several significant coefficients reported (examples): Net outflows * Mpru = 2.390** for capital category; Ln VIX * Mpru = 3.324*** for FX exposure.
  - Spillovers (Table 6): others’ MPru measures show significance in some cases (example: Others’ MPru = 4.296*); net outflows * others’ MPru often positive and significant (examples: 0.418**, 0.120***, 0.113**, 0.068**, 0.170***, 0.116***).

- Macroprudential regulation and monetary policy robustness (Tables 7–9):
  - Baseline domestic policy-rate regressions (Table 7): key coefficients include US policy rate = 0.597*** (column 1), Ln VIX = 1.396***, Net outflows = 0.404***, MPru = 0.220***, US policy rate * MPru = -0.013***, Ln VIX * MPru = -0.116***.
  - Table 8 (reverse causality): US policy rate * MPru = -0.014***; -0.018***; -0.005 (third loses significance); Ln VIX * MPru = -0.116***; -0.129***; -0.124***. Observations = 1,241; 1,211; 1,250; Countries = 25.
  - Table 9 (omitted variables): US policy rate * MPru and Ln VIX * MPru remain negative and significant in most specifications (examples): -0.012***; -0.012***; -0.008*; -0.017***; -0.021***; -0.008**; -0.013***; -0.005; -0.016*** and Ln VIX * MPru = -0.111***; -0.107***; -0.097***; -0.099***; -0.088***; -0.074**; -0.104***; -0.091***; -0.064***.
  - One exception: augmenting with official reserves where US policy rate interaction loses 10 percent significance.

- Comparison with capital controls (Tables 10–13):
  - Measures used: Chinn and Ito (2008), Fernández et al. (2015), Quinn and Toyoda (2008), Pasricha et al. (2018).
  - Dampening on GDP:
    - No systematic evidence that stricter capital controls (CC) dampen VIX or capital outflow effects on GDP; most CC interaction coefficients insignificant; some negative significant (possible exacerbation).
    - Positive and weakly significant interactions only in limited columns (examples): Table 10 column (3) against VIX and column (7) against net outflows.
    - Distinguishing inflow vs outflow controls (Table 11) shows no systemic pattern; mixed significant coefficients.
  - Effect on monetary policy:
    - Regressions replacing MPru with CC do not find evidence that stricter CC support more countercyclical monetary response.
    - Some CC * US policy rate coefficients are positive and significant (examples): 0.279***; 0.274***, implying more procyclical monetary response in some specifications.
    - Differentiating inflow/outflow controls (Table 13) yields similar mixed results (examples of CC * US policy rate: 0.219***; 0.258***; Fernández et al. net outflows * CC = 0.398***; 0.384***).
  - Interpretation: CC stringency does not systematically substitute for MPru in damping shocks or enabling countercyclical monetary policy; possible leakage/circumvention and focus on flexible-exchange-rate countries may explain null results. Caveat: CC may still help in extreme shocks or when optimally adjusted.

- Conclusions from robustness:
  - Main findings confirmed: MPru significantly dampens macroeconomic impacts of global financial shocks on EMs; effects symmetric; not driven by narrow instruments; no negative cross-country spillovers; MPru enables more countercyclical monetary policy.
  - Policy implication: a broad macroprudential framework helps EMs strengthen resilience; capital controls are not a valid substitute in the broad sample and specifications used.

---

### Appendix A. Data — sample, sources, classification, measures
- Sample and country coverage:
  - Sample consists of 38 EMs (April 2020 WEO classification), from 2000Q1 to 2016Q4.
  - Selection criteria: (i) population > 1 million; (ii) at least 10 years of GDP data; (iii) at least 5 years of net capital inflows data; (iv) MPru data from iMaPP (Alam et al. 2019).
  - Country list (Table A.2): 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.
- Data sources and variables (Table A.1):
  - Capital Flow Measures: Fernandez and others (2016).
  - Commodity terms of trade: Gruss and Kebhaj (2019).
  - Exchange rate regime: Ilzetzki and others (2019).
  - Expected inflation: Consensus forecast, Haver Analytics, IMF staff calculations.
  - Gross capital inflows/outflows: IMF, Balance of Payment Statistics.
  - Gross public debt and public FX debt: IMF, World Economic Outlook.
  - Inflation: Haver Analytics.
  - Inflation expectation anchoring index: Bems and others (2018).
  - Institutional quality: Worldwide Governance Indicators.
  - Official reserves: IMF, Balance of Payment Statistics.
  - Macroprudential regulation: iMaPP dataset, Alam and others (2019).
  - Net capital inflows: IMF, Balance of Payment Statistics.
  - Nominal effective exchange rate, real effective exchange rate, and other computed series: IMF staff calculations and Haver Analytics.
  - Policy rates: Bank for International Settlements; Haver Analytics; IMF, International Financial Statistics.
  - VIX: Haver Analytics.
- Exchange rate classification:
  - Uses Ilzetzki et al. (2019) coarse classification.
  - Flexible regimes include bands, crawls, managed floats (categories 2, 3, 4).
  - Fixed regimes include hard pegs, currency boards, horizontal bands, de facto pegs (category 1).
  - Freely falling regimes (category 5) are excluded.
- Macroprudential measures, aggregation, and measurement:
  - iMaPP provides information on 17 macroprudential tools (Alam et al., 2019).
  - Measures grouped into overall index and five subcategories: bank capital, credit demand, credit supply, foreign exchange positions, liquidity.
  - Stringency measured by cumulating net tightening actions since 1990; cumulated indexes rescaled to be positive to accommodate squared terms.
- Categories of macroprudential measures (Table A.3) — variable names in iMaPP in parentheses:
  - Capital: Capital Requirements (capital); Leverage Limits (LVR); Loan Loss Requirements (LLP); Countercyclical Capital Buffer (CCB); Capital Conservation Buffer (Conservation); SIFI measures (SIFI).
  - Credit Demand: Loan-To-Value ratio (LTV); Debt-Service-to-Income ratio (DSTI); Tax on Transactions (tax).
  - Credit Supply: Limits on Credit Growth (LCG); Loan Restrictions (LoanR).
  - FX Exposure: Limit on Foreign Currency (LFC); Limit on gross open FX positions (LFX); Reserve Requirements on FC assets (RRFCD).
  - Liquidity: Reserve Requirements (RRdom); Liquidity Measures (liquidity); Loan-to-Deposit ratio (LTD).
- Econometric and construction notes:
  - Cumulated MPru indexes computed since 1990 and rescaled to be positive.
  - Exchange rate regime classification follows Ilzetzki et al. (2019); category 5 (freely falling) excluded.

*Source: wpiea2020106-print-pdf (IMF).*

### 1. US policy rate and VIX

### 1. US policy rate and VIX

### Data and measures
- Global financial shocks included:
  - shocks to US monetary policy identified by Iacoviello and Navarro (2019).
  - Chicago Board Options Exchange’s Volatility Index (VIX).
  - net capital inflows (normalized by the HP-trend of GDP and instrumented to isolate global push factors).
- Macroprudential regulation index:
  - Constructed by cumulating macroprudential tightening actions net of loosening actions reported in the iMaPP database from 1990 to 2016.
  - The index is rescaled upward across all countries so values are always positive because it enters quadratically in regressions.
- Sample timing and series references visible in figures and notes:
  - Quarterly axes shown for 2000q1, 2005q1, 2010q1, 2015q1.
  - US policy rate is the federal funds rate except during the zero lower bound (ZLB) period where the implied rate from Wu and Xia (2015) is used.
- Data sources cited: Bank for International Settlements; IMF, International Financial Statistics database; Haver Analytics; Wu and Xia (2015); IMF, Balance of Payments; iMaPP database; authors’ calculations.

### Macroprudential policy evolution
- Evidence from the iMaPP database:
  - Macroprudential regulation has been considerably tightened over the years, especially since 2005.
  - The global financial crisis led to a temporary loosening but emerging markets returned to tighten regulation during the subsequent recovery.
  - There is substantial cross-country dispersion in the level of macroprudential regulation.
- Panel measurements reported in the source:
  - Panel 1: Mean of net tightening actions (time series).
  - Panel 2: Cumulative net tightening actions with interdecile range shading.

### Empirical strategy
- Regression specification (as in the source):
  - Y_{i,t} = α_i + β S_{i,t} + γ(S_{i,t}·MPru_{i,t}) + δ(S_{i,t}·MPru^2_{i,t}) + ζ MPru_{i,t} + θ MPru^2_{i,t} + κ C_{i,t} + ε_{i,t}
  - Y_{i,t} denotes quarterly real GDP growth for country i at time t; α_i is a country fixed effect.
  - S_{i,t} is the vector of global financial shocks; MPru_{i,t} is the level of macroprudential regulation.
  - Specification includes quadratic interactions to allow for non-linear effects.
- Controls (vector C_{i,t}) include:
  - lagged GDP growth, lagged log of real GDP per capita, institutional quality, a linear trend, lagged output gap, and commodity terms of trade.
- Identification and estimation:
  - Net capital inflows instrumented with gross inflows to other emerging markets following Blanchard et al. (2017).
  - Two-stage least squares procedure with Driscoll and Kraay (1998) correction to standard errors to address cross-sectional dependence, autocorrelation, and heteroscedasticity.
  - Time fixed effects are excluded in baseline to permit estimation of common global financial shocks; robustness checks include time fixed effects.

### Main empirical findings
- Dampening effects on GDP growth:
  - A more stringent level of macroprudential regulation significantly reduces the sensitivity of GDP growth in emerging markets to global financial shocks.
  - The dampening effects display decreasing marginal returns: at higher levels of macroprudential regulation, additional tightening is less effective.
  - Dampening properties are not driven by a single tool; a broad range of measures contribute (tools that boost bank capital and liquidity, limit foreign exchange exposures, and avert risky credit).
  - Dampening effects are heterogeneous across types of global financial shock.
  - Macroprudential regulation leads to symmetric dampening effects: it supports GDP growth when global financial shocks are adverse but can lower economic activity when global financial conditions are favorable.
- Spillovers:
  - No evidence of negative spillovers from one country tightening macroprudential regulations; instead, a higher level of macroprudential regulation in one country tends to enhance macroeconomic stability in other countries.
- Interaction with monetary policy:
  - Macroprudential regulation allows for a more countercyclical monetary policy response to global financial shocks.
  - At low levels of macroprudential regulation central banks tend to respond procyclically (increasing rates when global financial conditions tighten).
  - At more stringent levels of regulation, monetary policy response becomes countercyclical (declining policy rates when global financial conditions tighten).
- Comparison with capital controls:
  - Using capital control indicators from Chinn and Ito (2008), Fern ández et al. (2015), Quinn and Toyoda (2008), and Pasricha et al. (2018), the analysis does not find evidence that tighter capital controls provide similar benefits to those of tighter macroprudential regulation.
  - Findings align with Forbes et al. (2015) and Frost et al. (2020) that macroprudential tools, especially FX-based measures, are more effective than capital controls in influencing macroeconomic outcomes.

### Caveats, robustness, and limitations
- Measurement limitations:
  - Index of macroprudential regulation is subject to measurement drawbacks: initial cross-country differences in 1990 and the dummy-type nature of policy action indicators that do not capture intensity (iMaPP records tighten/loosen but not intensity, except LTV limits).
  - Measurement imprecision is likely to bias results toward underestimating effects rather than producing spurious significant impacts.
- Endogeneity concerns:
  - Robustness tests exploit both time-series and cross-sectional identification, but residual omitted-variable bias cannot be entirely ruled out.
  - Any omitted variable that biases results would need to comove over time and across countries with macroprudential regulation and affect macroeconomic resilience.
- Robustness checks:
  - Results robust to a broad set of endogeneity tests, inclusion of time fixed effects in robustness section, and both time-series and cross-sectional approaches.

### Relation to existing literature
- Confirms and extends literature on macroprudential effectiveness:
  - Prior findings: borrower-based and financial macroprudential tools affect credit growth in emerging markets (Cerutti et al., 2017), and macroprudential policies can curb credit and house price growth (Fendoğlu, 2017; Akinci and Olmstead-Rumsey, 2018).
  - Micro-level evidence shows effectiveness at bank/sector level (Saurina, 2009; Jiménez et al., 2017; Aiyar et al., 2016; Dassatti et al., 2019).
  - Some studies find temporary GDP declines following macroprudential tightening (Kim and Mehrotra, 2018; Eickmeier et al., 2018; Richter et al., 2019), while others find longer-term growth benefits (Boar et al., 2017; Agénor et al., 2018; Neanidis, 2019).
- Distinct contribution:
  - This paper focuses on whether the level (not changes) of macroprudential regulation affects the transmission of global financial shocks to domestic economies, reducing endogeneity concerns relative to analyses of regulatory changes.

*Source: wpiea2020106-print-pdf (IMF).*

### 2.1    Baseline results

### 2.1    Baseline results

### Baseline regression findings
- Table 1 regression results:
  - Column (1): An increase in the VIX and an outflow of capital have negative and highly statistically significant effects on economic growth.
  - Column (2): US monetary shocks have a detrimental effects on growth if the regression does not control for other global financial shocks, but they lose significance when controlling for the VIX and capital flows.
  - Interpretation: Changes in US monetary policy affect emerging markets through changes in risk premia proxied by the VIX and effects on the supply of foreign capital, rather than through changes in risk-free rates.
- Control variables:
  - The coefficient on the lag of the output gap is negative and significant, suggesting deviations from potential growth tend to be reduced over the following quarter.
- Instrumentation:
  - The instrumentation approach for net capital flows appears reliable since the F-statistic is well above the conventional threshold.
  - Reported statistic: Kleibergen-Paap rk WaldF-statistic (appropriate with more than one endogenous regressor and non-iid errors).
- Diagnostic tests:
  - Lagrange Multiplier tests point to the existence of serial correlation.
  - Modified Wald test for group-wise heteroskedasticity indicates heteroscedasticity.
  - Pesaran test, Frees test, and Friedman test all reject the null hypothesis of cross-sectional independence.

### Dampening effects of macroprudential regulation and nonlinearity
- Inclusion of macroprudential regulation (column (3)):
  - The coefficients on the VIX and net outflows remain significant.
  - The coefficients on the interaction terms between the shocks and macroprudential regulation are negative and highly statistically significant, implying a more stringent level of macroprudential regulation dampens the effects of global financial shocks on GDP growth.
  - Column (4): Results robust to excluding periods with fixed exchange rates.
- Nonlinear specification (column (5)):
  - Regression extended to include the squared level of macroprudential regulation and its interactions with shocks.
  - Findings corroborate contractionary effects of VIX increases and capital outflows and the dampening effects of macroprudential regulation.
  - Interaction terms of the VIX and capital flows with the squared level of regulation are negative and statistically significant.
  - Conclusion: Dampening effects of macroprudential regulation are subject to decreasing marginal returns; they weaken as the level of regulation tightens (consistent with diminishing marginal effectiveness and possible circumvention into shadow or international credit).
- Derivative of GDP growth with respect to shock s_j (from equation (2)):
  - ∂Y_i,t/∂S_j,t = β + γ_j MPru_i,t + δ_j MPru^2_i,t
  - This is a nonlinear function of the level of macroprudential regulation.

### Quantitative illustrations (Figure 3 findings)
- At the lowest level of macroprudential regulation in the sample:
  - A doubling of the VIX (100 percent increase) leads to a decline in quarterly GDP growth by 1.8 percentage points.
  - A net outflow worth three percent of GDP triggers similar effects.
- Median-level macroprudential regulation in the sample:
  - The same shocks produce a GDP decline of only 0.5 percentage points.
- If macroprudential regulation is sufficiently tight:
  - The VIX and net capital outflows no longer have statistically significant effects on emerging markets’ GDP.
- Confidence intervals:
  - Shaded areas correspond to 90 percent confidence intervals computed with Driscoll-Kraay standard errors.
- Notes:
  - Net capital outflows are scaled by the HP-trend of GDP.
  - Distribution of macroprudential regulation shown for 2000–2016 and at end-2016.

### Robustness to reverse causality and omitted variables
- Concern: Level of macroprudential regulation may respond to changes in GDP growth (reverse causality).
  - Persistence and lower volatility of the macroprudential index: it is obtained by cumulating past tightening and loosening actions and is largely predetermined to shock realizations and GDP responses.
  - Narrative evidence: Richter et al. (2019) find only three of 92 changes in loan-to-value ratios were motivated by GDP, inflation, or other real variables.
- Table 2 robustness checks for reverse causality (specifications and results):
  - Excluding observations with negative GDP growth (column 1).
  - Lagging macroprudential regulation by one quarter (column 2).
  - Lagging macroprudential regulation by one year (column 3).
  - Replacing time-varying levels with time-invariant country averages for 2000—2016 (column 4) to rely exclusively on cross-country heterogeneity.
  - Across all these specifications, the coefficients on the interaction terms between macroprudential regulation and the VIX or net outflows remain positive and highly statistically significant; dampening effects appear robust to reverse causality.
- Omitted variable tests (Table 3):
  - Augmented regressions include interactions of global shocks with:
    - Country structural characteristics: institutional quality and financial development (IMF’s Financial Development Index).
    - Fiscal variables: gross public debt in percent of GDP, gross public debt in foreign currency in percent of total public debt, cyclically-adjusted fiscal balance in percent of GDP.
    - Monetary policy variables: monetary policy rate and anchoring of inflation expectations.
    - Exchange rate regime (fixed vs floating), stringency of capital controls, stock of official reserves in percent of GDP.
  - Results: Dampening effects of macroprudential regulation are robust to all omitted variable tests.
  - Instrumentation caveat: When instrumenting three variables (net outflows, interaction with level of macroprudential regulation, and interaction with relevant test variable), the F-statistic falls below ten for institutional quality, public debt, and inflation expectation anchoring index.
- Time fixed effects and sensitivity:
  - Column (11) in Table 3: results robust to inclusion of time fixed effects (cannot separately estimate impact of common shocks like US monetary shocks and the VIX, but can estimate interactions with country-level macroprudential regulation).
  - Interactions with the VIX and capital outflows remain positive and statistically significant.
  - Results robust to excluding any country and to dropping the period of the global financial crisis.

### Symmetric dampening effects (positive and negative shocks)
- Extended specification (equation (3)) differentiates positive and negative realizations of global shocks using dummies D^+_i,t and D^-_i,t and estimates separate dampening coefficients γ^+ and γ^-.
- Table 4 results:
  - γ^+ and γ^- coefficients for both the VIX (column 1) and capital outflows (column 2) are statistically significant and positive.
  - Coefficients γ^+ and γ^- are quite similar in size; a Wald test confirms they are not statistically different.
  - Interpretation: Macroprudential regulation entails symmetric dampening effects against both positive and negative global financial shocks.
    - A tighter level of macroprudential regulation supports growth in case of negative financial shocks.
    - The same tighter regulation lowers economic activity when global financial shocks are positive.
- Illustration (Figure 4):
  - Growth differential between a country with macroprudential regulation at the 75th percentile and one at the 25th percentile over 2000–2016.
  - Example: Higher macroprudential regulation would have increased quarterly GDP growth by about 0.6 percent between the fourth quarter of 2008 and the second quarter of 2009.
  - Conversely, macroprudential regulation would have lowered economic growth considerably in years before the global financial crisis when global financial conditions were buoyant.
  - Shaded area: 90 percent confidence interval computed with Driscoll-Kraay standard errors.
- Policy implication:
  - Maintaining a high level of macroprudential regulation is not costless: it forgoes growth opportunities when global financial conditions are favorable.
  - Tightening regulation only when conditions are adverse would be improper because constraining excessive risk-taking when conditions are loose is a key channel to ensure resilience during distress.
  - Calls for further analysis on how to optimally adjust macroprudential regulation to maximize protection against negative shocks without unduly constraining activity when conditions are supportive.

### Categories of macroprudential measures
- Disaggregation: The overall index of macroprudential regulation is decomposed into categories:
  - Bank capital and liquidity.
  - Credit demand (such as loan-to-value ratios).
  - Credit supply (such as limits on credit growth).
  - Foreign currency exposure.
  - (Mapping from individual tools to these categories described in Table A.3; analysis cannot be run on individual tools due to data sparsity.)
- Table 5 and Figure 5 results (coefficients on interaction terms with shocks; level of regulation divided by 10 for visualization):
  - VIX (panel 1): Measures targeted at credit demand, FX exposure, and liquidity dampen the impact on GDP growth arising from the VIX.
  - Net capital outflows (panel 2): Measures targeted at bank capital, credit demand, and credit supply strengthen resilience against net capital outflows.
  - Positive and statistically significant coefficients denote dampening effects; vertical lines correspond to 90 percent confidence intervals with Driscoll-Kraay standard errors.
- Interpretation:
  - Dampening effects are not driven by a narrow set of measures; they are not limited to FX-exposure measures that operate similarly to capital controls.
  - Ensuring adequate bank capital and liquidity and preventing excessive risk-taking in credit provision also play crucial roles in fostering resilience to global financial shocks.
  - Policy suggestion: Countries seeking resilience to both VIX and capital flow shocks should adopt a well-rounded macroprudential framework rather than focusing narrowly on a few measures.

*Source: Authors’ calculations, as reported in the chapter "2.1 Baseline results" of the provided content.*

### 2.5    Cross-country spillovers

### 2.5    Cross-country spillovers

### Objective and conceptual framework
- Examines whether macroprudential regulation generates cross-country spillovers: tighter domestic macroprudential regulation could
  - expose other countries to greater volatility through relocation of risky financial activities, or
  - generate positive spillovers by making the regulated country more resilient and thereby stabilizing trade and financial flows.
- References arguments linking similar concerns to capital flow management measures.

### Empirical specification and identification
- Extended regression (equation (4)) estimates:
  - Yi,t = αi + β Si,t + γ(Si,t MPrui,t) + γ(Si,t MPrūi,t) + ζ MPrui,t + κ Ci,t + εi,t
  - Includes interaction of global financial shocks Si,t with:
    - country i’s level of macroprudential regulation MPrui,t, and
    - the average level of regulation in other emerging markets MPrūi,t (weighted by average size of gross capital inflows).
- MPrūi,t is computed in alternative ways to capture likely channels of spillovers:
  - Column (1): average level of regulation in all other emerging markets.
  - Column (2): average in countries within the same geographical region.
  - Columns (3) and (4): differentiated by whether GDP per capita is above or below the median (column (3) uses quarterly GDP so groups can change over time; column (4) uses average GDP over the 2002/2016 sample so group assignment is time invariant).
  - Columns (5) and (6): differentiated by political/economic/financial/institutional risk class using a composite risk index from the Political Risk Service.

### Main findings on spillovers
- The analysis does not find evidence of negative spillovers from macroprudential regulation.
- Spillovers tend to be positive vis-à-vis net outflows:
  - The coefficient on the interaction between net outflows and the average level of macroprudential regulation in other emerging markets is positive and significant regardless of how MPrūi,t is computed.
  - Interpretation: a country becomes more resilient to capital flow shocks if other emerging markets have a higher level of macroprudential regulation.
- Magnitudes:
  - Across columns (2) to (6) the coefficients on the interaction of net capital outflows with MPrūi,t are similar to those on the interaction with each country’s own level of macroprudential regulation.
  - This implies that countries experience similar dampening effects from their own level of macroprudential regulation and from the average level in other similar countries.
  - Caveat: the coefficients on the interaction with MPrūi,t capture the effect of a one-unit increase in the average level of macroprudential regulation in all other emerging markets within the same group — a larger tightening than a one-unit increase of an economy’s own level of macroprudential regulation.

### Implications
- No empirical support for the hypothesis that tighter macroprudential measures in one emerging market create negative spillovers that increase other countries’ susceptibility to global financial shocks.
- Instead, higher average macroprudential regulation across peer emerging markets mitigates the adverse impact of net capital outflows on a given country’s GDP growth.

_Italic: Source: wpiea2020106-print-pdf - 2.5    Cross-country spillovers_

### 3.2    Robustness

### 3.2    Robustness

### Reverse causality tests
- Purpose: Check whether countries move macroprudential regulation in response to policy rate changes (reverse causality).
- Specifications tested (Table 8):
  - MPru = one-quarter lag of MPru.
  - MPru = one-year lag of MPru.
  - MPru = country average of MPru between 2000–2016 (estimation relies solely on cross-country variation).
- Sample and methodology:
  - Estimations based on a sample of EM from 2000Q1 to 2016Q4.
  - All specifications include country fixed effects.
  - Net inflows (in percent of trend GDP) for each country are instrumented using gross inflows to other EMs (in percent of trend GDP).
  - The US policy rate is the effective federal funds rate except during the zero lower bound period where the implied policy rate from Wu and Xia (2015) is used.
- Findings:
  - Across lag specifications, macroprudential regulation continues to support a more countercyclical response of monetary policy to global financial conditions.
  - When using country average MPru, regulation supports a more countercyclical response to capital flow shocks rather than to changes in US monetary policy.
  - Table 8 reports observations = 1,241; 1,211; 1,250 and countries = 25; 25; 25 for columns (1)–(3) respectively.
  - F-statistics reported: 22.97; 22.15; 20.53.

### Omitted variable tests
- Purpose: Address omitted variable bias by augmenting baseline regressions with additional structural and policy variables and their interactions with MPru (Table 9).
- Variables included in separate specifications:
  - Institutional quality.
  - Financial development.
  - Gross public debt.
  - Public debt in foreign currency (government FX debt).
  - Cyclically adjusted balance.
  - Inflation expectation anchoring.
  - Capital flow measures.
  - Official reserves (percent of trend GDP).
  - Time fixed effects.
- Sample and methodology:
  - Estimations based on a sample of EM from 2000Q1 to 2016Q4.
  - MPru is divided by 10 to ease visualization.
  - Driscoll-Kraay standard errors reported in parentheses.
- Findings (Table 9, columns (1)–(9)):
  - The coefficients on the interaction of MPru and the VIX remain negative and significant in all tests (reported as Ln VIX * MPru coefficients: e.g., -0.111***, -0.107***, -0.097***, -0.099***, -0.088***, -0.074**, -0.104***, -0.091***, -0.064***).
  - The interaction coefficients with the US policy rate are negative and significant in most specifications (US policy rate * MPru: -0.012***, -0.012***, -0.008*, -0.017***, -0.021***, -0.008**, -0.013***, -0.005, -0.016***), except the specification augmented with official reserves where the interaction remains negative but loses statistical significance at the 10 percent level.
  - Column (9) shows the interaction coefficients with US policy rates and the VIX remain negative and significant even with time fixed effects.
  - Observations vary by specification (e.g., 1,250; 1,239; 1,157; 1,129; 1,250; 1,236; 1,250; 1,236; 1,250) and countries vary (e.g., 25; 25; 24; 24; 18; 25; 25; 24; 25). F-statistics reported across columns include 3.95; 26.90; 11.87; 12.45; 5.79; 2.74; 20.18; 2.77; 5.46.

### Robustness to reverse causality for GDP dampening effects
- Purpose: Test whether dampening effects on GDP growth are robust when excluding negative GDP growth observations and when using lagged or average MPru (Table 2).
- Specifications (Table 2 columns (1)–(4)):
  - Excluding negative GDP growth.
  - MPru = one-quarter lag of MPru.
  - MPru = one-year lag of MPru.
  - MPru = country average of MPru.
- Sample and methodology:
  - Estimations based on EM from 2000Q1 to 2016Q4; country fixed effects; MPru divided by 10.
- Findings:
  - Across these specifications MPru continues to dampen the effects of global financial shocks on GDP growth.
  - Example coefficients (MPru and interactions): MPru = -0.639** (column 1); MPru = -1.286*** (column 2); MPru = -1.534*** (column 3).
  - Interaction terms: Ln VIX * MPru = 0.259**; 0.598***; 0.683***; 0.444** across columns (1)–(4). Net outflows * MPru = 0.067***; 0.111***; 0.125***; 0.149**.
  - Observations: 1,846; 2,235; 2,153; 2,260. Countries = 38 across columns. F-statistics: 35.1; 32.9; 32.6; 29.7.

### Tests for omitted variables in GDP regressions
- Purpose: Augment GDP regressions with controls and interactions to test robustness of MPru dampening effects (Table 3).
- Controls included (each in turn as X and interacted with shocks): institutional quality; financial development; gross public debt; gross public debt in foreign currency; cyclically adjusted balance; monetary policy rate; inflation expectation anchoring; fixed ER regime; capital flow measures; official reserves; time fixed effects.
- Findings:
  - MPru main effect remains negative and statistically significant in all specifications (e.g., MPru = -1.412***; -1.285***; -1.439***; -1.428***; -1.402***; -1.399***; -1.679**; -1.348***; -1.357***; -1.128***; -1.090***).
  - Ln VIX main effect remains negative and significant (e.g., Ln VIX = -1.623***; -2.169***; -1.265*; -1.372***; -1.294***; -1.999***; -1.555***; -1.540***; -1.996***; -1.592***).
  - Net outflows main effect remains negative and significant (e.g., -0.412***; -0.410***; -0.552***; -0.491***; -0.251***; -0.302**; -0.358**; -0.419***; -0.321***; -0.525***; -0.108***).
  - Interaction patterns between MPru and shocks (Ln VIX * MPru and Net outflows * MPru) remain positive and significant across specifications (e.g., Ln VIX * MPru = 0.644***; 0.627***; 0.620***; 0.641***; 0.559***; 0.652***; 0.816**; 0.615***; 0.575***; 0.522***; 0.449***; Net outflows * MPru = 0.117***; 0.124***; 0.131***; 0.101***; 0.065**; 0.094**; 0.097*; 0.102***; 0.091***; 0.064**; 0.025*).

### Symmetry and instrument heterogeneity
- Symmetric dampening (Table 4):
  - The dampening effects of MPru are symmetric against positive and negative movements of Ln VIX and net outflows.
  - MPru * D+ = -1.101** and MPru * D- = -1.721*** (column 1); MPru * D+ = -1.324*** and MPru * D- = -1.346*** (column 2).
  - Wald tests for equality between coefficients on (S * D+ * MPru) and (S * D- * MPru): p-values 0.421 and 0.943 indicating no rejection.
- Macroprudential categories (Table 5):
  - Different MPru categories included: capital, credit demand, credit supply, FX exposure, liquidity.
  - Results show heterogeneous effects across categories; several statistically significant coefficients reported (e.g., Net outflows * Mpru = 2.390** for capital category; Ln VIX * Mpru = 3.324*** for FX exposure).
- Spillovers (Table 6):
  - Others’ MPru measures (averages across EMs, regions, income class, risk class) included.
  - Others’ MPru coefficients show some significance (e.g., Others’ MPru = 4.296* in column (1)), and net outflows * others’ MPru often positive and significant (e.g., 0.418**, 0.120***, 0.113**, 0.068**, 0.170***, 0.116***).
  - Wald tests reported (p-values e.g., 0.063; 0.993; 0.384; 0.078; 0.511; 0.739).

### Macroprudential regulation and monetary policy robustness
- Baseline domestic policy rate regressions (Table 7):
  - Key coefficients: US policy rate = 0.597*** (column 1 baseline), Ln VIX = 1.396***, Net outflows = 0.404***, MPru = 0.220***, US policy rate * MPru = -0.013***, Ln VIX * MPru = -0.116***.
  - Sample: Observations = 1,360; 1,262; 1,250; 1,250; 1,250 across specifications; Country FE included; sample excludes pegged and freely falling exchange rates.
- Robustness to reverse causality for domestic policy rates (Table 8):
  - Using lagged and average MPru, MPru main effects remain positive and interactions preserve pattern: US policy rate * MPru = -0.014***; -0.018***; -0.005 (third loses significance); Ln VIX * MPru = -0.116***; -0.129***; -0.124***.
  - Observations in Table 8: 1,241; 1,211; 1,250; Countries = 25 across columns.
- Robustness to omitted variables for domestic policy rates (Table 9):
  - Across specifications augmenting with X variables, MPru main effect remains positive and US policy rate * MPru and Ln VIX * MPru remain negative and significant in most columns (e.g., US policy rate * MPru = -0.012***; -0.012***; -0.008*; -0.017***; -0.021***; -0.008**; -0.013***; -0.005; -0.016***; Ln VIX * MPru = -0.111***; -0.107***; -0.097***; -0.099***; -0.088***; -0.074**; -0.104***; -0.091***; -0.064***).
  - One exception: specification augmented with official reserves where US policy rate interaction loses statistical significance at 10 percent level.
  - Column (9) with time fixed effects shows continued negative significant interactions.
  - Observations vary by column; F-statistics reported across columns (e.g., 3.95; 26.90; 11.87; 12.45; 5.79; 2.74; 20.18; 2.77; 5.46).

### Comparison with capital controls
- Purpose: Assess whether capital controls (CC) provide similar benefits as macroprudential regulation in damping global financial shocks and enabling countercyclical monetary policy.
- Measures used: indexes from Chinn and Ito (2008), Fernández et al. (2015), Quinn and Toyoda (2008), Pasricha et al. (2018).
- Dampening effects on GDP (Table 10 and Table 11):
  - Across specifications, increases in the VIX or capital outflows have negative effects on GDP growth.
  - No systematic evidence that more stringent capital controls dampen these effects: most interaction coefficients are insignificant; some are negative and significant (suggesting possible exacerbation).
  - Positive and weakly significant interactions only in column (3) against the VIX and column (7) against net capital outflows in Table 10.
  - Distinguishing inflow vs outflow controls (Table 11) does not reveal a systemic pattern; most coefficients insignificant; three positive significant cases balanced by three negative coefficients.
  - F-statistics and sample sizes vary across specifications (e.g., Table 10 Observations = 2,220; 2,220; 1,925; 1,925; 1,856; 1,856; 918; 918; Countries vary).
- Effect on monetary policy (Table 12 and Table 13):
  - Regressions replacing MPru with CC do not find evidence that stricter capital controls support a more countercyclical monetary policy response.
  - Most CC interaction coefficients with shocks in Table 12 columns (1)–(4) are insignificant. Only CC * US policy rate in column (1) and (2) are positive and significant (0.279***; 0.274***), and in column (4) CC * US policy rate = -0.019* (weakly significant). Several instances in columns (1) and (2) show positive and strongly significant coefficients suggesting tighter CC associated with a more procyclical monetary response.
  - When differentiating inflow and outflow controls (Table 13), similar findings: CC * US policy rate often positive and significant (e.g., 0.219***; 0.258***), and CC * Net outflows mostly positive and significant for Fernández et al. measures (e.g., 0.398***; 0.384***), while Pasricha measures show mixed or insignificant interactions.
  - Observations and countries vary across these tables (e.g., Table 12 Observations = 1,250; 1,250; 1,171; 806; Countries = 25; 25; 23; 15).
- Interpretation:
  - Overall, the stringency of capital controls does not systematically dampen the impact of global financial shocks on GDP in emerging markets, nor does it support a more countercyclical monetary policy response.
  - Possible reasons: post-2000 greater international capital market integration allowing leakage/circumvention; the analysis focuses on countries with flexible exchange rates (which already enjoy monetary independence).
  - Caveat: lack of significant effects should not dismiss potential gains from capital controls in extreme shocks or when optimally adjusted in response to shocks.

### Conclusions (from robustness section and adjoining discussion)
- Robustness checks confirm the main findings:
  - Macroprudential regulation significantly dampens macroeconomic impacts of global financial shocks on emerging markets.
  - A tighter level of regulation reduces sensitivity of GDP growth to fluctuations in risk premia (VIX) and changes in supply of international capital flows.
  - Dampening effects are symmetric (apply to both positive and negative shocks).
  - Effects are not driven by a narrow set of instruments; measures targeting liquidity, capital, foreign exchange exposures, and risky credit contribute to resilience.
  - No evidence of negative cross-country spillovers; higher MPru in one country may strengthen resilience in others.
  - A key channel: MPru allows monetary policy to respond more countercyclically to global financial shocks (at low MPru central banks increase policy rates when conditions tighten; at higher MPru central banks tend to lower policy rates when conditions tighten).
- Policy implication emphasized:
  - A sound macroprudential regulatory framework, using a broad range of measures, may substantially help emerging markets strengthen resilience against global financial shocks.
  - Imposing capital controls does not appear to be a valid substitute for adopting a solid macroprudential framework.

*Source: IMF authors’ calculations and tables in the provided content.*

### Appendix A.  Data

### Appendix A.  Data

### Sample and country coverage
- Sample consists of 38 EMs, based on the April 2020 World Economic Outlook classification, from the first quarter of 2000 to the last quarter of 2016.
- Country and period selection criteria:
  - (i) a population larger than one million,
  - (ii) at least 10 years of GDP data,
  - (iii) at least 5 years of data for net capital inflows,
  - (iv) data on macroprudential regulation from the iMaPP databases of Alam et al. (2019).
- Table A.2: Country Coverage
  - 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

### Data sources and variables (Table A.1)
- Capital Flow Measures: Fernandez and others (2016)
- Commodity terms of trade: Gruss and Kebhaj (2019)
- Exchange rate regime: Ilzetzki and others (2019)
- Expected inflation: Consensus forecast, Haver Analytics, IMF staff calculations
- Gross capital inflows: IMF, Balance of Payment Statistics
- Gross capital outflows without FX reserves: IMF, Balance of Payment Statistics
- Gross public debt in foreign currency to gross public debt: IMF, World Economic Outlook
- Gross public debt to GDP: IMF, World Economic Outlook
- Inflation: Haver Analytics
- Inflation expectation anchoring index: Bems and others (2018)
- Institutional quality: Worldwide Governance Indicators
- Official reserves: IMF, Balance of Payment Statistics
- Macroprudential regulation: iMaPP dataset, Alam and others (2019)
- Net capital inflows: IMF, Balance of Payment Statistics
- Nominal effective exchange rate: IMF staff calculations
- Nominal gross domestic product: Haver Analytics
- Policy rates: Bank of International Settlement; Haver Analytics; and IMF, International Financial Statistics
- Population: IMF, World Economic Outlook
- Real credit: Bank of International Settlement; Haver Analytics; and IMF, International Financial Statistics
- Real effective exchange rate: IMF staff calculations
- Real GDP: Haver Analytics
- Real house prices: IMF, Global House Watch
- Real investment: IMF, World Economic Outlook
- Real private consumption: IMF, World Economic Outlook
- Reserve assets: IMF, Balance of Payment Statistics
- VIX: Haver Analytics

_Source:  IMF staff compilation._

### Exchange rate classification
- Uses the coarse classification in Ilzetzki et al. (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 are excluded from the analysis (category 5).

### Macroprudential measures, aggregation, and measurement
- Data on macroprudential measures are from the IMF’s integrated Macroprudential Policy (iMaPP) that provides information on 17 macroprudential tools (Alam et al., 2019).
- Measures are grouped into:
  - an overall index, and
  - five subcategories targeting bank capital, credit demand, credit supply, foreign exchange positions, and liquidity.
- The stringency of macroprudential regulation is measured by cumulating the net tightening actions for each country since 1990, the first year in the iMaPP database.
- In the econometric analysis, the cumulated macroprudential indexes are rescaled across all countries so that values are always positive. This is because the regression framework includes squared values of these indexes.

### Categories of macroprudential measures (Table A.3)
- Capital
  - Capital Requirements (capital) — Including risk weights
  - Leverage Limits (LVR)
  - Loan Loss Requirements (LLP) — Including dynamic and sector-specific provisioning (e.g. housing)
  - Countercyclical Capital Buffer (CCB)
  - Capital Conservation Buffer (Conservation)
  - Measures targeted at SIFIs (SIFI) — Including capital and liquidity surcharges
- Credit Demand
  - Loan-To-Value ratio (LTV) — Mostly targeted at housing loans
  - Debt-Service-to-Income ratio (DSTI)
  - Tax on Transactions (tax) — Including stamp duties and capital gain taxes
- Credit Supply
  - Limits on Credit Growth (LCG) — Including penalties for values
  - Loan Restrictions (LoanR) — Tailored LCG conditional on loan and bank characteristics, or other factors
- FX Exposure
  - Limit on Foreign Currency (LFC) — Limits on foreign currency lending
  - Limit on gross open FX positions (LFX) — Including currency mismatch regulations
  - Reserve Requirements on FC assets (RRFCD)
- Liquidity
  - Reserve Requirements (RRdom) — On domestic currency assets
  - Liquidity Measures (liquidity) — Including liquidity coverage ratios, liquid asset ratios, net stable funding ratios, core funding ratios, and ext. debt restrictions
  - Loan-to-Deposit ratio (LTD) — Including penalties for high values

Notes: The name of the variable in the iMaPP dataset is reported in parentheses.

### Econometric and construction notes
- Cumulated macroprudential indexes are computed since 1990 (first year in iMaPP).
- Cumulated indexes are rescaled to be always positive to accommodate squared terms in regressions.
- Exchange rate regime classification follows Ilzetzki et al. (2019) coarse categories, with category 5 (freely falling) excluded.

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020106-print-pdf.pdf_