## ch2onlineannex - EXECUTIVE SUMMARY ## Source details **Canonical URL:** [ch2onlineannex - EXECUTIVE SUMMARY](https://www.imf.org/-/media/files/publications/gfsr/2021/april/english/ch2onlineannex.pdf) ## Other formats - [Markdown version](/-/media/files/publications/gfsr/2021/april/english/ch2onlineannex.pdf.md) - [Structured JSON version](/-/media/files/publications/gfsr/2021/april/english/ch2onlineannex.pdf.json) --- ### Data Sources and Leverage by Region - Aggregate macrofinancial indicators and sectoral leverage series: - Nonfinancial Corporations Leverage: debt-to-GDP ratio, country-specific or weighted global average, percent of GDP, quarterly — Institute of international finance, downloaded in mid-November 2020. - Household Leverage: debt-to-GDP ratio, country-specific or weighted global average, percent of GDP, quarterly — Institute of international finance, downloaded in mid-November 2020. - Output Gap: Output gap, constant prices in national currency, percent, annual — IMF, World Economic Outlook. - Real Gross Domestic Product: Gross domestic product, constant prices in national currency, quarterly — IMF, World Economic Outlook. - Consumer Price Index: Consumer price index, quarterly — IMF, World Economic Outlook. - Financial Conditions Indices (FCIs): quarterly — IMF Calculations (see Online Annex 1.1 of October 2018 GFSR). - Short-Term Interest rate: 3-month Treasury Bill rate, 3-month Money Market Rate, 3-month Deposit Rate, quarterly — Haver Analytics. - Monetary Policy Rate: Shadow Short Rate (SSR) estimates by Krippner for US, euro area, Japan, and UK; other countries from BIS. Quarterly — BIS; Leo Krippner. - Macroprudential Policy: 17 types of macroprudential measures, quarterly — Integrated Macroprudential Policy (iMaPP) database. - Sample for regional leverage charts includes 52 economies (list provided in source). ### Financial Conditions and Leverage: Empirical Specification and Key Findings - Baseline linear regression (equation (1)) details: - Dependent variable y_{i,t,h} = h-quarter change in sector-specific debt-to-GDP ratio. - Regressors: country fixed effects, fci_{i,t}, controls X_{i,t} (lagged inflation (y/y), real GDP growth (q/q), short-term interest rate, lagged one-quarter change in sector debt-to-GDP), and global time effects (GFC dummy for 2008:Q4–2009:Q4 or time fixed effects). - Standard errors: Driscoll-Kraay (1998) approach. - Sample: 19 advanced economies and 10 emerging economies; period 1996:Q1 to 2020:Q3. - Regime-dependent regression (equation (2)): - Θ_{i,t} = credit boom dummy: country-specific fci below median and eight-quarter change in credit-to-GDP in the top three deciles. - Robustness checks (non-exhaustive): - Residualized (purged) FCIs, global FCI, VIX, time fixed effects, inflation-adjusted percent changes in debt, alternative controls (output gap, unemployment, government budget balance), alternative credit-boom definitions, inclusion of fiscal controls. - Conclusion: results qualitatively robust to these alternatives. - Asymmetry analysis (equation (3)) — loosening vs tightening episodes: - Nonfinancial corporate debt: - A one-unit loosening in financial conditions leads to a 6 percentage point of GDP increase in non-financial corporate debt after eight quarters. - A one-unit tightening leads to a 1 ½ percentage point of GDP decrease in non-financial corporate debt after eight quarters. - Household debt: - A one-unit loosening leads to a 3 percentage point of GDP increase after three years. - A one-unit tightening leads to a 1 percentage point of GDP decrease after three years. - Visualization: Online Annex Figure 2.2.1 shows cumulative responses (one to sixteen quarters ahead) with 90 percent confidence intervals; blue line = loosening effects, red line = tightening effects. ### Growth-at-Risk (GaR) Analysis - Panel quantile local projection model (equation (4)): - Dependent variable y_{i,t,h}^{τ} = year-on-year growth rate of real GDP in quarter t+h. - Regressors: fci_{i,t}, 8-quarter changes in nonfinancial corporate leverage and household leverage, controls X_{i,t} (lagged y/y real GDP growth, lagged y/y inflation, global time trend). - Estimation at τ = 10th percentile; panel quantile method follows Machado and Santos Silva (2019); standard errors bootstrapped. - Sample: 19 advanced economies and 10 emerging market economies; period 1996:Q1 to 2020:Q3. - Regime-dependent GaR (equation (5)): interactions with credit-boom dummy Θ_{i,t} produce separate coefficient vectors for credit-boom and non-credit-boom regimes. - Robustness exercises: - Replace domestic fci with (i) global FCI at t, (ii) purged FCI at t, or (iii) FCIs over t to t+h. - Replace leverage changes by shocks obtained via Cholesky decomposition of the PVAR. - Replace time-trend with post-GFC dummy (zero before 2008:Q4, one from 2008:Q4 onward). - Include fiscal policy controls such as government balance to GDP ratio. - Use bandpass filter (Baxter and King 1999) to remove cyclical components (6–32 quarters) from real GDP (log), leverage, and FCI. ### Role of Macroprudential Policy - Macroprudential specification (equation (6)): - Dependent variable: changes in sector-specific debt-to-GDP ratio from t to t+h. - Macroprudential variable MMPru_{i,t} = net number of instruments tightened within a given category in quarter t (number tightened minus number loosened) — IMF iMaPP database. - Controls X_{i,t}: lagged inflation (y/y), real GDP growth (q/q), domestic FCIs, short-term interest rate, lagged one-quarter change in sector debt-to-GDP; GFC dummy for 2008:Q4–2009:Q4 (or time fixed effects). - Standard errors: Driscoll-Kraay approach. - Macroprudential categories analyzed: - Borrower-based: DSTI limits and LTV instruments. - Bank Liquidity: reserve requirements (RR), liquidity requirements, limits to loan-deposit ratio (LTD). - Bank FX exposure: limits on foreign currency lending (LFC), limits on gross open FX positions (LFX), reserve requirements on foreign currency assets (RR_FCD). - Robustness: results hold using time fixed effects and using inflation-adjusted percent changes in debt-to-GDP as dependent variables. - Macroprudential Tightening and Growth-at-Risk (equation (7)): - y_{i,t,h}^{τ} regression includes SSM_17_{i,t}, discrete variable equal to net number of macroprudential tightenings in quarter t (tightenings minus loosenings). - Quantile coefficient β_{4,h}^{τ} indicates direct impact of one additional net tightening action on the left-tail of future growth distribution. - Regime dummy Θ_{i,t}: equals 1 if sum of SSM_17_{i,t} in past 4 quarters is positive; else 0. - Augmented regime specification (equation (8)) includes ∆NNNS_8q (8-quarter change in nonfinancial sector debt-to-GDP). - Estimation: panel quantile models per Machado and Santos Silva (2019); standard errors bootstrapped. - Sample: 19 advanced economies and 10 emerging market economies; period 1996:Q1 to 2020:Q3. - Example: if a country tightened LTV twice and loosened RR once in quarter t, SSM_17_{i,t} = 1. ### Literature Review on Leverage and Financial Vulnerability (Online Annex Box 2.1) - Key themes: - High levels or rapid increases in nonfinancial sector leverage predict downside risks to growth and financial stress. - Distinction between beneficial credit deepening and disruptive private credit booms (Verner 2019). - Most instances of financial distress are preceded by credit booms (Schularick and Taylor 2012). - High leverage associated with dampened business cycle volatility but more pronounced crashes (Jordà and others 2016). - High corporate leverage associated with weaker post-crisis recovery due to debt overhang (Kalemli-Ozcan and others 2020). - Household debt accumulation particularly related to financial stability risk and boom-bust dynamics (Mian, Sufi, and Verner 2017). - Adrian and others (forthcoming): loose financial conditions plus rapid aggregate credit growth reduces near-term downside risk but increases medium-run downside risk from a GaR perspective. - Additional note: household debt tends to be concentrated in low-wealth quantiles, amplifying COVID-19 shock effects on household sector vulnerabilities (Institute of International Finance, 2020). ### System Approach: PVAR (Online Annex Box 2.2) — Model and Key Impulse Responses - Purpose: estimate joint dynamics of economic activity, household leverage, non-financial corporate leverage, and FCI in a system treating all variables as endogenous. - Model features: - Panel Vector Autoregressive (PVAR) with recursive identification and dynamic time dummies to separate national from global leverage. - Impulse response functions relate to mean responses (contrast with GaR low-quantile analysis). - Key PVAR results: - An unexpected loosening of financial conditions by one standard deviation leads to an increase in household leverage by about 0.2 percentage points after 4 years. - The PVAR suggests the increase in household leverage is about half the size of the increase in non-financial corporate leverage. - PVAR impulse response evidence (as described): - Looser financial conditions increase household debt and have an even larger impact on non-financial corporate debt. - A one-standard-deviation loosening in FCI leads to an increase in real GDP by about 0.2 to 0.3 percent in the 4 years after the shock, with most of the boost frontloaded in the first year. - The positive effect of looser financial conditions on output tends to fade after about 6 quarters. - An unexpected increase in household leverage by one standard deviation leads, on average, to a fall in real GDP of about 0.2 percent after 4 years; the drag is immediate and builds across horizons. - A shock to non-financial corporate leverage reduces GDP by about 0.1 percent within the first 4 years after the shock. ### Macroprudential Policy: Transmission Channels and Policy Implications - Purpose: Macroprudential policy—aimed at mitigating the materialization of systemic risk to achieve financial stability (ESRB 2014)—can tame leverage buildups and reduce risks to macrofinancial stability. - Tools and roles: - Borrower-based tools (LTV and DSTI) constrain borrowing eligibility from the demand side and provide affordability buffers to improve resiliency of banks. - Supply-based tools include credit growth limits, capital- (capital adequacy and loan loss provisions), and liquidity-oriented tools (liquidity coverage, net stable funding, and loan-to-deposit ratios). - Sector-based tools include FX exposure limits and sector-specific loan restrictions. - Broad-based tools include countercyclical capital buffers to lean against the buildup phase of the credit cycle. - Effects: - These tools can help tame leverage buildups and financial imbalances and reduce systemic risks. - They may help moderate growth volatility directly (e.g., by reducing default risk) or indirectly by dampening the impact of financial conditions on growth. - Policy cautions and implementation notes: - Policymakers should be mindful of leakages, such as cross-border circumvention and domestic spillover effects (e.g., credit demand shifts from banks to nonbank financial institutions). - Communications with timely announcements for implementation have been shown to enhance policy effectiveness. *Source: ch2onlineannex - EXECUTIVE SUMMARY (Online annexes to "Loose Financial Conditions, Rising Leverage, and Risks to Macro-Financial Stability")* ### EXECUTIVE SUMMARY ### ch2onlineannex - EXECUTIVE SUMMARY ### Data Sources and Leverage by Region - Aggregate macrofinancial indicators and sectoral leverage series: - Nonfinancial Corporations Leverage: debt-to-GDP ratio, country-specific or weighted global average, percent of GDP, quarterly — Institute of international finance, downloaded in mid-November 2020. - Household Leverage: debt-to-GDP ratio, country-specific or weighted global average, percent of GDP, quarterly — Institute of international finance, downloaded in mid-November 2020. - Output Gap: Output gap, constant prices in national currency, percent, annual — IMF, World Economic Outlook. - Real Gross Domestic Product: Gross domestic product, constant prices in national currency, quarterly — IMF, World Economic Outlook. - Consumer Price Index: Consumer price index, quarterly — IMF, World Economic Outlook. - Financial Conditions Indices (FCIs): quarterly — IMF Calculations (see Online Annex 1.1 of October 2018 GFSR). - Short-Term Interest rate: 3-month Treasury Bill rate, 3-month Money Market Rate, 3-month Deposit Rate, quarterly — Haver Analytics. - Monetary Policy Rate: Policy rate series for US, euro area, Japan, and UK are Shadow Short Rate (SSR) estimates by Krippner; policy rates for other countries from BIS. Quarterly — BIS; Leo Krippner. - Macroprudential Policy: 17 types of macroprudential measures, quarterly — Integrated Macroprudential Policy (iMaPP) database. - Sample for regional leverage charts includes 52 economies (list provided in source). ### Financial Conditions and Leverage: Empirical Specification and Key Findings - Baseline linear regression (equation (1)) specification: - Dependent variable y_{i,t,h} = h-quarter change in sector-specific debt-to-GDP ratio. - Regressors include country fixed effects, a measure of financial conditions (fci_{i,t}), controls X_{i,t} (lagged inflation (y/y), real GDP growth (q/q), short-term interest rate, and lagged one-quarter change in sector debt-to-GDP), and global time effects (GFC dummy for 2008:Q4–2009:Q4 or time fixed effects in robustness checks). - Standard errors: Driscoll-Kraay (1998) approach. - Sample: 19 advanced economies and 10 emerging economies; period 1996:Q1 to 2020:Q3. - Regime-dependent regression (equation (2)): - Θ_{i,t} is a credit boom dummy: country-specific fci below median and eight-quarter change in credit-to-GDP in the top three deciles. - Robustness checks performed (non-exhaustive): - Use residualized (purged) FCIs, global FCI, VIX, time fixed effects, inflation-adjusted percent changes in debt, alternative controls (output gap, unemployment, government budget balance), alternative credit-boom definitions, inclusion of fiscal controls. - Conclusion: results qualitatively robust to these alternatives. - Asymmetry analysis (equation (3)) — loosening vs tightening episodes: - A one-unit loosening in financial conditions leads to a 6 percentage point of GDP increase in non-financial corporate debt after eight quarters. - A one-unit tightening in financial conditions leads to a 1 ½ percentage point of GDP decrease in non-financial corporate debt after eight quarters. - For household debt: a one-unit loosening leads to a 3 percentage point of GDP increase after three years; a one-unit tightening leads to a 1 percentage point of GDP decrease after three years. - Visualization: Online Annex Figure 2.2.1 shows cumulative responses (one to sixteen quarters ahead) with 90 percent confidence intervals; blue line = loosening effects, red line = tightening effects. ### Growth-at-Risk (GaR) Analysis - Panel quantile local projection model (equation (4)): - Dependent variable y_{i,t,h}^{τ} = year-on-year growth rate of real GDP in quarter t+h. - Regressors include country financial condition index fci_{i,t}, 8-quarter changes in nonfinancial corporate leverage and household leverage, and controls X_{i,t} (lagged y/y real GDP growth, lagged y/y inflation, global time trend). - Estimation at τ = 10th percentile; panel quantile method follows Machado and Santos Silva (2019); standard errors bootstrapped. - Sample: 19 advanced economies and 10 emerging market economies; period 1996:Q1 to 2020:Q3. - Regime-dependent GaR (equation (5)): interactions with credit-boom dummy Θ_{i,t} (as defined in Online Annex 2.2) produce separate coefficient vectors for credit-boom and non-credit-boom regimes. - Robustness exercises for GaR: - Replace domestic fci with (i) global FCI at t, (ii) purged FCI at t, or (iii) FCIs over t to t+h to mitigate endogeneity/reverse causality concerns. - Replace leverage changes by shocks obtained via Cholesky decomposition of the PVAR. - Replace time-trend with post-GFC dummy (zero before 2008:Q4, one from 2008:Q4 onward). - Include fiscal policy controls such as government balance to GDP ratio. - Use bandpass filter (Baxter and King 1999) to remove cyclical components (6–32 quarters) from real GDP (log), leverage, and FCI. ### Role of Macroprudential Policy - Macroprudential specification (equation (6)): - Dependent variable: changes in sector-specific debt-to-GDP ratio from t to t+h. - Macroprudential variable: MMPru_{i,t} = net number of instruments tightened within a given category in quarter t (number tightened minus number loosened) — from IMF iMaPP database. - Controls X_{i,t}: lagged inflation (y/y), real GDP growth (q/q), domestic FCIs, short-term interest rate, lagged one-quarter change in sector debt-to-GDP; GFC dummy for 2008:Q4–2009:Q4 (or time fixed effects in robustness checks). - Standard errors: Driscoll-Kraay approach. - Macroprudential categories analyzed: - Borrower-based: DSTI limits and LTV instruments. - Bank Liquidity: reserve requirements (RR), liquidity requirements, limits to loan-deposit ratio (LTD). - Bank FX exposure: limits on foreign currency lending (LFC), limits on gross open FX positions (LFX), reserve requirements on foreign currency assets (RR_FCD). - Robustness: - Results hold using time fixed effects and using inflation-adjusted percent changes in debt-to-GDP as dependent variables. - Macroprudential Tightening and Growth-at-Risk (equation (7)): - y_{i,t,h}^{τ} regression includes SSM_17_{i,t}, a discrete variable equal to net number of macroprudential tightenings in quarter t (tightenings minus loosenings). - Quantile coefficient β_{4,h}^{τ} indicates direct impact of one additional net tightening action on the left-tail of future growth distribution (values reported in Chapter Figure 2.8 panels referenced in source). - Regime dummy Θ_{i,t}: equals 1 if sum of SSM_17_{i,t} in past 4 quarters is positive; else 0. - Augmented regime specification given in equation (8) with ∆NNNS_8q (8-quarter change in nonfinancial sector debt-to-GDP) included. - Estimation: panel quantile models per Machado and Santos Silva (2019); standard errors bootstrapped. - Sample: 19 advanced economies and 10 emerging market economies; period 1996:Q1 to 2020:Q3. - Note on SSM_17 discrete variable example: if a country tightened LTV twice and loosened RR once in quarter t, SSM_17_{i,t} = 1. ### Literature Review on Leverage and Financial Vulnerability (Online Annex Box 2.1) - Key themes from cited literature: - High levels or rapid increases in nonfinancial sector leverage are predictors of downside risks to growth and financial stress. - Distinction between beneficial credit deepening and disruptive private credit booms (Verner 2019). - Most instances of financial distress are preceded by credit booms (Schularick and Taylor 2012). - High leverage associated with dampened business cycle volatility but more pronounced crashes (Jordà and others 2016). - High corporate leverage associated with weaker post-crisis recovery due to debt overhang (Kalemli-Ozcan and others 2020). - Household debt accumulation particularly related to financial stability risk and boom-bust dynamics (Mian, Sufi, and Verner 2017). - Adrian and others (forthcoming): loose financial conditions plus rapid aggregate credit growth reduces near-term downside risk but increases medium-run downside risk from a GaR perspective. - Additional note: household debt tends to be concentrated in low-wealth quantiles, amplifying COVID-19 shock effects on household sector vulnerabilities (Institute of International Finance, 2020). ### System Approach: PVAR (Online Annex Box 2.2) - Purpose: estimate relationship between economic activity, household leverage, non-financial corporate leverage, and FCI in a single system treating all variables as endogenous. - Model features: - Panel Vector Autoregressive (PVAR) with recursive identification and dynamic time dummies to separate national from global leverage. - Impulse response functions relate to mean responses (contrast with GaR low-quantile analysis). - Key PVAR result reported (supporting main analysis): - An unexpected loosening of financial conditions by one standard deviation leads to an increase in household leverage by about 0.2 percentage points after 4 years. - The PVAR suggests the increase in household leverage is about half the size of the increase in non-financial corporate leverage (figure referenced in source). *Source: ch2onlineannex - EXECUTIVE SUMMARY (Online annexes to "Loose Financial Conditions, Rising Leverage, and Risks to Macro-Financial Stability")* ### 0.4 percentage points. ### ch2onlineannex - 0.4 percentage points. ### Findings on Financial Conditions and Growth - PVAR results suggest economic activity expands after a loosening in financial conditions, but contracts after an increase in leverage. - A one standard-deviation loosening in FCI leads to an increase in real GDP by about 0.2 to 0.3 percent in the 4 years after the shock, with most of the boost frontloaded in the first year. - The positive effect of looser financial conditions on output tends to fade after about 6 quarters. ### Impact of Leverage on Economic Activity - An unexpected increase in household leverage by one standard deviation leads, on average, to a fall in real GDP of about 0.2 percent after 4 years. - The drag on GDP is immediate and continues to build across all horizons. - These results contrast with previous findings (Mian, Sufi and Verner, 2017) that suggest a short-run boost to GDP following a shock to household leverage. - The difference is partly explained by the inclusion of quarter dummies, which filter out global leverage dynamics. Excluding the dummies leads to a marginally significant boost to output in the first year after the shock. - A shock to non-financial corporate leverage reduces GDP by about 0.1 percent within the first 4 years after the shock. - The impact of non-financial corporate leverage on output is smaller compared to household leverage. - Overall, estimating the relationship between financial conditions, leverage and economic activity in a system fully supports the results obtained through other methods in the chapter. ### PVAR Impulse Response Evidence (as described) - Online Annex Box Figure 2.2.1: Looser financial conditions increase household debt and have an even larger impact on non-financial corporate debt. - Panel descriptions: responses shown to a one-standard-deviation loosening in financial conditions; dotted lines show +/- 2 standard deviation error bands. - Online Annex Box Figure 2.2.2: - Panel 1: Impulse response of real GDP following a one-standard-deviation loosening in financial conditions (Percent, quarters). - Panel 2: Impulse response of real GDP following a one-standard-deviation increase in household leverage (Percent, quarters). - Panel 3: Impulse response of real GDP following a one-standard-deviation increase in non-financial corporate leverage (Percent, quarters). - Note: Dotted lines show +/- 2 standard deviation error bands. ### Macroprudential Policy: Transmission Channels and Implications - Purpose: Macroprudential policy—aimed at mitigating the materialization of systemic risk to achieve financial stability (ESRB 2014)—can tame leverage buildups and reduce risks to macrofinancial stability. - Types of tools and their roles: - Borrower-based tools (LTV and DSTI) constrain borrowing eligibility from the demand side and provide affordability buffers to improve resiliency of banks. - Supply-based tools include credit growth limits, capital- (capital adequacy and loan loss provisions), and liquidity-oriented tools (liquidity coverage, net stable funding, and loan-to-deposit ratios). - Sector-based tools include FX exposure limits and sector-specific loan restrictions. - Broad-based tools include countercyclical capital buffers to lean against the buildup phase of the credit cycle. - Effects: - These tools can help tame leverage buildups and financial imbalances and reduce systemic risks. - They may help moderate growth volatility directly (e.g., by reducing default risk) or indirectly by dampening the impact of financial conditions on growth. - Policy cautions and implementation notes: - Policymakers should be mindful of leakages, such as cross-border circumvention (e.g., increase in foreign provision of financing to domestic corporates) and domestic spillover effects (e.g., credit demand shifts from banks to nonbank financial institutions). - Communications with timely announcements for implementation have been shown to enhance policy effectiveness. *Source: ch2onlineannex - 0.4 percentage points.* --- _Source: https://www.imf.org/-/media/files/publications/gfsr/2021/april/english/ch2onlineannex.pdf_