## wpiea2020123-print-pdf ## Source details **Canonical URL:** [wpiea2020123-print-pdf](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020123-print-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020123-print-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020123-print-pdf.pdf.json) --- ### I. Introduction and research question; methodological approach - Focus: assess costs and benefits of using different policy tools (macroprudential, monetary, foreign-exchange intervention, capital flow measures) to dampen buildup of financial vulnerabilities from changes in domestic and global financial conditions. - Key empirical innovation: - Estimate policy effects on the entire distributions of future real GDP growth and inflation. - Evaluate net benefits using quadratic loss functions and alternative (including asymmetric) loss specifications. - Data and sample: quarterly data for financial conditions, economic growth and policy indicators over the period 1990 to 2016. - Identification: exploit policy surprises constructed as deviations from estimated policy rules. - Main steps: - Step 1: fixed-effects quantile regressions (FE-QR) for horizons h=1...H and quantiles q=0.05...0.95 to estimate conditional quantiles QY_{i,t+h}(q | Z_{i,t}). - Step 2: fit empirical conditional quantiles to a Skewed-Normal distribution (Azzalini 1985) to obtain full probability distributions of future GDP growth. - Step 3: compute expected discounted losses under alternative policy scenarios using period quadratic loss l_t = ω_y (y_t - y^*)^2 + ω_π π_t^2 and compare (L_1 - L_0)/L_0. ### Empirical estimator and inference - Conditional quantile specification: - QY_{i,t+h}(q | Z_{i,t}) = α_{0i}^h(q) + β_{1}^h(q) f_{i,t} + β_{2}^h(q) P_{i,t} + β_{3}^h(q) P_{i,t} × f_{i,t} + x_{i,t} Γ^h - Z ≡ [f_{i,t}, P_{i,t}, x_{i,t}]. - Variables: - Y_{i,t+h} ≡ (y_{i,t+h} - ȳ_{i,t+h}) - (y_{i,t} - ȳ_{i,t}) — detrended log real GDP. - f_{it} is the financial condition index (FCI); higher FCI = looser financial conditions. - P_{it} is the policy variable (assumed symmetric effects for tightening and loosening). - x includes controls: current GDP growth, inflation, credit growth. - Hypothesis testing: - For large h and low q, expect β_{1} negative; policy effectiveness requires β_{3} positive. - Inference: - Standard errors and confidence bands via Hagemann’s (2017) cluster-robust bootstrap (country clusters), bootstrap uses 400 replications; percentile bootstrap for confidence bands. ### Loss-function weighting choices - Weights considered: - Output-only: ω_y = 1, ω_π = 0. - Debertoli and others’ (2018) weights when activity measured as deviation from trend: ω_y = 0.542 and ω_π = 1. - Net benefit of policy P defined as (L_1 - L_0)/L_0 (discounted expected loss differential). --- ### II. Data, policy-shock construction, and key methodological details - Macroeconomic data: real GDP growth and inflation from IMF’s World Economic Outlook database. - Financial conditions: - Financial conditions index based on IMF (2018); FCI constructed from up to 10 financial variables via iterative PCA with missMDA imputation; for most countries 2 to 3 principal components are optimal. - Global FCI proxy: United States FCI used as proxy for global financial conditions. - Policy shocks — identification: - Macroprudential shocks (iMaPP index): categorical indicators I_{mpp} ∈ {-2,-1,0,1,2}; shocks ε̂_{mpp} = I_{mpp} - E_{t-1}[I_{mpp}] from ordered probit with country fixed effects; controls include credit-to-GDP gap (cgap) and house price gap (hgap) using Hamilton (2018) detrending. - Monetary shocks: residual of country-level Taylor-type rule; policy rate from IFS except shadow short rates for euro area, Japan, United Kingdom, United States (Krippner 2015); residuals standardized country-by-country (unit shock = one standard deviation). - FXI shocks: FX interventions measured as central bank net foreign assets change (percent of GDP) where available or proxy; FXI shock = residual from OLS rule per country with covariates including egap and σ(e). - CFM shocks: Baba and others’ index from AREAER; ordered probit with country fixed effects; shocks = actual minus predicted categorical indicator. Key variable-definition highlights (excerpt): - I_iit_mpm_A / I_iit_mpm_B / I_iit_mpm_F: categorical macroprudential indicators (0, ±1, ±2). - ccm_iit: credit-to-GDP gap (BIS data). - hccm_iit: house price gap (BIS data). - FXI_iit: foreign exchange interventions as percent of GDP (actual or proxy). - ε̂_iit_r: monetary policy shock (standardized Taylor-rule residual). --- ### III. Main empirical findings — policy effects on the intertemporal tradeoff - No-policy benchmark: - Looser domestic financial conditions today → greater downside risks to growth in the medium term (confirms Adrian and others 2018). - For inflation, no apparent intertemporal tradeoff. - Macroprudential policy: - Tightening macroprudential measures is associated with large loss reductions on average. - Borrower-based tools (caps on LTV and DSTI) particularly effective in reducing losses. - Financial-institutions-based tools (capital and liquidity tools) effective at building resilience even when vulnerabilities are modest. - Tightening borrower-based tools when vulnerabilities (credit aggregates or asset prices) are high is more decisively associated with loss reductions than when vulnerabilities are low. - Tightening mitigates the tradeoff for output growth: tail-risk deterioration due to FCI loosening is less marked in the medium term and this mitigation appears to come at little short-term cost (tightening does not offset the stimulating effects of easier financial conditions on output in the short term). Many estimates are not statistically significant; macroprudential tightening itself has a contractionary effect on output growth when examined as impulse responses, but magnitude not large. - Monetary policy: - Tightening monetary policy does not improve the trade-off and on net is associated with higher losses over the policy horizon (short-term economic costs outweigh stability benefits in this framework). - Monetary tightening tends to worsen tail risk statistically significantly in the medium term; effects on inflation are ambiguous. - Global financial conditions and cross-border tools: - For an easing of global financial conditions, tightening macroprudential policies is the most effective response. - Leaning with tighter monetary policy or FX intervention yields very small (but positive) effects over the policy horizon. - Tightening CFMs to counter loose global financial conditions brings only a weak reduction in the loss function. - FX purchases and tightening CFMs entail very small effects on distribution of output, never statistically different than zero on average. --- ### IV. Quantitative loss-function results (selected exact estimates) - Domestic FCI (Domestic Shock) — reported coefficients (effect on loss function; negative = loss reduction): - MPM All: -0.089***; -0.085***; -0.083*** - MPM Borrower-Based: -0.100***; -0.068***; -0.065*** - MPM FI-Based: -0.053**; -0.036**; -0.035** - MP: 0.121***; 0.115***; 0.111*** - HF MP: -0.011; -0.011; -0.011 - External FCI (External Shock) — reported coefficients: - MPM All: -0.112***; -0.107***; -0.104*** - MPM Borrower-Based: -0.107***; -0.101***; -0.096*** - MPM FI-Based: -0.068***; -0.067***; -0.065*** - MP: 0.038*; 0.036*; 0.036* - FXI: -0.022; -0.021; -0.021 - CFM: -0.039; -0.034; -0.030 - HF MP: -0.025; -0.023; -0.022 - Vulnerability-split (Table 4) — selected contrasts (Domestic/External variants; Low vs High vulnerabilities): - MPM Borrower-Based (Low Credit): -0.033; -0.032; -0.031 - MPM Borrower-Based (High Credit): -0.083***; -0.078***; -0.075*** - MPM FI-Based (Low Credit): -0.076**; -0.072**; -0.070** - MPM FI-Based (High Credit): -0.028; -0.027; -0.026 - MP (Low Credit): 0.137***; 0.132***; 0.129*** - MP (High Credit): 0.126***; 0.120***; 0.115*** - Coordinated policy changes (Table 5) — MPM tightening combined with looser monetary policy: - MPM All (coordinated): -0.162; -0.157; -0.153 - MPM Borrower-Based (coordinated): -0.189; -0.180; -0.173 - MPM FI-Based (coordinated): -0.140; -0.136; -0.133 - If MPM tightens while MP tightens, benefits largely erased. - Alternative loss functions (Table 6) — asymmetric and linear-quadratic: - MPM All: -0.080***; -0.076***; -0.087***; -0.100***; -0.095***; -0.109*** - MPM Borrower-Based: -0.092***; -0.087***; -0.098***; -0.097***; -0.089***; -0.100*** - MPM FI-Based: -0.046**; -0.045**; -0.053*; -0.060**; -0.058**; -0.067*** - MP: 0.134***; 0.126***; 0.124***; 0.046**; 0.044**; 0.040* - FXI (external): -0.029; -0.027*; -0.024 - CFM (external): -0.040; -0.033; -0.041 - Time profile (cumulated loss differential ΔL for H=1,...,14 quarters; ωy=1, ωπ=0): - Macroprudential policy: reduces losses uniformly, benefits peak around 10 quarters ahead; benefits start accruing in short term and realized in medium term. - Monetary policy tightening: may reduce losses initially but induces increases in losses from about 5 quarters out. - Borrower-based macroprudential has more persistent beneficial effect than financial-institution-based; FXI, CFM, FI-based MPM: about half reduction up to 10 quarters reversed after 14 quarters. --- ### V. Policy interactions, conditionality, and country-group heterogeneity - Interactions between macroprudential and monetary policy: - Tightening macroprudential accompanied by looser monetary policy produces larger reduction in central bank loss function: combined effect reported as around -16 percent (compared to about -9 percent for macroprudential tightening alone). - Tightening both MPM and MP tends to erase benefits of macroprudential leaning. - Policy intuition: accommodative monetary policy cushions adverse output effects of macroprudential tightening, increasing net welfare gains. - Conditional effectiveness by vulnerabilities: - When credit or asset prices are high: - Borrower-based MPM tightening more effective at reducing GDP tail risks and losses (e.g., borrower-based loss reductions between 9.6 and 10.7 percent depending on weights in some tables). - Financial-institution-based tools less effective. - When credit is low: - Financial-institution-based tools can deliver larger benefits while avoiding efficiency costs of borrower-based tools. - Advanced vs emerging economies: - Group-specific results broadly support benchmark findings: macroprudential policies reduce losses more than other policies. - Financial-institution-based MPM and CFMs perform relatively better in emerging markets than in advanced economies. - FX purchases do not reduce losses in either group. --- ### VI. Robustness, extensions, and caveats - Robustness checks: - Alternative estimators accounting for common factors and unobserved heterogeneity (Koenker’s penalized fixed effects; Ando and Bai frequentist panel quantile with five factors) produce qualitatively similar results: macroprudential policies reduce losses; borrower-based measures particularly effective; monetary policy does not reduce losses. - High-frequency monetary policy surprises (sample of 27 countries): qualitative results unchanged; leaning with monetary policy associated with very small reduction in loss function: 1 percent for loose domestic FCI and at most 2.5 percent for loose global FCI. For sample with HF MP data, MPM leaning reduces losses between 3.5 (global FCI) and 5 percent (domestic FCI). - Asymmetric preferences: - Linex loss (a = -1) and fourth-order Taylor expansion show expected linex loss depends on conditional variance, skewness, and kurtosis; empirical results with asymmetric preferences confirm standard quadratic-loss findings. - Caveats and further research: - FXI and CFMs may produce benefits in certain circumstances (e.g., mitigate adverse shocks) not captured by leaning-against-the-wind design. - Cross-country differences (financial system structure, international financial linkages) warrant further exploration. - Results reflect historical correlations; drawing definitive policy prescriptions subject to Lucas (1976) critique on changing policy regimes. --- ### VII. Appendix A — selected data summary and policy-shock size statistics (exact values) - Detrended output: Observations 2,876; Average -0.03; Standard Deviation 1.11; Median 0.02; Minimum -8.08; Maximum 7.21 - Quarterly inflation: Observations 2,876; Average 0.66; Standard Deviation 0.87; Median 0.55; Minimum -3.07; Maximum 9.84 - Financial conditions index: Observations 2,876; Average -0.07; Standard Deviation 0.94; Median -0.20; Minimum -2.74; Maximum 6.11 - Real GDP growth*: Observations 2,876; Average 2.97; Standard Deviation 4.68; Median 2.86; Minimum -22.29; Maximum 26.89 - YoY Inflation: Observations 2,876; Average 2.72; Standard Deviation 2.51; Median 2.23; Minimum -6.32; Maximum 16.37 - Growth in credit*: Observations 2,876; Average 0.64; Standard Deviation 2.46; Median 0.50; Minimum -7.60; Maximum 10.60 - Credit to GDP (logs): Observations 2,876; Average 4.78; Standard Deviation 0.58; Median 4.93; Minimum 3.00; Maximum 5.99 - House prices (logs): Observations 2,876; Average 5.03; Standard Deviation 0.67; Median 4.90; Minimum 3.16; Maximum 8.45 - Output gap: Observations 2,808; Average 0.00; Standard Deviation 0.04; Median 0.00; Minimum -0.20; Maximum 0.33 - MPM All: Observations 2,876; Average 0.01; Standard Deviation 0.48; Median -0.05; Minimum -2.51; Maximum 2.24 - MPM Borrower-Based: Observations 2,876; Average 0.00; Standard Deviation 0.23; Median -0.01; Minimum -2.07; Maximum 1.99 - MPM FI-Based: Observations 2,876; Average 0.00; Standard Deviation 0.40; Median -0.03; Minimum -2.58; Maximum 2.17 - MP: Observations 2,852; Average -0.03; Standard Deviation 0.95; Median -0.05; Minimum -5.74; Maximum 9.05 - FXI: Observations 2,259; Average 0.00; Standard Deviation 0.01; Median 0.00; Minimum -0.09; Maximum 0.14 - CFM: Observations 2,259; Average -0.02; Standard Deviation 0.52; Median 0.03; Minimum -2.00; Maximum 2.30 - Note: * Winsorized at the top and bottom 0.5 percent. - Policy shock size by country (one standard deviation) — selected entries (exact): - China 42: MPM All 1.23; MPM Borrower-Based 0.55; MPM FI-Based 0.41; MP 1.04; FXI 0.05; CFM 0.96 - India 27: MPM All 0.97; MPM Borrower-Based 0.34; MPM FI-Based 0.43; MP 1.03; FXI 0.14; CFM 1.06 - United States 10: MPM All 0.32; MPM Borrower-Based 0.10; MPM FI-Based 0.17; MP 1.05; FXI 0.06; CFM 0.03 - Brazil 5: MPM All 0.96; MPM Borrower-Based 0.13; MPM FI-Based 0.47; MP 1.01; FXI 0.17; CFM 0.95 - Turkey 2: MPM All 1.00; MPM Borrower-Based 0.29; MPM FI-Based 0.50; MP 1.03; FXI 0.19; CFM 0.64 --- *Italic: Source: wpiea2020123-print-pdf (excerpt: Introduction, methodology, Sections II–IV, Appendix A and Appendix B, and selected tables and findings).* ### References ............................................................................................................. ### wpiea2020123-print-pdf - References ............................................................................................................. ### I. Introduction and research question - Focus: assess costs and benefits of using different policy tools (macroprudential, monetary, foreign-exchange intervention, capital flow measures) to dampen buildup of financial vulnerabilities from changes in domestic and global financial conditions. - Key empirical innovation: estimate policy effects on the entire distributions of future real GDP growth and inflation, then evaluate net benefits using quadratic loss functions. - Data and sample: quarterly data for financial conditions, economic growth and policy indicators over the period 1990 to 2016. - Identification: exploit policy surprises constructed as deviations from estimated policy rules. ### Methodological approach - Step 1: use quantile regressions to estimate how policies affect the trade-off between current financial conditions and the expected distribution of future GDP growth and inflation. - Step 2: fit empirical conditional quantiles to a known distribution function (skewed-Normal) to obtain full probability distributions of future GDP growth. - Step 3: feed the resulting distributions into a standard loss function and compare loss function values with and without policy actions. - Loss-function evaluation includes quadratic loss and alternative specifications (including an asymmetric loss that penalizes downward deviations more than upward ones). ### Relation to existing literature and conceptual framing - Builds on Adrian, Boyarchenko, and Giannone (2019) and others documenting that easy financial conditions boost short-term output but increase medium-term downside risks to growth. - Addresses limitations of prior approaches: - Moves beyond binary “crisis vs no crisis” frameworks to evaluate the whole future distribution of outcomes. - Directly measures effects on ultimate objectives (tail risks to output) rather than only intermediate targets (e.g., credit growth). - Avoids ad-hoc transmission assumptions for macroprudential tools by letting empirical estimates capture both credit-dynamics and resilience effects. ### Main empirical findings - Macroprudential policy - Tightening of macroprudential measures is associated with large loss reductions on average. - Tightening of borrower-based tools (caps on LTV and DSTI) is particularly effective in reducing losses. - Financial-institutions-based tools (capital and liquidity tools) are effective at building resilience even when vulnerabilities are modest. - Tightening borrower-based macroprudential policies when vulnerabilities (as measured by credit aggregates or asset prices) are high is more decisively associated with a reduction in losses than when vulnerabilities are low. - Monetary policy - Tightening monetary policy does not improve the trade-off between looser current financial conditions and future tail risks to GDP growth. - On net, tightening monetary policy is associated with higher losses over the policy horizon, indicating the short-term economic costs outweigh stability benefits in this framework. - Global financial conditions and policy responses - For an easing of global financial conditions, tightening macroprudential policies is the most effective response. - Leaning with tighter monetary policy or foreign-exchange intervention yields very small (but positive) effects over the policy horizon. - Tightening capital flow management measures (CFMs) to counter loose global financial conditions brings only a weak reduction in the loss function. ### Interactions and policy combinations - Joint use of macroprudential and monetary policy - A tightening of macroprudential policy accompanied by looser monetary policy produces a larger reduction in the central bank’s loss function than macroprudential tightening alone. - This supports the idea that accommodative monetary policy can cushion adverse output effects of macroprudential tightening, increasing overall welfare gains. - Robustness: results broadly robust to alternative loss-function specifications, including asymmetric loss functions. ### Contributions and advances - Extends macroprudential evaluation from intermediate outcomes to ultimate objectives—containment of tail risks to real economic outcomes. - Replaces binary crisis frameworks with distributional assessment across all states of the world. - Empirical strategy captures both credit-dynamics and resilience channels of macroprudential policy without imposing stylized transmission mechanisms. _Italic: Source: wpiea2020123-print-pdf - References (content excerpt including Introduction, methodology, literature context, and summarized results)._ ### Section III presents a first comparison of results comparing no-policy outcomes with ### wpiea2020123-print-pdf - Section III presents a first comparison of results comparing no-policy outcomes with ### Empirical approach — Quantile regressions - Objective: - Investigate whether and how different types of policies can ameliorate the intertemporal tradeoff: easy financial conditions boost short-term output but increase medium-term downside risks (Adrian and others, 2018). - Estimator and specification: - Use fixed-effects quantile regressions (FE-QR) for each horizon h=1...H and quantiles q=0.05...0.95, applying the FE-QR estimator described in Kato et al. (2012). - Conditional quantile specification: - QY_{i,t+h}(q | Z_{i,t}) = α_{0i}^h(q) + β_{1}^h(q) f_{i,t} + β_{2}^h(q) P_{i,t} + β_{3}^h(q) P_{i,t} × f_{i,t} + x_{i,t} Γ^h - Z ≡ [f_{i,t}, P_{i,t}, x_{i,t}]. - Variables: - Y_{i,t+h} ≡ (y_{i,t+h} - ȳ_{i,t+h}) - (y_{i,t} - ȳ_{i,t}) — detrended log real GDP (output gap as deviation from a linear time trend); analogous specification for inflation. - f_{it} is the financial condition index (FCI); higher FCI represents looser financial conditions. - P_{it} is the policy variable; specification assumes symmetric effects for tightening and loosening. - x includes controls: current GDP growth, inflation, credit growth. - Hypothesis testing and inference: - For large h and low q (e.g., q=0.1), expect β_{1} negative (looser conditions → fatter left tail). Policy effectiveness requires β_{3} positive at same horizon/quantiles. - Account for serial correlation by calculating standard errors and confidence bands with Hagemann’s (2017) cluster-robust bootstrap (country clusters). ### Integrating distributions and computing loss functions - Rationale: - Policies can change left and right tails and shift entire conditional distributions; aggregate effects evaluated via a policymaker loss function. - Period quadratic loss function used (New Keynesian standard): - l_t = ω_y (y_t - y^*)^2 + ω_π π_t^2 (quadratic in output deviation and inflation). - Discounted expected losses and net benefit: - Compare expected discounted value of future losses under no-policy and policy use; net benefit of policy P is (L_1 - L_0)/L_0. - Weights considered: - Output-only: ω_y = 1, ω_π = 0. - Debertoli and others’ (2018) optimal weights when activity is measured as deviation from trend output: ω_y = 0.542 and ω_π = 1. - Estimation of distribution moments: - Fit empirical conditional quantiles (q = 5,...,95) from FE-QR to a Skewed-Normal distribution (Azzalini 1985) via a minimum distance estimator for computational ease. - Parameter vector θ_h estimated as θ̂_h = argmin_{θ ∈ Θ} ∑_q (EQF_h(q) - SkewNQF_θ(q))^2 for each h=1...H. - Skewed-Normal chosen over Skewed-t to reduce the number of parameters (location, scale, skewness); excess kurtosis is a function of skewness. - Inference: - Confidence intervals for (L_1 - L_0)/L_0 calculated with a country-level cluster bootstrap of the entire sample (Y_1...Y_H and all left-hand-side variables). - Bootstrap uses 400 replications; confidence bands calculated with the percentile bootstrap. ### Data — Financial and policy measures - Macroeconomic data: - Real GDP growth and inflation from the IMF’s World Economic Outlook database. - Financial conditions: - Financial conditions index based on IMF (2018); focus on the price of risk; other vulnerabilities (leverage, credit growth) considered explicitly elsewhere. - Policy shocks — identification approach: - For each policy instrument (monetary, macroprudential, foreign exchange interventions (FXI), capital flow management measures (CFMs)), estimate country-specific reaction functions conditioning on economic and financial variables; use residuals as policy shocks to remove endogenous responses. - Summary statistics of each policy shock shown in Tables 1 and 2 (not reproduced here). - Macroprudential policy shocks: - Indicator from iMaPP database (Alam and others, 2019); three categories: overall, borrower-based, financial-institution-based. - Ordered probit with country fixed effects: latent m* used to generate categorical indicator I_{mpp} ∈ {-2,-1,0,1,2} (net more than one loosening, one loosening, no change, one tightening, more than two tightenings in quarter t). - Controls include credit-to-GDP gap (cgap) and house price gap (hgap) using Hamilton (2018) detrending with eight quarter lags. - Policy shock ε̂_{mpp} = I_{mpp} - E_{t-1}[I_{mpp}] = I_{mpp} - ∑_{k=-2}^{2} m̂_k(x_{t-1}) · k. - Monetary policy shocks: - Policy rate from IFS, except for euro area, Japan, United Kingdom, United States where Krippner’s (2015) shadow short rates used. - Monetary shock: residual of country-level Taylor-type rule (regression (7)) with quarter-on-quarter changes ∆r, ∆y, ∆p, ∆neer; includes E_t ∆y_{t+12} and E_t ∆p_{t+12} from Consensus Forecasts. - Residual standardized country-by-country; a unit shock = one standard deviation in each country. - Foreign exchange interventions (FXIs): - FXIs measured as central bank net foreign assets change (adjusted for valuation and interest) when intervention data limited; FXI in percent of GDP (positive for net purchases of foreign currency). - FXI shock = residual from OLS linear rule per country (equation (8)); covariates include those in Forbes and Klein (2015) first column, egap (exchange rate deviation using Hamilton 2018) and σ(e) (quarterly NEER volatility from daily data). - Capital flow management measures (CFMs): - Use Baba and others’ changes index (based on AREAER) capturing tightening/easing of capital controls. - Ordered probit with country fixed effects for categorical CFM indicator I_{CFM} ∈ {-2,-1,0,1,2} (more than one tightening/loosening, one tightening/loosening, no change). - Controls same as FXI specification. - CFM shock = actual minus predicted categorical indicator. ### Results — Effects of policies on the intertemporal tradeoff - Strategy: - Examine estimated effects of each policy in response to easy financial conditions on conditional quantiles of future detrended GDP growth and inflation: β_{1}^h(q) + β_{3}^h(q) · P. - Present changes in the 10th percentile after a one standard deviation loosening in financial conditions under no policy change (blue line) and under a tightened policy (red line). - No-policy benchmark (confirming stylized facts): - Findings confirm Adrian and others (2018): looser domestic financial conditions today → greater downside risks to growth in the medium term (intertemporal tradeoff). - For inflation, there is no apparent intertemporal tradeoff. - Macroprudential policy tightening: - Tightening macroprudential policy mitigates (but does not eliminate) the tradeoff for output growth: tail-risk deterioration due to FCI loosening is less marked in the medium term. - This mitigation appears to come at little short-term cost: tightening does not offset the stimulating effects of easier financial conditions on output in the short term. - However: - These estimates are mostly not statistically significant. - The policy tightening itself has a contractionary effect on output growth (and prices to some extent) when examining impulse responses to a macroprudential tightening shock (β_{2}^h(q) + β_{3}^h(q) · f); effects reported in Figure B.2 in Appendix B. - The magnitude of contractionary effects does not appear large, consistent with the literature. - Monetary policy tightening: - Monetary tightening tends to worsen the underlying tradeoff: further worsening of tail risk statistically significantly in the medium term. - Policy effects on inflation are ambiguous. - Distributional considerations: - Results presented for the left tail (10th percentile); complementary results for median growth and other quantiles referenced (see Figure B.1 for median), indicating analysis extends beyond tails. *Italic: Source — wpiea2020123-print-pdf (excerpt: Sections II–IV).* ### Appendix B), we find that the intertemporal tradeoff documented by Adrian and others ### wpiea2020123-print-pdf - Appendix B), we find that the intertemporal tradeoff documented by Adrian and others ### Intertemporal tradeoff and quantile evidence - The intertemporal tradeoff documented by Adrian and others (2018) is absent for median output: "the effect of easing financial conditions on median output is positive through the entire horizon." - Tighter macroprudential policy does not reduce that positive effect of easing financial conditions on median economic activity. - Tighter monetary policy appears, "if anything, to worsen the median outlook," consistent with the view that macroprudential policy is effective in dampening downside risks to GDP growth whereas monetary policy is a blunt tool. - For most countries in the sample, inflation expectations are well-anchored, which could explain "the finding of no intertemporal tradeoff for inflation." ### Net benefits of policies (loss-function results) - Benchmark loss function: quadratic loss function (1), measuring costs and benefits using output and price volatility. - Domestic financial conditions (Table 3, left panel): - Tightening macroprudential policies when domestic financial conditions loosen reduces losses by about 9 percent (quadratic policymaker preferences). - Monetary policy tightening "seems to increase losses by about 12 percent." - Borrower-based macroprudential measures: estimated net benefits between -6.5 and -10 percent. - Financial institution-based macroprudential measures: at most -5.3 percent reduction. - Statistical precision caveat: "our estimates are not sufficiently precise to establish that these two types of policies have different effects at the usual levels of statistical significance." - Mechanism: contrast in net benefits mainly from opposite effects on output volatility; "all policies have little effects on price volatility." - Tighter macroprudential policy: distribution of future output gaps shrinks; increase in its 10th percentile. - Leaning with monetary policy: increases output volatility. ### Global financial shocks and cross-border tools - For global financial conditions (proxied with U.S. financial conditions): - Effect of macroprudential policy tightening conditional on easing global financial conditions is "by and large unchanged" relative to domestic case. - Monetary policy now "appears to reduce, rather than increase, downside tail risks for GDP at the 10th percentile" (see Figure B.3 in Appendix B). - FX interventions (buy foreign currency) and capital flow management (CFM): - Intervening in foreign exchange markets to buy foreign currency does not affect the intertemporal tradeoff; the effect of looser global financial conditions on tail risks to GDP does not change. - Adopting new measures impeding capital inflows shows similar non-effect. - Loss reductions from tightening macroprudential measures remain substantial (Table 3, right panel). - Borrower-based macroprudential loss reductions range between 9.6 and 10.7 percent (depending on loss-function weights). - Financial institution-based tightening brings smaller gains (at most a 6.8 percent reduction). - FX purchases and tightening CFMs "entail very small effects on the distribution of output, which are never statistically different than zero." ### Time profile of cumulated loss changes (horizon dynamics) - Computed loss differential ΔL=(L1(Θ)-L0(Θ))/L0(Θ) for H=1,...,14 quarters; exercise uses period loss with ωy=1 and ωπ=0 for global financial conditions. - Three main patterns emerge (Figure 3): - Macroprudential policy reduces losses in a uniform way until benefits peak around 10 quarters ahead; benefits start accruing in the short term and are realized fully in the medium term. - Monetary policy tightening appears to reduce losses initially, but "starts to induce increases in losses from about 5 quarters out," suggesting initial benefits are eroded by increased output volatility over the medium term. - Borrower-based macroprudential policy has a more persistent beneficial effect than financial-institution-based macroprudential policy. - For FXI, CFM, and financial-institution-based macroprudential policy, about half the reduction in the loss function up to 10 quarters ahead is reversed after 14 quarters — gains are largely temporary, whereas borrower-based tool benefits are longer lasting. ### Effects conditional on financial vulnerabilities - Vulnerability measures used: credit-to-GDP ratio (leverage) and BIS house price index (both in logs). - Quantile results: - Tighter macroprudential policies more strongly associated with reductions in GDP tail risks (10th percentile) when credit levels or house prices are high. - Using monetary policy to lean against the wind is not associated with a reduction in GDP at risk, "irrespective of the level of credit." - Loss-function results (Table 4) — nuanced effectiveness: - Tightening borrower-based macroprudential policies is more strongly associated with reductions in losses when credit and asset prices are high (high vulnerabilities) than when vulnerabilities are low. - Tightening financial-institutions-based macroprudential policies appears to have larger benefits when credit or asset prices are low. - Interpretation: - When credit or asset prices are high, financial-institutions-based tools (e.g., capital requirements) that affect credit supply indirectly are insufficient; borrower-based tools that directly increase borrower resilience are required. - When credit is low, measures that reduce further credit build-up by affecting financial institutions can be effective while avoiding efficiency costs of borrower-based tools. - Overall: evidence is "suggestive of the dual role of macroprudential policy in both leaning against the wind and building resilience." - Monetary policy effects appear independent of vulnerabilities: quadratic loss functions increase by similar amounts when monetary policy is tightened regardless of credit levels; differences between low and high vulnerability cases are not statistically different than zero. ### Interaction between monetary and macroprudential policies - Two schools of thought: - Opposite-direction use: tighten macroprudential while loosening monetary policy to offset negative effects on credit and output (e.g., Collard and others 2017). - Same-direction use: both policies move the same way to reinforce reduction in credit provision (Stein 2013), but may have adverse output effects. - Empirical tests via expanded quantile regressions: - Combining macroprudential tightening with looser monetary policy yields larger reduction in central bank loss function: -16 percent (compared to about -9 percent for macroprudential tightening alone). - Tightening both macroprudential and monetary policies "basically erases the benefits of leaning against the wind from macroprudential policy alone." - Policy intuition: - Macroprudential tightening stabilizes the economy; monetary tightening is counterproductive and costly, especially in the medium term. - Combining tight macroprudential with easy monetary policy offsets undesirable output effects of macroprudential tightening and achieves larger net benefits. - Conclusion: "when responding to a loosening of financial conditions, it is useful to move macroprudential and monetary policy in opposite directions." *Source: wpiea2020123-print-pdf - Appendix B), we find that the intertemporal tradeoff documented by Adrian and others* ### introduction of a level (linear) term of output in the policymaker’s preferences can be ### introduction of a level (linear) term of output in the policymaker’s preferences can be ### Loss-function extensions: linear (level) term - A level (linear) term of output in the policymaker’s preferences can be micro-founded with a New Keynesian model with a distorted steady state (Benigno and Woodford 2005). - The loss function becomes: - 26 () () 2 2 . tythtthtt yy yy π ωωπ ++  = −− −+  l - Including a linear term of output yields the same ordering of loss reductions among policies as pure quadratic losses (Table 6), because the linear-quadratic loss can be rewritten as a pure quadratic loss involving only a shift of the loss function (Benigno and Woodford 2005). ### Asymmetric preferences: linex loss and higher moments - Policymakers may exhibit loss aversion; a flexible loss that accommodates asymmetry is the linex loss (Varian 1975; Zellner 1986). - Focusing only on output (disregarding inflation), the period loss is: - ℓ ii = �e e a a y y � t t+ h − a a y y � t t+ h −1� a a 2 , where y y � i i+ h ≡ ( y y i i+ h − y ȳ ) (11) - The linex loss includes the quadratic loss (equation (4)) as a special case when a→0 and covers loss aversion when a<0. For simplicity, the authors use a = -1 and assume potential output grows at a constant rate. - A fourth-order Taylor expansion is used to express the expected linex loss as a function of the moments of the conditional distribution of future GDP growth: - 퐸퐸 ii (ℓ ii+ h ) = 푙푙(0) + 푙푙′(0)퐸퐸 ii (푦푦 � ii+ h ) + 1 2 푙푙′′(0)퐸퐸 ii (푦푦 � ii+ h ) 2 + 1 3! 푙푙′′′(0)퐸퐸 ii (푦푦 � ii+ h ) 3 + 1 4! 푙푙′′′′(0)퐸퐸 ii (푦푦 � ii+ h ) 4 + ⋯ ≅ 1 2 퐸퐸 ii (푦푦 � ii+ h ) 2 − 1 3! 푐푐퐸퐸 ii (푦푦 � ii+ h ) 3 + 1 4! 푐푐 2 퐸퐸 ii (푦푦 � ii+ h ) 4 - The expression shows that, when a is not too small, the linex loss can be written as a function of the conditional variance, skewness, and kurtosis of detrended GDP at each horizon h. A footnote clarifies that when a is small, the linex approximates the quadratic term and preferences are approximately symmetric. ### Empirical results with asymmetric preferences - Results with asymmetric preferences confirm findings from standard quadratic losses (Table 6): - Borrower-based macroprudential policies remain the most effective in reducing the loss functions. - Monetary policy remains ineffective. - FX purchases and CFM have mostly insignificant effects. ### Alternative measures of monetary policy (high-frequency surprises) - Monetary policy shocks are difficult to identify; measuring monetary policy surprises via asset-price responses at announcement has been used (Kuttner 2001; Gürkaynak and others 2005). - For many emerging markets, requisite intraday data are absent or poor quality; robustness checks use high-frequency measures from a variety of sources (see Appendix A). - The authors gathered such data for 27 of the 37 countries in the sample, noting important shortcomings (Appendix). - Qualitative results are unchanged: macroprudential policy preferred to monetary policy by the loss-function metric. - Leaning against the wind with monetary policy is associated with a very small reduction in the loss function: - 1 percent for loose domestic FCI - at most 2.5 percent for loose global FCI - For the sample with monetary policy surprise data, leaning against the wind with macroprudential policy reduces losses between 3.5 (global FCI) and 5 percent (domestic FCI). The authors prefer the benchmark results for comparison because of the larger sample. ### Advanced vs. emerging market economies - Pooling advanced and emerging economies could confound results due to: - Different intertemporal tradeoffs between current financial conditions and future downside risks. - Different prevalences of macroprudential tools (e.g., mortgage-related measures more common in advanced economies; CFMs more common in emerging economies). - Group-specific results (Table 7) broadly support benchmark findings: - Macroprudential policy generally associated with larger reductions in the loss function than other policies. - Financial-institution-based macroprudential policies and, to a smaller extent, CFMs appear to work better in emerging markets than in advanced economies. - FX purchases again do not reduce losses. ### Common factors and unobserved heterogeneity - Global macroeconomic and financial common factors are captured via time fixed effects (two-way error structure) in addition to country fixed effects. - Under fixed N, Kato and others’ (2012) FE-QR estimator is inconsistent with two-way fixed effects; authors estimate country fixed effects as dummies and, assuming time fixed effects constant across quantiles, use Koenker’s (2004) penalized fixed effects estimator. - Results (Table 8 panel A) are qualitatively similar to Section IV: macroprudential policy reduces losses and monetary policy increases them. - To address heterogeneous slopes across countries, the authors use Ando and Bai’s (2019) frequentist approach to panel quantile models with unobserved heterogeneity and common factors. - The estimated model includes quantile-specific unobservable factors (φ) and factor loadings (λ). - The authors use five factors for all policy shocks. - Because Ando and Bai’s approach requires country-by-country estimation of conditional quantiles, the sample is restricted to countries with at least 30 observations, reducing country coverage from 37 to 31 and reducing the number of quantiles from 19 to 9. - Results (Table 8 panel B) are broadly in line with benchmark results: macroprudential policy reduces losses; borrower-based measures reduce them more; monetary policy does not produce similar reductions. ### Conclusions: principal findings and policy-relevant implications - Methodology: novel approach evaluating effects of “leaning against the wind” by assessing impacts of different policies on the entire future probability distributions of growth and inflation and evaluating them using standard and nonstandard loss functions. - Main findings: - Overall, trade-offs favor macroprudential policies in response to changes in domestic and global financial conditions. - When credit is high, tightening borrower-based measures is more advantageous than tightening financial-institution-based macroprudential measures. - When credit is low, financial-institution-based measures perform relatively better. - Tightening macroprudential policies has stronger benefits when accompanied by looser monetary policy. - Tightening monetary policy alone to lean against the wind is generally unfavorable and associated with higher losses. - Foreign-exchange interventions and CFMs produce small reductions in the loss function, considerably smaller than macroprudential policy; effects on future output growth are not statistically significant on average. - Caveats and avenues for further research: - FXI and CFMs may produce benefits in certain circumstances or when used to mitigate adverse shocks rather than to lean against easing financial conditions. - Further work should explore cross-country differences (financial system structure, depth of international financial linkages). - Results are first insights using a novel approach and are not definitive; drawing direct policy prescriptions is subject to Lucas’s (1976) critique that policy effects may depend on how policies are used. - Relation to existing literature: - The paper aligns with macroeconometric policy evaluation traditions (Sims 1980, 1982; Jordà 2005; Angrist and Kuersteiner 2011), maintaining the assumption that regression methods applied to past data can inform policy choices (Kocherlakota 2019). *Source: wpiea2020123-print-pdf - introduction of a level (linear) term of output in the policymaker’s preferences can be* ### REFERENCES ### REFERENCES ### Monetary and macroprudential policy literature - Adam, Klaus, and Michael Woodford. 2018. “Leaning Against Housing Prices as Robustly Optimal Monetary Policy,” NBER Working Paper No. 24629 - Agénor, Pierre-Richard, Koray Alper, and Luiz Pereira da Silva. 2013. “Capital Regulation, Monetary Policy, and Financial Stability,” International Journal of Central Banking 9(3): 198-243. - Agur, Itai. 2019. “Monetary and Macroprudential Policy Coordination Among Multiple Equilibria.” Journal of International Money and Finance 96: 192-209. - Ajello, Andrea, Thomas Laubach, David López-Salido, and Taisuke Nakata. 2019. “Financial Stability and Optimal Interest Rate Policy,” International Journal of Central Banking 15(1): 279-326. - Alpanda, Sami, and Sarah Zubairy. 2017. “Addressing Household Indebtedness: Monetary, Fiscal or Macroprudential Policy?” European Economic Review 92: 47-73. - Alpanda, Sami, Gino Cateau, and Cesaire Meh. 2018. “A policy model to analyze macroprudential regulations and monetary policy.” Canadian Journal of Economics, 51 (3), 828-863 - Angelini, Paolo, Stefano Neri, and Fabio Panetta. 2014. “The Interaction between Capital Requirements and Monetary Policy.” Journal of Money, Credit, and Banking 46 (6): 1073-1112. - Aoki, Kasuke, Gianluca Benigno, and Nobuhiro Kiyotaki, 2016. “Monetary and Financial Policies in Emerging Markets,” unpublished manuscript. - Benigno, Pierpaolo, and Michael Woodford. 2005. “Inflation stabilization and welfare: The case of a distorted steady state.” Journal of the European Economic Association 3, no. 6: 1185-1236. - Collard, Fabrice, Harris Dellas, Behzad Diba, and Olivier Loisel. 2017. “Optimal Monetary and Prudential Policies." American Economic Journal: Macroeconomics 9, No. 1: 40-87. - Debortoli, Davide, Jinill Kim, Jesper Lindé, and Ricardo Nunes. 2018. “Designing a Simple Loss Function for Central Banks: Does a Dual Mandate Make Sense?” The Economic Journal 129, no. 621: 2010-2038. - Jiménez, Gabriel, Steven Ongena, José-Luis Peydró, and Jesús Saurina. 2017. “Macroprudential Policy, Countercyclical Bank Capital Buffers, and Credit Supply: Evidence from the Spanish Dynamic Provisioning Experiments.” Journal of Political Economy 125(6): 2126-2177. - Mendicino, Caterina, Kalin Nikolov, Javier Suarez, and Dominik Supera, 2019. "Bank Capital in the Short and in the Long Run." Journal of Monetary Economics. - Rubio, Margarita, and José A. Carrasco-Gallego. 2014. “Macroprudential and monetary policies: Implications for financial stability and welfare.” Journal of Banking & Finance, vol. 49: 326–336. - Stein, Jeremy C., 2012. “Monetary Policy as Financial Stability Regulation.” Quarterly Journal of Economics 127:57–95. - Stein, Jeremy C., 2013. “Overheating in Credit Markets: Origins, Measurement, and Policy Responses.” Speech at the Restoring Household Financial Stability after the Great Recession: Why Household Balance Sheets Matter research symposium sponsored by the Federal Reserve Bank of St. Louis, St. Louis, Missouri, February 17, 2013. https://www.federalreserve.gov/newsevents/speech/stein20130207a.htm - International Monetary Fund (IMF). 2013. “Key Aspects of Macroprudential Policy.” IMF Policy Paper. Washington, D.C. - ____________________________. 2015. “Monetary Policy and Financial Stability.” IMF Policy Paper. Washington, D.C. - ____________________________. 2017. “Are Countries Losing Control of Domestic Financial Conditions?” in Global Financial Stability Report—Getting the Policy Mix Right, October. Washington, D.C. - ____________________________. 2018. “A Decade after the Global Financial Crisis: Are We Safer?” in Global Financial Stability Report— A Decade after the Global Financial Crisis: Are We Safer? October. Washington, D.C. ### Leaning against the wind, asset bubbles, and financial stability tradeoffs - Filardo, Andrew, and P. Rungcharoenkitkul. 2016. “A Quantitative Case for Leaning Against the Wind.” BIS Working Papers 594, Bank for International Settlements. - Gambacorta, Leonardo, and Federico M. Signoretti. 2014. “Should monetary policy lean against the wind?: An analysis based on a DSGE model with banking.” Journal of Economic Dynamics and Control 43: 146-174. - Gourio, Francois, Anil K. Kashyap, and Jae W. Sim. 2018. “The Tradeoffs in Leaning Against the Wind.” IMF Economic Review 66:70-115. - Schularick, Moritz, Lucas Ter Steege, and Felix Ward, 2020. “Leaning Against the Wind and Crisis Risk.” CEPR Discussion Paper DP14797. - Svensson, Lars EO, 2017. “Cost-benefit analysis of leaning against the wind.” Journal of Monetary Economics 90: 193-213. - Benigno, Pierpaolo, and Michael Woodford. 2005. “Inflation stabilization and welfare: The case of a distorted steady state.” Journal of the European Economic Association 3, no. 6: 1185-1236. - Biljanovska, Nina, Lucyna Gornicka, and Alexandros Vardoulakis. 2019. “Optimal Macroprudential Policy and Asset Price Bubbles.” IMF Working Paper No. 19/184. International Monetary Fund. - Richter, Björn, Moritz Schularick, and Ilhyock Shim. 2019. “The Costs of Macroprudential Policy.” Journal of International Economics 118: 263-282. ### Foreign exchange intervention, capital flows, and reserve management - Adler, Gustavo, Noemie Lisack, and Rui Mano. 2019. “Unveiling the effects of foreign exchange intervention: A panel approach.” Emerging Markets Review, Vol. 40: 100620. - Avdjiev, Stefan, Valentina Bruno, Catherine Koch, and Hyun Song Shin. 2019. “The Dollar Exchange Rate as a Global Risk Factor: Evidence from Investment." IMF Economic Review 67, no. 1: 151-173. - Baba, Chikako, Ricardo Cervantes, Salim Darbar, Annamaria Kokenyne, and Viktorya Zotova. Forthcoming. “New Measures of Capital Flow Restrictions — AREAER Indices.” IMF Working Paper. (Washington: International Monetary Fund). - Cavallino, Paolo, 2019. “Capital Flows and Foreign Exchange Intervention.” American Economic Journal: Macroeconomics, Vol. 11(2), 127-70. - Chamon, Marcos, David Hofman, Sergi Lanau, Umang Rawat, and Miklos Vari. 2019. “The Effectiveness of Intervention,” Chapter 4: 43-61, in Foreign Exchange Intervention in Inflation Targeters in Latin America, edited by Marcos Chamon, David Hofman, Nicolás E. Magud, and Alejandro Werner. International Monetary Fund. - Chang, Roberto. 2018. “Foreign Exchange Intervention Redux,” NBER Working Paper No. 24463. - Choi, Jae Hoon. 2020. “Capital Controls and Foreign Exchange Market Intervention.” Journal of International Money and Finance 101: 102098. - Dominguez, Kathryn ME, 2012. “Foreign Reserve Management During the Global Financial Crisis.” Journal of International Money and Finance 31, no. 8: 2017-2037. - Cukierman, Alex, 2019. “Forex Intervention and Reserve Management in Switzerland and Israel since the Financial Crisis: Comparison and Policy Lessons.” Open Economies Review No. 30: 403–424 - Henning, C. Randall, 2006. “Organizing Foreign Exchange Intervention in the Euro Area.” JCMS: Journal of Common Market Studies 45, no. 2: 315-342. - Larrain, Catalina and Diego Saravia, 2019. “Interventions in Chile,” Chapter 8: 115-35, in Foreign Exchange Intervention in Inflation Targeters in Latin America, edited by Marcos Chamon, David Hofman, Nicolás E. Magud, and Alejandro Werner. International Monetary Fund. - Neely, Christopher J. 2011. “A Foreign Exchange Intervention in an Era of Restraint.” Federal Reserve Bank of St. Louis Review 93, no. 5: 303-24. - Obstfeld, Maurice, 2015. “Trilemmas and Trade-offs: Living with Financial Globalization,” BIS Working Papers No. 480. - Korinek, Anton. 2018. “Regulating Capital Flows To Emerging Markets: An Externality View.” Journal of International Economics 111: 61-80. - Rebucci, Alessandro, and Chang Ma. 2019. “Capital Controls: A Survey of the New Literature,” NBER Working Paper No. 26558. - Chinn, Menzie D. and Hiro Ito, 2006. “What Matters for Financial Development? Capital Controls, Institutions, and Interactions.” Journal of Development Economics, Volume 81, Issue 1, Pages 163-192. - Avdjiev, Stefan, Valentina Bruno, Catherine Koch, and Hyun Song Shin. 2019. “The Dollar Exchange Rate as a Global Risk Factor: Evidence from Investment." IMF Economic Review 67, no. 1: 151-173. ### Quantile methods, factor models, and statistical/econometric methods - Amengual, Dante and Mark W Watson. 2007. “Consistent Estimation of the Number of Dynamic Factors in a Large N and T Panel,” Journal of Business & Economic Statistics, 25(1): 91-96. - Ando, Tomohiro, and Jushan Bai. 2019. “Quantile co-movement in financial markets: A panel quantile model with unobserved heterogeneity.” Journal of the American Statistical Association, article in press. - Azzalini, Adelchi. 1985. “A Class of Distributions Which Includes the Normal Ones.” Scandinavian Journal of Statistics 12, No. 2: 171-178. - ______________. 2013. The Skew-Normal and Related Families. Vol. 3. Cambridge University Press. - Bai, Jushan and Serena Ng. 2002. “Determining the Number of Factors in Approximate Factor Models,” Econometrica, 70: 191–221. - Canay, Ivan A., 2011. “A Simple Approach to Quantile Regression for Panel Data.” The Econometrics Journal 14, no. 3: 368-386. - Hagemann, Andreas. (2017). “Cluster-Robust Bootstrap Inference in Quantile Regression Models.” Journal of the American Statistical Association, 112(517), 446-456. - Hallin, Marc and Roman Liška. 2007. “Determining the Number of Factors in the General Dynamic Factor Model,” Journal of the American Statistical Association, 102(478): 603-617. - Jolliffe, Ian. 2002. Principal Component Analysis, 2nd edition. Springer, New York. - Josse, Julie and François Husson. 2012a. “Handling missing values in exploratory multivariate data analysis methods.” Journal de la Société Française de Statistique 153, no. 2: 79-99. - ____________________________. 2012b. “Selecting the number of components in PCA using cross-validation approximations.” Computational Statistics and Data Analysis 56 (6): 1869-1879. - ____________________________. 2016. “missMDA: A Package for Handling Missing Values in Multivariate Data Analysis.” Journal of Statistical Software, 70(1), 1-31. - Kato, Kengo, Antonio F. Galvão Jr, and Gabriel V. Montes-Rojas. 2012. “Asymptotics for panel quantile regression models with individual effects.” Journal of Econometrics 170, no. 1: 76-91. - Koenker, Roger, 2004. "Quantile Regression for Longitudinal Data." Journal of Multivariate Analysis 91, no. 1: 74-89. - Koenker, Roger, 2019. Quantreg: Quantile Regression, R package version 5.40. - Krippner, Leo. 2015. Zero Lower Bound Term Structure Modeling: A Practitioner’s Guide. Berlin: Springer. - Hamilton, James D. 2018. “Why you should never use the Hodrick-Prescott filter.” Review of Economics and Statistics 100, no. 5: 831-843. - Canay, Ivan A., 2011. “A Simple Approach to Quantile Regression for Panel Data.” The Econometrics Journal 14, no. 3: 368-386. ### Financial conditions, sentiment, and credit channels - Adrian, Tobias, Nina Boyarchenko, and Domenico Giannone. 2019. “Vulnerable Growth.” American Economic Review, Vol. 109(4): 1263-89. - Adrian, Tobias, Fernando Duarte, Nellie Liang, and Pawel Zabcyk. Forthcoming. “Monetary and Macroprudential Policy with Endogenous Risk.” IMF Working Paper. - Adrian, Tobias, Federico Grinberg, Nellie Liang, and Sheheryar Malik. 2018. “The Term Structure of Growth-at-Risk,” IMF Working Paper No. 18/180. (Washington: International Monetary Fund). - Adrian, Tobias, and Nellie Liang. 2018. “Monetary Policy, Financial Conditions, and Financial Stability.” International Journal of Central Banking 14(1): 73-131. - López-Salido, David, Jeremy C. Stein, and Egon Zakrajšek, 2017. “Credit-Market Sentiment and the Business Cycle,” Quarterly Journal of Economics, 1373–1426. - Hatzius, Jan, Peter Hooper, Frederic S. Mishkin, Kermit L. Schoenholtz, and Mark W. Watson. 2010. Financial conditions indexes: A fresh look after the financial crisis. U.S. Monetary Policy Forum (Chicago: Chicago Booth Initiative on Global Markets) pp. 3-59. - Bernanke, Ben S., Mark Gertler, and Simon Gilchrist. 1999. “The Financial Accelerator in a Quantitative Business Cycle Framework." Handbook of Macroeconomics 1: 1341-1393. - López-Salido, David, Jeremy C. Stein, and Egon Zakrajšek, 2017. “Credit-Market Sentiment and the Business Cycle,” Quarterly Journal of Economics, 1373–1426. - Gourio, Francois, Anil K. Kashyap, and Jae W. Sim. 2018. “The Tradeoffs in Leaning Against the Wind.” IMF Economic Review 66:70-115. ### Methodological and foundational contributions in macroeconomics and finance - Sims, Christopher A., 1980. “Macroeconomics and Reality.” Econometrica 48, No. 1: 1-48. - Sims, Christopher A., 1982. “Policy Analysis with Econometric Models.” Brookings Papers on Economic Activity, no. 1 (1982): 107-164. - Clarida, Richard, Jordí Gali and Mark Gertler, 2000. “Monetary Policy Rules and Macroeconomic Stability: Evidence and Some Theory”, The Quarterly Journal of Economics, Vol 115 (1), 147-180. - Lucas, Robert E. 1976. “Econometric policy evaluation: A critique.” In Carnegie-Rochester Conference Series on Public Policy, vol. 1, no. 1: 19-46. 1976. - Romer, Christina D., and David H. Romer. 2004. “A New Measure of Monetary Shocks: Derivation and Implications." American Economic Review 94, no. 4: 1055-1084. - Ramey, Valerie A. 2016. “Macroeconomic Shocks and Their Propagation.” In Handbook of Macroeconomics, vol. 2: 71-162. Elsevier. - Sharpe, William F, 1964. “Capital Asset Prices: A Theory of Market Equilibrium Under Conditions of Risk.” The Journal of Finance 19, no. 3: 425-442. - Zellner, Arnold, 1986. “Bayesian Estimation and Prediction Using Asymmetric Loss Functions.” Journal of the American Statistical Association 81, no. 394: 446-451. - Varian, Hal R., 1975. “A Bayesian Approach to Real Estate Assessment," in Studies in Bayesian Econometrics and Statistics in Honor of Leonard J. Savage, eds. Stephen E. Fienberg and Arnold Zellner, Amsterdam: North-Holland, pp. 195-208. *Italic: Source — wpiea2020123-print-pdf - REFERENCES (https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020123-print-pdf.pdf)* ### Appendix A for details. ### Appendix A — Data Summary, Policy Shock Sizes, and Regression Results ### Data summary statistics (sample sizes and univariate moments) - Detrended output: Observations 2,876; Average -0.03; Standard Deviation 1.11; Median 0.02; Minimum -8.08; Maximum 7.21 - Quarterly inflation: Observations 2,876; Average 0.66; Standard Deviation 0.87; Median 0.55; Minimum -3.07; Maximum 9.84 - Financial conditions index: Observations 2,876; Average -0.07; Standard Deviation 0.94; Median -0.20; Minimum -2.74; Maximum 6.11 - Real GDP growth*: Observations 2,876; Average 2.97; Standard Deviation 4.68; Median 2.86; Minimum -22.29; Maximum 26.89 - YoY Inflation: Observations 2,876; Average 2.72; Standard Deviation 2.51; Median 2.23; Minimum -6.32; Maximum 16.37 - Growth in credit*: Observations 2,876; Average 0.64; Standard Deviation 2.46; Median 0.50; Minimum -7.60; Maximum 10.60 - Credit to GDP (logs): Observations 2,876; Average 4.78; Standard Deviation 0.58; Median 4.93; Minimum 3.00; Maximum 5.99 - House prices (logs): Observations 2,876; Average 5.03; Standard Deviation 0.67; Median 4.90; Minimum 3.16; Maximum 8.45 - Output gap: Observations 2,808; Average 0.00; Standard Deviation 0.04; Median 0.00; Minimum -0.20; Maximum 0.33 - MPM All: Observations 2,876; Average 0.01; Standard Deviation 0.48; Median -0.05; Minimum -2.51; Maximum 2.24 - MPM Borrower-Based: Observations 2,876; Average 0.00; Standard Deviation 0.23; Median -0.01; Minimum -2.07; Maximum 1.99 - MPM FI-Based: Observations 2,876; Average 0.00; Standard Deviation 0.40; Median -0.03; Minimum -2.58; Maximum 2.17 - MP: Observations 2,852; Average -0.03; Standard Deviation 0.95; Median -0.05; Minimum -5.74; Maximum 9.05 - FXI: Observations 2,259; Average 0.00; Standard Deviation 0.01; Median 0.00; Minimum -0.09; Maximum 0.14 - CFM: Observations 2,259; Average -0.02; Standard Deviation 0.52; Median 0.03; Minimum -2.00; Maximum 2.30 - Note: * Winsorized at the top and bottom 0.5 percent. ### Policy shock size by country (one standard deviation of each policy shock) - Australia 10: MPM All 0.38; MPM Borrower-Based 0.00; MPM FI-Based 0.19; MP 1.03; FXI 0.13; CFM 0.58 - Austria 10: MPM All 0.29; MPM Borrower-Based 0.00; MPM FI-Based 0.14; MP 0.93; FXI 0.09; CFM 0.22 - Belgium 9: MPM All 0.34; MPM Borrower-Based 0.01; MPM FI-Based 0.17; MP 0.95; FXI 0.09; CFM 0.27 - Brazil 5: MPM All 0.96; MPM Borrower-Based 0.13; MPM FI-Based 0.47; MP 1.01; FXI 0.17; CFM 0.95 - Canada 10: MPM All 0.53; MPM Borrower-Based 0.34; MPM FI-Based 0.14; MP 0.98; FXI 0.19; CFM 0.50 - Chile 5: MPM All 0.13; MPM Borrower-Based 0.13; MPM FI-Based 0.00; MP 1.04; FXI 0.19; CFM 0.69 - China 42: MPM All 1.23; MPM Borrower-Based 0.55; MPM FI-Based 0.41; MP 1.04; FXI 0.05; CFM 0.96 - Colombia 7: MPM All 0.43; MPM Borrower-Based 0.23; MPM FI-Based 0.11; MP 0.95 - Denmark 10: MPM All 0.24; MPM Borrower-Based 0.17; MPM FI-Based 0.17; MP 1.04; FXI 0.11; CFM 0.03 - Finland 10: MPM All 0.34; MPM Borrower-Based 0.14; MPM FI-Based 0.17; MP 0.95; FXI 0.09; CFM 0.25 - France 10: MPM All 0.22; MPM Borrower-Based 0.00; MPM FI-Based 0.14; MP 0.97; FXI 0.09; CFM 0.31 - Germany 10: MPM All 0.28; MPM Borrower-Based 0.00; MPM FI-Based 0.10; MP 0.95; FXI 0.10; CFM 0.22 - Hong Kong SAR 10: MPM All 0.56; MPM Borrower-Based 0.47; MPM FI-Based 0.21; MP 1.05; FXI 0.14; CFM 0.46 - Hungary 7: MPM All 0.50; MPM Borrower-Based 0.35; MPM FI-Based 0.00; MP 1.00; FXI 0.15; CFM 0.52 - India 27: MPM All 0.97; MPM Borrower-Based 0.34; MPM FI-Based 0.43; MP 1.03; FXI 0.14; CFM 1.06 - Indonesia 5: MPM All 0.61; MPM Borrower-Based 0.27; MPM FI-Based 0.13; MP 1.02; FXI 0.16; CFM 0.80 - Ireland 7: MPM All 0.30; MPM Borrower-Based 0.16; MPM FI-Based 0.16; MP 0.94; FXI 0.16; CFM 0.09 - Italy 10: MPM All 0.35; MPM Borrower-Based 0.10; MPM FI-Based 0.17; MP 0.96; FXI 0.10; CFM 0.16 - Japan 10: MPM All 0.24; MPM Borrower-Based 0.00; MPM FI-Based 0.10; MP 1.22; FXI 0.12; CFM 0.27 - Kazakhstan 3: MPM All 0.42; MPM Borrower-Based 0.16; MPM FI-Based 0.22; MP 1.05 - Korea 10: MPM All 0.74; MPM Borrower-Based 0.58; MPM FI-Based 0.19; MP 1.04; FXI 0.20; CFM 0.98 - Malaysia 10: MPM All 0.58; MPM Borrower-Based 0.22; MPM FI-Based 0.14; MP 1.02; FXI 0.19; CFM 0.82 - Mexico 4: MPM All 0.41; MPM Borrower-Based 0.00; MPM FI-Based 0.26; MP 1.00; FXI 0.09; CFM 0.57 - Netherlands 10: MPM All 0.39; MPM Borrower-Based 0.25; MPM FI-Based 0.10; MP 0.94; FXI 0.15; CFM 0.03 - Norway 10: MPM All 0.43; MPM Borrower-Based 0.22; MPM FI-Based 0.32; MP 1.05; FXI 0.13; CFM 0.22 - Peru 7: MPM All 0.75; MPM Borrower-Based 0.12; MPM FI-Based 0.17; MP 1.01 - Philippines 3: MPM All 0.40; MPM Borrower-Based 0.18; MPM FI-Based 0.18; MP 0.79; FXI 0.09; CFM 0.70 - Poland 6: MPM All 0.59; MPM Borrower-Based 0.24; MPM FI-Based 0.21; MP 1.04; FXI 0.16; CFM 0.51 - Russia 5: MPM All 0.89; MPM Borrower-Based 0.00; MPM FI-Based 0.50; MP 0.90; FXI 0.19; CFM 0.91 - Singapore 10: MPM All 0.52; MPM Borrower-Based 0.33; MPM FI-Based 0.10; MP 1.05; FXI 0.14; CFM 0.48 - South Africa 10: MPM All 0.24; MPM Borrower-Based 0.00; MPM FI-Based 0.19; MP 1.03; FXI 0.11; CFM 0.85 - Spain 10: MPM All 0.36; MPM Borrower-Based 0.10; MPM FI-Based 0.22; MP 0.98; FXI 0.08; CFM 0.22 - Sweden 10: MPM All 0.34; MPM Borrower-Based 0.14; MPM FI-Based 0.26; MP 1.03; FXI 0.11; CFM 0.25 - Switzerland 10: MPM All 0.31; MPM Borrower-Based 0.00; MPM FI-Based 0.24; MP 1.03; FXI 0.11; CFM 0.31 - Turkey 2: MPM All 1.00; MPM Borrower-Based 0.29; MPM FI-Based 0.50; MP 1.03; FXI 0.19; CFM 0.64 - United Kingdom 10: MPM All 0.33; MPM Borrower-Based 0.10; MPM FI-Based 0.26; MP 1.02; FXI 0.09; CFM 0.03 - United States 10: MPM All 0.32; MPM Borrower-Based 0.10; MPM FI-Based 0.17; MP 1.05; FXI 0.06; CFM 0.03 ### Effect of policy changes on loss functions (Table 3 — expected loss coefficients) - Loss function: ℓ = ωy (y_{t+h} − ȳ_{t+h})^2 + ωπ π_{t+h}^2; quantile regressions for cumulative GDP growth up to 14 quarters ahead. Exercise conditions: tightening of macroprudential policy, monetary policy, capital flow measures, or FX interventions (purchase of foreign currency) when financial conditions are loose. MPM All = Alam et al. (2019) index of 17 macroprudential measures. MP = monetary policy shock (residual of estimated Taylor rule). CFM = Baba et al.’s index of capital-flow-management measures. HF MP = high-frequency monetary policy surprise measure. - Domestic FCI (Domestic Shock) results (ωy=1, ωp=0; ωy=1, ωp=1; ωy=0.542, ωp=1): - MPM All: -0.089***; -0.085***; -0.083*** - MPM Borrower-Based: -0.100***; -0.068***; -0.065*** - MPM FI-Based: -0.053**; -0.036**; -0.035** - MP: 0.121***; 0.115***; 0.111*** - HF MP: -0.011; -0.011; -0.011 - External FCI (External Shock) results (ωy=1, ωp=0; ωy=1, ωp=1; ωy=0.542, ωp=1): - MPM All: -0.112***; -0.107***; -0.104*** - MPM Borrower-Based: -0.107***; -0.101***; -0.096*** - MPM FI-Based: -0.068***; -0.067***; -0.065*** - MP: 0.038*; 0.036*; 0.036* - FXI: -0.022; -0.021; -0.021 - CFM: -0.039; -0.034; -0.030 - HF MP: -0.025; -0.023; -0.022 ### Effect of policy changes on loss functions by vulnerability (Table 4) - Setup: Quantile regressions with vulnerability interaction (Vul = credit-to-GDP in Panel A; house-price index in Panel B). Results shown for low vs high vulnerability and low vs high valuations with loose financial conditions. - Panel A (Low Credit / High Credit; Domestic and External FCI variants): - MPM All: -0.089**; -0.086**; -0.084** (Low Credit); -0.099**; -0.094**; -0.090** (High Credit) - MPM Borrower-Based: -0.033; -0.032; -0.031 (Low Credit); -0.083***; -0.078***; -0.075*** (High Credit) - MPM FI-Based: -0.076**; -0.072**; -0.070** (Low Credit); -0.028; -0.027; -0.026 (High Credit) - MP: 0.137***; 0.132***; 0.129*** (Low Credit); 0.126***; 0.120***; 0.115*** (High Credit) - Panel B (Low Valuations / High Valuations; Domestic and External FCI variants): - MPM All: -0.152***; -0.144***; -0.139*** (Low Valuations); -0.072**; -0.068**; -0.066** (High Valuations) - MPM Borrower-Based: -0.123**; -0.114**; -0.107** (Low Valuations); -0.074**; -0.071**; -0.068** (High Valuations) - MPM FI-Based: -0.085**; -0.082**; -0.080** (Low Valuations); -0.031; -0.031; -0.030 (High Valuations) - MP: 0.150***; 0.145***; 0.141*** (Low Valuations); 0.216***; 0.208***; 0.202*** (High Valuations) ### Effect of coordinated policy changes (Table 5) - Setup: Quantile regressions assessing MPM tightening with monetary policy expanding or contracting (ωy variants reported). - When MPM tightens and MP expands/contrasts (ωy=1, ωp=0; ωy=1, ωp=1; ωy=0.542, ωp=1): - MPM All: -0.162; -0.157; -0.153 - MPM Borrower-Based: -0.189; -0.180; -0.173 - MPM FI-Based: -0.140; -0.136; -0.133 - Coordinated exercise (MP expands vs contracts reported separately): - MPM All: -0.026; -0.023; -0.021 - MPM Borrower-Based: 0.025; 0.024; 0.023 - MPM FI-Based: 0.023; 0.023; 0.024 ### Effect of policy changes with alternative loss functions (Table 6) - Loss functions: linear-quadratic and asymmetric policymaker preferences; quantile regressions for domestic and external shocks under loose financial conditions. MP measured as residual of Taylor rule; FXI and CFM included for external shocks where noted (selling foreign currency). - Asymmetric and Linear-quadratic results (columns show ωy and ωp variants): - MPM All: -0.080***; -0.076***; -0.087***; -0.100***; -0.095***; -0.109*** - MPM Borrower-Based: -0.092***; -0.087***; -0.098***; -0.097***; -0.089***; -0.100*** - MPM FI-Based: -0.046**; -0.045**; -0.053*; -0.060**; -0.058**; -0.067*** - MP: 0.134***; 0.126***; 0.124***; 0.046**; 0.044**; 0.040* - FXI (external shock columns): -0.029; -0.027*; -0.024 - CFM (external shock columns): -0.040; -0.033; -0.041 ### Effect of policy changes on loss functions by group of countries (Table 7) - Setup: Separate quantile-regression estimates for groups (advanced economies vs emerging economies) under loose financial conditions. (Table content present in source but numeric rows not reproduced verbatim in appendix extract provided.) ### Effect of policy changes on loss functions accounting for common factors (Table 8) - Panel A (Common factors; quantile regressions with country and quarter fixed effects estimated via Koenker’s penalized estimator): - MPM All: -0.073; -0.070; -0.067 - MPM Borrower-Based: -0.038; -0.035; -0.033 - MPM FI-Based: -0.053; -0.051; -0.050 - MP: 0.142; 0.135; 0.130 - Panel B (Unobserved heterogeneity; quantile regressions with country-specific slopes and five unobserved common factors estimated via Ando and Bai’s procedure): - MPM All: -0.104; -0.108; -0.110 - MPM Borrower-Based: -0.477; -0.336; -0.228 - MPM FI-Based: -0.129; -0.124; -0.120 - MP: 0.063; 0.060; 0.057 ### Financial Conditions Index (FCI) — data sources and construction - Purpose: FCI approximates unobservable financial conditions by summarizing information from financial variables that influence and predict future economic activity. - Underlying data: up to 10 financial variables (depending on availability): the real short-term interest rate, the interbank spread, the spreads on local currency and U.S. dollar-denominated sovereign and corporate debt, equity returns (in local currency), the implied volatility of equity returns, the change in a debt-weighted exchange rate index, and the year-on-year growth in real house prices. - Missing data treatment: Use Josse and Husson’s (2012a) iterative principal component analysis algorithm (implemented in R package missMDA) to impute missing values. - Number of components: Optimal number p chosen via Josse and Husson’s (2012b) cross-validation approximation method; the FCI is the sum of the first p principal components, each weighted by its share of total variance. For most countries, 2 to 3 factors are optimal. - Global FCI proxy: FCI in the United States used as a proxy for global financial conditions (supported by Arregui et al. (2018)). ### Variable definitions (excerpt) - FCI_{i,t}: Financial conditions index. Calculated as the weighted sum of the first N principal components of a set of up to 11 financial variables for each country. N is chosen using Josse and Husson’s (2012b) cross-validation method. Each principal component is weighed by its share of the total variance. Missing values of the underlying financial variables are treated with Josse and Husson’s (2012a) imputation method. The following financial variables are measured at the country level, when available: the real short-term interest rate, the interbank spread, the term spread, the spread on sovereign local-currency debt, the spread on sovereign dollar-denominated debt, the spread on corporate local-currency debt, the spread on corporate dollar-denominated debt, the average price-to-book ratio of the MSCI country index, the historical or implied volatility of country stock returns, debt-weighted exchange rate returns, and year-on-year growth of house prices (deflated by the CPI). *Source: Appendix A (data tables and methodology) from the provided IMF working paper materials.* ### Annex 1.1 of IMF (2018) for a detailed description of sources. For the analysis of global FCI, the US FCI is used ### Annex 1.1 — Variable definitions and methodological details ### Variable definitions and shock construction - yy_iit: The natural log of real GDP. Real GDP data is from the IMF’s World Economic Outlook database. - π_iit+h: Inflation. The h-quarter difference in the natural log of the consumer price index (CPI). The CPI data is from the IMF’s World Economic Outlook database. - I_iit_mpm_A: Index of changes in all macroprudential measures. Takes the value of 0 for no change, 1 (-1) for one tightening (loosening) action, and 2 (-2) for more than one tightening (loosening) actions, in net terms, in each quarter and in each country. Data source: iMaPP database by Alam and others (2019). - I_iit_mpm_B: Index of changes in borrower-based macroprudential measures (subset of I_iit_mpm_A). Same coding as above; subcategory consists of limits to loan-to-value ratios and debt-service-to-income ratios. Data source: iMaPP (Alam and others, 2019). - I_iit_mpm_F: Index of changes in financial institutions-based macroprudential measures (subset of I_iit_mpm_A). Same coding as above; subcategory consists of (1) countercyclical buffers, (2) conservation buffers, (3) capital requirements, (4) leverage limits, (5) loan loss provisions, (6) limits to credit growth, (7) loan restrictions, (8) limits to foreign currency loans, (9) liquidity requirements, (10) limits to the loan to deposit ratio, (11) limits to FX positions, (12) reserve requirements, and (13) measures on systemically important financial institutions. Data source: iMaPP (Alam and others, 2019). - ε̂_iit_mpm_A: Aggregate macroprudential policy shock. Calculated as the difference between I_iit_mpm_A and its prediction estimated with an ordered probit model with the credit-to-GDP and house price gaps as explanatory variables. - ε̂_iit_mpm_B: Borrower-based macroprudential shock. Calculated as above using I_iit_mpm_B. - ε̂_iit_mpm_F: Financial institutions-based shock. Calculated as above using I_iit_mpm_F. - ccm_iit: Credit-to-GDP gap. Residuals of country-level regressions of credit-to-GDP on four lags of itself. Data from the Bank for International Settlements. - hccm_iit: House price gap. Residuals of country-level regressions of a house price index on four lags of itself. Data from the Bank for International Settlements. - eccm_iit: Exchange rate gap. Residuals of country-level regressions of country i's dollar exchange rate on four lags of itself. Data from the IMF’s International Financial Statistics. - σ(e)_iit: The quarterly nominal effective exchange rate volatility calculated from daily data. Data from the IMF’s International Financial Statistics. - ε̂_iit_r: Monetary policy shock. Calculated as the standardized residual of an estimated Taylor-type rule. The rule is estimated with the quarterly change of the monetary policy rate on the left-hand-side and 12-month ahead expected inflation and GDP growth (from Consensus Forecasts) and the lagged policy rate. - FXI_iit: Foreign exchange interventions as a percent of GDP. Actual FXI data are used for a specified list of countries as reported in central bank websites and literature; for all countries and periods interventions are approximated with Adler and others’ (2019) measure of valuation-adjusted changes to central banks’ net foreign assets when actual data are unavailable. - ε̂_iit_FXINT: Foreign exchange intervention shock. Calculated as the residual of an estimated FXI rule (OLS) with FXI percent of GDP on the lag of exchange rate gap (egap), exchange rate volatility (σ(e)), and covariates used in the first column of Table 2 in Forbes and Klein (2015) (log VIX, change in the U.S. policy interest rate, change in capital inflows as a percentage of GDP, inflation, commodity price index, Economist All-Commodity Dollar index interacted with a major commodity exporter dummy, reserves as percent of GDP, Chin and Ito’s index of capital account openness, log of average of six World Bank governance indexes, rate of exchange rate depreciation vs. U.S. dollar, and the interest rate differential against the U.S.’s policy rate). For the United States, exchange rate and interest rate differential are measured against the euro. - I_iit_CFM_CF: Index of changes in capital flow management measures. Takes value 0 for no change, 1 (-1) for tightening (loosening) actions, and 2 (-2) for more than one tightening (loosening) actions, in net terms, in each quarter in each country. Data from Baba and others (forthcoming). - ε̂_iit_CFM_CF: CFM shock. Calculated as the difference between I_iit_CFM_CF and its prediction estimated with an ordered probit model with lag of exchange rate gap (egap), exchange rate volatility (σ(e)), and covariates used in the first column of Table 2 in Forbes and Klein (2015). ### Macroprudential policy: data construction and estimation - Data source: iMaPP database (Alam and others, 2019) provides dummy-type indices for 17 categories of macroprudential policy instruments (January 1990 through December 2016). - Categorical indicators constructed: I_iit_mpm_A (overall), I_iit_mpm_B (borrower-based), I_iit_mpm_F (financial-institution-based). Each indicator takes values 0, ±1, ±2 in net quarterly terms. - Shock extraction approach: Panel ordered probit regressions with country fixed effects (specification 6) are estimated for each categorical indicator; shocks are residuals (observed minus fitted probabilities) following the procedure in Section III.B. Ordered Probit regression results (Table A.2; specification columns (1)(2)(3)) - Credit gap coefficients: - Column (1): 0.0066** (0.011) - Column (2): 0.0124*** (0.001) - Column (3): 0.0051* (0.075) - House price gap coefficients: - Column (1): 0.0004 (0.293) - Column (2): 0.0006 (0.261) - Column (3): 0.0002 (0.587) - Past policy actions coefficients: - Column (1): 0.1080*** (0.000) - Column (2): 0.2221*** (0.000) - Column (3): 0.1195*** (0.000) - Thresholds (α): - α_1: -2.4806***, -3.1003***, -2.6595*** (all (0.000)) - α_2: -1.6860***, -2.3596***, -1.8421*** (all (0.000)) - α_3: 1.4329***, 2.0822***, 1.5476*** (all (0.000)) - α_4: 2.0945***, 2.8356***, 2.2820*** (all (0.000)) - Sample and fit: - Observations: 2,935 - Country fixed effects: YES - Number of countries: 37 - Pseudo R2: 0.0422 (col 1), 0.0572 (col 2), 0.0540 (col 3) ### Monetary policy: baseline and robustness measures - Baseline monetary policy shock: residual of a country-level Taylor-type rule (standardized residual). - Robustness: high-frequency monetary policy surprises around announcements (asset-price responses). These are used as robustness checks and for some small open economies the US or euro area surprises are used as proxies (e.g., use US for Hong Kong SAR and Singapore; use euro area for Denmark and Hungary). - Limitations of high-frequency surprises noted: - Shorter time series compared with Taylor-rule shocks (from 25 quarters for Switzerland to 108 quarters for the United States). - Missing data concentrated in emerging markets (which tend to use macroprudential policy more often). - Heterogeneous data sources and uneven data quality; true intra-day high-frequency data mainly for major central banks in advanced economies; many emerging markets use daily-frequency surprises. - High-frequency data sources and sample sizes (Table A.3) — selected entries: - Australia: 45 observations, Start year 2005, End year 2019, source Kearns, Schrimpf, and Xu (2018), based on intra-day asset price movements. - Brazil: 56 observations, Start year 2003, End year 2016, Barbone-Gonzalez (2019), based on daily asset price movements. - Hong Kong SAR: 108 observations, Start year 1990, End year 2016, Jarocinski and Karadi (2019), use US shock. - Singapore: 108 observations, Start year 1990, End year 2016, Jarocinski and Karadi (2019), use US shock. - United States: 108 observations, Start year 1990, End year 2016, Jarocinski and Karadi (2019), based on intra-day asset price movements. - Many countries (e.g., China, India, Kazakhstan, Norway, Peru, Poland) have "No data available." ### Foreign exchange interventions (FXIs): measurement and proxy construction - Actual FXI data used where publicly available (Table A.4 lists countries with actual data). - Proxy FXI variable (percent of GDP) for other countries: - FXI_iit = (ΔNFA_iit − ΔadjNFA_iit) / GGDP_iit - ΔNFA_iit: quarter-on-quarter change in central bank’s net foreign assets (in U.S. dollars). - ΔadjNFA_iit: adjustment term for valuation changes and income flows. - GGDP_iit: annual GDP (U.S. dollars) for country i in quarter t. - Adjustment-term assumptions: - Adjust only two NFA components: securities (S_iit) and currencies & deposits (CG_iit). ΔadjNFA_iit = ΔadjS_iit + ΔadjCG_iit. - Assume no valuation changes and no income flows for other NFA components. - Assume uniform currency composition across NFA components due to limited granularity. - Assume securities are mostly 10-year government bonds. - For currencies and deposits, assume valuation changes are zero and interest income based on the 3-month interbank rate. - Computation of adjustment terms: - ΔadjS_iit = Σ_{c ∈ C} [(R_ic / R_i,c−1 − 1) * S_iit−1 * ω_i,c−1] - ΔadjCG_iit = Σ_{c ∈ C} [i_ic−1 * CG_iit−1 * ω_i,c−1] - Where R_ic / R_i,c−1 − 1 is growth rate of the treasury’s total return index (Thomson Reuters Datastream), i_ic−1 is the 3-month interbank rate (Haver), and ω_i,c is currency-c asset share from IMF COFER. Currency set C: U.S. dollar, Australian dollar, Canadian dollar, British pound, Japanese yen, Swiss Franc, Chinese yuan and Euro. - Correlations between proxy FXI and actual FXI (Table A.4; selected values): - Brazil (EM): 0.39; Group Average EM: 0.56 - Chile (EM): 0.30 - Colombia (EM): 0.65 - India (EM): 0.83 - Russian Federation (EM): 0.88 - Australia (AE): -0.04; Group Average AE: 0.22 - Denmark (AE): 0.84 - Switzerland (AE): 0.86 - United States (AE): 0.01 - Notes: Correlations calculated for actual FXI data and the proxy FXI variable, both in percent of annual GDP. ### Capital flow management measures (CFMs): estimation - Policy shocks computed as in Section III.B using ordered probit regressions. - Ordered probit regression results for CFM (Table A.5): - Coefficients (standard errors / p-values in parentheses when reported): - VIX: 0.0150*** (0.001) - Δ FFR: 0.1072 (0.177) - Commodity price index: 0.0366 (0.751) - Commodity producer dummy x Commodity price index: 0.0568 (0.610) - Δ inflation: 0.0150 (0.409) - Real GDP growth: 0.0134** (0.019) - FX Reserves (percent of GDP): 0.0009 (0.411) - Institutions: 0.4234 (0.137) - Capital account openness: 0.2199*** (0.000) - Inward capital flows (percent of GDP): -0.0005 (0.906) - Outward capital flows (percent of GDP): 0.0002 (0.964) - Δ Exchange rate: -0.9428 (0.132) - FX volatility: -0.9128* (0.079) - Δ Exchange rate x FX volatility: 2.4946** (0.045) - Δ IR: 0.0123*** (0.005) - Thresholds (α): - α_1: -0.5279 (0.532) - α_2: -0.1394 (0.869) - α_3: 3.2341*** (0.000) - α_4: 3.6070*** (0.000) - Sample and fit: - Observations: 2,675 - Country fixed effects: YES - Number of countries: 37 - Pseudo R2: 0.0978 ### Appendix B — Additional results (figures and inference notes) - Figures B.1–B.3 present median responses of GDP growth and inflation, and tail risk (10th percentile of future detrended GDP growth), to domestic and global FCI loosening shocks, conditional on presence or absence of policy changes (macroprudential or monetary). - Inference: - A square marker denotes effects significantly different from zero at least at the 10 percent significance level. - Standard errors clustered at country level using Hagemann’s (2017) wild bootstrap approach. - Horizontal axis: number of quarters since the time of a loosening shock to domestic or global financial conditions. - Note on global FCI: For the analysis of global FCI, the US FCI is used as a proxy for global financial conditions. *Source: wpiea2020123-print-pdf — Annex 1.1 (IMF, 2018) — variable definitions, data sources, and estimation details.* --- _Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020123-print-pdf.pdf_