## wpiea2021256-print-pdf

## Source details

**Canonical URL:** [wpiea2021256-print-pdf](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021256-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021256-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021256-print-pdf.pdf.json)

---

### Calibration approach and key assumptions
- Steady-state leverage ratio: ad-hoc target set below historical observations to reflect bank buffer behavior.
- Persistence and responsiveness parameters chosen to match data.
- Standard deviations of shocks: generally calibrated unless indicated as estimated.
- Calibration purpose: obtain reasonable dynamics of gaps and trends and ensure model stability.
- Rule for setting the leverage target is calibrated.

### Model parameters related to credit and macroprudential instrument blocks (Table IV.1)
- Demand for credit – equation (3)
  - 휌휌1: 0.5 — Bayesian estimation
  - 훼훼1: 0.14 — Bayesian estimation
  - 훼훼2: 0.12 — Bayesian estimation
- Credit premium – equation (7)
  - 휌휌4: 0.8 — Bayesian estimation
  - 훼훼3: 0.2 — Bayesian estimation
  - 훼훼4: 0.001 — Calibration
- Credit growth trend – equation (4)
  - 휌휌2: 0.75 — Calibration
  - Δ푛푛푐푐푐푐�����푠푠푠푠: 0 — Calibration
- Real monetary condition index – equation (12)
  - 푏푏4: 0.7 — Calibration
  - 푏푏5: 0.5 — Calibration
- 1Y rate – equation (5)
  - 휌휌3: 0.5 — Calibration
  - 푏푏1: 0.6 — Calibration
- Aggregate demand
  - 푏푏2: 0.2 — Calibration
  - 푏푏3: 0.6 — Calibration
  - 푏푏푐푐푆푆푆푆푇푇퐴퐴푇푇: 11 — Calibration
  - 푏푏6: 0.25 — Calibration
- Target for the leverage ratio – equation (16)
  - 휌휌5: 0.975 — Calibration
  - 휙휙: (value not specified in provided excerpt)

### Model properties and impulse-response analysis
- Focus: impulse responses to financial-system shocks; new credit and macroprudential blocks allow tracking credit responses without altering business-cycle IRFs from Guo et al. 2019.
- Lending rate specification: depends on expected policy rate; with leverage constraint, lending rate and real lending rate gap incorporate additional effects.
- All IRF shocks: 1 percent magnitude and unexpected.
- Identification issues: 푏푏5 and 휌휌4 are difficult to distinguish from other parameters.
- Fisher matrix: property of the model; SVD used to assess parameter identifiability (see Appendix C summary).

### Credit demand shock — mechanism and counterfactuals
- Shock: positive shock to credit gap (demand above trend controlling for business cycle and cost).
- Macroprudential reaction: raises required leverage ratio to moderate credit growth and safeguard bank resilience.
- Bank responses:
  - Accumulate capital to meet higher required leverage ratio.
  - Lending margins increase (via 푓푓표표(푏푏푐푐푡푡 − 푏푏푐푐푡푡푇푇퐴퐴푇푇) > 0 when gap < 0).
  - Dividend payments cut to zero (consistent with 푓푓1(...) close to zero under gap < 0).
- Counterfactual comparison:
  - Active policy (raising leverage target) vs passive policy (unchanged target): active policy induces larger increase in lending rate, mitigates credit boom, and leads to milder monetary tightening.
- Additional implementation notes:
  - Shock implemented as shock minus control; variables reported as deviations from steady state.
  - Bank lending modeled as demand determined; no supply constraint assumed.
  - Deviations of dividend payoffs from zero in simulation are very small (possible numerical errors).
  - Leverage target dynamics follow equation (16) and can be recalibrated.

### Bank profitability shocks — mechanism and counterfactuals
- Negative bank profitability shock (e.g., higher default rates):
  - Macroprudential authority cuts leverage ratio target to ease stress.
  - Banks with leverage below lowered target withhold dividends and increase lending margin.
  - Higher lending rate reduces credit demand → economic slowdown → negative output gap → inflation drops → monetary policy becomes accommodative.
  - Characteristic: policy and lending rates move in opposite directions.
- Counterfactual: keeping leverage ratio target unchanged results in higher stress, larger declines in output gap, and larger required monetary accommodation compared with active macroprudential easing.
- Positive bank income shock:
  - Macroprudential policy may not react unless building buffers is desired.
  - Banks may pay dividends; leverage rises above target.
  - Under excess capital, spread function 푓푓표표(...) = 0 and dividend payoff 푓푓1(...) > 0; lending rate, output gap, and inflation remain unchanged.

### Non-linearities and policy inference
- Non-linear macroprudential block generates asymmetric IRFs that help distinguish positive vs adverse profitability shocks.
- Countercyclical macroprudential policy can reduce reliance on monetary policy to address excessive credit booms or drops in bank profitability.
- Findings align with literature advocating countercyclical macroprudential tools to address endogenous financial procyclicality and risk-taking.

### Forecast evaluation (in-sample) — 2002Q1–2019Q4
- Forecast design: forecasts from 2001Q4 for subsequent six quarters, rolling to 2019Q4; assume known current unobserved variables and external variables over horizon.
- Forecasting focus: credit block forecasting performance tested assuming spread, leverage ratio, and ROAs observed; macroprudential policy block not historically set.
- RMSE ratio interpretation: values < 1 mean model RMSE lower than random walk; > 1 opposite; 1 equal.
- Selected RMSE ratios (model vs random walk)
  - Business Cycle Block (2002-2019)
    - Inflation, y-o-y: 0.5, 0.6, 0.7, 0.8, 0.9, 0.9 (t+1 to t+6)
    - Policy interest rate: 2.1, 2.1, 2.0, 2.0, 1.9, 1.9
    - Real GDP growth, y-o-y: 0.5, 0.5, 0.5, 0.5, 0.6, 0.7
    - Nom. exchange rate depreciation, y-o-y: 0.6, 0.6, 0.5, 0.5, 0.5, 0.5
    - Output gap: 0.9, 0.9, 0.8, 0.8, 0.8, 0.9
  - Business Cycle Block (2010-2019)
    - Inflation, y-o-y: 0.6, 0.7, 0.8, 0.8, 0.7, 0.6
    - Policy interest rate: 1.5, 1.4, 1.3, 1.2, 1.0, 0.8
    - Real GDP growth, y-o-y: 0.5, 0.4, 0.4, 0.6, 0.7, 0.8
    - Nom. exchange rate depreciation, y-o-y: 0.5, 0.4, 0.4, 0.5, 0.5, 0.6
    - Output gap: 1.0, 0.9, 0.8, 0.7, 0.8, 0.9
  - Credit block (2002-2019)
    - Credit flow/GDP gap: 1.0, 1.0, 1.0, 1.1, 1.3, 1.3
    - Outstanding credit/GDP: 0.8, 0.8, 0.7, 0.7, 0.7, 0.6
    - Lending rate: 1.5, 1.4, 1.5, 1.5, 1.5, 1.5
    - Credit premium: 0.8, 0.7, 0.7, 0.7, 0.7, 0.7
  - Credit block (2010-2019)
    - Credit flow/GDP gap: 1.1, 1.0, 1.1, 1.1, 1.0, 1.1
    - Outstanding credit/GDP: 0.6, 0.6, 0.6, 0.6, 0.6, 0.5
    - Lending rate: 2.7, 1.1, 1.7, 1.4, 1.3, 1.2
    - Credit premium: 1.1, 1.0, 1.0, 1.0, 1.0, 0.9
- Interpretation of forecast performance:
  - Model generally outperforms random walk for business cycle variables, except for policy rate.
  - Evidence on credit cycle forecasting is mixed: credit premium forecast reasonable; lending rate and credit gap forecasts impaired by policy rate forecast errors.
  - Policy rate forecasting problems stem from model-implied stronger responsiveness to inflation and output gap than observed data; model implied deep negative rate recommendations during global financial crisis and counterfactual drops in 2015–2016 not reflected in actual policy rates.
  - Policy rate forecast performance improves after 2010.

### Historical interpretation using the QPM (Kalman filter)
- Filtration purpose: estimate trends and structural shocks fitting historical Philippine data.
- Observed data groups used:
  1. Business cycle block: headline CPI (core and food-and-energy), real GDP, GDP deflator, export/import shares, nominal exchange rate, policy interest rate; foreign variables: global CPI, GDP, oil and food prices, U.S. interest rates.
  2. Credit and macroprudential blocks: outstanding bank credit to firms, lending rate, VIX.
  3. Bank-related: total bank assets, ROAs, total banks’ capital.
- Filtering approach and practical adjustments:
  - Linear Kalman filter used despite non-linear model; expert tunes applied to address critical nonlinearities (dividends and spread) by restricting them to be non-negative.
  - Two-run approach: first run no restrictions; second run set dividends and spread to zero where negative in first run.
  - Measurement errors assumed absent; variables not de-trended but seasonally adjusted; VIX de-meaned.
  - Observation of 1Y benchmark rate not available; term premium assumed constant.
- Key filtration findings — credit and financial variables:
  - Credit gap history:
    - Positive credit gap opened in 2016, peaking at about 0.5 percent in 2018 (smoothed credit gap).
    - Positive credit gap closed rapidly in 2019.
    - Model-based filtration indicates credit gap moderate at peaks and not a significant financial stability threat historically.
  - Credit gap decomposition (equation (3)):
    - Real lending rate important determinant.
    - Credit shocks pronounced and sometimes offset real lending rate gap effects (e.g., 2008, 2015, 2018).
    - Output gap has lower impact on credit gap relative to real lending rate or credit demand shock.
  - Structural shock contributions:
    - Domestic and foreign business cycle shocks and monetary policy shocks important.
    - Domestic developments were main factor closing positive credit gap in 2018–2019; monetary policy and exchange rate helped smooth credit cycle since 2016.
  - Lending rate decomposition:
    - Driven by 1Y rate, term premium (constant), credit premium, and spread.
    - Spread contributions small, mainly early in filtration period.
  - Credit premium decomposition:
    - Credit premium close to neutral level in 2019.
    - Determinants: expected future business cycle developments, VIX (world financial conditions), and other structural shocks.
    - Credit premium assumed persistent (equation 8); lag term substituted out in decomposition.
  - Real lending rate and natural rate:
    - Real lending rate fluctuated around 3 percent since 2008.
    - Equilibrium level equals natural real interest rate adjusted for constant term premium; strong correlation between policy and lending rate in gap terms.
  - Banking sector variables:
    - Bank data available from 2008Q2 onward; prior periods estimated by filter.
    - Leverage ratio fluctuated around 12 percent since 2011, with a significant spike in 2012–2013.
    - Implicit target of 11 percent assumed in steady state calibration (ad-hoc and below historical average observed).
    - Banks’ lending spread triggered by regulatory requirements estimated close to zero on average; dividend payments positive.
    - ROA relatively stable around 1.5 percent; predicted ROA hovers around average ROA in data.
  - Financial sector feedback to business cycle:
    - Credit feedback to output gap appeared limited in historical sample.
    - Financial shocks played a limited role shaping the output gap historically; caveat: limited historical financial stress in sample.

### Conclusions and policy implications
- Model description:
  - Practical Quarterly Projection Model (QPM) with endogenous financial-real interactions, applied to the Philippines.
  - Extension of Guo et al. (2019) via credit block and macroprudential block focused on leverage-ratio targets.
  - Parameterization via calibration and Bayesian estimation prioritizing simplicity and tractability for routine forecasting and policy analysis.
- Key substantive findings:
  - Macroprudential measures can meaningfully affect credit dynamics and interact with monetary policy.
  - In the historical sample, the business cycle tends to lead the financial cycle; limited financial-to-real feedback observed (possibly due to absence of large crises).
  - Countercyclical macroprudential policy can mitigate the need for monetary policy adjustments and stabilize leverage and credit cycles.
  - Non-linearities are crucial to capture asymmetric bank responses to positive vs adverse profitability shocks and to design appropriate macroprudential reactions.
- Policy recommendations and uses:
  - Use model for joint forecasting and policy analysis of policy interest rate, inflation, output gap, exchange rate, credit growth, and macroprudential measures.
  - Calibrate macroprudential authority responses to the underlying shock type (different responses for excess credit booms vs adverse profitability shocks).
  - Incorporate sector-specific financial risks and evaluate targeted macroprudential measures (e.g., borrower-based tools like LTV caps, targeted exposure caps).
- Alignment with literature:
  - Results support consensus for countercyclical macroprudential policy to address endogenous financial sector procyclicality.
  - Monetary policy often an inefficient tool for financial-stability concerns; macroprudential policy can buffer financial shocks and reduce costly monetary interventions.

### Appendix C — calibrated coefficients (selected)
- Real GDP
  - 푏푏1: 0.6
  - 푏푏2: 0.2
  - 푏푏3: 0.6
  - 푏푏4: 0.7
  - 푏푏5: 0.5
  - 푏푏6: 0.25
  - 푐푐̅푠푠푠푠: 1
  - 푤푤푛푛: 0.425
  - 푤푤퐹퐹: 0.14
- Monetary policy: Interest rate
  - 푔푔1: 0.75
  - 푔푔2: 0.5
  - 푔푔3: 0.2
  - 휌휌휋∗: 0.8
  - 휋휋�푆푆푆푆∗: 2
  - 휌휌푑푑∗: 0.8
- Credit Cycle and Lending Rate (selected)
  - 휌휌푦푦�∗: 0.8
  - 휌휌푛푛푛푛푛푛�: 0.5
  - 휌휌푛푛̅∗: 0.9
  - 훼훼1푛푛푛푛푛푛�: 0.14
  - 훼훼1푛푛푛푛푛푛� (second listing): 0.12
  - 휌휌푑푑푑푑푣푣: 0.25
  - 휌휌푛푛푛푛푛푛�����: 0.75
  - 푐푐̅푠푠푠푠∗: 1
- Macroprudential Policy Block (selected)
  - 푡푡푡푡푤푤푝푝푤푤푤푤 푝푝: 30
  - 휌휌푝푝�: 0.75
  - 휌휌푏푏푛푛TTTTTT: 0.975
  - 푏푏푐푐푆푆푆푆 TTAA TT: 11
  - 휙휙: 0.5

### Identification, Bayesian estimation, and SVD findings (Appendix C)
- SVD of Fisher matrix indicates parameter groups identifiable together; parameters 휌휌1, 훼훼1, 훼훼2, 휌휌4, and 훼훼3 can be estimated jointly while others face weak identification.
- Weak identification noted specifically for 휌휌4 and 푏푏5.
- Bayesian estimation setup:
  - Algorithm: Metropolis-Hasting
  - Number of draws: 100 thousand
  - Sample period used for estimation: 2008–2019
  - Priors: persistency parameters (beta), standard deviations (inverse gamma), others (gaussian)
- Posterior/mode results (maximized posterior values)
  - alpha 1: 0.14 (Prior: Normal =0.2, =0.05)
  - alpha 2: 0.12 (Prior: Normal =0.1, =0.05)
  - rho 1: 0.52 (Prior: Beta =0.6, =0.2)
  - std of shock to credit gap: 1.02 (Prior: InvGamma =1.4 =0.3)
  - rho 4: 0.77 (Prior: Beta =0.75 =0.1)
  - alpha 3: 0.18 (Prior: Normal =0.3 =0.1)
  - std of shock to credit premium: 0.67 (Prior: InvGamma =1 =0.15)

*Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021256-print-pdf.pdf*

### Appendix D for details). Other parameters in the second group are calibrated to ensure

### Appendix D: Model calibration for credit and macroprudential instrument blocks

### Calibration approach and key assumptions
- The calibration of the steady-state level of the leverage ratio is ad-hoc; because U/KBs usually hold a buffer above the regulatory requirements to avoid a penalty and stigmatization, the target is set below the historical observations of the leverage ratio.
- The choice of persistence and responsiveness to the credit gap are selected to match the data.
- Standard deviations of shocks are generally calibrated, unless indicated as estimated. Calibration is done to obtain reasonable dynamics of gaps and trends when the calibrated model is applied to filter historical data.
- Other parameters in the second group are calibrated to ensure reasonable dynamic properties of the model.
- The rule for setting the leverage target is calibrated.

### Table IV.1: Model parameters related to credit and macroprudential instrument blocks
- Demand for credit – equation (3)
  - 휌휌1: 0.5 — Bayesian estimation
  - 훼훼1: 0.14 — Bayesian estimation
  - 훼훼2: 0.12 — Bayesian estimation
- Credit premium – equation (7)
  - 휌휌4: 0.8 — Bayesian estimation
  - 훼훼3: 0.2 — Bayesian estimation
  - 훼훼4: 0.001 — Calibration
- Credit growth trend – equation (4)
  - 휌휌2: 0.75 — Calibration
  - Δ푛푛푐푐푐푐�����푠푠푠푠: 0 — Calibration
- Real monetary condition index – equation (12)
  - 푏푏4: 0.7 — Calibration
  - 푏푏5: 0.5 — Calibration
- 1Y rate – equation (5)
  - 휌휌3: 0.5 — Calibration
  - 푏푏1: 0.6 — Calibration
- Aggregate demand
  - 푏푏2: 0.2 — Calibration
  - 푏푏3: 0.6 — Calibration
  - 푏푏푐푐푆푆푆푆푇푇퐴퐴푇푇: 11 — Calibration
  - 푏푏6: 0.25 — Calibration
- Target for the leverage ratio – equation (16)
  - 휌휌5: 0.975 — Calibration
  - 휙휙: (value not specified in provided excerpt)

### Notes on methodology
- Parameters shown as "Bayesian estimation" are estimated using Bayesian methods; other parameters are set by calibration.
- Calibration choices emphasize matching observed dynamics of credit and macro-financial variables, and ensuring model stability and realistic gap-trend behavior.

*Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021256-print-pdf.pdf*

### 0.5 Calibration

### 0.5 Calibration

### Model Properties—Analyzing IRFs
- Focus: impulse responses to shocks originating in the financial system; new credit and macro-prudential blocks do not alter impulse responses to common business cycle shocks in Guo et al. 2019, but they allow the model to track the response of credit.
- Lending rate specification: depends only on the expected path of the policy interest rate in the baseline; with leverage constraint, the lending rate (and real lending rate gap) incorporates additional effects beyond the policy rate.
- All shocks in impulse response analysis are 1 percent in magnitude and are unexpected.
- Identification and estimation notes:
  - 푏푏5 and 휌휌4 are difficult to distinguish from other parameters (Appendix D).
  - The Fisher matrix is a property of the model and is not related to the data.
  - When applied in practice, the rule for the leverage target may need recalibration to reflect current policymakers’ preferences rather than historical data.

### Credit demand shock
- Shock definition: positive shock to demand for credit measured as a shock to the credit gap (increase in demand for credit above trend controlling for business cycle and cost of credit).
- Macroprudential authority reaction: raises the required leverage ratio to moderate credit growth and safeguard bank soundness and resilience (Figure IV.1).
- Bank responses:
  - Banks accumulate capital to meet the higher required leverage ratio.
  - For a given level of bank capital, the leverage ratio drops below target when credit demand rises, prompting banks to boost capital by increasing lending margins (consistent with equation (17): 푓푓표표(푏푏푐푐푡푡 − 푏푏푐푐푡푡푇푇퐴퐴푇푇) > 0 under 푏푏푐푐푡푡 − 푏푏푐푐푡푡푇푇퐴퐴푇푇 < 0).
  - Banks cut dividend payments to zero (consistent with equation (18): 푓푓1(푏푏푐푐푡푡 − 푏푏푐푐푡푡푇푇퐴퐴푇푇) close to zero under 푏푏푐푐푡푡 − 푏푏푐푐푡푡푇푇퐴퐴푇푇 < 0).
  - Both higher margins and zero dividends raise capital to meet leverage target.
- Counterfactual (Figure IV.2):
  - Active countercyclical macroprudential policy: raising leverage ratio in response to the credit boom (solid lines).
  - Passive policy: keeping the ratio unchanged (dotted lines).
  - Results: raising the leverage ratio induces a larger increase in the lending rate compared to unchanged leverage ratio, mitigating the credit boom; monetary policy tightening is milder with active macroprudential policy.
- Additional notes:
  - Shock implementation: shock to the credit gap is implemented as a shock minus control; variables reported as deviations from steady state.
  - Clarification: adjustments to the required leverage ratio are hypothetical; BSP has not revised the 5 percent minimum Basel III leverage ratio or the 5 percent total capital to total assets ratio under Prompt Corrective Action since adoption.
  - Bank lending is demand determined; no supply constraint assumed.
  - Deviations of dividend payoffs from zero in simulation are very small (possibly numerical errors from non-linear simulation).
  - Leverage ratio target dynamics follow model equation (16) and can be modified by parameter calibration.
- Mechanism summary:
  - Leverage target stabilizing and countercyclical during demand booms by preventing significant decline of capital relative to assets.
  - Monetary policy reacts to inflation deviations and output gap (standard Taylor Rule formulation) and does not directly react to credit growth; higher credit growth affects output gap and inflation, indirectly triggering policy rate hikes.

### Bank profitability shocks
- Non-linearity in macroprudential block is key to capturing banks' behavior facing regulatory requirements.
- Negative shock to bank profits (e.g., increase in default rates) — Figure IV.3:
  - Regulatory authority cuts leverage ratio target to ease bank stress (reaction function in equation (16)).
  - Leverage ratio drops below the lowered target, triggering banks to withhold dividends and increase lending margin.
  - Higher lending rate reduces credit demand → slowdown in economic activity → negative output gap → inflation drops below target → monetary policy becomes accommodative.
  - Characteristic: policy and lending rates move in opposite directions in this shock.
- Counterfactual (Figure IV.4):
  - Keeping leverage ratio target unchanged results in higher stress: shallower drop in actual leverage ratio, more adverse effects on credit creation, and larger declines in the output gap compared to active policy.
  - Monetary policy accommodation of the adverse shock is correspondingly higher under active macroprudential policy.
- Positive shock to banks’ income (Figure IV.5):
  - Macroprudential policy may not need to react unless building additional capital buffers is desired.
  - Under higher income, banks may pay dividends; leverage ratio increases above target.
  - Under excess capital (푏푏푐푐푡푡 − 푏푏푐푐푡푡푇푇퐴퐴푇푇 > 0), spread function 푓푓표표(푏푏푐푐푡푡 − 푏푏푐푐푡푡푇푇퐴퐴푇푇) = 0 and dividend payoff function 푓푓1(푏푏푐푐푡푡 − 푏푏푐푐푡푡푇푇퐴퐴푇푇) > 0.
  - Lending rate does not change; output gap is zero; inflation stays at target.
- Simulation conclusions:
  - Macroprudential authority must react based on the underlying shock type; responses differ for excessive credit boom vs adverse bank profitability shock.
  - Countercyclical macroprudential policy could mitigate the need for monetary policy reactions in cases of excessive credit booms and drops in bank profitability.
  - Non-linearities produce asymmetric impulse responses that help regulatory authorities distinguish positive vs adverse profitability shocks.
  - These conclusions align with literature calling for countercyclical macroprudential policy to address endogenous financial sector procyclicality and risk-taking.

### Forecast evaluation (in-sample) — 2002Q1–2019Q4
- Forecast setup:
  - Start from 2001Q4 and run pure model-based forecasts for subsequent six quarters, rolling forward one quarter until 2019Q4.
  - Forecasts assume known: (i) estimates of unobserved variables for current quarter (based on entire sample information); (ii) all external variables over forecasting horizon.
  - Because macroprudential authority did not set leverage ratio historically, assume spread, leverage ratio, and ROAs are observed. Test forecasting performance of credit block only, not macroprudential policy block.
- RMSE comparison: model-based forecasts versus random walk (Table V.1). Note: values < 1 mean model RMSE lower than random walk; > 1 the opposite; 1 means equal RMSEs.
  - Business Cycle Block (2002-2019): Inflation, y-o-y: 0.5, 0.6, 0.7, 0.8, 0.9, 0.9 (t+1 to t+6); Policy interest rate: 2.1, 2.1, 2.0, 2.0, 1.9, 1.9; Real GDP growth, y-o-y: 0.5, 0.5, 0.5, 0.5, 0.6, 0.7; Nom. exchange rate depreciation, y-o-y: 0.6, 0.6, 0.5, 0.5, 0.5, 0.5; Output gap: 0.9, 0.9, 0.8, 0.8, 0.8, 0.9.
  - Business Cycle Block (2010-2019): Inflation, y-o-y: 0.6, 0.7, 0.8, 0.8, 0.7, 0.6; Policy interest rate: 1.5, 1.4, 1.3, 1.2, 1.0, 0.8; Real GDP growth, y-o-y: 0.5, 0.4, 0.4, 0.6, 0.7, 0.8; Nom. exchange rate depreciation, y-o-y: 0.5, 0.4, 0.4, 0.5, 0.5, 0.6; Output gap: 1.0, 0.9, 0.8, 0.7, 0.8, 0.9.
  - Credit block (2002-2019): Credit flow/GDP gap: 1.0, 1.0, 1.0, 1.1, 1.3, 1.3; Outstanding credit/GDP: 0.8, 0.8, 0.7, 0.7, 0.7, 0.6; Lending rate: 1.5, 1.4, 1.5, 1.5, 1.5, 1.5; Credit premium: 0.8, 0.7, 0.7, 0.7, 0.7, 0.7.
  - Credit block (2010-2019): Credit flow/GDP gap: 1.1, 1.0, 1.1, 1.1, 1.0, 1.1; Outstanding credit/GDP: 0.6, 0.6, 0.6, 0.6, 0.6, 0.5; Lending rate: 2.7, 1.1, 1.7, 1.4, 1.3, 1.2; Credit premium: 1.1, 1.0, 1.0, 1.0, 1.0, 0.9.
- Interpretation:
  - Model outperforms random walk forecasts for business cycle variables overall, with notable exception of the policy rate.
  - Evidence on credit cycle forecasting is mixed.
  - Quarter-by-quarter comparisons indicate: model fails to predict policy rate evolution (improves post-2010); matches credit premium well; falls short on lending rate; mixed results for credit gap.
- Policy rate forecasting issues:
  - In-sample policy rate forecast is more volatile than actual (Figure V.1).
  - Policy rule (Appendix A, equation B.13) implied deep negative interest rate recommendation during the global financial crisis and counterfactual drops in 2015-2016, not reflected in actual data.
  - Actual policy rate appears less responsive to inflation and output gap than assumed in the model.
  - Forecasts of policy rate seem to improve after 2010.
- Lending rate and credit premium:
  - Lending rate equals sum of 1Y benchmark rate, credit premium, and regulatory margin (equation (7)); term premium and margin assumed fixed.
  - Forecast bias in lending rate primarily driven by errors in policy rate forecast; credit premium forecast generally reasonable.
  - Errors in policy rate forecasts cause errors in lending rate forecasts, which impair credit gap forecasts.

### Historical interpretation using the QPM (Kalman filter)
- Purpose: use calibrated QPM and Kalman filter to estimate trends and structural shocks that best fit historical Philippine data and interpret historical developments.
- Observed data groups informing filtration:
  1. Business cycle block (headline CPI including core and food-and-energy, real GDP, GDP deflator, shares of exports and imports in GDP, nominal exchange rate, policy interest rate; foreign variables: global CPI, GDP, oil and food commodity prices, U.S. interest rates).
  2. Credit cycle and macroprudential policy blocks (outstanding bank credit to firms, lending rate, VIX).
  3. Bank-related data (total assets of banks, ROAs, total banks’ capital).
- Kalman filter application notes:
  - Linear Kalman filter used despite non-linear model blocks because most solution software uses first-order approximations and non-linear filters are computationally demanding and often require stationarity.
  - Use of expert tunes to address critical nonlinearities (dividends and spread) by restricting dividends and bank spread values to be non-negative; two-run approach:
    - First run: no expert restrictions.
    - Second run: dividends and spread set to zero in periods where found negative in first run.
  - Assumptions: measurement errors assumed not to exist; variables not de-trended but seasonally adjusted where applicable; VIX de-meaned; observations of 1Y benchmark rate not available (term premium assumed constant).
- Kalman filtration findings:
  - Credit gap history:
    - Credit flow and trend estimated; positive credit gap opened in 2016, peaking at about 0.5 percent in 2018 (smoothed credit gap, black line). Positive credit gap closed rapidly in 2019.
    - Model-based filtration suggests credit gap relatively moderate in peaks; credit did not significantly deviate from equilibrium trend and did not significantly threaten financial stability.
  - Credit gap decomposition (equation (3), Figure VI.3):
    - Real lending rate has been an important determinant (orange bars).
    - Credit shocks (purple bars) pronounced and sometimes offset real lending rate gap effects (e.g., 2008, 2015, 2018).
    - Output gap (red bars) has somewhat lower impact on credit gap compared with real lending rate or credit demand shock.
  - Structural shock contributions to credit gap (Figure VI.4):
    - Domestic and foreign business cycle shocks and monetary policy shocks played important roles.
    - Domestic developments were main factor closing positive credit gap in 2018-2019; monetary policy and exchange rate developments helped smooth credit cycle by partly offsetting domestic demand and supply shocks since 2016.
  - Lending rate decomposition (Figure VI.5):
    - Lending rate driven by 1Y rate, term premium (constant), credit premium, and spread.
    - Term premium contribution constant by model assumption.
    - Spread contributions small, mainly in first half of filtration period.
  - Credit premium decomposition (Figure VI.6):
    - Credit premium close to neutral level in 2019.
    - Determinants: expected future business cycle developments (red bars), world financial conditions approximated by VIX (orange bars), and structural shocks capturing other factors (purple bars).
    - Credit premium assumed persistent in the model (equation 8); lag term substituted out in decomposition.
  - Real lending rate and natural rate (Figures VI.7–VI.8):
    - Real lending rate fluctuated around 3 percent since 2008.
    - Equilibrium level equals natural real interest rate adjusted for constant term premium; strong correlation between policy and lending rate in gap terms.
  - Banking sector variables:
    - Actual data on bank balance sheet, capital, and ROAs available from 2008Q2 onward; prior periods estimated by filter.
    - Leverage ratio fluctuated around 12 percent since 2011, with a significant spike in 2012-2013.
    - Implicit target of 11 percent assumed in steady state calibration, ad-hoc and below historical average observed.
    - Banks’ lending spread triggered by regulatory requirements estimated close to zero on average (Figure VI.11); dividend payments positive (Figure VI.10); both spread and dividend payout ratios unobserved in data.
    - ROA relatively stable at around 1.5 percent (Figure VI.12); predicted ROA from filtration hovers around average ROA in data.
  - Financial sector feedback to business cycle:
    - Credit feedback to output gap (financial sector feedback) does not appear to have affected the gap significantly in historical sample (purple bars in Figure VI.13).
    - Decomposition of output gap to structural shocks (Figure VI.14) similarly shows financial shocks played limited role shaping output gap historically.
    - Caveat: limited historical financial stress or crisis in sample may explain limited feedback observed.

### Conclusions and policy implications
- Model summary:
  - Practical Quarterly Projection Model incorporating endogenous interactions between financial sector and macroeconomy, applied to the Philippines.
  - Extends Guo et al. (2019) by adding a credit block and a macroprudential block focused on leverage-ratio targets.
  - Parameterization via calibration and Bayesian estimation; emphasis on relative simplicity and tractability for routine forecasting and policy analysis.
- Key substantive findings:
  - Impulse response functions show macroprudential measures can meaningfully affect credit dynamics and interact with monetary policy.
  - In historical sample, business cycle seems to lead financial cycle in the Philippines; limited feedback from financial shocks to broader economy observed (possible reason: absence of large crises in sample).
  - Countercyclical macroprudential policy can mitigate the need for monetary policy adjustments and can stabilize leverage and credit cycles.
  - Non-linearities are important to capture asymmetric bank responses to positive vs adverse profitability shocks and to guide appropriate macroprudential responses.
- Policy recommendations and uses:
  - Model useful for joint forecasting and policy analysis of policy interest rate, inflation, output gap, exchange rate, credit growth, and macroprudential measures.
  - Model can help identify underlying shocks and guide macroprudential authority’s calibrated responses (different responses needed for excess credit booms vs adverse profitability shocks).
  - Flexible structure allows incorporation of sector-specific financial risks (e.g., growing real estate credit, unhedged corporate foreign-currency borrowing, house prices, long-term household debt imbalances, FX exposure) and evaluation of targeted macroprudential measures (e.g., borrower-based tools like LTV caps, targeted exposure caps).
- Literature alignment:
  - Conclusions support consensus calling for countercyclical macroprudential policy to address endogenous financial sector procyclicality and risk-taking.
  - Considerable literature suggests monetary policy is often an inefficient and costly tool to address financial stability concerns; macroprudential policy can buffer financial shocks and reduce macroeconomic disequilibria that would otherwise require larger monetary policy reactions.

*Source: Authors’ computations and analyses in the supplied PDF content.*

### REFERENCES

### wpiea2021256-print-pdf - REFERENCES

### References cited
- Acosta-Ormaechea, S. and L. Martinez, 2021, “A Guide and Tool for Projecting Public Debt and Fiscal Adjustment Paths with Local- and Foreign-Currency Debt”, Institute for Capacity Development, IMF Technical Notes and Manuals 2021/005.
- Adler, G., C. Casas, L. Cubeddu, G. Gopinath, N. Li, S. Meleshchuk, C. Osorio Buitron, D. Puy, and Y. Timmer, 2020, “Dominant Currencies and External Adjustment,” IMF Staff Discussion Note, SDN 20/05.
- Adrian, T., and N. Liang, 2018, “Monetary Policy, Financial Conditions, and Financial Stability.” International Journal of Central Banking 14(1): 73-131.
- Adrian, T., C. Erceg, J. Linde, P. Zabczyk, and J. Zhou, 2020a, “Quantitative Model of the Integrated Policy Framework.” IMF Working Paper 20/122.
- Adrian, T., 2020b, ““Low for Long” and Risk-Taking,” IMF-MCM Departmental Paper Note 20/15.
- Adrian, T., F. Duarte, N. Liang, and P. Zabczyk, 2020c, “NKV: A New Keynesian Model with Vulnerability.” AEA Papers and Proceedings 110: 470–76.
- Aikman, D., A.G. Haldane, and B.D. Nelson, 2015, “Curbing the Credit Cycle,” Economic Journal, 125(585), 1072–109.
- Alam, Z., A. Alter, J. Eiseman, G. Gelos, H. Kang, M. Narita, E. Nier, and N. Wang, 2019, “Digging Deeper—Evidence on the Effects of Macroprudential Policies from a New Database,” IMF Working Paper 19/66.
- Andrle, M., 2010, “A Note on Identification Patterns in DSGE Models,” ECB Working Paper no. 1235.
- Basel Committee on Banking Supervision (BCBS), 2010, Guidance for National Authorities Operating the Countercyclical Capital Buffer.
- Baksa, D., A. Bulíř, and D. Heng, 2020, “A Simple Macrofiscal Model for Policy Analysis: An Application to Cambodia,” IMF Working Paper 20/202.
- Baksa, D., A. Bulíř, and R. Cardarelli, 2021, “A Simple Macrofiscal Model for Policy Analysis: An Application to Morocco,” IMF Working Paper 21/190.
- Bulíř, A., D. Baksa, J.S. Corrales, A. González-Gomez, D. Heng, and D. Rodríguez-Guzman, “Using Macroeconomic Frameworks to Analyze the Impact of COVID-19: An Application to Colombia and Cambodia,” Institute for Capacity Development, IMF Technical Notes and Manuals 2021/001.
- Bernanke, B.S., Gertler, M. and Gilchrist, S., 1999, “The Financial Accelerator in a Quantitative Business Cycle Framework”, in: John B. Taylor and Michael Woodford, eds., Handbook of Macroeconomics, Volume 1C.
- Brandao-Marques, L., G. Gelos, M. Narita, and E. Nier, 2020, “Leaning Against the Wind: A Cost-Benefit Analysis for an Integrated Policy Framework.” IMF Working Paper 20/123.
- Byggs, M. and T. Mayer, 2013, “Bring Credit Back into the Monetary Policy Framework,” PEFM Policy Brief.
- Chen, J., D. Finocchiaro, J. Lindé, and C.R. Nunes, 2020, “The Costs of Macroprudential Deleveraging in a Liquidity Trap,” IMF Working Paper 20/89.
- Guo, S., P. Karam, and J. Vlcek, 2019, “Decomposing the Inflation Dynamics in the Philippines,” IMF Working Paper 19/153.
- International Monetary Fund, 2015, “Monetary Policy and Financial Stability.” IMF Policy Paper.
- International Monetary Fund, 2020, “Towards an Integrated Policy Framework,” (month?).
- Juillard, M., C. Freedman, D. Korshunov, D. Laxton, O. Kamenik, I. Carabenciov, I. Ermolaev, and J. Laxton, 2008, “A Small Quarterly Multi-Country Projection Model with Financial-Real Linkages and Oil Prices,” IMF Working Paper 08/280.
- Kalemli-Ozcan, S., 2019, “U.S. Monetary Policy and International Risk Spillovers,” Jackson Hole Symposium Proceedings.
- Kim, M., 2019, “Financial Development, Exchange Rate Fluctuations and Debt Dollarization: A Firm-Level Evidence,” IMF Working Paper 19/168.
- Nier, E., T. T. Olafsson, and Y. G. Rollinson, 2020, “Exchange Rates and Domestic Credit—Can Macroprudential Policy Reduce the Link?” IMF Working Paper 20/187.
- Simon, D., 2009, “Kalman Filtering with State Constraints: A Survey of Linear and Nonlinear Algorithms,” IET Control Theory and Applications.
- Svensson, L.E.O., 2017, “Cost-Benefit Analysis of Leaning Against the Wind,” Journal of Monetary Economics 90: 193-213.
- Van den Hof, P.M.J., J.F.M. Van Doren, and S.G. Douma, 2009, “Identification of Parameters in Large Scale Physical Model Structures, for the Purpose of Model-Based Operations,” In: Hof P., Scherer C., Heuberger P. (eds) Model-Based Control. Springer, Boston, MA.
- Vlcek, J., M. Pranovich, P. Hitayezu, B. Mwenese, and C. Nyalihama, 2020, “Quarterly Projection Model for the National Bank of Rwanda,” IMF Working Paper 20/295.

### Appendix A — Model variables (inventory)
- Core macro variables:
  - 푦푦: Real GDP, 100*ln
  - 푦푦�: Output gap, percent
  - 푦푦�: Potential GDP, 100*ln
  - Δ푦푦: Real GDP growth, annualized q-o-q, percent
  - 훥훥4푦푦: Real GDP growth, y-o-y, percent
  - Δ푦푦�: Real GDP Trend growth, annualized q-o-q, percent
  - 훥훥4푦푦�: Real GDP Trend growth, y-o-y, percent
  - 퐸퐸푡푡푦푦�푡푡+1: Expected output gap at t+1, percent
- Inflation and price components:
  - 푐푐푝푝푣푣: Headline CPI, 100*ln
  - 휋휋: Headline CPI inflation, annualized q-o-q, percent
  - 휋휋4: Headline CPI inflation, y-o-y, percent
  - 푐푐푝푝푣푣퐶퐶: Core CPI, 100*ln; 휋휋퐶퐶: Core CPI inflation (annualized q-o-q)
  - Subcomponents: food (퐹퐹) and energy (푇푇) CPI levels and inflation measures and their real marginal costs (푐푐푚푚푐푐…)
- Exchange rates and external:
  - 푠푠: Nominal exchange rate FRW to USD, 100*ln
  - Δ푠푠: Nominal ER depreciation, annualized q-o-q, percent; 훥훥4푠푠: y-o-y percent
  - 푧푧: Real exchange rate, 100*ln; 푧푧̅: Equilibrium real exchange rate, 100*ln; 푧푧̂: Real effective exchange rate gap, percent
- Interest rates and policy:
  - 푣푣푇푇푀푀: Nominal policy interest rate, percent p.a.
  - 푐푐: Real interest rate, percent
  - 푐푐̅: Equilibrium real interest rate, percent
  - 푐푐̂: Real interest rate gap, percent
  - 푐푐푠푠1푦푦: 1-year benchmark interest rate, percent p.a.; 푐푐푠푠퐿퐿: Lending interest rate, percent p.a.
- Financial sector and macroprudential:
  - 푛푛푐푐푐푐: Ratio of nominal credit to nominal GDP, percent
  - 푏푏푐푐: Leverage ratio: total banks’ capital over total assets, percent
  - 푏푏푐푐푇푇퐴퐴푇푇: Target leverage ratio, percent
  - 푡푡푡푡: Macroprudential policy instrument indicator
  - 푑푑푣푣: Banks’ dividends: ratio of total dividends to assets, percent
  - 푠푠푝푝: Spread (margin) banks include in lending rate to accelerate capital accumulation, percent
- External prices:
  - 푝푝푤푤표표…: World food price index, world oil price index, levels and inflation measures
- US / external variables:
  - 푦푦�∗: US output gap, percent
  - 푣푣∗: US nominal interest rate, percent p.a.
  - 푐푐∗: US real interest rate, percent p.a.
  - 휋휋∗: US CPI and inflation measures
- Expectations and shocks:
  - 퐸퐸 expectations operators for inflation and subcomponents (e.g., 퐸퐸푡푡휋휋푡푡+1퐶퐶)
  - Extensive set of shocks with notation such as 휀휀 with subscripts indicating variable and process (e.g., 휀휀푡푡푦푦� for shock to output gap)

### Appendix B — Model equations (structure and key blocks)
- Scope and notation:
  - Only dynamic equations and definitions for trends and gaps are reported; all variables are in natural logs unless noted. Δ denotes q-o-q annualized growth rates; Δ4 denotes y-o-y growth rates. Shocks have zero mean and are not autocorrelated.
- Real GDP block:
  - Identity: 푦푦푡푡 = 푦푦�푡푡 + 푦푦�푡푡 (B.1)
  - Trend dynamics: 푦푦�푡푡 = 푏푏1푦푦�푡푡−1 − 푏푏2푚푚푐푐푣푣푡푡 + 푏푏3푦푦�푡푡∗ + 푏푏6휀휀푡푡−1푛푛푛푛푛푛� + 휖휖푡푡푦푦� (B.2)
  - Growth process for trend growth Δ푦푦� (B.3)
- Real monetary conditions index:
  - Definition combining policy and exchange rate gap: 푚푚푐푐푣푣푡푡 = 푏푏4(푏푏5푐푐푐푐�푡푡 + (1−푏푏5)푐푐푐푐�퐿퐿) + (1−푏푏4)(−푧푧̂푡푡) (B.4)
- Inflation block:
  - Headline constructed from weighted subcomponents (B.5)
  - Core inflation dynamics: 휋휋푡푡퐶퐶 = 푡푡11퐸퐸푡푡휋휋푡푡+1퐶퐶 + (1−푡푡11)휋휋푡푡−1퐶퐶 + 푡푡21푐푐푚푚푐푐푡푡퐶퐶 + 휀휀푡푡휋휋퐶퐶 (B.6)
  - Subcomponent marginal cost specifications (B.7), and analogous equations for food (B.8–B.9) and energy (B.10–B.11).
- Exchange rate:
  - Dynamics include persistence, expected future value, and risk premia/country risk terms (B.12).
- Monetary policy rule:
  - Nominal policy rate evolution with gradualism and response to expected inflation and output gap: 푣푣푡푡푇푇푀푀 = 푔푔1푣푣푡푡−1푇푇푀푀 + (1−푔푔1)�푣푣푡푡푁푁푝푝푁푁… + 푔푔2퐸퐸푡푡휋휋4푡푡+3퐷퐷푝푝푑푑 + 푔푔3푦푦�푡푡� + 휀휀푡푡푑푑푀푀푀푀 (B.13)
  - Link between policy rate, neutral rate, and inflation expectations (B.14–B.16)
- Real exchange rate and equilibrium dynamics:
  - Real exchange rate identity: 푧푧푡푡 = 푠푠푡푡 + 푐푐푝푝푣푣푡푡∗ − 푐푐푝푝푣푣푡푡 (B.19)
  - Decomposition into equilibrium and gap: 푧푧푡푡 = 푧푧̅푡푡 + 푧푧̂푡푡 (B.20–B.21)
- UIP and external sector:
  - Equilibrium UIP specification for country risk premium and external real interest differentials (B.22)
  - External blocks for US inflation, output gap, and interest rates (B.64–B.69)
- Relative prices and trends:
  - Relative price definitions and AR processes for equilibrium relative price growth (B.23–B.31)
- Deflator and nominal GDP:
  - GDP deflator AR(1) and relationship to headline inflation (B.32–B.34)
- Credit cycle, lending rate, and macroprudential block:
  - Credit-to-GDP decomposition and AR dynamics for credit gap and flow-to-GDP (B.34–B.37)
  - Benchmark 1-year rate and lending rate relations including credit premium and spreads (B.37–B.41)
  - Macroprudential instrument and leverage dynamics, including equations for target leverage, capital accumulation, spreads, and dividend adjustments (B.43–B.48)
- Trade and external prices:
  - Exports and imports decomposed into cyclical and equilibrium components (B.49–B.55)
  - World oil and food price relative price processes and AR specifications for cyclical and trend components (B.56–B.63)
- External sector AR processes and shocks:
  - Processes for external inflation, output, interest rates, and volatility (B.64–B.70)
- Shocks:
  - Comprehensive shock notation covers output, trends, inflation subcomponents, exchange rate, monetary policy, US shocks, long-term real exchange rate trends, country risk premium, oil and food cyclical and trend shocks, credit demand and trend shocks, benchmark rate shocks, credit premium shocks, bank balance sheet shocks, capital accumulation shocks, margin and dividend shocks, VIX shocks, export/import equilibrium and gap shocks, and measurement errors.

*Content derived from the REFERENCES, APPENDIX A, and APPENDIX B sections of wpiea2021256-print-pdf.*

### APPENDIX C. Calibration of the Model Coefficients

### APPENDIX C. Calibration of the Model Coefficients

### Calibrated coefficients (by block)
- Real GDP
  - 푏푏1: 0.6
  - 푏푏2: 0.2
  - 푏푏3: 0.6
  - 푏푏4: 0.7
  - 푏푏5: 0.5
  - 푏푏6: 0.25
  - 푐푐̅푠푠푠푠: 1
  - 푤푤푛푛: 0.425
  - 푤푤퐹퐹: 0.14

- Real interest rate
  - 휌휌푛푛̅: 0.8
  - 휌휌Δ푧푧̅: 0.8
  - Δ푧푧̅푠푠푠푠: -1
  - 휌휌Δ푦푦�: 0.8
  - Δ푦푦�푠푠푠푠: 5
  - 훼훼푝푝푛푛푝푝푝푝: 0.5
  - 훼훼푝푝푧푧: 0.1
  - 훼훼푝푝푦푦: 0.4
  - 훼훼푝푝표표푑푑푁푁: 0.005

- Real exchange rate
  - 휌휌푝푝푛푛푝푝푝푝: 0.5
  - 푤푤푛푛: 0.425 (also listed under Real GDP)
  - 휌휌Δ푝푝푦푦: 0.25
  - 휌휌휀Δ푝푝푦푦: 0.85

- Equilibrium version of UIP
  - 휌휌푝푝: 0.5
  - 푝푝1: 0.45
  - 훼훼푝푝푛푛푝푝푝푝퐶퐶: 0.2
  - 훽훽푝푝푛푛푝푝푝푝퐶퐶: 0.001

- Inflation Deflator and Nominal GDP
  - 휌휌Δ푝푝푦푦: 0.25
  - 휌휌휀Δ푝푝푦푦: 0.85

- Inflation – Core inflation (relative prices)
  - 푡푡11: 0.5
  - 푡푡21: 0.4
  - 푡푡31: 0.8
  - 푡푡12: 0.5
  - 푡푡22: 0.2
  - 푡푡32: 0.25
  - 휌휌푛푛푝푝����퐶퐶: 0.75
  - Δ푐푐푝푝����푠푠푠푠퐶퐶: -0.75
  - 휌휌푛푛푝푝����퐹퐹: 0.75
  - Δ푐푐푝푝����푠푠푠푠퐹퐹: 0.75

- Inflation – Energy price inflation
  - 훥훥푐푐푝푝���푠푠푠푠푤푤표표푑푑푁푁: 5
  - 훥훥푐푐푝푝���푠푠푠푠푤푤푤푤 표표표표푤푤: 0

- World oil price
  - 휌휌푛푛푝푝����푤푤표표표표표표: 0.7
  - 휌휌훥훥푛푛푝푝����푤푤표표표표표표: 0.85

- World food price
  - 휌휌푛푛푝푝����푤푤푓푓표표표표푓푓: 0.7
  - 휌휌훥훥푛푛푝푝����푤푤표표표표표표: 0.85

- Exchange rate
  - 훥훥푐푐푝푝���푠푠푠푠푤푤푤푤 표표표표푤푤: 0

- Monetary policy: Interest rate
  - 푔푔1: 0.75
  - 푔푔2: 0.5
  - 푔푔3: 0.2
  - 휌휌휋∗: 0.8
  - 휋휋�푆푆푆푆∗: 2
  - 휌휌푑푑∗: 0.8

- Credit Cycle and Lending Rate
  - 휌휌푦푦�∗: 0.8
  - 휌휌푛푛푛푛푛푛�: 0.5
  - 휌휌푛푛̅∗: 0.9
  - 훼훼1푛푛푛푛푛푛�: 0.14
  - 훼훼1푛푛푛푛푛푛� (second listing): 0.12
  - 휌휌푑푑푑푑푣푣: 0.25
  - 휌휌푛푛푛푛푛푛�����: 0.75
  - 푐푐̅푠푠푠푠∗: 1

- Exports and Imports
  - Δ푛푛푐푐푐푐�����푠푠푠푠: 0
  - 휌휌푣푣̅: 0.85
  - 휌휌푛푛푠푠1푦푦: 0.75
  - 푥푥̅푠푠푠푠: 35
  - 푝푝푐푐 푝푝푚푚 TTTTTT: 2.75
  - 휌휌푝푝�: 0.85
  - 휌휌푝푝푛푛푝푝푝푝퐶퐶: 0.8
  - 푚푚�푠푠푠푠: 40
  - 훼훼푝푝푛푛푝푝푝푝퐶퐶: 0.2
  - 휌휌푣푣�: 0.5
  - 훼훼푣푣푧푧: 0.25
  - 훼훼푣푣푦푦: 0.6

- Macroprudential Policy Block
  - 푡푡푡푡푤푤푝푝푤푤푤푤 푝푝: 30
  - 휌휌푝푝�: 0.75
  - 휌휌푏푏푛푛TTTTTT: 0.975
  - 푏푏푐푐푆푆푆푆 TTAA TT: 11
  - 휙휙: 0.5

### Key statistics called out in the appendix text
- The SVD decomposition of the Fisher matrix is presented as Table A.1 showing normalized base vectors for groups of parameters.
- The analysis suggests that parameters 휌휌1, 훼훼1, 훼훼2, 휌휌4, and 훼훼3 can be estimated together while other parameters may face weak identification.
- The text specifically notes weak identification in particular for 휌휌4 and 푏푏5 parameters.

### Identification (SVD of Fisher matrix) — summary of Table A.1 findings
- Table A.1 lists normalized base vectors (columns 1–12) for parameters 푏푏1 through 훼훼4.
- Column structure indicates groups of parameters that can be identified together; values closer to one indicate stronger identification, values closer to zero indicate weaker identification.
- The selected largest identifiable group (column 10 in Table A.1) covers most parameters in the new credit and bank blocks; an alternative (column 8) was not selected because it includes 푏푏6, which affects the output gap equation and feedback from credit to business cycles.

### Bayesian estimation — setup and outcomes
- Estimation method and sample
  - Algorithm: Metropolis-Hasting
  - Number of draws: 100 thousand
  - Sample period used: 2008-2019 (the whole sample was not used)
  - Rationale: dynamics of financial variables appear to have changed after the world financial crisis

- Priors and distributions
  - Persistency parameters: beta distribution
  - Standard deviations of shocks: inverse gamma distribution
  - Other parameters: gaussian distribution

- Posterior / maximized posterior values (from Figures A.1 and A.2)
  - alpha 1 (Prior: Normal =0.2, =0.05): Maximized Poster: 0.14
  - alpha 2 (Prior: Normal =0.1, =0.05): Maximized Poster: 0.12
  - rho 1 (Prior: Beta =0.6, =0.2): Maximized Poster: 0.52
  - std of shock to credit gap (Prior: InvGamma =1.4 =0.3): Maximized Poster: 1.02
  - rho 4 (Prior: Beta =0.75 =0.1): Maximized Poster: 0.77
  - alpha 3 (Prior: Normal =0.3 =0.1): Maximized Poster: 0.18
  - std of shock to credit premium (Prior: InvGamma =1 =0.15): Maximized Poster: 0.67

- Figure legend note (as in source figures)
  - Blue dotted line – prior
  - Yellow line – posterior distribution
  - Red line with circle – mode of the posterior
  - Magenta with triangles – parameter bounds

*Source: APPENDIX C and APPENDIX D, authors’ computations in the supplied PDF content.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021256-print-pdf.pdf_
