## Annex IV. Schematic Description of Model

## Source details

**Canonical URL:** [Annex IV. Schematic Description of Model](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021010-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021010-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021010-print-pdf.pdf.json)

---

### Objective and context
- Focus: develop a framework to understand the role of the banking sector as a transmission channel for external shocks in small open (Asian) economies and to assess macro-financial policy responses.
- Motivation: large and volatile cross-country capital flows (international financial cycle) can produce surges and sudden stops with macrofinancial consequences for emerging market economies (EMs), especially in Asia.
- Key features of Asian banking systems highlighted:
  - Banks are the dominant source of financing for investment and consumption.
  - Gross foreign assets and liabilities are sizeable.
  - Several Asian EMs maintain large net foreign asset positions; foreign assets tend to be more volatile than liabilities and both are more volatile in Asian EMs than in advanced economies.
  - Home deposits in domestic currency; foreign borrowing and many liabilities denominated in foreign currency; banks hold FX-denominated safe assets (e.g., US Treasuries).

### Empirical strategy (Panel VAR) — specification and data
- Method: panel vector autoregression (panel VAR) following Abrigo and Love (2015) using system-GMM.
- Sample (quarterly): 2002Q1 to 2019Q4 for ten Asian economies:
  - Advanced: Australia, Korea, Japan, New Zealand.
  - Emerging: China, India, Indonesia, Malaysia, the Philippines, Thailand.
- Monthly robustness check: 2001M12 to 2020M2 (uses IIP in lieu of investment; specification has a 12-period lag).
- Model specification:
  - Y_it = Y_it−1 A1 + Y_it−2 A2 + ⋯ + Y_it−T A_T + X_it B + u_it + e_it
  - Endogenous vector Y_it: growth in foreign assets, growth in foreign liabilities, change in the real exchange rate, growth in credit to the private sector, growth in real investment.
  - Exogenous X_it: US GDP growth, inflation, terms of trade, external financing shock.
- Lag length: T = 2 chosen by MBIC, MAIC, MQIC criteria.
- Identification: Cholesky ordering — real exchange rate, foreign liabilities, foreign assets, credit, investment.
- External financial shock proxy: exogenous US monetary policy shock (residual of one-year government bond rate regression with 30-min changes of one-month ahead Fed funds futures around FOMC announcements). Alternatives tested: excess bond premium, VIX, US federal funds rate.

### Empirical findings — role of bank balance sheets and transmission channels
- Three headline findings from the panel VAR:
  1. The impact on credit and investment through bank balance sheets from an external financial shock is more pronounced than from a real shock.
  2. Banks (particularly in EMs) appear to have an incentive to hold foreign assets as safe assets; foreign assets and liabilities impact credit independently.
  3. The real exchange rate plays a key role in the ultimate impact of bank funding shocks on credit; the exchange rate channel is stronger in EM Asia than in advanced economies.
- Specific empirical responses:
  - External financial shock (tightening): reduces foreign liabilities and assets, decreases bank credit supply, lowers investment for both advanced and emerging Asia.
  - External demand shock (higher US growth): increases foreign liabilities and assets, but leads to a small depreciation in the real exchange rate (given US dollar appreciation), producing a smaller net increase in credit than in the financial-shock scenario.
  - Positive shock to foreign liabilities (capital inflow): raises foreign assets, leads to real exchange rate appreciation, and increases credit and investment; effect on exchange rate larger in advanced than emerging Asia.
  - Positive shock to foreign assets: evidence of a lagged increase in foreign liabilities followed by credit and investment, especially in EM subsample — consistent with banks holding foreign “safe” assets to attract foreign funding and relax constraints.
  - Real exchange rate appreciation (controlling for bank balance sheet): affects credit but with wider confidence intervals on foreign assets and liabilities.
- Variance decomposition:
  - Banking financial shocks (foreign liabilities/assets) and external financing shocks account for a larger share of forecast error variance for foreign assets/liabilities, credit, and investment than real shocks (US growth, terms of trade).
- Robustness:
  - Monthly VAR confirms directional findings for VIX and FFR shocks (with larger error bands for real activity measures).
  - Including nonfinancial corporate FX borrowing: bank balance sheet channels remain significant under capital inflow shocks; corporate FX debt has larger standard errors.

### Model structure (canonical New-Keynesian small open economy with banking)
- Base: extends Gertler and Karadi (2011) and Aoki, Benigno and Kiyotaki (2018) framework to include:
  - Banking sector that accesses domestic and foreign funds.
  - Banks hold FX-denominated assets as safe assets.
  - Traded goods priced in foreign currency (U.S. dollar invoicing).
  - Non-tradable goods prices with Rotemberg adjustment costs.
  - Policy instruments: monetary policy interest rate, foreign exchange intervention (FXI), macroprudential measures (MPMs), capital flow measures (CFMs).
- Agents and key mechanisms:
  - Representative household with Greenwood-Hercowitz-Huffman preferences and external habit; substitution between traded and non-traded goods governed by st and α_T.
  - Producers: monopolistic competition for intermediate non-traded goods, Cobb-Douglas production, Rotemberg price adjustment, capital accumulation with adjustment cost Φ(·).
  - Banks (bankers):
    - Bank balance sheet: s_t b*_jt + b_jt + Q_t^K k_jt = d_jt + s_t d*_jt + n_t.
    - Participation constraint (moral hazard): investors’ belief γ(x) = θ_0 exp(−θ_1 x) where x = s_t b*_jt / n_t; holding foreign safe assets relaxes funding constraints.
    - Banker value linear in net worth: V(n_jt) = ψ_t n_jt; participation multiplier λ_t appears in first-order conditions.
  - Foreign investor: representative risk-neutral foreign investor supplies foreign credit with quadratic cost 1/(2χ_d,t) (D*_t)^2, yielding D*_t = χ_d,t (R*_t − R_t^{B*}). Two capital flow types: passive (foreign interest rate shock R^{B*}) and active (χ_d,t shock).
  - Government: rule-based monetary policy (Taylor rule), fiscal lump-sum tax for budget financing, sterilized FX intervention rule:
    - FX reserves steady state chosen as 1.5 times GDP (ten-year average for EM Asia); FX reserve volatility target matches data (8.9 percent of GDP).
    - FX intervention: F_t = F̄ + ρ_FS (s_t − s̄) + ρ_FD (D*_t − D̄*).
  - Market equilibrium: resource constraints and balance of payments (equations preserved).

### Calibration and quantitative experiments
- Benchmark: Asian EM average calibrated to post-crisis low interest rate period (2009 to 2018). Frequency: quarterly.
- Banking sector parameters (σ, ξ, θ_0, θ_1) calibrated to match average interest rate differential and banking sector gross FX assets and liabilities in EM Asia dataset.
- Steady-state normalization: total output = 1.
- FX reserves steady state: chosen as 1.5 times GDP; volatility of change in FX reserves calibrated to 8.9 percent of GDP.
- Capital flow shock calibration:
  - Size of one standard deviation of primitive shock χ chosen so that resulting one standard deviation of banks’ gross external assets over GDP matches data.
  - Simulation results to a temporary capital outflow shock (no intervention baseline):
    - Bank external borrowing drops by 27 percent relative to steady state.
    - Bank external assets drop by 29 percent relative to steady state.
    - Liquidity value falls by 52 percent.
    - Banks cut investment to the real sector by 7.4 percent.
    - Output falls (percent not provided in supplied excerpt).

### Policy comparisons and implications (from empirical results and model simulations)
- Policy instruments compared: monetary policy interest rate, FX intervention (FXI), macroprudential measures (MPMs — reserve requirement on bank foreign liabilities), capital flow measures (CFMs — cyclical tax on foreign currency borrowing).
- Key model-based insights:
  - Effectiveness of FXI depends on the relative importance of financial versus non-financial shocks:
    - If external financial shocks dominate, FXI can improve overall welfare by mitigating sharp exchange rate declines, preserving bank balance sheets, credit, and investment; competitiveness gains via depreciation are muted due to dollar price invoicing.
    - If non-financial shocks dominate, a standard Taylor rule yields better welfare outcomes.
  - Cyclical MPMs and CFMs can generate welfare gains by stabilizing bank balance sheets and external exposures, thereby allowing monetary policy to concentrate on macro objectives.
  - Using the policy interest rate alone to lean against the wind (LATW) may be limited for small open economies because it can trigger capital flow movements and worsen tradeoffs between macro- and financial stability; MPMs and CFMs serve as critical complements.

### Effects of a capital outflow (financial) shock
- Real sector contraction: reported fall of 0.6 percent.
- Real economy channels affected: labor, wages, the price of capital and marginal cost all significantly drop, leading to deflation.
- Monetary response: a central bank conducting conventional monetary policy lowers the interest rate, but this is not enough to restore output.
- Exchange rate dynamics: real exchange rate depreciated by 2.2 percent in the model simulation, exacerbating representative consumer welfare losses.
- Central bank FX intervention (FXI) response:
  - Intervention rule implies the central bank should sell reserves to support domestic financial sector.
  - Central bank sells 8.9 percent of FX reserves to the economy.
  - Resulting balance-sheet effects: gross FX liability falls further than without intervention, while gross FX assets decline only by half.
  - Outcome: financial conditions loosen relative to no intervention; decline in investment and output is lower than without FXI.
- Policy interpretation: sterilized FXI can be an effective tool to target exchange rate depreciation caused by financial-sector shocks, particularly in Asian emerging markets with large FX positions and limited hedging opportunities.

### Productivity (TFP) shock
- Shock size and macro response: TFP shock of size 1 percent leads to output falling by 0.96 percent; investment, labor, and output all fall.
- Financial-sector response is limited compared to the capital outflow shock:
  - Bank gross FX assets fall by 5 percent, which is about 1.8 percent of GDP.
  - Gross FX liabilities fall by 3 percent.
- Exchange rate: appreciates and acts as a shock absorber.
- FXI under TFP shock:
  - If the central bank buys reserves to offset appreciation (i.e., implements FXI), output response is lower than without FXI (FXI worsens the recession).
  - Conclusion: FXI can be costly when the exchange rate is a shock absorber because it switches off the risk-sharing channel.
- Structural drivers: the high FX exposures on bank balance sheets and limited hedging availability in Asian EMs make exchange-rate effects on balance sheets nontrivial; different balance-sheet structures would yield different results.

### Alternative policies to FXI
- Capital Flow Measures (CFMs):
  - Modeled as a tax on foreign funding: d_t^* = χ_d,t ((1−τ_d,t) R_t^* − R_t^{B*}).
  - Rule-based tax: τ_{d,t} = ρ_{ {τ_s} } (s_t − s̅) + ρ_{ {τ_d} } (D_t^* − D̅^*).
  - Government imposes higher tax when exchange rate is appreciating and/or FX liabilities are larger.
- Augmented Monetary Policy:
  - Augmented Taylor rule: i_t = i̅ + ρ_π (π_t − π̅) + ρ_y (Y_{N t} − {Y_N}̅) + ρ_{id} (D_t^* − D̅^*) + ρ_{is} (s_t − s̅).
  - Central bank raises interest rate following depreciation and with lower FX liabilities; lowers rate following appreciation and with greater FX liabilities.
  - Mechanism: stimulates economy (raises inflation), households work more under sticky prices; welfare cost arises via increased labor hours.
- Macroprudential Measures (MPMs) — reserve requirement on FX borrowing:
  - Bank balance sheet becomes: B_t + s_t B_t^* + Q_t K_t = D_t + s_t (1−τ_{m,t}) D_t^* + N_t.
  - Rule-based reserves: τ_{m,t} = ρ_{ms} (s_t − s̅) + ρ_{md} (D_t^* − D̅^*).
  - Government imposes higher reserve requirements when exchange rate appreciates and with larger FX liabilities.

### Comparison across policies (calibrated to match baseline FXI output response)
- General finding: all alternative policies can smooth recession from a capital outflow shock when calibrated to match baseline FXI output response.
- FXI vs MPM:
  - Responses are quite similar because both target banking sector–foreign investor relationship.
  - Key difference: MPM distorts domestic banks’ demand for foreign funds via marginal propensity to receive FX borrowing; consumption loss relative to FXI is moderate.
- CFM:
  - Distorts foreign investor incentives (supply of foreign funding).
  - Reducing tax on foreign investors acts like a subsidy shifting foreign funding supply right; observed drop in foreign borrowing is about half of that in FXI case.
  - Subsidy origin: consumer welfare; cost to consumption.
- Augmented monetary policy:
  - Different mechanism: stimulates economy via inflation and labor supply.
  - Side effects: increase in labor hours is a direct welfare cost.
- Practical considerations: economies rarely face a single identifiable shock; coordination or combination of policies may be optimal. FXI may distort monetary signaling and credibility; persistent sterilization may impede financial market development.

### Optimal policy (two-stage problem)
- Policy objective: central bank minimizes L = min E[ y_t′ W y_t ] where y includes deviations from steady state and W is weighting matrix.
- Implementation assumptions: monetary policy used solely for domestic conditions; exchange rate policy used in response to external shocks.
- Optimal monetary policy (targeting inflation and output with external shocks set to zero, only productivity shocks active):
  - Optimal parameters: ρ_y = 0.08 and ρ_π = 1.2.
- Optimal exchange rate policy evaluated in three shock environments: only external shock (D*), only productivity shock (A), and both shocks simultaneously (Both).
- Table 2: Comparison of Loss Function Loadings by Type of Shock (values presented as in source)
  - Loading on D*:
    - FXI: 0.938 0.96 0.879
    - CFM: 0.24 0.09 0.24
    - MPM: 0.06 -0.04 -1.20
    - Augmented MP: 0.09 0.036 0.049
  - Loading on S:
    - FXI: -10.1 27 -12.7
    - CFM: 0.8 -0.06 1.17
    - MPM: -2.01 -3.84 -0.25
    - Augmented MP: 0.78 -0.68 1.28
  - Loss:
    - FXI: 0.007 9.0e-06 0.007
    - CFM: 0.017 0.001 0.016
    - MPM: 0.040 0.018 0.022
    - Augmented MP: 0.009 0.001 0.006
  - Note: Three shock types considered: productivity shocks (A), capital flow shocks (D*) and both productivity and capital flow shocks simultaneously (“Both”).
- Interpretation:
  - FXI performs very well when only external shock (D*) is present (loss nearly 0).
  - FXI performs poorly under productivity shock (A); applying FXI when exchange rate is a shock absorber raises losses.
  - When both shocks occur, FXI alone is insufficient; CFMs and MPMs perform worse due to added distortions.
  - Policymakers face identification challenges in practice; complementary policies and weighing operational costs are necessary.

### Conclusion (key takeaways)
- Banking systems with large FX exposures can propagate external shocks to the real economy; exchange rate can act as a shock amplifier under adverse external financial shocks (e.g., capital outflows).
- FXI can mitigate external financial shocks by providing FX funding and dampening exchange-rate effects on bank balance sheets, but is not a universal solution.
- For nonfinancial shocks where exchange rate is a shock absorber, standard monetary policy (Taylor rule) improves welfare; FXI can be counterproductive.
- Cyclical macroprudential policies (reserve requirements on foreign liabilities) and CFMs (cyclical tax on foreign borrowing) can help stabilize bank balance sheets and external exposures, but may introduce distortionary welfare costs.
- Optimal policy choice depends heavily on economy-specific characteristics and the nature of shocks; operational costs, credibility effects, and financial market development implications must be weighed.

### Annex I — Summary statistics (selected points)
- Note: All variables expressed in m/m growth (monthly: 2001M12-2020M2) or q/q growth (quarterly: 2001Q4-2019Q4).
- Quarterly panel (selected entries presented exactly as in source table):
  - Emerging: Foreign assets mean 3.05 St. Dev 17.11 Freq 414; Foreign liabilities mean 2.59 St. Dev 9.77 Freq 414; REER mean 0.27 St. Dev 2.48 Freq 438; Credit to private sector mean 2.87 St. Dev 3.96 Freq 414; IIP mean 1.82 St. Dev 3.15 Freq 428.
  - Advanced: Foreign assets mean 3.15 St. Dev 10.20 Freq 240; Foreign liabilities mean 2.08 St. Dev 6.86 Freq 240; REER mean 0.13 St. Dev 3.79 Freq 291; Credit to private sector mean 1.68 St. Dev 5.44 Freq 240; Investment mean 0.70 St. Dev 2.20 Freq 284.
  - Country example: CHN—Foreign assets mean 2.87 St. Dev 6.76 Freq 55; Foreign liabilities mean 3.13 St. Dev 11.46 Freq 55; REER mean 0.32 St. Dev 2.24 Freq 73; Credit to private sector mean 3.45 St. Dev 3.06 Freq 55; Investment mean 2.52 St. Dev 2.19 Freq 71.

### ANNEX II: PANEL VAR ESTIMATES — key methodological points and selected coefficients
- Methodology: quarterly panel VAR estimated by GMM.
- Endogenous variables: g_fl, g_fa, g_reer, g_credit, g_inv (quarter-on-quarter growths).
- Exogenous variables: inflation (inf), g_us, g_tot, external financing shock e_mp.
- Sample partitions and observations:
  - Full Sample: 487
  - Emerging Economies: 308
  - Advanced Economies: 179
- Selected estimated coefficients (preserving reported numeric values and significance markers):
  - Lagged real effective exchange rate (L.g_reer): 0.0842*; 0.127**; -0.104*
  - Second lag real effective exchange rate (L2.g_reer): -0.0715; -0.0721*
  - Lagged growth in foreign liabilities (L.g_fl): -0.0299**; 0.0311; 0.0538
  - Second lag growth in foreign liabilities (L2.g_fl): -0.0148; 0.105**; 0.130*; -0.0332***
  - Lagged growth in foreign assets (L.g_fa): 0.00451; 0.0144*; -0.145*
  - Second lag growth in foreign assets (L2.g_fa): 0.0169*; 0.0195**; 0.376***
  - Lagged credit growth (L.g_credit): 0.255***; 0.197*; -0.381*; 0.475***
  - Second lag credit growth (L2.g_credit): 0.0922**; 0.173***
  - Lagged investment growth (L.g_inv): -0.542*; 0.126
  - Second lag investment growth (L2.g_inv): 0.165**
  - Exogenous monetary policy shock (e_mp): 0.0516; -1.588*; -1.510***; -0.365*; -1.855**; 0.88; -1.163**; -0.302
  - Terms of trade change (g_tot): 0.0920**; 0.125***
  - US growth (g_us): 0.865***; 1.939; 1.104**
  - Inflation (inf): 0.332**; 0.327**; 1.520**
- Reporting conventions: coefficients reported with p-values in parentheses in source; significance indicated as *** p<0.01, ** p<0.05, * p<0.1.

*Source: wpiea2021010-print-pdf - Annex IV. Schematic Description of Model; Annex I and ANNEX II extracts.*

### Annex I. Summary Statistics.............................................................................................

### Annex I. Summary Statistics

### Annex listings and page references
- Annex I. Summary Statistics ................................................................................................... 29
- Annex II: Panel VAR Estimates ............................................................................................. 30
- Annex III. Panel VAR Impulse Response Functions.............................................................. 31

*Source: wpiea2021010-print-pdf - Annex I. Summary Statistics.*

### Annex IV. Schematic Description of Model ........................................................................... 35

### Annex IV. Schematic Description of Model

### Objective and context
- Focus: develop a framework to understand the role of the banking sector as a transmission channel for external shocks in small open (Asian) economies and to assess macro-financial policy responses.
- Motivation: large and volatile cross-country capital flows (international financial cycle) can produce surges and sudden stops with macrofinancial consequences for emerging market economies (EMs), especially in Asia.
- Key features of Asian banking systems highlighted:
  - Banks are the dominant source of financing for investment and consumption.
  - Gross foreign assets and liabilities are sizeable (Figures referenced).
  - Several Asian EMs maintain large net foreign asset positions; foreign assets tend to be more volatile than liabilities and both are more volatile in Asian EMs than in advanced economies.
  - Home deposits in domestic currency; foreign borrowing and many liabilities denominated in foreign currency; banks hold FX-denominated safe assets (e.g., US Treasuries).

### Empirical strategy (Panel VAR) — specification and data
- Method: panel vector autoregression (panel VAR) following Abrigo and Love (2015) using system-GMM.
- Sample (quarterly): 2002Q1 to 2019Q4 for ten Asian economies:
  - Advanced: Australia, Korea, Japan, New Zealand.
  - Emerging: China, India, Indonesia, Malaysia, the Philippines, Thailand.
- Monthly robustness check: 2001M12 to 2020M2 (uses IIP in lieu of investment; specification has a 12-period lag).
- Model specification (equation form preserved):
  - Y_it = Y_it−1 A1 + Y_it−2 A2 + ⋯ + Y_it−T A_T + X_it B + u_it + e_it
  - Endogenous vector Y_it: growth in foreign assets, growth in foreign liabilities, change in the real exchange rate, growth in credit to the private sector, growth in real investment.
  - Exogenous X_it: US GDP growth, inflation, terms of trade, external financing shock.
- Lag length: T = 2 chosen by MBIC, MAIC, MQIC criteria (Andrews and Lu (2001)).
- Identification: Cholesky ordering — real exchange rate, foreign liabilities, foreign assets, credit, investment.
- External financial shock proxy: exogenous US monetary policy shock (residual of one-year government bond rate regression with 30-min changes of one-month ahead Fed funds futures around FOMC announcements). Alternatives tested: excess bond premium, VIX, US federal funds rate.

### Empirical findings — role of bank balance sheets and transmission channels
- Three headline findings from the panel VAR:
  1. The impact on credit and investment through bank balance sheets from an external financial shock is more pronounced than from a real shock.
  2. Banks (particularly in EMs) appear to have an incentive to hold foreign assets as safe assets; foreign assets and liabilities impact credit independently.
  3. The real exchange rate plays a key role in the ultimate impact of bank funding shocks on credit; the exchange rate channel is stronger in EM Asia than in advanced economies.
- Specific empirical responses:
  - External financial shock (tightening): reduces foreign liabilities and assets, decreases bank credit supply, lowers investment for both advanced and emerging Asia.
  - External demand shock (higher US growth): increases foreign liabilities and assets, but leads to a small depreciation in the real exchange rate (given US dollar appreciation), producing a smaller net increase in credit than in the financial-shock scenario.
  - Positive shock to foreign liabilities (capital inflow): raises foreign assets, leads to real exchange rate appreciation, and increases credit and investment; effect on exchange rate larger in advanced than emerging Asia.
  - Positive shock to foreign assets: evidence of a lagged increase in foreign liabilities followed by credit and investment, especially in EM subsample — consistent with banks holding foreign “safe” assets to attract foreign funding and relax constraints.
  - Real exchange rate appreciation (controlling for bank balance sheet): affects credit but with wider confidence intervals on foreign assets and liabilities.
- Variance decomposition:
  - Banking financial shocks (foreign liabilities/assets) and external financing shocks account for a larger share of forecast error variance for foreign assets/liabilities, credit, and investment than real shocks (US growth, terms of trade).
- Robustness:
  - Monthly VAR confirms directional findings for VIX and FFR shocks (with larger error bands for real activity measures).
  - Including nonfinancial corporate FX borrowing: bank balance sheet channels remain significant under capital inflow shocks; corporate FX debt has larger standard errors.

### Model structure (canonical New-Keynesian small open economy with banking)
- Base: extends Gertler and Karadi (2011) and Aoki, Benigno and Kiyotaki (2018) framework to include:
  - Banking sector that accesses domestic and foreign funds.
  - Banks hold FX-denominated assets as safe assets.
  - Traded goods priced in foreign currency (U.S. dollar invoicing).
  - Non-tradable goods prices with Rotemberg adjustment costs.
  - Policy instruments: monetary policy interest rate, foreign exchange intervention (FXI), macroprudential measures (MPMs), capital flow measures (CFMs).
- Agents and key mechanisms preserved:
  - Representative household with Greenwood-Hercowitz-Huffman preferences and external habit; substitution between traded and non-traded goods governed by st and α_T.
  - Producers: monopolistic competition for intermediate non-traded goods, Cobb-Douglas production, Rotemberg price adjustment, capital accumulation with adjustment cost Φ(·).
  - Banks (bankers):
    - Bank balance sheet: s_t b*_jt + b_jt + Q_t^K k_jt = d_jt + s_t d*_jt + n_t.
    - Participation constraint (moral hazard): investors’ belief γ(x) = θ_0 exp(−θ_1 x) where x = s_t b*_jt / n_t; holding foreign safe assets relaxes funding constraints.
    - Banker value linear in net worth: V(n_jt) = ψ_t n_jt; participation multiplier λ_t appears in first-order conditions.
  - Foreign investor: representative risk-neutral foreign investor supplies foreign credit with quadratic cost 1/(2χ_d,t) (D*_t)^2, yielding D*_t = χ_d,t (R*_t − R_t^{B*}). Two capital flow types: passive (foreign interest rate shock R^{B*}) and active (χ_d,t shock).
  - Government: rule-based monetary policy (Taylor rule), fiscal lump-sum tax for budget financing, sterilized FX intervention rule:
    - FX reserves steady state chosen as 1.5 times GDP (ten-year average for EM Asia); FX reserve volatility target matches data (8.9 percent of GDP).
    - FX intervention: F_t = F̄ + ρ_FS (s_t − s̄) + ρ_FD (D*_t − D̄*).
  - Market equilibrium: resource constraints and balance of payments (equations preserved).

### Calibration and quantitative experiments
- Benchmark: Asian EM average calibrated to post-crisis low interest rate period (2009 to 2018). Frequency: quarterly.
- Banking sector parameters (σ, ξ, θ_0, θ_1) calibrated to match average interest rate differential and banking sector gross FX assets and liabilities in EM Asia dataset.
- Steady-state normalization: total output = 1.
- FX reserves steady state: chosen as 1.5 times GDP; volatility of change in FX reserves calibrated to 8.9 percent of GDP.
- Capital flow shock calibration:
  - Size of one standard deviation of primitive shock χ chosen so that resulting one standard deviation of banks’ gross external assets over GDP matches data.
  - Simulation results to a temporary capital outflow shock (no intervention baseline):
    - Bank external borrowing drops by 27 percent relative to steady state.
    - Bank external assets drop by 29 percent relative to steady state.
    - Liquidity value falls by 52 percent.
    - Banks cut investment to the real sector by 7.4 percent.
    - Output falls (percent not provided in supplied excerpt beyond “output falls by” — model text truncated).

### Policy comparisons and implications (from empirical results and model simulations)
- Policy instruments compared: monetary policy interest rate, FX intervention (FXI), macroprudential measures (MPMs) — exemplified as a reserve requirement on bank foreign liabilities, and capital flow measures (CFMs) — exemplified as a cyclical tax on foreign currency borrowing.
- Key model-based insights:
  - Effectiveness of FXI depends on the relative importance of financial versus non-financial shocks:
    - If external financial shocks dominate, FXI can improve overall welfare by mitigating sharp exchange rate declines, preserving bank balance sheets, credit, and investment; competitiveness gains via depreciation are muted due to dollar price invoicing.
    - If non-financial shocks dominate, a standard Taylor rule (monetary policy interest rate response) yields better welfare outcomes.
  - Cyclical MPMs and CFMs can generate welfare gains by stabilizing bank balance sheets and external exposures, thereby allowing monetary policy to concentrate on macro objectives.
  - Using the policy interest rate alone to lean against the wind (LATW) may be limited for small open economies because it can trigger capital flow movements and worsen tradeoffs between macro- and financial stability; MPMs and CFMs serve as critical complements.

*Source: wpiea2021010-print-pdf - Annex IV. Schematic Description of Model; https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021010-print-pdf.pdf*

### 0.6 percent. All sides of the real sector are affected. Labor, wages, the price of capital and

### wpiea2021010-print-pdf

### Effects of a capital outflow (financial) shock
- Real sector contraction: reported fall of 0.6 percent.
- Real economy channels affected: labor, wages, the price of capital and marginal cost all significantly drop, leading to deflation.
- Monetary response: a central bank conducting conventional monetary policy lowers the interest rate, but this is not enough to restore output.
- Exchange rate dynamics: real exchange rate depreciated by 2.2 percent in the model simulation, exacerbating representative consumer welfare losses (contrary to conventional risk sharing predictions).
- Central bank FX intervention (FXI) response:
  - Intervention rule implies the central bank should sell reserves to support domestic financial sector.
  - Central bank sells 8.9 percent of FX reserves to the economy.
  - Resulting balance-sheet effects: gross FX liability falls further than without intervention, while gross FX assets decline only by half.
  - Outcome: financial conditions loosen relative to no intervention; decline in investment and output is lower than without FXI.
- Policy interpretation: sterilized FXI can be an effective tool to target exchange rate depreciation caused by financial-sector shocks, particularly in Asian emerging markets with large FX positions and limited hedging opportunities.

### Productivity (TFP) shock
- Shock size and macro response: TFP shock of size 1 percent leads to output falling by 0.96 percent; investment, labor, and output all fall.
- Financial-sector response is limited compared to the capital outflow shock:
  - Bank gross FX assets fall by 5 percent, which is about 1.8 percent of GDP.
  - Gross FX liabilities fall by 3 percent.
- Exchange rate: appreciates and acts as a shock absorber (consistent with conventional risk-sharing).
- FXI under TFP shock:
  - If the central bank buys reserves to offset appreciation (i.e., implements FXI), output response is lower than without FXI (FXI worsens the recession).
  - Conclusion: FXI can be costly when the exchange rate is a shock absorber because it switches off the risk-sharing channel.
- Structural drivers: the high FX exposures on bank balance sheets and limited hedging availability in Asian EMs make exchange-rate effects on balance sheets nontrivial; different balance-sheet structures would yield different results.

### Alternative policies to FXI
- Policy set considered: Capital Flow Measures (CFMs), Augmented Monetary Policy, Macroprudential Measures (MPMs—reserve requirements on FX borrowing).
- Capital Flow Measures (CFMs):
  - Modeled as a tax on foreign funding (tax on portfolio inflows).
  - Credit supply function: d_t^* = χ_d,t ((1−τ_d,t) R_t^* − R_t^{B*}).
  - Rule-based tax: τ_{d,t} = ρ_{ {τ_s} } (s_t − s̅) + ρ_{ {τ_d} } (D_t^* − D̅^*).
  - Government imposes higher tax when exchange rate is appreciating and/or FX liabilities are larger.
- Augmented Monetary Policy:
  - Augmented Taylor rule: i_t = i̅ + ρ_π (π_t − π̅) + ρ_y (Y_{N t} − {Y_N}̅) + ρ_{id} (D_t^* − D̅^*) + ρ_{is} (s_t − s̅).
  - Central bank raises interest rate following depreciation and with lower FX liabilities; lowers rate following appreciation and with greater FX liabilities.
  - Mechanism: stimulates economy (raises inflation), households work more under sticky prices, stimulus spills over to banks; welfare cost arises via increased labor hours.
- Macroprudential Measures (MPMs) — reserve requirement on FX borrowing:
  - Bank balance sheet becomes: B_t + s_t B_t^* + Q_t K_t = D_t + s_t (1−τ_{m,t}) D_t^* + N_t.
  - Bank reserves earn same return as central bank FX reserves.
  - Net worth evolution and government budget constraint adjusted to include τ_{m,t} terms (see model equations in source).
  - Rule-based reserves: τ_{m,t} = ρ_{ms} (s_t − s̅) + ρ_{md} (D_t^* − D̅^*).
  - Government imposes higher reserve requirements when exchange rate appreciates and with larger FX liabilities.

### Comparison across policies (calibrated to match baseline FXI output response)
- General finding: all alternative policies can smooth recession from a capital outflow shock when calibrated to match baseline FXI output response.
- FXI vs MPM:
  - Responses are quite similar because both target banking sector–foreign investor relationship.
  - Key difference: MPM distorts domestic banks’ demand for foreign funds via marginal propensity to receive FX borrowing; consumption loss relative to FXI is moderate.
- CFM:
  - Distorts foreign investor incentives (supply of foreign funding).
  - Reducing tax on foreign investors acts like a subsidy shifting foreign funding supply right; observed drop in foreign borrowing is about half of that in FXI case.
  - Subsidy origin: consumer welfare; cost to consumption.
- Augmented monetary policy:
  - Different mechanism: stimulates economy via inflation and labor supply.
  - Side effects: increase in labor hours is a direct welfare cost.
- Practical considerations: economies rarely face a single identifiable shock; coordination or combination of policies may be optimal. FXI may distort monetary signaling and credibility; persistent sterilization may impede financial market development.

### Optimal policy (two-stage problem)
- Policy objective: central bank minimizes L = min E[ y_t′ W y_t ] where y includes deviations from steady state and W is weighting matrix.
- Implementation assumptions: monetary policy used solely for domestic conditions; exchange rate policy used in response to external shocks.
- Optimal monetary policy (targeting inflation and output with external shocks set to zero, only productivity shocks active):
  - Optimal parameters: ρ_y = 0.08 and ρ_π = 1.2.
- Optimal exchange rate policy evaluated in three shock environments: only external shock (D*), only productivity shock (A), and both shocks simultaneously (Both).
- Table 2: Comparison of Loss Function Loadings by Type of Shock (values presented as in source)
  - Columns: FXI | CFM | MPM | Augmented MP, with subcolumns for shocks: Both, D*, A
  - Loading on D*:
    - FXI: 0.938 0.96 0.879
    - CFM: 0.24 0.09 0.24
    - MPM: 0.06 -0.04 -1.20
    - Augmented MP: 0.09 0.036 0.049
  - Loading on S:
    - FXI: -10.1 27 -12.7
    - CFM: 0.8 -0.06 1.17
    - MPM: -2.01 -3.84 -0.25
    - Augmented MP: 0.78 -0.68 1.28
  - Loss:
    - FXI: 0.007 9.0e-06 0.007
    - CFM: 0.017 0.001 0.016
    - MPM: 0.040 0.018 0.022
    - Augmented MP: 0.009 0.001 0.006
  - Note: Three shock types considered: productivity shocks (A), capital flow shocks (D*) and both productivity and capital flow shocks simultaneously (“Both”).
- Interpretation of optimal policy results:
  - FXI performs very well when only external shock (D*) is present (loss nearly 0).
  - FXI performs poorly under productivity shock (A); applying FXI when exchange rate is a shock absorber raises losses.
  - When both shocks occur, FXI alone is insufficient; CFMs and MPMs perform worse due to added distortions.
  - Policymakers face identification challenges in practice; complementary policies and weighing operational costs are necessary.

### Conclusion (key takeaways)
- Banking systems with large FX exposures can propagate external shocks to the real economy; exchange rate can act as a shock amplifier under adverse external financial shocks (e.g., capital outflows).
- FXI can mitigate external financial shocks by providing FX funding and dampening exchange-rate effects on bank balance sheets, but is not a universal solution.
- For nonfinancial shocks where exchange rate is a shock absorber, standard monetary policy (Taylor rule) improves welfare; FXI can be counterproductive.
- Cyclical macroprudential policies (reserve requirements on foreign liabilities) and CFMs (cyclical tax on foreign borrowing) can help stabilize bank balance sheets and external exposures, but may introduce distortionary welfare costs.
- Optimal policy choice depends heavily on economy-specific characteristics and the nature of shocks; operational costs, credibility effects, and financial market development implications must be weighed.

### Annex I — Summary statistics (selected points from monthly and quarterly panels)
- Note: All variables expressed in m/m growth (monthly: 2001M12-2020M2) or q/q growth (quarterly: 2001Q4-2019Q4).
- Quarterly panel (selected entries presented exactly as in source table):
  - Emerging: Foreign assets mean 3.05 St. Dev 17.11 Freq 414; Foreign liabilities mean 2.59 St. Dev 9.77 Freq 414; REER mean 0.27 St. Dev 2.48 Freq 438; Credit to private sector mean 2.87 St. Dev 3.96 Freq 414; IIP mean 1.82 St. Dev 3.15 Freq 428.
  - Advanced: Foreign assets mean 3.15 St. Dev 10.20 Freq 240; Foreign liabilities mean 2.08 St. Dev 6.86 Freq 240; REER mean 0.13 St. Dev 3.79 Freq 291; Credit to private sector mean 1.68 St. Dev 5.44 Freq 240; Investment mean 0.70 St. Dev 2.20 Freq 284.
  - Country examples (quarterly): CHN—Foreign assets mean 2.87 St. Dev 6.76 Freq 55; Foreign liabilities mean 3.13 St. Dev 11.46 Freq 55; REER mean 0.32 St. Dev 2.24 Freq 73; Credit to private sector mean 3.45 St. Dev 3.06 Freq 55; Investment mean 2.52 St. Dev 2.19 Freq 71.
  - (Additional country-level means and standard deviations are reported in the source tables for AUS, IDN, IND, JPN, KOR, MSY, NZL, PHL, THA.)

*Source: wpiea2021010-print-pdf (IMF).*

### ANNEX II: PANEL VAR ESTIMATES

### ANNEX II: PANEL VAR ESTIMATES

### Methodology and variables
- A quarterly panel VAR is estimated by GMM.
- Reported numbers show the coefficients of regressing the column variables on the row variables. Coefficients and p-values are reported.
- Endogenous variables:
  - g_fl: growth in foreign liabilities
  - g_fa: growth in foreign assets
  - g_reer: change in the real effective exchange rate, positive values are an appreciation
  - g_credit: growth in credit to the private sector
  - g_inv: growth in real investment (GFCF)
- All growth rates are computed quarter on quarter.
- Exogenous variables:
  - inflation: CPI inflation (inf)
  - g_us: US growth
  - g_tot: change in the terms of trade index
  - external financing shock: the exogenous monetary policy shock e_mp shown in this table (excess bond premium, the VIX, and the US federal funds rate are used as alternative measures).
- L and L2 denote the first and second period lags, respectively.
- Sample partitions reported: Full Sample, Emerging Economies, Advanced Economies.
- Observations:
  - Full Sample: 487
  - Emerging Economies: 308
  - Advanced Economies: 179
- pval in parentheses
  - Significance legend: *** p<0.01, ** p<0.05, * p<0.1

### Key estimated coefficients (selected, as reported in table)
- Lagged real effective exchange rate (L.g_reer)
  - 0.0842*
  - 0.127**
  - -0.104*
- Second lag real effective exchange rate (L2.g_reer)
  - -0.0715
  - -0.0721*
- Lagged growth in foreign liabilities (L.g_fl)
  - -0.0299**
  - 0.0311
  - 0.0538
- Second lag growth in foreign liabilities (L2.g_fl)
  - -0.0148
  - 0.105**
  - 0.130*
  - -0.0332***
- Lagged growth in foreign assets (L.g_fa)
  - 0.00451
  - 0.0144*
  - -0.145*
- Second lag growth in foreign assets (L2.g_fa)
  - 0.0169*
  - 0.0195**
  - 0.376***
- Lagged credit growth (L.g_credit)
  - 0.255***
  - 0.197*
  - -0.381*
  - 0.475***
- Second lag credit growth (L2.g_credit)
  - 0.0922**
  - 0.173***
- Lagged investment growth (L.g_inv)
  - -0.542*
  - 0.126
- Second lag investment growth (L2.g_inv)
  - 0.165**
- Exogenous monetary policy shock (e_mp)
  - 0.0516
  - -1.588*
  - -1.510***
  - -0.365*
  - -1.855**
  - 0.88
  - -1.163**
  - -0.302
- Terms of trade change (g_tot)
  - 0.0920**
  - 0.125***
- US growth (g_us)
  - 0.865***
  - 1.939
  - 1.104**
- Inflation (inf)
  - 0.332**
  - 0.327**
  - 1.520**

(Note: the table reports many coefficients across dependent-variable columns and sample partitions; the above lists preserve the exact numeric values and significance markers as they appear in the table.)

### Sample sizes and partitioning
- Full Sample observations: 487
- Emerging Economies observations: 308
- Advanced Economies observations: 179

### Reporting conventions
- All coefficients reported with p-values in parentheses (not shown in this extract).
- Significance indicated as:
  - *** p<0.01
  - ** p<0.05
  - * p<0.1

*Source: ANNEX II: PANEL VAR ESTIMATES (from provided content unit).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021010-print-pdf.pdf_
