## wpiea2025072-print-pdf

## Source details

**Canonical URL:** [wpiea2025072-print-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025072-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025072-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025072-print-pdf.pdf.json)

---

### Key empirical findings on UIP premium and debt composition
- The UIP premium disappears for firms that were eligible for the government guarantees program; this disappearance is not observed for ineligible firms.
- The reduction in the UIP premium for eligible firms is mainly due to an average reduction in the domestic interest rate on local currency debt, not due to:
  - an increase in the foreign-lender and/or domestic-lender interest rate on dollar loans, or
  - a depreciation of the local currency that increases liquidity/convenience yields on the dollar.
- Eligible firms both substitute foreign lenders with domestic ones and substitute foreign currency debt with local currency debt from domestic lenders.
- The reduction of the UIP premium cannot be attributed to dollar appreciation effects because there was upward pressure on interest rates of dollar-denominated loans due to domestic lenders’ difficulty obtaining dollar funding abroad.
- Interpretation: credit support policies reduced credit risk for local currency debt, even if currency risk increased.

### Model structure and properties
- Model type and key ingredients:
  - Two-period small open economy model with heterogeneous firms.
  - Real model without an explicit currency dimension.
  - Firms can borrow domestically and abroad and face different collateral constraints in each lending market, `a la Caballero and Krishnamurthy (2001).
- Financial intermediation:
  - Domestic credit supply comes from financial intermediaries who lend what they obtain from households and the Central Bank.
  - A critical element is financial intermediaries’ risk aversion.
- Three key properties delivered by the model:
  - Endogenous firms’ finance mix between domestic and foreign debt, responsive to shocks and policies.
  - Larger firms are more leveraged and issue more debt abroad relative to smaller firms, consistent with Chilean microdata.
  - An endogenous interest rate wedge between domestic and foreign debt that stems from differential collateral constraints, effectively segmenting markets.
- Model dynamics and policy interaction:
  - Sudden stop shock raising cost of borrowing abroad → firms shift from foreign debt to domestic debt (domestic debt share rises).
  - Without domestic credit support, domestic interest rates increase and UIP deviation rises.
  - Introducing credit support policies can endogenously generate declining UIP deviations as observed.
  - Policy counterfactuals:
    - The credit line facility alone cannot fully offset the domestic rate increase.
    - Government guarantees alone cannot fully offset the shock because relaxing collateral constraints boosts credit demand and raises domestic rates.
    - When both policies are active, the model can reproduce observed behavior: debt volumes restored to pre-shock levels while borrowing costs are lower than they would be under the shock alone (debt volumes at higher and borrowing costs at lower levels than before the shock).

### Credit-support policies in Chile (implemented at COVID onset)
- Two main unconventional policies studied:
  1. FCIC: Central Bank credit line facility to commercial banks conditional on credit issuance to small and medium firms.
     - Started March 2020 as a credit line to commercial banks for four years at a fixed interest rate equal to the MPR.
     - Most credits were given at the effective lower bound of the MPR (0.5%).
     - First stage: USD 24 billion, about 8.4% of Chile’s 2019 GDP.
     - Banks could access up to 15% of loans in their balance sheets, with 3% unconditional access.
     - Access required collateral (partly reserves at the Central Bank; the rest other assets).
     - Access open for six months; 95% used.
     - Second phase (June 2020): nearly USD 16 billion available for eight months (FCIC-2); use was 30%.
     - Remaining 70% used in FCIC-3 (triggered March 2021) tied to FOGAPE Reactiva.
  2. FOGAPE-COVID: sovereign credit guarantees on commercial banks’ loans to firms below a pre-determined size for working-capital purposes.
     - Program dates from 1980; eligibility depends on yearly sales defined in UF.
     - Before November 2019: firms with yearly sales below 25,000 UF were eligible.
     - October 2019 expansion increased eligibility threshold to 350,000 UF; by January 2020 capitalized with USD 100 million.
     - On April 25, 2020, FOGAPE-COVID recapitalized the fund by USD 3 billion, guaranteeing up to USD 24 billion in credits.
     - Covered only new and working-capital loans; guarantees between 60% and 85% of each credit depending on firm size.
     - FOGAPE-COVID institutional changes:
       - Increased cutoff to access typical FOGAPE credit from 350,000 UF to 1 million UF.
       - Interest rate ceiling of MPR plus 300 basis points (i.e., ceiling of MPR + 300 basis points).
       - Increased fraction of loan guaranteed and maximum FOGAPE loan for all firm sizes.
     - Eligibility for FOGAPE-COVID based on past sales from 2019.
- Identification leverage: cutoff-based eligibility (1 million UF) allowed a Regression Discontinuity Design (RDD) because the sales threshold is exogenous and based on 2019 sales.

### Data sources and descriptive statistics
- Merged administrative datasets (April 2012–December 2020):
  1. Deudex: foreign debt dataset (stocks and flows) including loan characteristics (interest rates, maturity, currency) — April 2012 to December 2020.
  2. D32: credit registry on firm-to-domestic bank new loans and conditions; complemented with firm-to-bank FOGAPE-COVID loans during 2020.
  3. D10: consolidated debt stocks of firms with the domestic banking system.
  4. Domestic Bond Issuance: records firm bond issuance values in domestic bond market.
  5. F29: firms’ total monthly sales from value-added tax records.
- Data custodians:
  - Deudex primary source: Central Bank of Chile.
  - D32, D10, Domestic Bond Issuance: Chilean Financial Markets Commission.
  - F29: Chilean IRS.
- Sample filters and features:
  - Monthly frequency between April 2012 and December 2020.
  - For firms borrowing abroad: kept only non-trade credit loans and bond issuance.
  - Kept foreign credits in US Dollars, Euros, Japanese Yen, or Chilean Pesos (representing more than 98% of external borrowing).
  - Kept only credits with positive spreads.
- Representative descriptive facts:
  - Mean domestic peso loan size ≈ USD 150 thousand (using spot exchange rate).
  - Mean foreign loan ≈ USD 40 million.
  - Mean interest rate on a domestic loan in pesos: 13.2%.
  - Mean interest rate for foreign loans in dollars: 3.3%.
  - Correcting foreign loans by ex-post UIP yields mean of 10.2%.
  - Yearly debt stock-to-GDP ratio: 34.6% for domestic loans and 31.13% for foreign loans.
  - Firm counts (bottom panel of Table II): out of total 284,090 firms:
    - 282,922 borrow only domestically,
    - 465 only abroad,
    - 703 in both markets.
  - Mean yearly sales of all firms: 157.7% of GDP; represent on average 72.3% of total sales in tax records before filters.
  - By January 31st, 2019, 1 UF = 34.5 USD.

### Debt composition and interest-rate behavior during COVID
- Debt composition (April–July 2020 dynamics):
  - Pre-FOGAPE-COVID (April 2020) domestic debt share by firm size:
    - Small firms: 75% domestic debt share.
    - Medium firms: 66% domestic debt share.
    - Mega firms: 40% domestic debt share.
  - Between April and July 2020, new debt issuance shifted significantly toward domestic debt issuance.
  - Increased domestic debt issuance concentrated in FOGAPE-eligible groups:
    - Small-medium firms: domestic debt issuance share increased to 99%.
    - Large firms: domestic debt issuance share increased to 95%.
    - Mega firms (not eligible for FOGAPE-COVID): domestic debt share remained at 40%.
  - Between April and July 2020, about 80% of credit flows were in pesos and 20% in dollars.
- Interest rate behavior:
  - Mean domestic interest rate fell to 5% between March and May 2020, from 15.9% in same period of 2019.
  - Mean foreign interest rate for newly issued debt in dollars fell from 4.3% to 3.5%.
  - Mean foreign interest rate measured in Chilean pesos (ex-post UIP corrected) increased from 11.5% to 22.6%.
  - Removing FOGAPE-COVID loans from sample raises the average domestic interest rate from 5% to close to 9%.
  - Observed CEMBI spread increased from 2.5% to 5.1%.
- UIP deviation series:
  - UIP deviation between domestic pesos debt and dollar debt increases at onset of COVID in March 2020 and remains high until May 2020; drops to pre-COVID levels after credit support policies implemented.

### Empirical identification strategy and RDD implementation
- Identification:
  - Regression Discontinuity Design (RDD) exploiting exogenous eligibility cutoff of 1 million UF in 2019 sales to determine FOGAPE-COVID eligibility.
  - Treatment: Eligible = 1 if 2019 sales < 1 million UF.
  - Local RDD with triangular kernel; estimated using degree-0 and degree-1 polynomials. Also used Epanechnikov kernel as alternative.
  - Outcome: domestic debt share = Domestic debt / Total debt (domestic + foreign), where foreign debt converted to dollars at spot exchange rate; outcome averaged between May and July 2020.
  - Analysis window: May–July 2020 (avoiding August 2020 onward due to additional policies like pension withdrawals and direct transfers).
- RDD specification:
  - D_domestic_i / D_total_i = ω_0 + ω_1 Log(sales2019_i) + ε Eligible_i + θ_i
  - ε is the coefficient of interest: average treatment effect at cutoff.

### RDD results and economic magnitude
- Sample around cutoff: 665 firms (442 left of cutoff, 223 right of cutoff).
- Causal estimate:
  - Baseline (degree-0 triangular kernel): becoming eligible for FOGAPE-COVID increases domestic debt share by 9.4 percentage points for firms around the cutoff (reported as -0.09422**, Standard Error 0.05115, Observations 665; sign convention reflects specification).
  - All reported specifications are statistically significant at the 10% level; baseline and one alternative significant at 5%.
- Macroeconomic relevance:
  - Total sales of firms that became eligible represent 18% of GDP and 8% of total sales in F29 database.
  - The increase in domestic credit by these firms at the beginning of the crisis reached about 1% of 2020’s GDP.
- Decomposition evidence:
  - The change in finance mix was driven by a considerable increase in domestic liabilities relative to total, not just foreign debt falling.

### Mechanism: cost of capital and UIP premium estimation
- Estimation approach:
  - Following di Giovanni et al. (2021) approach: estimate UIP premium via regression of nominal loan interest rates controlling for firm-by-bank fixed effects, trend, foreign-currency dummy, firm and bank controls, and macro controls.
  - Regression structure:
    - i_f,b,d,m = π_f,b + ρTrend_m + ε FX_f,b,d,m + #1 X_f,m + #2 Z_b,m + #3 Macro_m→1 + θ_f,b,d,m
    - FX dummy = 1 if loan in foreign currency (restricted to dollars for domestic credits).
    - Firm/bank controls: value-added, market share (2-digit sector), leverage.
    - Macro controls: price of copper, MPR, monthly indicator of economic activity.
  - Standard errors clustered at the firm level.
- UIP premium estimates:
  - April 2012 – September 2019 period: UIP premium estimated at 3.95 percentage points (relative to average domestic rate in pesos of 13.2%).
  - Comparative literature context: di Giovanni et al. (2021) find UIP premium of 6.9 p.p for Turkey; Gutierrez et al. (2023) find 2 p.p for Peru.
- Crisis-period estimates (March–July 2020):
  - During March–July 2020 the coefficient on FX becomes statistically insignificant, suggesting the UIP premium disappears on average.
  - Estimating with an interaction for FOGAPE-COVID eligibility:
    - For firms ineligible for FOGAPE-COVID, the UIP premium reappears, though one order of magnitude smaller than in normal times.
    - For firms eligible for FOGAPE-COVID, the UIP premium disappears; evidence: positive and significant estimate on FX·eligible interaction (Fx·eligible coefficient 0.0117***, standard error 0.00239; Table V column 3).
  - The reduction in the UIP premium for eligible firms is mainly due to an average reduction in the domestic interest rate, rather than an increase in the foreign interest rate.

### Exact reported interest-rate and lending statistics (selected)
- Mean domestic interest rate (simple mean by loan) decreased from 8.7% to 5.9% between March–July 2019 and the same period in 2020.
- Mean foreign interest rate (simple mean by loan) dropped from 4.4% to 3% between March–July 2019 and the same period in 2020.
- Average interest rate banks faced on foreign dollar-denominated debt was 2.8% between May–July 2019 and fell to 1.3% in the same period of 2020.
- New external borrowing in dollars: 2019 = 6 billion USD; 2020 = 4.5 billion USD.
- Total new lending reported: USD 42.2 billion, with FCIC representing more than two thirds of the new credit.
- Eligibility cutoff for FOGAPE-COVID: 1 million UF.
- Manipulation test (Cattaneo et al. 2020) result: p-value = 0.68 (95% confidence bands) — reject the null hypothesis of manipulation in the running variable (log of sales).

### Robustness evidence and alternative explanations
- RDD validity and robustness:
  - Cattaneo et al. (2020) manipulation test finds no evidence of manipulation around the 1 million UF cutoff (p-value = 0.68).
  - Placebo RDD (May–July 2019): estimate of ε is not significant — no evidence of discontinuity in absence of treatment.
  - Fixed-effects specifications (bank-by-firm, bank-by-firm + firm-by-month, firm-by-month, bank-by-month, firm-month-bank, firm-by-month + bank-by-month) all show a UIP premium in normal times and a considerable fall in the crisis period driven by FOGAPE-COVID eligibility.
  - Inclusion of foreign loans yields the same qualitative result: UIP premium present in normal times and considerably falls during the crisis due to FOGAPE-COVID eligibility.
- Rejected alternative hypothesis: foreign credit dry-out
  - Evidence against dry-out:
    - Net increase in foreign borrowing (bonds and loans) was similar in May–July 2020 and May–July 2019.
    - New external borrowing in 2020 remained significant (4.5 vs 6 billion USD).
    - Banks’ average interest rate on foreign dollar-denominated debt fell from 2.8% to 1.3% (May–July 2019 vs same period 2020), inconsistent with a credit dry-out raising rates.
  - Conclusion: substitution appears driven by policies (FCIC and FOGAPE-COVID) and changes in domestic interest rates, not by foreign credit dry-out.

### Mechanism interpretation: how policies drove debt substitution
- FOGAPE-COVID (sovereign guarantees) increased firms’ access to domestic credit by effectively increasing φ_d (share of firms’ output available as domestic collateral).
- FCIC (Central Bank liquidity provision) increased e_1,CB (Central Bank supply of funds to banks).
- Reduced-form credit-supply parameter ↽ captures financial intermediaries’ behavior and risk aversion:
  - A global shock (rise in risk premium, increasing R_ω) increases ↽, reducing the share of Central Bank funds that flow to firms and contracting total credit supply e_1,T.
  - FOGAPE-type policies reduce ↽ (facilitating pass-through of Central Bank funds), thereby complementing FCIC and increasing total credit supply.
- Empirical linkage:
  - Reduction in UIP premium mainly due to a larger drop in domestic interest rates relative to rates in dollars, making dollar borrowing less attractive given exchange-rate risk.
  - Selection channel: smaller firms (riskier) shifted to local debt markets because foreign borrowing costs were too high; larger/better firms continued borrowing abroad.

### Quantitative model results and policy counterfactuals
- Baseline parametrization highlights (Table VIII):
  - Gross foreign interest rate Rω = 1
  - Firms’ productivity A2 = 3
  - Concavity of the technology π = 1/2
  - First-best capital kω = 2.25
  - Upper bound on international collateral ̄ρkω = 0.2
  - Pledgeable share of output φd = 0.25
  - Initial credit supply e1,T = 1.4781
  - Central Bank supply of credit e1,CB = 0.5
  - Responsiveness of financial intermediaries’ risk-aversion to FOGAPE ς = 24
  - FCIC size $e1,CB = 0.05
  - FOGAPE size $φd = 0.02
- Numerical equilibrium (Table IX summary):
  - Pre-shock: Rω = 1.00, R2 = 1.10, Credit = 1.48.
  - After risk-premium shock (Post-Shock No policies): Rω = 1.10, R2 = 1.20, Credit = 1.44.
  - FCIC only: Rω = 1.10, R2 = 1.15, Credit = 1.50.
  - FOGAPE only: Rω = 1.10, R2 = 1.12, Credit = 1.63.
  - FCIC and FOGAPE jointly: Rω = 1.10, R2 = 1.10, Credit = 1.67.
  - Mapping: joint deployment restores R2 to pre-COVID level and expands credit beyond pre-shock levels; UIP premium disappears in the joint policy scenario (consistent with empirical evidence).
- Sensitivity analysis:
  - Explored φ_d between 0.25 and 0.29 and e_1,CB between 0 and 0.8.
  - For intermediate φ_d and e_1,CB (covering most parameter space), joint implementation of FOGAPE-COVID and FCIC generates a fall in domestic interest rate; size depends on initial state.
  - Non-monotonic relationship between φ_d and R_2 for extreme e_1,CB; monotonic decline of R_2 with e_1,CB.
  - Example: increase in φ_d from 0.25 to 0.27 leads to a fall in R_2 from point C = 1.195 to point D_↑ = 1.12 under the parametrization.

### Appendix: microfoundation for financial intermediaries and equilibrium characterization
- Appendix provides a micro foundation for financial intermediaries `a la Curdia and Woodford (2011), featuring loan origination costs decreasing in FOGAPE and FCIC.
- Key implication: credit supply increases when the two policies are jointly implemented.
- Equilibrium price R_2 derived by equating firms' demand for domestic credit to total credit supply e_1,T (Equation 19), separating constrained and unconstrained firms via threshold ρ̂.
- Graphical equilibrium analysis clarifies demand and supply effects of the shock and policy moves (points A, B, C, D in figures referenced).
- Microfoundation first-order conditions link origination costs to spreads and show both FOGAPE and FCIC complement each other in increasing credit supply.

### Main conclusions and policy-relevant takeaways
- Empirical strategy: regression discontinuity design using comprehensive administrative data on Chilean firms identifies causal effects of credit lines and government-backed credit guarantees during the COVID sudden stop.
- Key empirical findings:
  - Firms eligible for the programs increased borrowing from domestic lenders at a relatively lower cost.
  - Policies reduced the cost of domestic currency debt relative to foreign currency debt, lowering the relative cost of domestic capital in the short term.
  - The effect is conditional on selection at firm and bank levels: only policy-eligible firms benefited from lower credit costs from the same lender that non-eligible firms also borrow from.
- Contributions:
  1. Causal identification of government debt guarantees on firm credit expansion from domestic lenders in local currency.
  2. Establishes mechanism: decline in relative cost of local currency borrowing from domestic lenders vs foreign or foreign-currency borrowing from foreign lenders.
  3. Provides a model of heterogeneous firms with distinct financial frictions in foreign and domestic financing that matches observed higher domestic debt from domestic lenders at a lower cost after policies.
- Policy-relevant conclusion: government policies (FOGAPE-COVID sovereign guarantees and FCIC Central Bank credit lines), particularly when jointly implemented, can incentivize domestic lenders to replace financing lost from international markets during stress events in the short term, expanding domestic credit and lowering domestic borrowing costs.

*Italic: Source — content from wpiea2025072-print-pdf (selected sections provided).*

### introduction of the policies, we find that the UIP premium disappears for firms that were

### wpiea2025072-print-pdf - introduction of the policies, we find that the UIP premium disappears for firms that were

### Key empirical findings on UIP premium and debt composition
- The UIP premium disappears for firms that were eligible for the government guarantees program; this disappearance is not observed for ineligible firms.
- The reduction in the UIP premium for eligible firms is mainly due to an average reduction in the domestic interest rate on local currency debt, not due to:
  - an increase in the foreign-lender and/or domestic-lender interest rate on dollar loans, or
  - a depreciation of the local currency that increases liquidity/convenience yields on the dollar.
- Eligible firms both substitute foreign lenders with domestic ones and substitute foreign currency debt with local currency debt from domestic lenders.
- The reduction of the UIP premium cannot be attributed to dollar appreciation effects because there was upward pressure on interest rates of dollar-denominated loans due to domestic lenders’ difficulty obtaining dollar funding abroad.
- Interpretation: credit support policies reduced credit risk for local currency debt, even if currency risk increased.

### Model structure and properties
- Model type:
  - Two-period small open economy model with heterogeneous firms.
  - Real model without an explicit currency dimension.
  - Firms can borrow domestically and abroad and face different collateral constraints in each lending market, `a la Caballero and Krishnamurthy (2001).
- Financial intermediation:
  - Domestic credit supply comes from financial intermediaries who lend what they obtain from households and the Central Bank.
  - A critical element is financial intermediaries’ risk aversion.
- Three key properties delivered by the model:
  - Endogenous firms’ finance mix between domestic and foreign debt, responsive to shocks and policies.
  - Larger firms are more leveraged and issue more debt abroad relative to smaller firms, consistent with Chilean microdata.
  - An endogenous interest rate wedge between domestic and foreign debt that stems from differential collateral constraints, effectively segmenting markets.
- Model dynamics and policy interaction:
  - Sudden stop shock raising cost of borrowing abroad → firms shift from foreign debt to domestic debt (domestic debt share rises).
  - Without domestic credit support, domestic interest rates increase and UIP deviation rises (local currency debt becomes more expensive than foreign debt).
  - Introducing credit support policies can endogenously generate declining UIP deviations as observed.
  - Policy counterfactuals:
    - The credit line facility alone cannot fully offset the domestic rate increase.
    - Government guarantees alone cannot fully offset the shock because relaxing collateral constraints boosts credit demand and raises domestic rates.
    - When both policies are active, the model can reproduce observed behavior: debt volumes restored to pre-shock levels while borrowing costs are lower than they would be under the shock alone (debt volumes at higher and borrowing costs at lower levels than before the shock).

### Credit-support policies in Chile (implemented at COVID onset)
- Two main unconventional policies studied:
  1. FCIC: a credit line facility from the Central Bank to commercial banks conditional on the growth of credit issuance to small and medium firms.
     - FCIC initial facts:
       - Started March 2020 as a credit line to commercial banks for four years at a fixed interest rate equal to the MPR.
       - Most credits were given at the effective lower bound of the MPR (0.5%).
       - First stage: USD 24 billion, about 8.4% of Chile’s 2019 GDP.
       - Banks could access up to 15% of loans in their balance sheets, with 3% unconditional access.
       - Access required collateral (partly reserves at the Central Bank; the rest other assets).
       - Access open for six months; 95% used.
       - Second phase (June 2020): nearly USD 16 billion available for eight months (FCIC-2); use was 30%.
       - Remaining 70% used in FCIC-3 (triggered March 2021) tied to FOGAPE Reactiva.
  2. FOGAPE-COVID: sovereign credit guarantees on commercial banks’ loans to firms below a pre-determined size for working-capital purposes.
     - FOGAPE program facts:
       - Program dates from 1980; eligibility depends on yearly sales defined in UF (inflation-indexed unit).
       - Before November 2019: firms with yearly sales below 25,000 UF were eligible.
       - October 2019 expansion increased eligibility threshold to 350,000 UF; by January 2020 capitalized with USD 100 million.
       - On April 25, 2020, FOGAPE-COVID recapitalized the fund by USD 3 billion, guaranteeing up to USD 24 billion in credits.
       - Covered only new and working-capital loans; guarantees between 60% and 85% of each credit depending on firm size.
       - FOGAPE-COVID institutional changes:
         - Increased cutoff to access typical FOGAPE credit from 350,000 UF to 1 million UF.
         - Interest rate ceiling of MPR plus 300 basis points (i.e., ceiling of MPR + 300 basis points).
         - Increased fraction of loan guaranteed and maximum FOGAPE loan for all firm sizes.
       - Eligibility for FOGAPE-COVID based on past sales from 2019.
- Implementation detail leveraged for identification: the cutoff-based eligibility (1 million UF) allowed a Regression Discontinuity Design (RDD) because the sales threshold is exogenous and based on 2019 sales.

### Data sources and descriptive statistics
- Merged administrative datasets (April 2012–December 2020):
  1. Deudex: foreign debt dataset (stocks and flows) including loan characteristics (interest rates, maturity, currency) — April 2012 to December 2020.
  2. D32: credit registry on firm-to-domestic bank new loans and conditions; complemented with firm-to-bank FOGAPE-COVID loans during 2020.
  3. D10: consolidated debt stocks of firms with the domestic banking system.
  4. Domestic Bond Issuance: records firm bond issuance values in domestic bond market.
  5. F29: firms’ total monthly sales from value-added tax records.
- Data custodians:
  - Deudex primary source: Central Bank of Chile.
  - D32, D10, Domestic Bond Issuance: Chilean Financial Markets Commission.
  - F29: Chilean IRS.
- Sample filters and features:
  - Monthly frequency between April 2012 and December 2020.
  - For firms borrowing abroad: kept only non-trade credit loans and bond issuance.
  - Kept foreign credits in US Dollars, Euros, Japanese Yen, or Chilean Pesos (representing more than 98% of external borrowing).
  - Kept only credits with positive spreads.
- Representative descriptive facts:
  - Mean domestic peso loan size ≈ USD 150 thousand (using spot exchange rate).
  - Mean foreign loan ≈ USD 40 million.
  - Mean interest rate on a domestic loan in pesos: 13.2%.
  - Mean interest rate for foreign loans in dollars: 3.3%.
  - Correcting foreign loans by ex-post UIP yields mean of 10.2%.
  - Yearly debt stock-to-GDP ratio: 34.6% for domestic loans and 31.13% for foreign loans.
  - Firm counts (bottom panel of Table II): out of total 284,090 firms:
    - 282,922 borrow only domestically,
    - 465 only abroad,
    - 703 in both markets.
  - Mean yearly sales of all firms: 157.7% of GDP; represent on average 72.3% of total sales in tax records before filters.
  - By January 31st, 2019, 1 UF = 34.5 USD.

### Debt composition and interest rate behavior during COVID
- Debt composition (April–July 2020 dynamics):
  - Pre-FOGAPE-COVID (April 2020) domestic debt share by firm size:
    - Small firms: 75% domestic debt share.
    - Medium firms: 66% domestic debt share.
    - Mega firms: 40% domestic debt share.
  - Between April and July 2020, new debt issuance shifted significantly toward domestic debt issuance.
  - Increased domestic debt issuance concentrated in FOGAPE-eligible groups:
    - Small-medium firms: domestic debt issuance share increased to 99%.
    - Large firms: domestic debt issuance share increased to 95%.
    - Mega firms (not eligible for FOGAPE-COVID): domestic debt share remained at 40%.
  - Between April and July 2020, about 80% of credit flows were in pesos and 20% in dollars.
- Interest rate behavior (Table III highlights):
  - Mean domestic interest rate fell to 5% between March and May 2020, from 15.9% in same period of 2019.
  - Mean foreign interest rate for newly issued debt in dollars fell from 4.3% to 3.5%.
  - Mean foreign interest rate measured in Chilean pesos (ex-post UIP corrected) increased from 11.5% to 22.6%.
  - Removing FOGAPE-COVID loans from sample raises the average domestic interest rate from 5% to close to 9%.
  - Observed CEMBI spread increased from 2.5% to 5.1%.
- UIP deviation series (Figure I notes):
  - Two average UIP deviations plotted monthly since January 2019:
    1) domestic debt in pesos vs foreign debt in dollars;
    2) domestic debt in pesos vs domestic debt in dollars.
  - UIP deviation between domestic pesos debt and dollar debt increases at onset of COVID in March 2020 and remains high until May 2020; drops to pre-COVID levels after credit support policies implemented.

### Empirical identification strategy and RDD implementation
- Identification:
  - Regression Discontinuity Design (RDD) exploiting exogenous eligibility cutoff of 1 million UF in 2019 sales to determine FOGAPE-COVID eligibility.
  - Treatment: Eligible = 1 if 2019 sales < 1 million UF (i.e., eligible for FOGAPE-COVID loans), 0 otherwise.
  - Local RDD with triangular kernel; estimated using degree-0 and degree-1 polynomials. Also used Epanechnikov kernel as alternative.
  - Outcome: domestic debt share = Domestic debt / Total debt (domestic + foreign), where foreign debt converted to dollars at spot exchange rate; outcome averaged between May and July 2020.
  - Analysis window: May–July 2020 (avoiding August 2020 onward due to additional policies like pension withdrawals and direct transfers).
- RDD specification (as presented):
  - D_domestic_i / D_total_i = ω_0 + ω_1 Log(sales2019_i) + ε Eligible_i + θ_i
  - ε is the coefficient of interest: average treatment effect at cutoff.

### RDD results and economic magnitude
- Sample around cutoff: 665 firms (442 left of cutoff, 223 right of cutoff).
- Causal estimate:
  - Baseline (degree-0 triangular kernel): becoming eligible for FOGAPE-COVID increases domestic debt share by 9.4 percentage points for firms around the cutoff.
  - All reported specifications are statistically significant at the 10% level; baseline and one alternative significant at 5%.
- Macroeconomic relevance:
  - Total sales of firms that became eligible represent 18% of GDP and 8% of total sales in F29 database.
  - The increase in domestic credit by these firms at the beginning of the crisis reached about 1% of 2020’s GDP.
- Decomposition evidence:
  - The change in finance mix was driven by a considerable increase in domestic liabilities relative to total, not just foreign debt falling.

### Mechanism: cost of capital and UIP premium estimation
- Estimation approach:
  - Following di Giovanni et al. (2021) approach: estimate UIP premium via regression of nominal loan interest rates controlling for firm-by-bank fixed effects, trend, foreign-currency dummy, firm and bank controls, and macro controls.
  - Regression (Equation 2) structure:
    - i_f,b,d,m = π_f,b + ρTrend_m + ε FX_f,b,d,m + #1 X_f,m + #2 Z_b,m + #3 Macro_m→1 + θ_f,b,d,m
    - FX dummy = 1 if loan in foreign currency (restricted to dollars for domestic credits).
    - Firm/bank controls: value-added, market share (2-digit sector), leverage.
    - Macro controls: price of copper, MPR, monthly indicator of economic activity.
  - Standard errors clustered at the firm level.
- UIP premium estimates (Table V excerpts):
  - April 2012 – September 2019 period: UIP premium estimated at 3.95 percentage points (relative to average domestic rate in pesos of 13.2%).
  - Comparative literature context: di Giovanni et al. (2021) find UIP premium of 6.9 p.p for Turkey; Gutierrez et al. (2023) find 2 p.p for Peru.

*Italic: Source — content from wpiea2025072-print-pdf (selected sections provided).*

### 2020. For this period, the coe”cient onFXbecomes statistically insignificant, suggesting

### wpiea2025072-print-pdf - 2020. For this period, the coe”cient onFXbecomes statistically insignificant, suggesting

### Key empirical findings on UIP premium and debt substitution
- During March–July 2020 the coefficient on FX becomes statistically insignificant, suggesting the UIP premium disappears and that, on average, during the beginning of the COVID-19 crisis, borrowing in dollars was not cheaper than borrowing in pesos.
- Estimating Equation 3 (March–July 2020) with E_f,m a dummy for FOGAPE-COVID eligibility and interacted with FX_f,b,d,m yields:
  - For firms ineligible for FOGAPE-COVID, the UIP premium reappears, though it is one order of magnitude smaller than in the normal-times period.
  - For firms eligible for FOGAPE-COVID, the UIP premium disappears, evidenced by the positive and significant estimate of ς on the interaction E_f,m * FX_f,b,d,m.
- The reduction in the UIP premium for eligible firms is mainly due to an average reduction in the domestic interest rate, rather than an increase in the foreign interest rate.

### Exact reported interest-rate and lending statistics
- Mean domestic interest rate (simple mean by loan) decreased from 8.7% to 5.9% between March–July 2019 and the same period in 2020.
- Mean foreign interest rate (simple mean by loan) dropped from 4.4% to 3% between March–July 2019 and the same period in 2020.
- Average interest rate banks faced on foreign dollar-denominated debt was 2.8% between May–July 2019 and fell to 1.3% in the same period of 2020.
- New external borrowing in dollars: 2019 = 6 billion USD; 2020 = 4.5 billion USD.
- Total new lending reported: USD 42.2 billion, with FCIC representing more than two thirds of the new credit.
- Eligibility cutoff for FOGAPE-COVID: 1 million UF (sales cutoff).
- Manipulation test (Cattaneo et al. 2020) result: p-value = 0.68 (95% confidence bands) — reject the null hypothesis of manipulation in the running variable (log of sales).

### Robustness evidence (RDD, fixed effects, and alternative samples)
- RDD validity:
  - Cutoff uses 2019 IRS sales and policy implemented in May 2020, making pre-policy manipulation unlikely.
  - Cattaneo et al. (2020) manipulation test finds no evidence of manipulation around the 1 million UF cutoff (p-value = 0.68).
  - Placebo RDD (May–July 2019) re-estimating Equation 1: estimate of ε is not significant under baseline and three alternative specifications — no evidence of discontinuity in absence of treatment.
- Fixed-effects robustness for Equations 2 and 3:
  - Specifications explored: bank-by-firm (π_f,b), bank-by-firm + firm-by-month (π_f,b + π_f,m), firm-by-month (π_f,m), bank-by-month (π_b,m), firm-month-bank (π_f,m,b), firm-by-month + bank-by-month (π_f,m + π_b,m).
  - Across all fixed-effects specifications there is always a UIP premium in normal times and a considerable fall in the crisis period driven by FOGAPE-COVID eligibility.
- Inclusion of foreign loans:
  - Re-estimating Equations 2 and 3 after adding foreign loans (assigning a unique lender identifier for foreign loans) yields the same qualitative result: UIP premium present in normal times and considerably falls during the crisis due to FOGAPE-COVID eligibility.
  - Note: introducing foreign loans removes bank-level controls because microeconomic information on foreign lenders is unavailable.

### Alternative explanations tested and rejected
- Hypothesis: external dollar credit dry-out for banks causing reduced domestic supply of dollar loans (raising their interest rate and lowering the UIP premium).
  - Evidence against dry-out:
    - Net increase in foreign borrowing (bonds and loans) was similar in May–July 2020 and May–July 2019.
    - New external borrowing in 2020, while lower (4.5 vs 6 billion USD), remained significant.
    - Banks’ average interest rate on foreign dollar-denominated debt fell from 2.8% to 1.3% (May–July 2019 vs same period 2020), inconsistent with a credit dry-out raising rates.
  - Conclusion: no support for a foreign credit dry-out; substitution appears driven by policies (FCIC and FOGAPE-COVID) and changes in domestic interest rates.

### Mechanism interpretation: how policies drove debt substitution
- FOGAPE-COVID (sovereign guarantees) increased firms’ access to domestic credit by effectively increasing φ_d (share of firms’ output available as domestic collateral).
  - The positive and significant interaction coefficient ς indicates that eligibility to FOGAPE-COVID is linked to the reduction in the UIP premium.
- FCIC (Central Bank liquidity provision) increased e_1,CB (Central Bank supply of funds to banks).
- Reduced-form credit-supply parameter ↽ captures financial intermediaries’ behavior and risk aversion:
  - A global shock (rise in risk premium, increasing R_ω) increases ↽, reducing the share of Central Bank funds that flow to firms and contracting total credit supply e_1,T.
  - FOGAPE-type policies reduce ↽ (facilitating pass-through of Central Bank funds), thereby complementing FCIC and increasing total credit supply.
- Empirical linkage:
  - Reduction in UIP premium mainly due to a larger drop in domestic interest rates relative to rates in dollars, making dollar borrowing less attractive given exchange-rate risk.
  - Selection channel: smaller firms (riskier) shifted to local debt markets because foreign borrowing costs were too high; larger/better firms continued borrowing abroad. The last row of Table III shows mean sales of firms that borrowed abroad during the crisis is higher than before the crisis.

### Model overview and role in rationalizing empirical results
- Model type: two-period small open economy with heterogeneous firms differing in international collateral ρ_i,2,f drawn from uniform[0, ̄ρ].
- Key ingredients:
  - Endogenous firms’ finance mix between domestic and foreign debt.
  - Firm-level heterogeneity: larger firms issue relatively more debt abroad; smaller firms borrow domestically.
  - Endogenous interest-rate wedge between domestic (R_2) and foreign (R_ω) debt.
- Notable model specifics and parameter identities preserved exactly as presented:
  - Time index t = 1, 2.
  - Utility: U(c_1, c_2) = c_2.
  - Productivity parameter: A_2 > 1.
  - Production function specification: A_2 (k_i1)^ε with π = 1/2.
  - Relationship assumed: ̄ρ < (A_2 π)^{1/(1→ω)} (as in Equation 5).
  - Foreign gross interest rate r_ω = 1 and R_ω = r_ω + risk premium.
  - Firms borrow domestically d_i1,d and from foreign financiers d_i1,f with interest rates R_2 and R_ω respectively.
  - Foreign collateral constraint: R_ω d_i1,f ↔ ρ_i2,f.
  - Domestic collateral constraint: R_2 d_i1,d ↔ φ_d A_2 (d_i1,d + d_i1,f)^ε + ρ_i2,f ↑ R_ω d_i1,f (as in Equation 9).
  - When R_2 > R_ω firms tap international markets first; foreign debt up to foreign collateral constraint: d_i1,f = ρ_i2,f / R_ω (Equation 11).
  - Domestic borrowing when slack: d_i1,d = k_ω − ρ_i2,f / R_ω (Equation 13).
  - Domestic borrowing when constrained: d_ω1,d(ρ_i2,f) given by quadratic solution in Equation 15.
- Equilibrium implications:
  - Firms’ total leverage (domestic + international debt over output) is increasing in output; constrained and unconstrained firm groups emerge depending on domestic collateral constraint.
  - Increase in φ_d (FOGAPE-COVID) increases firms’ access to domestic borrowing.
  - Credit supply e_1,T = e_1,CB + e_1,H where e_1,CB < 1 and ↽ encapsulates global shock and policy effects: ↽ = e^{R ε → 1 ↑ ς($φ_d)} (Equation 18) — an increase in R_ω increases ↽ and can reduce the share of Central Bank funds reaching firms.

*Source: wpiea2025072-print-pdf - 2020. For this period, the coe”cient onFXbecomes statistically insignificant, suggesting*

### Appendix provides a micro foundation for financial intermediaries `a laCurdia and Wood-

### wpiea2025072-print-pdf - Appendix provides a micro foundation for financial intermediaries `a la Curdia and Woodford (2011)

### Microfoundation for financial intermediaries
- Appendix builds a micro foundation for financial intermediaries `a la Curdia and Woodford (2011), featuring loan origination costs decreasing in FOGAPE and FCIC.
- Key implication: credit supply increases when the two policies are jointly implemented.
- Appendix provides further details on the derivations (integrals and optimization of intermediaries).

### Equilibrium characterization (analytical and graphical)
- The only equilibrium price in the model is R_2, found by equating firms' demand for domestic credit to total credit supply, e_1,T, via Equation 19 which separates demand by constrained and unconstrained firms using the endogenous threshold ρ̂ (Equation 14), dω_1,d (Equation 15), and e_1,T (Equations 17 and 18).
- Graphical analysis:
  - Pre-shock equilibrium at point A: vertical supply curve intersects negatively sloped demand curve (demand negative relationship verified from Equation 19).
  - COVID-type shock increases the interest rate at which firms borrow abroad (R_↑), producing two simultaneous effects:
    - Demand effect: demand for domestic credit increases as unconstrained firms substitute foreign for domestic credit, shifting demand upward and pushing domestic rate from point A to B.
      - Unconstrained firms can still finance first-best capital level; constrained firms face tighter domestic collateral constraints and reduced access to domestic credit.
      - If φ_d is high enough, unconstrained firms’ behavior dominates and total credit demand increases.
    - Supply effect: higher R_ω raises banks' aversion to lending (captured via ↽ and e^π_1,CB), shifting supply left and producing a new equilibrium C with a higher domestic rate and contracted total domestic credit.
  - Policy experiments (Figure VII):
    - Central Bank credit line (increase in e_1,CB, akin to FCIC): supply shifts right, equilibrium moves to point D — domestic interest falls relative to C but not to pre-shock level; credit volume rises but may not reach pre-shock level.
    - Sovereign guarantees (increase in φ_d, akin to FOGAPE): demand shifts up, domestic rates rise further but supply partly responds via reduced bank risk-aversion; can restore credit volume to pre-shock level while domestic interest remains much higher.
    - Joint deployment of both policies: complementarities can restore domestic rates to levels equal to (or below) pre-shock while credit volume increases considerably to point D — qualitatively consistent with observed Chilean experience.

### Quantitative analysis (parametrization and numerical equilibrium)
- Baseline parametrization notes (Table VIII summary in text):
  - Baseline risk premium equals zero; foreign interest rate R_ω = 1.
  - Upper bound on international collateral ρ̄ satisfies Equation 5; difference between k_ω and ρ̄,0.2, is arbitrary.
  - Pledgeable share of output φ_d is small enough to ensure increasing leverage between constrained and unconstrained firms.
- Example leverage calculations under Table VIII parametrization:
  - Unconstrained total leverage ↼_U = 0.5
  - Constrained firm with ρ = 1.22: ↼_C(ρ=1.22) = 0.2273 + 1.22/(A_2)(2.24) = 0.499 (satisfies ↼_U > ↼_C)
- Credit supply and policy parameters:
  - Baseline total credit supply e_1,T chosen so domestic interest rate is 10% (approximate pre-COVID average domestic-foreign difference in 2019 in Table III).
  - e_1,CB < 1.
  - Chosen values: ς = 24, FCIC funding = 0.05, increase in φ_d (FOGAPE) of 0.02 (from 0.25 to 0.27) to qualitatively match observed higher domestic credit and lower interest rate.
- Numerical results (Table IX summary in text):
  - Pre-shock: domestic interest rate = 10%, equilibrium credit = 1.48.
  - After risk-premium shock: foreign interest rate increases to 10%, domestic interest rate to 20%, credit contracts to 1.44.
  - FCIC only (fourth column): domestic rate decreases to 15%, total domestic credit = 1.5.
    - FCIC considered: a 10% increase in Central Bank credit supply; limited power to expand credit and lower interest rate.
  - FOGAPE only (fifth column): domestic credit increases to 1.63, interest rate lowers to 12% (still above pre-crisis 10%).
  - FCIC + FOGAPE jointly (sixth column): interest rate drops to pre-COVID level of 1.1, UIP premium disappears (consistent with empirical evidence in column (2) of Table V); equilibrium domestic credit expands to 1.67.

### Sensitivity analysis (parameter space and robustness)
- Parameter spaces explored:
  - φ_d between 0.25 and 0.29 (initial value 0.25).
  - e_1,CB between 0 and 0.8 (upper limits set so equilibrium remains real given Table VIII parametrization).
- Findings from Figure VIII (equilibrium domestic interest rate R_2 surface over φ_d and e_1,CB for R_ω = 1 and R_ω = 1.1):
  - Regardless of policy values, domestic interest rate increases with foreign interest rate (an increase in R_ω always yields an increase in R_2).
  - Monotonic relationship between e_1,CB and R_2: higher FCIC (e_1,CB) lowers the domestic interest rate (consequence of Equation 17 when ↽ > 0). The rate of decline is non-linear and depends on φ_d.
  - Non-monotonic relationship between φ_d and R_2:
    - For very low e_1,CB, increasing φ_d from 0.25 initially increases the interest rate (demand effect > supply effect); beyond a tipping point, supply effect dominates and interest rate decreases.
    - For very high e_1,CB a similar pattern holds: larger increases in φ_d required to lower R_2 below initial level.
  - For intermediate values of policy parameters (e_1,CB between 0.1 and 0.65), which cover most parameter space, an increase in φ_d always decreases R_2, but the marginal effect is non-linear in e_1,CB.
  - Example numeric movement: an increase in φ_d from 0.25 to 0.27 leads to a fall in R_2 from point C = 1.195 to point D_↑ = 1.12 under the parametrization.
- Summary conclusions from sensitivity analysis:
  1) For intermediate φ_d and e_1,CB (covering most of parameter space), joint implementation of FOGAPE-COVID and FCIC generates a fall in domestic interest rate; size depends on initial state.
  2) Given the risk-premium shock and implausibility of extreme initial e_1,CB values, joint implementation of both policies is required to attain equilibrium with more domestic credit and lower or equal domestic interest rate relative to pre-shock.
  3) There exist extreme initial states where sufficiently large increases in either φ_d or e_1,CB alone could achieve lower domestic rates and higher credit, but these require extremely high or low Central Bank liquidity—deemed unlikely given observed patterns during COVID-19.

### Empirical identification and main conclusions
- Empirical strategy: regression discontinuity design using comprehensive administrative data on Chilean firms to identify causal effects of credit lines and government-backed credit guarantees during the COVID sudden stop.
- Key empirical findings:
  - Firms eligible for the programs increased borrowing from domestic lenders at a relatively lower cost.
  - Policies reduced the cost of domestic currency debt relative to foreign currency debt, lowering the relative cost of domestic capital in the short term.
  - The effect is conditional on selection at firm and bank levels: only policy-eligible firms benefited from lower credit costs from the same lender that non-eligible firms also borrow from.
- Contributions relative to literature:
  1) Causal identification of government debt guarantees on firm credit expansion from domestic lenders in local currency.
  2) Establishes mechanism: decline in relative cost of local currency borrowing from domestic lenders vs foreign or foreign-currency borrowing from foreign lenders.
  3) Provides a model of heterogeneous firms with distinct financial frictions in foreign and domestic financing that matches observed higher domestic debt from domestic lenders at a lower cost after policies.
- Policy-relevant conclusion: government policies (FOGAPE-COVID sovereign guarantees and FCIC Central Bank credit lines), particularly when jointly implemented, can incentivize domestic lenders to replace financing lost from international markets during stress events in the short term, expanding domestic credit and lowering domestic borrowing costs.

*Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025072-print-pdf.pdf*

### References

### wpiea2025072-print-pdf - References

### Empirical evidence: Sudden stop in Chile and firm-level UIP premia
- Sudden stop timing: vertical line denotes February 2020 (the month before the first COVID case in Chile on March 7, 2020).
- Left-panel measures: fund flows’ EPFR measure and CEMBI spread for Chile (data sources: Informa PLC and Bloomberg).
- Right-panel measures: average firm-level International UIP Deviation (blue line) and Domestic UIP Premia (red dashed line). UIP Premium formula: UIPPf,m = if,m − i→f,m − (Em(em+12) − em)/em, where if,m and i→f,m are average interest rates of loans taken by the firm in domestic currency and dollars, and Em(em+12) is the one-year ahead expected nominal exchange rate from the Central Bank of Chile’s survey to financial operators; em is the average spot nominal exchange rate.
- Policy timing: sovereign guarantees policy implemented May 2020; vertical lines mark February 2020 (sudden stop) and May 2020 (sovereign guarantees).

### Policy instrument: FOGAPE (Chile) — design and COVID adjustments
- Fund capitalization: 1003,000 (USD Millions)
- Interest rate (CHP): MarketMPR+3%
- Max. annual sales eligibility threshold (UF): Jan 2020 = 350,000; April 2020 (FOGAPE-COVID) = 1,000,000
- Fraction guaranteed / maximum loan value — Sales range (UF):
  - 0 - 25,000: Jan-20 = 80% - 5,000 UF; May-20 = 85% - 6,250 UF
  - 25,000 - 100,000: Jan-20 = 50% - 15,000 UF; May-20 = 80% - 25,000 UF
  - 100,000 - 350,000: Jan-20 = 30% - 50,000 UF; May-20 = 70% - 150,000 UF
  - 350,000 - 600,000: Jan-20 = Non elegible; May-20 = 70% - 150,000 UF
  - 600,000 - 1,000,000: Jan-20 = Non elegible; May-20 = 60% - 250,000 UF
  - >1,000,000: Jan-20 = Non elegible; May-20 = Non elegible
- Note: FOGAPE-COVID was triggered at the very end of April 2020. (Sources: Chilean Financial Markets Commission and the Chilean Congress.)

### Descriptive statistics — merged dataset (April 2012 to December 2020 averages, as reported)
- Mean domestic loans: 150166 USD
- Mean foreign loans: 39530000 USD
- Mean domestic interest rate (CHP - %): 13.2
- Mean foreign interest rate (USD - %): 3.3
- Mean foreign interest rate (CHP Ex-Post UIP - %): 10.2
- Standard Deviation domestic loans: 1164683 USD
- Standard Deviation foreign loans: 184548000 USD
- Standard Deviation domestic interest rate (CHP - %): 8.8
- Standard Deviation foreign interest rate (USD - %): 2.3
- Standard Deviation foreign interest rate (CHP Ex-Post UIP - %): 9.1
- Total yearly loans (% of GDP): 34.59 (domestic), 32.13 (foreign)
- Number of loans: 197262 (domestic), 69872 (foreign)
- Total yearly sales (% GDP): Domestic loans only 122.2; Foreign loans only 2.8; Domestic and Foreign Debt 32.7; All firms 157.7
- Number of firms: Domestic loans only 282922; Foreign loans only 2465703; Domestic and Foreign Debt 284090

Notes: foreign interest rate in CHP is calculated using ex-post UIP such that it = iωt + (et+12 − et)/et, where t is the corresponding month. Ratios to GDP calculated yearly from 2013 to 2020 using Chile’s nominal GDP and averaged across years.

### Interest rate comparisons and pandemic period changes
- Table A.1 (Whole Sample / March - July 2019 / March - July 2020):
  - Meani(CHP Domestic Debt - %): Whole Sample 13.2; March - July 2019 15.95; March - July 2020 5.0
  - Meani(CHP Foreign Debt - %): Whole Sample 4.5; March - July 2019 3.8; March - July 2020 3.2
  - Meani(USD Domestic Debt - %): Whole Sample 4.7; March - July 2019 6.3; March - July 2020 5.5
  - Meani(USD Foreign Debt - %): Whole Sample 3.3; March - July 2019 4.3; March - July 2020 3.5

### Regression Discontinuity and interest-rate regression key estimates
- Regression Discontinuity (Table IV): Treatment estimate (effect of becoming eligible for FOGAPE-COVID on firm-level average domestic debt share between May and July 2020):
  - Baseline (degree 0, tri): -0.09422**, Standard Error 0.05115, Observations 665
  - Alternative 1 (degree 1, tri): -0.12271*, Standard Error 0.06666, Observations 665
  - Alternative 2 (degree 0, epa): -0.09773**, Standard Error 0.0505, Observations 665
  - Alternative 3 (degree 0, epa): -0.13589*, Standard Error 0.06699, Observations 665
- Interest Rate Regression (Table V): dependent variable = interest rate of a loan
  - April 2012 to Sept 2019 (column 1): Fx coefficient -0.0395*** (standard error 0.00345), Observations 5,929,453, R-squared 0.869
  - March 2020 to July 2020 (column 2): Fx coefficient 0.00115 (standard error 0.00131), Observations 348,550, R-squared 0.646
  - March 2020 to July 2020 with eligibility interaction (column 3): Fx coefficient -0.00377* (standard error 0.00215); Fx·elegible coefficient 0.0117*** (standard error 0.00239); Observations 348,550, R-squared 0.646
- Robustness and placebo results:
  - Placebo (May–July 2019): Treatment estimates near zero (e.g., Baseline -0.00131, Clustered SE 0.05025, Observations 652).
  - Extensive robustness checks (Table VII) show Fx negative and significant in Apr 2012–Sept 2019 across many specifications, with variation in magnitude; Fx in March–July 2020 is smaller in magnitude and sometimes only marginally significant.

Notes: *,**,*** denote significance at standard levels; standard errors clustered at the firm level for main regressions.

### Model — parametrization and equilibrium numerical results
- Model parameters (Table VIII) — selected entries:
  - Gross foreign interest rate Rω = 1
  - Firms’ productivity A2 = 3
  - Concavity of the technology π = 1/2
  - First-best capital kω = 2.25
  - Upper bound on international collateral ̄ρkω = 0.2
  - Pledgeable share of output φd = 0.25
  - Initial credit supply e1,T = 1.4781
  - Central Bank supply of credit e1,CB = 0.5
  - Responsiveness of financial intermediaries’ risk-aversion to FOGAPE ς = 24
  - FCIC size $e1,CB = 0.05
  - FOGAPE size $φd = 0.02
- Equilibrium numerical analysis (Table IX): values across scenarios (Pre-Shock; Post-Shock No policies; FCIC; FOGAPE; FCIC and FOGAPE)
  - Rω: 1.00 (Pre-Shock) → 1.10 (Post-Shock No policies) → 1.10 (FCIC) → 1.10 (FOGAPE) → 1.10 (FCIC and FOGAPE)
  - R2 (equilibrium domestic interest rate): 1.10 (Pre-Shock) → 1.20 (Post-Shock No policies) → 1.15 (FCIC) → 1.12 (FOGAPE) → 1.10 (FCIC and FOGAPE)
  - Credit (total): 1.48 (Pre-Shock) → 1.44 (Post-Shock No policies) → 1.50 (FCIC) → 1.63 (FOGAPE) → 1.67 (FCIC and FOGAPE)
  - Policy parameters across scenarios:
    - φd: 0.25 (Pre-Shock) → 0.25 → 0.25 → 0.27 → 0.27
    - e1,CB: 0.50 → 0.50 → 0.55 → 0.50 → 0.55
- Mapping to figures: equilibrium labels correspond to Figure VI (Pre-Shock A; Post-Shock C; D↑ in Figure VII.a for FCIC; D↑↑ in Figure VII.b for FOGAPE; D↑↑↑ in Figure VII.c for joint policies).

### Model derivations, extensions, and sensitivity analyses
- Domestic debt derivation: domestic collateral constraint manipulated into quadratic form (Equation 20) and solved with quadratic formula; negative root ruled out due to positivity conditions.
- Credit market equilibrium: integrated solution (Equation 21 → Equation 22) separates contributions from constrained and unconstrained firms.
- TFP shock analysis:
  - A negative TFP (A2) shock:
    - Decreases first-best capital and unconstrained firms’ demand for domestic debt → tends to lower domestic interest rate.
    - Has two effects on the share of constrained firms (ˆρ): tightens collateral constraints (increasing constrained share) but lower domestic interest rate slackens constraints (decreasing constrained share); the latter dominates in the reported exercise, reducing constrained share.
    - Constrained firms increase domestic debt; unconstrained firms decrease domestic debt; aggregate effect: total domestic debt share decreases.
- Microfoundation for credit supply:
  - Financial intermediaries choose L i1, s1, m1, eCB1 to maximize returns net of origination costs %(...), with reserves m1 and FCIC eCB1 included.
  - Loan origination cost function % depends negatively on FCIC and FOGAPE (guarantees reduce origination costs).
  - First-order conditions yield:
    - %L(...) = Rb2 − Rd2 = credit spread
    - %m(...) = Rd2 − Rm2 = spread between deposit and reserve rates
    - %L(...) = RCB2 for FCIC choice, linking private benefit of FCIC to its cost
  - In the microfoundation, credit supply increases with Rb2, φd, and eCB1; both FOGAPE and FCIC complement each other in increasing credit supply.
- Sensitivity analysis: numerical sensitivity of equilibria reported by varying e1,CB and φd; reported values restricted to parameter combinations for which real solutions exist given Table VIII parametrization.

*Firm Financing During Sudden Stops: Can Governments Substitute Markets? — Working Paper No. WP/2025/072*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025072-print-pdf.pdf_