## Monetary Policy Transmission to Lending Rates: Evidence from Brazil

## Source details

**Canonical URL:** [Monetary Policy Transmission to Lending Rates: Evidence from Brazil](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025152-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025152-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025152-source-pdf.pdf.json)

---

### Introduction and research questions
- Paper examines strength of monetary policy transmission to bank lending rates in Brazil using data for 2012–2025.
- Key questions:
  - How strong is the pass-through from policy rates to aggregate lending rates in Brazil?
  - Has the pass-through changed significantly since 2020?
  - How does pass-through vary across loan types and bank characteristics?
- Three methodological innovations:
  - Construct daily monetary policy shocks using forecast errors from Brazil’s Focus survey (2012–2025).
  - Use local projections with an instrumental variable (IV) application to estimate pass-through over a one-year horizon.
  - Leverage daily lending rates from about eighty Brazilian financial institutions to estimate pass-through by credit type.

### Data and empirical strategy
- Aggregate data:
  - Monthly lending rates of new loans for January 2012 – April 2025 used to estimate system-wide pass-through.
  - Separate estimates for market-oriented (non-earmarked) credit versus government-directed (earmarked) credit.
  - Government-directed credit accounted for 41.7 percent of total loans at end-2024; about 42 percent of total credit in 2024 noted elsewhere in the paper.
- Bank-level data:
  - Panel dataset with daily lending rates for individual banks by credit type from January 2012 – April 2025.
  - About eighty financial institutions (including some non-bank financial institutions such as fintech lenders and financing arms of auto dealers).
  - Database covers twenty-one credit types; analysis focuses on fifteen types covering more than 90 percent of market credit.
  - Fifteen types consolidated into six groups: three consumer (unsecured personal loans, payroll-backed loans, financing of cars and goods) and three corporate (working capital loans, guaranteed overdraft, discounted trade bills).
  - Lending rates are seven-day moving averages of interest rates for new credit operations.
- Monetary policy shocks:
  - Constructed as the difference between the actual SELIC rate announced by COPOM and the expected SELIC rate from the Focus survey one day before the announcement.
  - Focus survey contains forecasts made by 171 banks, asset managers, and other institutions.
  - Mean forecast error used as the monetary policy shock (median yields similar results).

### Identification and estimation
- Method:
  - Local projections with instrumental variables following the “cumulative multiplier” approach.
  - Change in SELIC instrumented with the cumulative monetary policy surprise from the Focus survey.
  - Cumulative outcomes and cumulative policy-rate changes are used; controls include two lags of the outcome and two lags of policy-rate changes; bank fixed effects included in panel analysis.
  - Endogeneity addressed by 2SLS where z_{t,h}^c instruments s_{t,h}^c.
  - To assess changes after 2020, regressions include an interaction between policy changes and a post-2020 indicator.
- Data cleaning:
  - Noisy observations above the 95th percentile in absolute value are deleted; small banks with highly volatile daily lending rates are dropped.

### First-stage diagnostics
- Constructed monetary policy shock z_{t,h}^c is a strong instrument:
  - Associated F-statistics are above 30.
  - First-stage slope coefficients range from 4 to 5.
- COPOM and shocks:
  - The resulting shocks range from -0.5 to 0.3 percentage points.
  - In the sample period from 2012-2025, there are 106 COPOM meetings, with an average interval of 46 days between consecutive meetings.
  - Meetings generally commence on Tuesday and conclude on Wednesday; the target rate takes effect from the subsequent business day.

### Main empirical findings — aggregate pass-through
- Aggregate pass-through estimates:
  - Pass-through to the rates of all new loans peaks at 70 percent after four months.
  - Implication: raising the average bank lending rate by one percentage point requires a policy rate increase of about 1.4 percentage points (one divided by 0.7).
- Market vs. government-directed credit:
  - Market (non-earmarked) lending rates respond one-for-one to policy rate changes (full pass-through after four months).
  - Government-directed (earmarked) credit responds weakly with pass-through estimated at about 20 percent.
  - Funding-cost sensitivity:
    - Funding cost of market credit responds one-for-one to policy rates.
    - Funding cost response for directed credit is about 40 percent.
- Time variation:
  - Aggregate pass-through peaks at around 60 percent for the 2012-19 period.
  - Aggregate pass-through strengthens to 100 percent after 2020.
  - Strengthening after 2020 driven mainly by more responsive corporate loan rates and improvements in BNDES lending following the 2018 reform that aligned BNDES rates with market-based TLP.

### Heterogeneity by credit type and bank characteristics (bank-level results)
- Pass-through by credit type (market credit):
  - Corporate working capital loans: pass-through peaks at 80 percent after two months.
  - Unsecured consumer loans (credit cards and non-payroll personal loans): pass-through peaks at 80 percent after four months.
  - Payroll-deducted loans: pass-through peaks at 40 percent after ten months; payroll-backed loans remain below 40 percent in many estimates due to interest rate caps.
  - Loans financing cars and other goods: pass-through 58 percent after nine months.
- Bank size and type:
  - Larger banks exhibit stronger pass-through.
    - When limiting panel to the largest five banks, lending rates for the largest five banks respond nearly one-for-one to policy rate changes for five out of six credit types.
  - Smaller banks demonstrate significantly weaker pass-through across credit types.
  - No systematic differences found for fintech lenders or public banks relative to private commercial banks.
- Asymmetry and regulatory caps:
  - No evidence of asymmetry between tightening cycles and easing cycles.
  - Interest rate caps have weakened pass-through for payroll-deducted loans:
    - Pension-backed loan ceiling at 2.14 percent per month since January 2022, despite policy rates increasing by 5.5 percentage points, limits banks’ ability to pass higher funding costs into lending rates when caps bind.

### Stylized facts and contextual findings on credit growth and market structure
- Credit growth and gaps:
  - Despite double-digit policy interest rates since 2022, credit growth and economic activity in Brazil remained strong.
  - Central Bank of Brazil’s broad credit gap expanded in 2024 to its highest level since 2016 at 5.4 percent of GDP.
- Market structure and drivers:
  - Expansion of the domestic bond market and issuance of tax-exempt debentures contributed to corporate bond market growth and stronger pass-through.
  - Fintech lenders and digital banks grew rapidly, increasing credit availability and competition:
    - Fintech lending accounts for one-quarter of credit card loans and 10 percent of nonpayroll personal loans.
  - Financial inclusion improved, with fintech making the most contribution.
- Cross-sectional lending-rate facts:
  - Considerable variation in monthly lending rates across credit types.
  - Credit card revolving lines average 13.5 percent per month (or 357 percent at an annualized rate) during the sample period.
  - Non-payroll personal loans average 7.1 percent per month and are mostly unsecured.
- Nonperforming loans:
  - NPL ratios are higher for market than for directed credit, and higher for household loans than for corporate loans.

### Key quantitative findings (selected)
- Shocks range: -0.5 to 0.3 percentage points.
- COPOM meetings in 2012-2025: 106.
- Average interval between COPOM meetings: 46 days.
- First-stage F-statistics: above 30.
- First-stage slope coefficients: range from 4 to 5.
- Aggregate pass-through to all new loans: 70 percent after four months.
- Policy increase required to raise average bank lending rate by one percentage point: 1.4 percentage points.
- Market credit pass-through: complete after four months.
- Directed credit pass-through: about 20 percent.
- Funding-cost response for directed credit: about 40 percent.
- Aggregate pass-through: around 60 percent for 2012-19; 100 percent after 2020.
- Consumer loan pass-through range: 40 percent to 80 percent.
  - Unsecured personal loans: 80 percent after four months.
  - Car and goods loans: 58 percent after nine months.
  - Payroll-backed loans: below 40 percent.
- Corporate loan pass-through: 60 percent to 80 percent, peaking after two months.
- Top five banks: near one-for-one pass-through for five of six credit types.
- Sample statistics (selected from bank-level Table 1, monthly non-annualized rates in percent):
  - Credit Card - financing (Unsecured consumer loans): N = 121180; Mean = 8.2; SD = 3.1; Min = 1.0; p25 = 5.9; p50 = 8.1; p75 = 10.1; Max = 21.1
  - Credit Card - revolving: N = 123334; Mean = 13.5; SD = 4.5; Min = 1.1; p25 = 10.8; p50 = 14.1; p75 = 16.9; Max = 31.1
  - Non-payroll loan: N = 224142; Mean = 7.1; SD = 5.1; Min = 1.0; p25 = 3.1; p50 = 5.1; p75 = 11.0; Max = 46.9
  - Pension-deducted loan (Payroll-backed Loans): N = 119576; Mean = 1.9; SD = 0.3; Min = 1.0; p25 = 1.7; p50 = 2.0; p75 = 2.1; Max = 3.9

### Conclusions and policy implications
- Monetary policy transmission to bank lending rates in Brazil remains effective, with aggregate pass-through at 70 percent after four months and full pass-through to market lending rates.
- Government-directed credit exhibits limited pass-through (about 20 percent), implying that policy-rate changes have muted effects on a large share of total credit (41.7 percent of loans at end-2024).
- Pass-through strengthened after 2020, especially for corporate loans, supported by domestic bond market expansion and BNDES reforms.
- Heterogeneity implies policy effects depend on bank size, credit type, and regulatory constraints (e.g., interest rate caps), suggesting targeted considerations for policymakers when assessing monetary policy stance and credit conditions.

### Suggestions for future research
- Deeper analysis of interest rate pass-through to the corporate bond market, recognizing bond yields’ sensitivity to market conditions, macroeconomics, credit risk, and liquidity premia—especially for debentures without collateral.
- Investigation of the impact of tax exemptions on monetary policy transmission, given that tax benefits increase when policy rates are elevated and issuance of infrastructure debentures surged in 2024 with compressed spreads.

*Source: IMF Working Paper WP/25/152, “Monetary Policy Transmission to Lending Rates: Evidence from Brazil” (Sections 1–3).*

### Section 1

### Monetary Policy Transmission to Lending Rates: Evidence from Brazil

### Introduction and research questions
- Paper examines strength of monetary policy transmission to bank lending rates in Brazil using data for 2012–2025.
- Key questions:
  - How strong is the pass-through from policy rates to aggregate lending rates in Brazil?
  - Has the pass-through changed significantly since 2020?
  - How does pass-through vary across loan types and bank characteristics?
- Three methodological innovations:
  - Construct daily monetary policy shocks using forecast errors from Brazil’s Focus survey (2012–2025).
  - Use local projections with an instrumental variable (IV) application to estimate pass-through over a one-year horizon.
  - Leverage daily lending rates from about eighty Brazilian financial institutions to estimate pass-through by credit type.

### Data and empirical strategy
- Aggregate data:
  - Monthly lending rates of new loans for January 2012 – April 2025 used to estimate system-wide pass-through.
  - Separate estimates for market-oriented (non-earmarked) credit versus government-directed (earmarked) credit.
  - Government-directed credit accounted for 41.7 percent of total loans at end-2024; about 42 percent of total credit in 2024 noted elsewhere in the paper.
- Bank-level data:
  - Panel dataset with daily lending rates for individual banks by credit type from January 2012 – April 2025.
  - About eighty financial institutions (including some non-bank financial institutions such as fintech lenders and financing arms of auto dealers).
  - Database covers twenty-one credit types; analysis focuses on fifteen types covering more than 90 percent of market credit.
  - Fifteen types consolidated into six groups: three consumer (unsecured personal loans, payroll-backed loans, financing of cars and goods) and three corporate (working capital loans, guaranteed overdraft, discounted trade bills).
  - Lending rates are seven-day moving averages of interest rates for new credit operations.
- Monetary policy shocks:
  - Constructed as the difference between the actual SELIC rate announced by COPOM and the expected SELIC rate from the Focus survey one day before the announcement.
  - Focus survey contains forecasts made by 171 banks, asset managers, and other institutions.
  - Mean forecast error used as the monetary policy shock (median yields similar results).

### Main empirical findings: aggregate pass-through
- Aggregate pass-through:
  - Estimated aggregate pass-through stabilizes at 70 percent after four months.
  - This aggregate incomplete pass-through reflects full pass-through for market-based lending rates and partial pass-through for government-directed credit.
  - Government-directed credit pass-through estimated at 20 percent.
  - Implication: raising average bank lending rates by one percentage point requires a monetary policy rate increase of about 1.4 percentage points (one divided by 0.7).
- Market vs. government-directed credit:
  - Market (non-earmarked) lending rates respond one-for-one to policy rate changes (full pass-through).
  - Government-directed (earmarked) credit responds weakly (20 percent pass-through), consistent with historical evidence and funding structures (special funds, federal lending, tax-exempt instruments, demand deposits).

### Heterogeneity by credit type and bank characteristics (bank-level results)
- Pass-through by credit type (market credit):
  - Corporate working capital loans: pass-through peaks at 80 percent after two months.
  - Unsecured consumer loans (credit cards and non-payroll personal loans): pass-through peaks at 80 percent after four months.
  - Payroll-deducted loans: pass-through peaks at 40 percent after ten months.
- Bank size and type:
  - Larger banks exhibit stronger pass-through; when limiting panel to the largest five banks, results indicate nearly complete pass-through for most loan types (except payroll-deducted loans).
  - Smaller banks demonstrate significantly weaker pass-through, potentially reflecting lower asset quality and higher credit risk.
  - No systematic differences found for fintech lenders or public banks relative to private commercial banks.
- Asymmetry and caps:
  - No evidence of asymmetry between tightening cycles and easing cycles.
  - Interest rate caps have weakened pass-through for payroll-deducted loans:
    - Pension-backed loan ceiling at 2.14 percent per month since January 2022, despite policy rates increasing by 5.5 percentage points, limits banks’ ability to pass higher funding costs into lending rates when caps bind.

### Changes since 2020 and drivers
- Pass-through has increased since 2020 for both market and government-directed credit.
  - For market credit to firms, expansion of the domestic bond market since 2020 provided alternative financing and may have encouraged banks to be more responsive to monetary policy changes.
  - For government-directed credit to firms, stronger pass-through reflects the 2018 reform of BNDES, which aligned its lending rates more closely with market-based rates (TLP).
- Strengthening mainly driven by more responsive corporate loan rates.

### Contextual findings on credit growth and market structure
- Despite double-digit policy interest rates since 2022, credit growth and economic activity in Brazil remained strong.
- Central Bank of Brazil’s broad credit gap (deviation of total private credit from historical trend) expanded in 2024 to its highest level since 2016 at 5.4 percent of GDP.
- Factors contributing to strong credit expansion despite tight policy:
  - Expansion of the corporate bond market aided by debentures offering tax exemption benefits.
  - Rapid growth of fintech lenders and digital banks increasing credit availability and competition, leading traditional banks to reduce lending rates and improve operational efficiency.
- Cross-sectional note on lending rates and risk:
  - Considerable variation in monthly lending rates across credit types.
  - Credit card revolving lines average 13.5 percent per month (or 357 percent at an annualized rate) during the sample period.
  - Nonpayroll personal loans are the second-highest average lending rates and are mostly unsecured.

### Methodological notes and scope
- Identification strategy:
  - Monetary policy shocks identified from Focus survey forecast errors; mean forecast error used as IV for changes in the monetary policy rate in local projection estimates.
- Time horizons:
  - Local projections estimate pass-through dynamics over a one-year horizon, allowing for multi-month adjustment patterns often overlooked by studies focusing on immediate (e.g., one-week) responses.
- Coverage limits:
  - Bank-level panel excludes government-directed credit provided by the government.
  - Some niche loan products have fewer reporters; sample size varies by credit type.

*Source: IMF Working Paper WP/25/152, “Monetary Policy Transmission to Lending Rates: Evidence from Brazil” (Section 1).

### Section 2

### Section 2

### COPOM schedule and constructed shocks
- The resulting shocks range from -0.5 to 0.3 percentage points.
- In the sample period from 2012-2025, there are 106 COPOM meetings, with an average interval of 46 days between consecutive meetings.
- Meetings generally commence on Tuesday and conclude on Wednesday, at which point the SELIC target rate is determined and announced. The target rate takes effect from the subsequent business day following the meeting until a new decision is rendered in the next meeting.

### Methodology — Local projections with instrumental variables
- Interest rate pass-through is estimated using local projections with instrumental variables, following the “cumulative multiplier” approach in Jordà and Taylor (2024) and Ramey and Zubairy (2018).
- The pass-through is calculated as the average effect per intervention to account for subsequent COPOM changes after an initial monetary policy shock.
- The change in the policy rate (SELIC) is instrumented with the surprise of monetary policy derived from the Focus survey.
- Cumulative outcomes and cumulative policy-rate changes are used (e.g., y_{i,t+h}^c = y_{i,t} + ... + y_{i,t+h} and s_{t,h}^c = s_t + ... + s_{t+h}).
- Controls x_t include two lags of the outcome variables (the lending rates) and two lags of the policy rate changes; bank fixed effects α_{i,h} are included for panel analysis. For time-series aggregate analysis, the bank subscript i is dropped.
- Endogeneity is addressed by 2SLS where z_{t,h}^c (the cumulative monetary policy shock from t−1 to t+h) instruments s_{t,h}^c.
- To assess changes after 2020, an interaction between s_{t,h}^c and a post-2020 indicator is included; post-2020 pass-through equals the sum of m(h)̂ and the coefficient on the interaction term.
- For bank-level analysis, noisy observations are removed: observations above the 95th percentile in absolute value are deleted; small banks with highly volatile daily lending rates (including sudden multi-percentage-point jumps) are dropped.

### First-stage results (instrument strength)
- The constructed monetary policy shock z_{t,h}^c is a strong instrument for the cumulative change in SELIC rates s_{t,h}^c.
- Associated F-statistics are above 30.
- First-stage slope coefficients range from 4 to 5.

### Aggregate pass-through estimates
- Second-stage pass-through estimates use horizons up to twelve months; point estimates shown with 90 percent confidence intervals and a reference line at unity for full pass-through.
- The pass-through to the rates of all new loans peaks at 70 percent after four months.
  - Implication: raising the average bank lending rate by one percentage point requires a policy rate increase of 1.4 percentage points (one divided by 0.7).
- The incomplete aggregate pass-through is driven by government-directed credit:
  - Pass-through to market credit is complete after four months.
  - Pass-through to directed credit is about 20 percent.
- Differential responses reflect funding-cost sensitivity:
  - Funding cost of market credit responds one-for-one to policy rates.
  - Funding cost response for directed credit is about 40 percent.
- Time variation:
  - Aggregate pass-through peaks at around 60 percent for the 2012-19 period.
  - Aggregate pass-through strengthens to 100 percent after 2020.
  - Both market operations and directed credit respond more after 2020.
  - Within market operations, corporate loans explain stronger and more immediate pass-through.
  - Within government-directed credit, improvement comes from BNDES loans.
- Institutional and market drivers:
  - Expansion of the domestic bond market and tax exemptions approved in 2011 contributed to corporate bond market growth, encouraging longer maturities and reduced spreads for issuers and strengthening pass-through.
  - The 2018 BNDES reform switched financing contracts from TJLP to TLP, aligning BNDES rates more closely with market rates and strengthening pass-through.

### Pass-through by credit types (bank-level panel analysis)
- Estimates are based on daily lending rates from around eighty banks for each credit type.
- Consumer loans (panel A):
  - Estimated pass-through ranges from 40 percent to 80 percent.
  - Unsecured personal loans: 80 percent after four months.
  - Loans financing cars and other goods: 58 percent after nine months.
  - Payroll-backed loans: below 40 percent (weak response reflecting interest rate caps imposed by INSS and other government agencies).
- Corporate loans (panel B):
  - Pass-through peaks after two months and ranges from 60 percent to 80 percent.
  - Faster corporate response may reflect greater financial sophistication, narrower spreads, and alternative financing (domestic bond market).
- Panel data versus aggregate results:
  - Panel results are incomplete for all six credit types, in contrast to full pass-through in aggregate data, because:
    - Aggregate pass-through is dominated by larger banks due to their substantial loan portfolios.
    - The panel approach gives equal weight to each bank.
  - When panel analysis is limited to the top five banks by loan portfolio size:
    - Pass-through estimates are larger than for smaller banks.
    - Lending rates for the largest five banks respond nearly one-for-one to policy rate changes for five out of six credit types.
  - For smaller banks, pass-through is smaller across credit types.
  - Payroll loans remain low in pass-through for all banks due to interest rate caps regardless of bank size.

### Key quantitative findings
- Shocks range: -0.5 to 0.3 percentage points.
- COPOM meetings in 2012-2025: 106.
- Average interval between COPOM meetings: 46 days.
- First-stage F-statistics: above 30.
- First-stage slope coefficients: range from 4 to 5.
- Aggregate pass-through to all new loans: 70 percent after four months.
- Policy increase required to raise average bank lending rate by one percentage point: 1.4 percentage points.
- Market credit pass-through: complete after four months.
- Directed credit pass-through: about 20 percent.
- Funding-cost response for directed credit: about 40 percent.
- Aggregate pass-through: around 60 percent for 2012-19; 100 percent after 2020.
- Consumer loan pass-through range: 40 percent to 80 percent.
  - Unsecured personal loans: 80 percent after four months.
  - Car and goods loans: 58 percent after nine months.
  - Payroll-backed loans: below 40 percent.
- Corporate loan pass-through: 60 percent to 80 percent, peaking after two months.
- Top five banks: near one-for-one pass-through for five of six credit types.

### Conclusion and implications
- The transmission of monetary policy to bank lending rates in Brazil remains effective.
- Aggregate interest rate pass-through is estimated at 70 percent after four months, with full pass-through to market lending rates and 20 percent pass-through to government-directed credit.
- Evidence suggests pass-through strengthened after 2020, particularly for corporate loan rates.
- Pass-through varies across credit types (40 percent to 80 percent) and across bank size (larger banks more responsive).
- Countervailing factors likely explain sustained credit growth despite high policy rates, including structural changes such as the rapid expansion of fintech lenders and capital deepening via tax-exempt debentures.

### Suggestions for future research
- Deeper analysis of interest rate pass-through to the corporate bond market, recognizing bond yields’ sensitivity to market conditions, macroeconomics, credit risk, and liquidity premia—especially for debentures without collateral.
- Investigation of the impact of tax exemptions on monetary policy transmission, given that tax benefits increase when policy rates are elevated and issuance of infrastructure debentures surged in 2024 with compressed spreads.

*IMF Working Paper — Monetary Policy Transmission to Lending Rates: Evidence from Brazil (Section 2).*

### Section 3

### Section 3

### Stylized facts: Brazil’s credit market
- Credit growth remains strong.
- The domestic bond market is raising the credit gap.
- Lending rates for market credit are around 50 percent, with higher rates for consumer loans than for firm loans.
- Lending rates on government-directed credit are lower and less responsive to policy rate changes.
- NPL ratios are higher for market than for directed credit, and higher for household loans than for corporate loans.
- After the reform in 2018, BNDES funding rates are more aligned with SELIC rates.

### Monetary policy shocks
- Figure labeled "Monetary Policy Shocks" reports shocks in percentage points.

### Summary statistics: Bank lending rates (highlights from Table 1)
Note: Table reports monthly (non-annualized) interest rates in percent as some of the loans have short maturities. Blue font indicates loans to firms; black font indicates loans to households (excluding mortgages which are government directed credit).

- Financing other: N = 125159; Mean = 3.7; SD = 2.1; Min = 0.2; p25 = 1.8; p50 = 3.6; p75 = 5.3; Max = 13.4
- Financing cars: N = 141291; Mean = 1.9; SD = 0.7; Min = 0.5; p25 = 1.4; p50 = 1.8; p75 = 2.3; Max = 5.2
- Working capital (<1 year, fixed): N = 137762; Mean = 2.4; SD = 1.3; Min = 0.5; p25 = 1.6; p50 = 2.1; p75 = 2.8; Max = 20.8
- Working capital (<1 year, float): N = 78739; Mean = 1.5; SD = 0.6; Min = 0.5; p25 = 1.1; p50 = 1.5; p75 = 1.8; Max = 12.7
- Working capital (>1 year, fixed): N = 124838; Mean = 2.0; SD = 0.8; Min = 0.5; p25 = 1.5; p50 = 1.9; p75 = 2.4; Max = 12.0
- Working capital (>1 year, float): N = 88084; Mean = 1.4; SD = 0.4; Min = 0.5; p25 = 1.1; p50 = 1.4; p75 = 1.6; Max = 4.9
- Credit Card - financing (Unsecured consumer loans): N = 121180; Mean = 8.2; SD = 3.1; Min = 1.0; p25 = 5.9; p50 = 8.1; p75 = 10.1; Max = 21.1
- Credit Card - revolving: N = 123334; Mean = 13.5; SD = 4.5; Min = 1.1; p25 = 10.8; p50 = 14.1; p75 = 16.9; Max = 31.1
- Non-payroll loan: N = 224142; Mean = 7.1; SD = 5.1; Min = 1.0; p25 = 3.1; p50 = 5.1; p75 = 11.0; Max = 46.9
- Discounted trade bills (DTB): N = 156731; Mean = 2.3; SD = 1.2; Min = 0.2; p25 = 1.5; p50 = 2.2; p75 = 3.0; Max = 26.7
- Guaranteed overdraft (fixed): N = 102572; Mean = 3.3; SD = 1.9; Min = 1.0; p25 = 2.2; p50 = 2.8; p75 = 3.9; Max = 14.8
- Guaranteed overdraft (float): N = 106824; Mean = 1.8; SD = 0.6; Min = 0.5; p25 = 1.5; p50 = 1.8; p75 = 2.1; Max = 11.8
- Pension-deducted loan (Payroll-backed Loans): N = 119576; Mean = 1.9; SD = 0.3; Min = 1.0; p25 = 1.7; p50 = 2.0; p75 = 2.1; Max = 3.9
- Private payroll loan: N = 158330; Mean = 2.7; SD = 0.8; Min = 1.0; p25 = 2.1; p50 = 2.6; p75 = 3.0; Max = 8.8
- Public payroll loan: N = 127991; Mean = 2.0; SD = 0.9; Min = 1.0; p25 = 1.6; p50 = 1.8; p75 = 2.1; Max = 9.0

### First stage: identification diagnostics
- Figure 3 reports:
  - Estimated slope coefficient for the relationship between the cumulative change in the SELIC policy rate and the cumulative monetary policy shock (the instrumental variable) together with 90 percent confidence bands for each horizon in terms of the number of COPOM meetings.
  - COPOM frequency: there are eight COPOM meetings each year with an average interval of 46 days.
  - F-statistic for the first stage is reported (panel B).

### Aggregate interest rate pass-through
- Figure 4 reports estimated responses of lending rates to monetary policy changes together with 90 percent confidence bands for:
  - Lending Rates: All New Loans
  - Lending Rates: Market vs. Directed Credit
  - Funding Cost: Market vs. Directed Credit

### Interest rate pass-through: 2012-19 vs. 2020-25
- Figure 5 reports estimated responses of lending rates to monetary policy changes together with 90 percent confidence bands for:
  - All New Loans
  - Market Lending Operations
  - Market Lending: Consumer Loans
  - Market Lending: Firms
  - Government-Directed Credit
  - Government-Directed Credit: BNDES

### Interest rate pass-through by credit types and banks
- Figure 6 reports estimated responses of daily lending rates to monetary policy changes using bank-level panel data for:
  - Consumer Loans (aggregate)
  - Firm Loans (aggregate)
  - Consumer Loans: Big Five
  - Firm Loans: Big Five
  - Consumer Loans: Other Banks
  - Firm Loans: Other Banks
- Note: “Big Five” indicates largest five banks based on loan volume.

### Drivers of credit growth
- Firms are increasingly relying on domestic bond market financing.
- Issuance of tax-exempt infrastructure bonds surged in 2024.
- Fintech lending (including digital banks) accounts for one-quarter of credit card loans and 10 percent of nonpayroll personal loans.
- Financial inclusion has also improved, with fintech making the most contribution.

*IMF WORKING PAPERS Monetary Policy Transmission to Lending Rates: Evidence from Brazil*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025152-source-pdf.pdf_