## Negative Interest Rates, Bank Balance Sheets, and Credit Supply (wpiea2019044)

## Source details

**Canonical URL:** [Negative Interest Rates, Bank Balance Sheets, and Credit Supply (wpiea2019044)](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019044.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019044.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019044.pdf.json)

---

### Context and potential confounding policies
- Credit growth to non-financial corporations in Italy:
  - May 2014: -4.7% (yearly basis).
  - End-June 2014 (immediately after NIRP introduction): -3.1%.
  - December 2014: -2.3%.
- Contemporaneous ECB measures and authors’ treatment:
  - TLTRO announced June 5, 2014; implemented starting in September 2014.
    - Authors control with a bank-level “capacity to borrow under the TLTRO” measure (share of outstanding loans and net lending to euro area NFCs and households excluding house purchase, over total assets, as of end-April 2014).
    - Benetton and Fantino (2018) show TLTRO reduced credit costs for Italian firms only after the second round (second and third quarter of 2015) — outside baseline window.
  - APP / LSAP announced January 22, 2015; implementation started March 9, 2015 — largely outside baseline window.
- Authors analyze LSAP QE and earlier policies to compare with NIRP and test whether—and why—NIRP is distinct.

### Data and institutional context (identification-relevant details)
- Double-matched administrative bank-firm monthly panel covering lending activities in Italy, 2013–2015.
  - Sources: loan-level credit register of the Bank of Italy (minimum loan size 30,000 euros); supervisory bank balance sheet data; yearly firm financials from CADS (Cerved Group).
  - Frequency: monthly (firm financials yearly; some bank balance sheet variables quarterly or semiannually).
  - Final regression sample: more than 167,000 firms borrowing from 95 banks.
  - Loan exposures: credit lines, short-term loans (maturity < 1 year), and long-term loans.
  - Interest rate data: Taxia section of the Credit Register (representative sample > 80% of total bank lending in Italy); robustness uses securities register and Thomson Reuters Eikon.
- Banking sector funding shares (March 2014 reference context):
  - Retail deposits: 69% of total funding.
  - Wholesale funding: 24% of total funding.
  - Eurosystem financing: 7% of total funding.
  - Loans to the non-financial private sector: 39% of total assets.
  - Loans to non-financial firms: 60% of the loan book.

### Identification strategy — portfolio rebalancing vs. retail deposits channel
- Theoretical channels after NIRP:
  - Portfolio rebalancing: lower yields on safer/liquid assets induce banks to shift toward loans.
  - Retail deposits channel: if negative short-term funding rates are not passed to retail deposits, margins compress and banks may reduce lending or raise fees.
- Bank-level pre-policy exposure measures (measured end-March 2014 unless otherwise noted):
  - Net interbank position = (interbank loans − interbank deposits) with maturity up to one week, divided by total assets.
    - Average March 2014: net interbank position is positive and accounts for 4% of assets.
  - Liquidity = securities divided by total assets.
    - Average March 2014: liquidity accounts for 29% of assets.
  - Net interbank position and liquidity are weakly negatively correlated; both show strong persistence and cross-sectional variation.
  - Excess reserves: average in June 2014 account for 0.001% of total assets (limited use as exposure).
- Empirical design:
  - Difference-in-differences comparing banks with different pre-policy exposures before and after June 2014 (policy implemented June 11, 2014; June observations dropped).
  - Collapsed pre/post windows: ±3 months and ±6 months (robustness includes ±1 month and ±12 months windows).
  - Main regression (collapsed pre/post):
    - ∆Loan_ib = α Net interbank position_b + β Liquidity_b + γ′ X_b + φ_i + e_ib
    - ∆Loan_ib = log difference in total loans granted by bank b to firm i between post- and pre-NIRP periods.
  - Controls:
    - Bank controls measured as of March 2014: Size (log total assets); Regulatory capital (Tier 1 / total assets); Non-performing loans (impaired loans / total assets).
    - Firm fixed effects φ_i capture credit demand shifts and unobserved firm shocks.
  - Estimator: OLS with standard errors double-clustered at bank and firm levels.
- Tests for retail deposits channel:
  - Cross-sectional evidence: banks with greater retail deposits obtain higher income from fees and commissions in months following NIRP (robust controlling for bank characteristics).
  - Direct tests (Table 2, column 4 and robustness): adding retail deposit share (% of assets) shows no evidence of a retail-deposit-driven contraction of lending; coefficients on net interbank position and liquidity remain unchanged.

### Main empirical results (baseline and magnitudes)
- Baseline windows: ±3 and ±6 months around June 2014.
- Baseline specification: jointly include net interbank position and liquidity (column 3 of Table 2 preferred).
- Net interbank position:
  - Coefficient positive, statistically significant and stable.
  - Economic magnitude: one standard deviation increase in net interbank position → increase of 1.3 percentage points in credit supply after 3 months (column 3, Table 2).
  - Effect increases to 1.8 percentage points after 6 months.
  - Table 2 reported coefficients (±3 months, column 3): net interbank position 0.1148* (SE 0.0582); (±6 months, column 3): 0.1679** (SE 0.0716).
- Bank liquidity:
  - Coefficient positive.
  - Economic magnitude: one standard deviation increase in liquidity → increase of 0.8 percentage points in credit supply after 3 months and 1.1 percentage points after 6 months.
  - Table 2 reported coefficients (±3 months, column 3): liquidity 0.0712*** (SE 0.0208); (±6 months, column 3): 0.0977*** (SE 0.0300).
- Comparison to aggregate credit trend:
  - Average growth rate of credit is -1.9%.
  - Estimated exposure effects are economically sizable relative to this -1.9% baseline.
- Interpretation:
  - Evidence supports a portfolio rebalancing channel under NIRP: more liquid banks and banks with larger net interbank positions expand credit supply.
  - This is opposite to the standard bank lending channel in normal (positive rate) times where less liquid banks respond more to easing.

### Robustness and falsification checks
- Additional liability-structure controls (secured repo, foreign funding, securities issued, interbank deposits) included — coefficients on net interbank position and liquidity remain statistically significant and similar in size (Table 3).
- Windfall gains from securities repricing:
  - Windfall gain measure (change in price end-June vs end-April 2014 × pre-announcement holdings / total equity) correlated with liquidity (correlation 0.56).
  - Including windfall gain does not alter positive significant coefficient on liquidity (Table 3, col 4).
- Alternative exposure timing:
  - Using end-June 2014 exposures (ex-post) yields similar main results (Table 3, col 5).
- Alternative estimators and windows:
  - WLS with log loan size weights produces similar point estimates (Table 3, col 6).
  - ±12 month window retains positive significant exposure effects (Table 3, col 7).
  - Accounting for delayed negative EONIA pass-through (EONIA turned negative in August): remove June–August and redefine window — baseline effects persist (Table 3, col 8).
  - Re-centering on May–June 2014 (excluding May and June observations) — results unaffected (Table 3, col 9).
- Parallel trends / placebo tests:
  - Re-center ±3 and ±6 month windows on months before NIRP — coefficients for net interbank position and liquidity are close to zero and statistically insignificant (Table 4), supporting parallel trends.
- Tests vs. other policy events (Panel A and B, Table 5):
  - July 11, 2012 rate cut (deposit facility to 0%): net interbank position and liquidity coefficients negative and significant → standard bank lending channel behavior.
  - July 2013 forward guidance: mixed and generally insignificant results.
  - January 2015 QE (APP): more liquid banks expand credit supply (portfolio rebalancing effect similar to NIRP); net interbank position coefficient smaller and not significant.
- TLTRO control:
  - Adding ex-ante TLTRO capacity control shows borrowing limits under TLTRO not associated with immediate loan supply changes after June 2014; main coefficients remain significant (Table 2, col 5).
- Narrow-window checks to rule out September 2014 announcements:
  - Recompute loan growth between July 2014 and April 2014 (drop May and June 2014 due to expectations) — portfolio rebalancing evidence remains (Table A6).

### Heterogeneity by firm risk and size, lending rates, and real effects
- Firm heterogeneity (Table 6):
  - More NIRP-exposed banks expand relatively more credit to:
    - Ex-ante riskier firms (low z-score / lower credit quality).
    - Smaller firms (below median total assets).
  - Coefficients (examples):
    - Net interbank position × risky: 0.1609** (SE 0.0778) in one specification.
    - Liquidity × risky: 0.0730*** (SE 0.0272).
    - Net interbank position × small: 0.1082** (SE 0.0532).
    - Liquidity × small: 0.0564*** (SE 0.0184).
  - No evidence that increased ex-ante risk-taking led to higher ex-post NPLs (Table A7; change in NPL ratio March 2014–March 2015 shows no deterioration for more NIRP-exposed banks).
- Pass-through to lending rates (Table 7):
  - Identification uses same firm borrowing from at least two banks with different exposures.
  - Quantitative pass-through after 6 months:
    - One standard deviation change in net interbank position → 6.9% reduction in loan rates (text summary; Table 7 coefficients: net interbank position ~ -0.0846 to -0.0996 depending on column; see table notes).
    - One standard deviation change in liquidity → 4.4% reduction in loan rates (Table 7 liquidity coefficients: -0.0452 to -0.0402 across columns).
  - Results robust to controls for retail deposits and TLTRO capacity; no evidence NIRP affected loan rates via retail deposit channel in Italy.
- Firm-level real outcomes (Table 8, ±6 month window; firm-level lender-exposure weighted averages):
  - Effects on total firm-level credit:
    - One standard deviation increase in a firm’s lenders’ net interbank position → 2.8 percentage point increase in total credit for that firm.
    - One standard deviation increase in lenders’ liquidity → 1.4 percentage point increase in total credit for that firm.
    - Average change in total firm-level credit in ±6 month window: 1.7 percent.
  - Manufacturing sample — investment and wage effects:
    - Net interbank position (one s.d.) → firm investment (growth of fixed assets) +5.6 percentage points; wage bill +3.5 percentage points.
    - Liquidity (one s.d.) → firm investment +5.2 percentage points.
  - Panel B (all firms): wage bill growth effects similar; investment effects stronger in manufacturing.
- Summary of heterogeneity:
  - NIRP-induced portfolio rebalancing concentrated toward smaller and riskier firms; pass-through to lower loan rates is economically meaningful; increased credit translates into higher investment and wage bills; no evidence of excessive deterioration in loan quality.

### Key numerical statistics and model outputs (selected exact figures)
- Sample and descriptive stats:
  - Final regression sample: more than 167,000 firms; 95 banks.
  - Table 1 bank-level March 2014 means:
    - Net interbank position, March 2014: Mean 4.200, St.Dev. 10.810, Median 1.862, Obs. 95
    - Liquidity, March 2014: Mean 28.670, St.Dev. 13.950, Median 25.940, Obs. 95
    - Retail deposits, March 2014: Mean 45.260, St.Dev. 16.120, Median 44.650, Obs. 95
    - TLTRO: Mean 35.670, St.Dev. 12.560, Median 36.370, Obs. 95
    - Windfall gain: Mean 1.378, St.Dev. 1.525, Median 0.919, Obs. 95
  - Loan-level variables (Table 1, Panel B):
    - ∆Loan: Mean -1.945, St.Dev. 20.086, Median 0.000, Obs. 495,942
    - Rate: Mean 12.515, St.Dev. 5.633, Median 11.642, Obs. 188,690
  - Firm-level variables (Table 1, Panel C):
    - ∆Loan: Mean -1.667, St.Dev. 21.840, Median -0.784, Obs. 142,302
    - Net investment: Mean 11.318, St.Dev. 75.497, Median -2.532, Obs. 127,101
    - Wage bill growth: Mean -1.045, St.Dev. 32.162, Median 1.272, Obs. 127,621
- Table 2 (baseline coefficients, ±3 and ±6 months; selected exact reported coefficients and SEs):
  - ±3 months (Observations 495,942), column 3:
    - Net interbank position: 0.1148* (0.0582)
    - Liquidity: 0.0712*** (0.0208)
    - R^2: ~0.3680
  - ±6 months (Observations 498,234), column 3:
    - Net interbank position: 0.1679** (0.0716)
    - Liquidity: 0.0977*** (0.0300)
    - R^2: ~0.3897
- Table 7 (lending rates, 6-month window, Observations 188,690):
  - Net interbank position coefficients: -0.0996*** (0.0282); -0.0842*** (0.0159); -0.0846*** (0.0169); -0.0806*** (0.0188) across columns
  - Liquidity coefficients: -0.0452*** (0.0132); -0.0427*** (0.0118); -0.0423*** (0.0139); -0.0402*** (0.0135)
  - R^2 range: 0.3959–0.3981
- Table 8 (firm-level, selected coefficients and observations preserved as reported):
  - Panel A (Manufacturing) Observations 4983; 1482; 574? (as reported)
    - Net interbank exposure (loan growth): 0.3258*** (0.0873)
    - Liquidity (loan growth): 0.0903*** (0.0224)
  - Panel B (All firms) Observations 14230; 2134960; 127621? (as reported)
    - Net interbank exposure (loan growth): 0.2627*** (0.0408)
    - Liquidity (loan growth): 0.1013*** (0.0199)

*Source: wpiea2019044 - Sections 3.2 and 4.1–4.4.*

### Section 3.2.)

### Section 3.2.

### Context and potential confounding policies
- Credit growth to non-financial corporations in Italy:
  - May 2014: -4.7% (yearly basis).
  - End-June 2014 (immediately after NIRP introduction): -3.1%.
  - December 2014: -2.3%.
- Other ECB policies around the sample window that could confound attribution to NIRP:
  - TLTRO announced June 5, 2014; implemented starting in September 2014.
    - If TLTRO had expansionary effects before implementation and bank take-up correlated with March 2014 net interbank and liquidity positions, estimates might be biased.
    - Authors control for TLTRO with a bank-level measure of ex-ante capacity to borrow under the TLTRO (based on supervisory data).
    - Benetton and Fantino (2018) show TLTRO reduced credit costs for Italian firms only after the second round, i.e. in the second and third quarter of 2015 (outside the baseline time window).
  - APP / LSAP announced January 22, 2015; implementation started March 9, 2015.
    - APP expanded purchases to include public-sector bonds for a total of 60 billion euros of monthly purchases.
    - Announcement and implementation largely outside baseline time window; APP is therefore unlikely to affect baseline estimates.
- The authors also analyze LSAP QE and earlier policies to compare with NIRP and to test whether—and why—NIRP is special.

### Data and institutional context (selected details relevant to identification)
- Dataset and coverage:
  - Double-matched administrative bank-firm monthly panel covering lending activities in Italy between 2013 and 2015.
  - Sources: (i) loan-level credit register of the Bank of Italy (minimum loan size 30,000 euros); (ii) supervisory bank balance sheet data; (iii) yearly firm financials from CADS (Cerved Group).
  - Data frequency: monthly (except firm financials yearly; some bank balance sheet variables quarterly or semiannually).
  - Final regression sample: more than 167,000 firms borrowing from 95 banks.
  - Loan exposures include credit lines, short-term loans (maturity < 1 year), and long-term loans.
  - Interest rate data: Taxia section of the Credit Register covering a representative sample > 80% of total bank lending in Italy (for robustness tests, securities register and Thomson Reuters Eikon used to construct bank-level windfall gains).
- Banking sector structure and shares:
  - Retail deposits: 69% of total funding.
  - Wholesale funding: 24% of total funding.
  - Eurosystem financing: 7% of total funding.
  - Loans to the non-financial private sector: 39% of total assets.
  - Loans to non-financial firms: 60% of the loan book.

### Identification strategy — portfolio rebalancing vs. retail deposits channel
- Theoretical channels after NIRP:
  - Portfolio rebalancing: lower yields on liquid assets shift banks’ asset allocation toward loans.
  - Retail deposits channel: if banks do not pass negative short-term funding rates to retail deposits, profit margins could shrink; banks may raise fees to compensate.
- Empirical suggestive evidence:
  - Banks with greater retail deposits obtain higher income from fees and commissions in months following NIRP introduction; correlation robust to controlling for other bank characteristics.
  - This suggestive finding implies the retail deposits channel may be ineffective in reducing lending via margin compression; the paper directly tests this in Section 4.

### Bank exposure variables used to identify the portfolio rebalancing channel
- Two pre-policy bank-level exposure measures (computed prior to policy):
  - Net interbank position:
    - Defined as (interbank loans − interbank deposits) with maturity up to one week, divided by total assets.
    - Rationale: one-week EURIBOR experienced the largest drop and became negative; interbank lending < one week represents about 70% of total interbank market activities.
    - Average in March 2014: net interbank position is positive and accounts for 4% of assets.
  - Liquidity (liquid balance sheet position):
    - Defined as securities divided by total assets.
    - Rationale: broader measure of balance sheet liquidity across maturities and asset types; downward shift in the yield curve after NIRP applied across maturities.
    - Average in March 2014: liquidity accounts for 29% of assets.
- Additional notes:
  - Net interbank position and liquidity are weakly negatively correlated in the sample.
  - Alternative exposures (e.g., excess reserve holdings) are of limited use: average excess reserves in June 2014 account for 0.001% of total assets for the typical bank in the sample.
  - Cross-sectional variation and persistence:
    - Figures (A1, A2) show strong persistence over time and significant cross-sectional variation in net interbank position and liquidity as of end-March 2014.
  - Empirical checks:
    - Multivariate bank-level regressions show banks with more liquid assets before NIRP reduce liquid assets in months after NIRP (March → September 2014).
    - Reduction in net interbank position driven by reduction in interbank loans rather than increase in interbank deposits.
    - The same tests run on the previous 6-month window (pre-NIRP) do not show similar reductions (Tables A2, A3).

### Portfolio rebalancing channel and the zero lower bound (ZLB)
- Mechanism specific to NIRP:
  - When short-term policy rates approach the ZLB, transmission of short-term cuts to long-term rates becomes weaker because markets thought negative rates unrealistic.
  - Introducing NIRP signaled that policy rates could go below zero, shifting downwards the distribution of expected short-term rates and flattening the forward curve.
  - Evidence:
    - Figure 3 shows a downward shift and flattening of the entire yield curve only after NIRP (June 2014), not after the July 2012 rate cut or the July 2013 forward guidance announcement.
    - A similar flattening occurs in January 2015 with the APP (LSAP) announcement.
    - The fall in yields of safer assets of all maturities widened the wedge between safer, more liquid assets and riskier assets (e.g., BB-rated European corporate bonds).
    - In the authors’ data, the difference in yields, before NIRP, between liquid assets and corporate loans is about 10 percentage points.

### Empirical specifications and estimation (summary)
- Difference-in-differences bank-firm approach comparing banks with different pre-policy exposure levels before and after June 2014.
- Collapsed pre- and post-NIRP periods using windows:
  - ±3 months and ±6 months around June 2014 (the policy was implemented on June 11, 2014; observations of June 2014 are dropped).
  - Robustness tests include a very short window of ±1 month.
- Bank-firm level loan growth dependent variable:
  - ∆Loan_ib = log difference in total loans granted by bank b to firm i between post- and pre-NIRP periods.
  - Estimated equation (as presented):
    - ∆Loan_ib = α Net interbank position_b + β Liquidity_b + γ′ X_b + φ_i + e_ib
- Key variables and controls:
  - Treatment variables: Net interbank position_b and Liquidity_b (both measured as of end-March 2014).
  - Bank controls (X_b), measured as of March 2014:
    - Bank size (log of total assets).
    - Regulatory capital (Tier 1 capital divided by total assets).
    - Non-performing loans (impaired loans divided by total assets).
  - Firm fixed effects (φ_i) capture credit demand shifts and unobserved firm shocks.
  - Authors report there is no systematic relationship between firm observables (size, distance from default, leverage, profitability) and bank exposure intensity (see Table A5; balancing tests).
- Estimation details:
  - Ordinary Least Squares (OLS).
  - Standard errors double-clustered at the bank and firm level to allow residual correlation within banks (treatment varies at bank level) and within firms (credit growth to the same firm may be correlated across banks).
- Extensions:
  - The same approach is used to test impacts on loan rates (change in loan rates applied by bank b to firm i between post- and pre-NIRP periods), focusing on the ±6 month window.
  - Firm-level real effects (total credit growth, investment (fixed assets growth), wage bill growth) are estimated using firm-level weighted averages of lender exposures and a specification that controls for firm credit demand ˆδ_i and province and industry fixed effects (ψ_p, φ_s).
  - Firm-level outcomes computed between end-2013 and end-2014 for the ±6 month window (firm balance sheet data are yearly); manufacturing subsample analyzed for real effects.

*Source: wpiea2019044 - Section 3.2.*

### 4.1    Main Results

### 4.1    Main Results

### Baseline Results
- Two time windows analyzed: ±3 and ±6 months around June 2014.
- Regression specifications: include net interbank position and liquidity separately (columns 1 and 2 of Table 2) and jointly (column 3).
- Net interbank position:
  - Coefficient is positive, statistically significant and stable across specifications.
  - Economic magnitude: one standard deviation increase in net interbank position is associated with an increase of 1.3 percentage points in credit supply after 3 months (based on column 3 of Table 2).
  - Effect increases to 1.8 percentage points after 6 months.
- Bank liquidity:
  - Coefficient is positive.
  - Economic magnitude: one standard deviation increase in liquidity is associated with an increase of credit supply by 0.8 percentage points after 3 months and 1.1 percentage points after 6 months.
- Comparison to average credit growth:
  - Average growth rate of credit is -1.9%.
  - The estimated effects of net interbank position and liquidity are economically sizable relative to this -1.9%.
- Interpretation:
  - Evidence supports a portfolio rebalancing channel of monetary policy transmission under NIRP: banks with greater liquidity shift toward expanding credit supply.
  - This is opposite to the standard bank lending channel in normal times, where less liquid banks react more to monetary policy (Kashyap and Stein, 2000).

### The Retail Deposits Channel
- Theoretical mechanism: negative policy rates could compress deposit margins and raise the cost of higher retail deposits if banks do not pass negative rates to retail depositors.
- Reasons the retail deposits channel may be inactive:
  - Retail deposit rate floor may not be strictly binding because of the cost of holding cash.
  - Empirical observation: banks with more retail-deposit-heavy funding (measured in % of assets in March 2014) increase fees on banking services after NIRP, possibly offsetting intermediation margin compression.
  - This correlation remains in cross-section controls for bank size, capital, liquidity, and non-performing loans (Table A1).
- Direct tests:
  - Column 4 of Table 2 adds banks’ reliance on retail deposits (share of total assets) as a control.
  - Results show no evidence of a retail deposits channel in either time window (±3, ±6 months).
  - Coefficients on net interbank position and liquidity remain unchanged when controlling for retail deposit reliance.
- Additional note: Section 4.2 and Table A6 provide further evidence inconsistent with the retail deposits channel.

### Potential Confounding Policies
- Concern: other ECB policies around the sample period (notably the TLTRO) may bias results if correlated with March 2014 net interbank and liquidity positions.
- TLTRO considerations:
  - Announced June 2014; first implemented September 2014 through 8 quarterly auctions.
  - TLTRO borrowing allowance: maximum threshold equal to 7% of the total amount of bank loans to non-financial corporations and households (excluding loans to households for house purchases) as of April 2014.
  - Constructed a bank-level measure of ex-ante “capacity to borrow under the TLTRO” as the share of outstanding loans and net lending to euro area non-financial corporations and households (excluding loans to households for house purchase) divided by total assets, as of end-April 2014.
  - Adding this TLTRO capacity control (last column of Table 2) shows differences in borrowing limits under the TLTRO are not associated with changes in loan supply immediately after June 2014.
  - Key coefficients (net interbank position and liquidity) remain statistically significant and similar to baseline.
- QE (APP) considerations:
  - QE announcement on January 22, 2015 — more than 3 months after baseline window — thus unlikely to affect baseline estimates.
  - Section 4.3 tests QE effects separately.
- Additional checks to rule out contemporaneous ECB measures (second deposit rate reduction, limited ABS and covered bond purchases announced September 2014):
  - Recompute loan growth using log difference between July 2014 and April 2014, dropping May and June 2014 (because NIRP was expected in May).
  - Table A6 shows evidence for portfolio rebalancing remains even in this narrowly defined window, ruling out effects from policies announced or executed in September 2014.

### Robustness of the Baseline (summary of Section 4.2)
- Preferred baseline: specification in column 3 of Table 2 (jointly include net interbank position and liquidity), focus on ±3 months around June 2014.
- Additional liability-structure controls added (secured repo funding, foreign funding, bank-issued securities, interbank deposits — all in % of total assets):
  - Inclusion leaves coefficients of interest statistically significant and almost identical in size (column 1 of Table 3).
  - In this saturated specification, coefficient on retail deposits is positive and statistically significant at the 10% level (contrary to retail deposits channel predictions).
- Further retail deposits exploration:
  - Column 2: remove net interbank position and leave retail deposits with standard controls — retail deposits coefficient positive and statistically insignificant.
  - Column 3: restrict sample to large banks (above-median total assets in March 2014) — retail deposits coefficient positive and statistically insignificant.
- Windfall gains from securities repricing:
  - Account for possible revaluation gains to ensure liquidity effect is not driven by repricing.
  - Windfall gain measure (change in price between end-June 2014 and end-April 2014 times pre-announcement holdings, expressed as % of total equity; restricted to “held for trading” and “available for sale” portfolios) is positively correlated with liquidity (correlation coefficient 0.56).
  - Including windfall gains does not affect the positive, significant coefficient on bank liquidity (column 4).
- Alternative exposure timing:
  - Using end-June 2014 exposure measures (ex-post) instead of end-March 2014 leaves main results unchanged (column 5).
- Alternative estimator:
  - Weighted Least Squares (WLS) with logarithm of loan size as weight yields similar point estimates for net interbank position and liquidity as OLS (column 6).
- Alternative time windows and timing adjustments:
  - Extending window to ±12 months around June 2014 still yields positive, significant effects of exposure variables (column 7).
  - Accounting for delayed pass-through of negative policy rates to overnight interbank rates (EONIA turned negative in August): remove June–August observations and redefine ±3 window around June–August 2014 — baseline effects remain statistically significant and similar (column 8).
  - Re-center analysis on ±3 months around May and June 2014 (excluding May and June observations) — results unaffected by shifting time window by one month (column 9).

### Falsification Tests (summary of Section 4.3)
- Parallel trends check:
  - Re-run baseline-style regressions centering ±3 and ±6 month windows on months before NIRP to detect pre-existing trends.
  - Results (Table 4): coefficients of net interbank position and liquidity (measured as of March 2014) centered on months before NIRP are close to zero and statistically insignificant, supporting the parallel trends assumption.
  - Note: coefficient of liquidity in column 1 of the top panel is significant but negative.
- Tests comparing NIRP to other monetary policy events:
  - Three events analyzed:
    - July 11, 2012 rate cut: deposit facility reduced by 25 basis points to 0%.
    - July 2013 forward guidance announcement.
    - January 2015 LSAP QE announcement (first QE).
  - Measurement windows: ±3 (panel A) and ±6 (panel B) months around each event.
  - Bank exposure measures taken before events: June 2012, June 2013, and December 2014 respectively.
  - Findings:
    - July 2012 rate cut:
      - Coefficients on net interbank position and liquidity are both negative and statistically significant.
      - Liquidity effect aligns with standard bank lending channel: monetary easing benefits banks with less liquidity (Kashyap and Stein, 2000).
    - July 2013 forward guidance:
      - In ±3 months, net interbank position coefficient negative; liquidity coefficient positive but much smaller than during NIRP.
      - Broader windows render these results statistically insignificant.
    - January 2015 QE:
      - More liquid banks expand credit supply, consistent with portfolio rebalancing evidence from QE programs.
      - Coefficient on net interbank position is smaller and not statistically significant.
- Overall interpretation from falsification and robustness:
  - NIRP activates a portfolio rebalancing mechanism: banks shift from liquid assets toward credit supply.
  - This rebalancing is not present in months before NIRP, nor after interest rate cuts in positive territory or after forward guidance announcements.
  - QE announcements generate portfolio rebalancing effects similar to NIRP in liquidity terms, but differ for short-term interbank positions versus broader liquid balance sheets.

*Source: wpiea2019044 - 4.1    Main Results*

### 4.4    Heterogeneity: Firm Risk and Size

### 4.4    Heterogeneity: Firm Risk and Size

### Question and approach
- Investigate whether the expansion of credit by NIRP-affected banks is homogeneous across firms or concentrated on particular firm types.
- Focus on two firm dimensions: riskiness and size.
  - Firms split around the median of total assets (end-March 2014) into small and large.
  - Firms split around the median of the Altman z-score (end-March 2014) into weak and strong credit ratings.
- Empirical strategy: include bank fixed effects and focus on interaction terms between firm characteristics and banks’ net interbank position and liquidity ratio. Results reported for ±3 month and ±6 month windows.

### Key findings on portfolio rebalancing across firms
- The relative expansion of credit supply by more NIRP-exposed banks (measured either by net interbank position or liquidity ratio) is significantly higher for:
  - ex-ante riskier firms (low credit rating, low z-score), and
  - smaller firms.
- Increase in ex-ante risk-taking by more NIRP-affected banks does not translate into higher ex-post NPLs.
  - Bank-level regressions (Table A7) show that more NIRP-exposed banks do not experience deterioration in loan quality (change in NPL ratio one year after NIRP introduction).

### Transmission to lending rates
- Identification compares loan rates for the same firm from at least two banks with different exposure to NIRP, before and after NIRP introduction.
- Negative policy rates pass through to lending rates via:
  - banks’ net interbank position, and
  - banks’ liquidity ratio.
- Quantitative pass-through (column 3, Table 7):
  - A change in one standard deviation in banks’ net interbank position leads to a 6.9% reduction of loan rates after 6 months.
  - A change in one standard deviation in banks’ liquidity leads to a 4.4% reduction of loan rates after 6 months.
- Results are robust to controls for banks’ reliance on retail deposits and “capacity to borrow under the TLTRO”.
  - No evidence that NIRP affects loan rates through the retail deposit channel in the case of Italy.

### Real effects on firms
- Firm-level regressions for the ±6 month window around June 2014 use as regressors the firm-level average of lenders’ net interbank position and liquidity (March 2014), weighted by each lender’s share of total credit to the firm.
- Controls include estimated firm fixed effects from firm-bank loan-growth regressions, and location and industry fixed effects.
- Effects on total firm-level credit (Table 8, full sample of manufacturing and services firms):
  - A one standard deviation change in a firm’s lenders’ net interbank position is associated with a 2.8 percentage point increase in total credit for that firm.
  - A one standard deviation change in a firm’s lenders’ liquidity is associated with a 1.4 percentage point increase in total credit for that firm.
  - Context: the average change in total firm-level credit in the ±6 month window is 1.7 percent.
- Effects on firm outcomes (manufacturing sample, Table 8):
  - For the net interbank position:
    - Firm investment (growth rate of fixed assets) increases by 5.6 percentage points for firms borrowing from banks with one standard deviation larger ex-ante net interbank position.
    - Wage bill (growth of total payroll) increases by 3.5 percentage points for similar firms.
  - For banks’ liquidity ratio:
    - A one standard deviation increase in banks’ liquidity is associated with a 5.2 percentage point increase in firm investment.
- Results for wage bill growth are similar in the full sample of firms (panel B). Investment effects are stronger in manufacturing firms.

### Summary conclusions
- NIRP (ECB mid-2014) induced banks to rebalance portfolios from liquid assets toward credit supply.
- This portfolio rebalancing is concentrated toward riskier and smaller firms.
- NIRP pass-through to lower loan rates is economically meaningful (6.9% via net interbank position; 4.4% via liquidity after 6 months).
- Increased credit from more NIRP-exposed banks translates into sizable firm-level real effects, including higher investment and wage bills.
- No evidence of “excessive” risk-taking as measured by subsequent NPLs.

*Source: wpiea2019044 - 4.4    Heterogeneity: Firm Risk and Size*

### References

### References (wpiea2019044)

### Major themes in the referenced literature
- Empirical and theoretical research on unconventional monetary policy, including:
  - Asset purchase programmes, quantitative easing, and large-scale asset purchases (e.g., Albertazzi et al. (2018); Altavilla et al. (2015); D‘Amico and King (2013); Krishnamurthy and Vissing-Jorgensen (2011, 2013); Rodnyansky and Darmouni (2017)).
  - Negative nominal interest rate policy (NIRP) and its bank-sector effects (e.g., Arce et al. (2018); Arseneau (2017); Arteta et al. (2018); Basten and Mariathasan (2018); Demiralp et al. (2017); Eggertsson et al. (2019); Eisenschmidt and Smets (2017); Heider et al. (2019); Lopez et al. (2018); Rostagno et al. (2016); Turk (2016); Wu and Xia (2017)).
  - Transmission of monetary policy through banks, the bank-lending channel, and credit supply responses (e.g., Acharya & Merrouche (2012); Adrian & Shin (2011); Chakraborty et al. (2016); Cingano et al. (2016); Dell’Ariccia et al. (2017); Jiménez et al. (2012, 2014, 2017, 2019); Peydró et al. (2017)).
  - Interaction of monetary policy with financial stability, risk-taking, and bank profitability (e.g., Altavilla et al. (2018); Borio & Zhu (2012); Brunnermeier & Koby (2018); Dell’Ariccia et al. (2018); Drechsler et al. (2018); Martinez-Miera & Repullo (2017); Stein (2013)).
  - Methodological contributions on identification and program evaluation in macroeconomics and microeconometrics (e.g., Imbens & Wooldridge (2009); McKay et al. (2016); Nakamura & Steinsson (2018)).

### Figures — captions and data notes
- Figure 1: Short-term Interbank Rates, January 2014–December 2014
  - Plots the ECB Deposit Facility rate and the Euribor 1 week interest rate (monthly, end-month values).
  - Solid vertical line corresponds to end June 2014, separating pre-NIRP (January 2014–June 2014) and post-NIRP (July 2014–December 2014).
  - Annotated: "ECB introduction of NIRP (-0.10%)".
  - Source: Thomson Reuters Eikon.

- Figure 2: Aggregate Credit Growth (year-on-year) to Non-Financial Corporation in Italy, 2014
  - Plots year-on-year growth rate of total bank credit to non-financial corporations.
  - Source: Bank of Italy.

- Figure 3: Forward Curves with Conventional and Unconventional Monetary Policy
  - Panels show yield curves before and after: July 2012 (rate cut), July 2013 (forward guidance), June 2014 (NIRP), and Jan. 2015 (QE/APP).
  - Source: Thomson Reuters Eikon.

- Figure 4: Negative Policy Rates and the Wedge Between Safer and Riskier Assets
  - Plots difference in yields between the BB European non-financial corporate 5-year bonds index and the European governments 20-year bonds benchmark (monthly data are period averages of daily returns).
  - Annotated: "ECB 1st decision on NIRP".
  - Source: Thomson Reuters Eikon.

### Tables — key descriptive statistics and regression estimates

- Table 1: Descriptive Statistics (values measured as of March 2014 unless otherwise noted)
  - Panel A: Bank-level variables (Mean / St.Dev. / Median / Obs.)
    - Net interbank position, March 2014: Mean 4.200, St.Dev. 10.810, Median 1.862, Obs. 95
    - Liquidity, March 2014: Mean 28.670, St.Dev. 13.950, Median 25.940, Obs. 95
    - Size: Mean 7.667, St.Dev. 2.310, Median 7.598, Obs. 95
    - Capital: Mean 8.533, St.Dev. 5.768, Median 7.079, Obs. 95
    - NPL: Mean 4.348, St.Dev. 3.555, Median 3.868, Obs. 95
    - Retail deposits, March 2014: Mean 45.260, St.Dev. 16.120, Median 44.650, Obs. 95
    - TLTRO: Mean 35.670, St.Dev. 12.560, Median 36.370, Obs. 95
    - Secured Repo: Mean 2.918, St.Dev. 8.172, Median 0.000, Obs. 95
    - Liabilities vis-a-vis non-resident: Mean 1.390, St.Dev. 2.240, Median 0.245, Obs. 95
    - Securities issued: Mean 14.520, St.Dev. 10.200, Median 14.560, Obs. 95
    - Interbank deposits: Mean 13.780, St.Dev. 9.785, Median 12.760, Obs. 95
    - Windfall gain: Mean 1.378, St.Dev. 1.525, Median 0.919, Obs. 95
    - Income fees over assets (%), June 2014: Mean 0.124, St.Dev. 0.107, Median 0.099, Obs. 83
  - Panel B: Loan-level variables
    - ∆Loan: Mean -1.945, St.Dev. 20.086, Median 0.000, Obs. 495,942
    - Rate: Mean 12.515, St.Dev. 5.633, Median 11.642, Obs. 188,690
  - Panel C: Firm-level variables
    - ∆Loan: Mean -1.667, St.Dev. 21.840, Median -0.784, Obs. 142,302
    - Net investment: Mean 11.318, St.Dev. 75.497, Median -2.532, Obs. 127,101
    - Wage bill growth: Mean -1.045, St.Dev. 32.162, Median 1.272, Obs. 127,621

- Table 2: Negative Policy Rates and Bank Credit Supply — Baseline regressions (OLS estimates of model 1)
  - Dependent variable: loan growth at the bank-firm-month level (log difference between post- and pre-NIRP).
  - Columns report results for two windows: ±3 month and ±6 month around June 2014.
  - Window: ±3 month around June 2014 (Observations 495,942)
    - Net interbank position: coefficients by column:
      - (1) 0.1354** (0.0585)
      - (2) 0.1176** (0.0575)
      - (3) 0.1148* (0.0582)
      - (4) 0.1130* (0.0608)
    - Liquidity:
      - (1) 0.0664*** (0.0194)
      - (2) 0.0632*** (0.0195)
      - (3) 0.0712*** (0.0208)
      - (4) 0.0688*** (0.0224)
    - Retail Deposits:
      - (4) 0.0145 (0.0209)
      - (5) 0.0216 (0.0241)
    - TLTRO:
      - (5) -0.0147 (0.0305)
    - Bank controls: Yes in all columns
    - Firm FE: Yes in all columns
    - R^2: 0.3679–0.3681
  - Window: ±6 month around June 2014 (Observations 498,234)
    - Net interbank position:
      - (1) 0.1966*** (0.0746)
      - (2) 0.1731** (0.0716)
      - (3) 0.1679** (0.0716)
      - (4) 0.1630** (0.0738)
    - Liquidity:
      - (1) 0.0875*** (0.0276)
      - (2) 0.0827*** (0.0270)
      - (3) 0.0977*** (0.0300)
      - (4) 0.0908*** (0.0334)
    - Retail Deposits (columns 4–5): 0.0269 (0.0290) and 0.0471 (0.0332)
    - TLTRO (column 5): -0.0414 (0.0428)
    - Bank controls: Yes in all columns
    - Firm FE: Yes in all columns
    - R^2: 0.3896–0.3898

- Table 3: Negative Policy Rates and Bank Credit Supply — Robustness (OLS; column 6 is WLS)
  - Dependent variable: loan growth at bank-firm-month level. Pre- and post-NIRP periods mostly 3 months.
  - Key coefficients (selected)
    - Net interbank position:
      - (1) 0.1210** (0.0535)
      - (2) 0.1178* (0.0665)
      - (3) 0.1111* (0.0573) — sample excludes firms in bottom half by total assets
      - (4) 0.2169** (0.0925)
      - (5) Net interbank position, June 2014: 0.1148** (0.0517)
      - (6) 0.1599** (0.0636) — WLS (log loan size weight)
    - Liquidity:
      - (1) 0.0839*** (0.0251)
      - (2) 0.0757*** (0.0205)
      - (3) 0.0581** (0.0233)
      - (4) 0.0632*** (0.0190)
      - (5) 0.0559*** (0.0196)
      - (6) 0.0954** (0.0373)
      - (7) 0.0817** (0.0376) — 12-month window
      - (9) 0.0814*** (0.0181) — window centered around May-June 2014
      - Liquidity, June 2014 in column 5: 0.0703*** (0.0175)
    - Retail Deposits:
      - (1) 0.0605* (0.0354)
      - (2) 0.0168 (0.0206)
      - (3) 0.0123 (0.0224)
    - Secured Repo (col 4): -0.0446 (0.0682)
    - Liabilities vis-a-vis non-resident (col 4): 0.3288** (0.1285)
    - Securities issued (col 4): 0.0927** (0.0381)
    - Interbank deposits (col 4): 0.0215 (0.0424)
    - Windfall gain (col 4): 0.0088 (0.3050)
  - Observations vary by column: examples include 495,942; 269,511; 503,462; 483,648; 490,397
  - Bank controls and Firm FE included in reported columns
  - R^2 reported by column: e.g., 0.3686, 0.3679, 0.3233, 0.3681, 0.4173, 0.3783, 0.3728

- Table 4: Negative Policy Rates and Bank Credit Supply — Parallel Trends (design notes)
  - Dependent variable: loan growth at the bank-firm-month level (log difference between post- and pre- monetary policy announcement).
  - Panel A: pre/post length 3 months centered around December 2013, January 2014, and February 2014 (with the month of policy announcement excluded).
  - Panel B: pre/post length 6 months centered around September 2013, October 2013, and November 2013 (with the month of policy announcement excluded).
  - Net interbank position measured as ratio of interbank loans minus interbank deposits with maturity up to one week, over total assets, as of end-March (measurement date noted).

_Italic: Source: wpiea2019044 - References (wpiea2019044) — content and figures/tables notes as provided._

### 2014.  Liquidity is the ratio of securities over total assets as of March 2014.  Bank control variables include:  i) Siz

### wpiea2019044 — Negative Interest Rates, Bank Balance Sheets, and Credit Supply (selected tables and notes)

### Key variables and measurement
- Liquidity: ratio of securities over total assets, measured as of March 2014 (and June/December 2014 in alternate specifications).
- Net interbank position: ratio of interbank loans minus interbank deposits with maturity up to one week, over total assets (measured as of end-March 2014, or June/2012/June 2013/December 2014 depending on table).
- Bank controls (measured as of March/June/December 2014 as specified):  
  - Size: logarithm of total assets.  
  - Capital: ratio of TIER 1 capital over total assets.  
  - NPL: non-performing loans scaled by total assets.
- Other bank-level measures: Retail deposits (share of total assets, March 2014), TLTRO (ratio of total loans to euro area non-financial corporations and households—excluding loans to households for house purchase—over total assets, as of April 2014).
- Standard errors: double clustered at the bank and firm level (unless otherwise noted). Significance notation: *** p<0.01, ** p<0.05, * p<0.1.

### Table 5 — Other monetary policy announcements and bank credit supply (OLS, loan growth at bank-firm-month level)
- Panel A: ±3 months windows centered on policy announcements
  - Windows: Sep 2013–Mar 2014; Oct 2013–Apr 2014; Nov 2013–May 2014 (columns 1–3 in first excerpt)
    - Net interbank position coefficients: -0.0083; 0.0187; 0.0556 (SEs: (0.0449), (0.0503), (0.0512))
    - Liquidity coefficients: -0.0847***; -0.0542; -0.0052 (SEs: (0.0282), (0.0357), (0.0280))
    - Observations: 50,878; 45,067; 45,073? (source lists: 50878 45067 45073 18 — preserve as in source: Observations508784506734507318). R2: 0.3717; 0.3698; 0.3684
  - Windows centered around: Previous Rate Cut (July 2012), Forward Guidance (July 2013), QE (APP) (January 2015)
    - Net interbank position: -0.2042***; -0.1188*; 0.0713 (SEs: (0.0548), (0.0622), (0.0548))
    - Liquidity: -0.0701*; 0.0420**; 0.0779*** (SEs: (0.0396), (0.0186), (0.0183))
    - Observations: 56035; 25273; 35479977? (preserve as in source: Observations560352527335479977). R2: 0.3699; 0.3704; 0.3612
- Panel B: ±6 months windows
  - Net interbank position: -0.2709***; -0.09890; 0.1291 (SEs: (0.0715), (0.1359), (0.0812))
  - Liquidity: -0.1046*; 0.0146; 0.1248*** (SEs: (0.0532), (0.0286), (0.0348))
  - Observations: 56285; 27994; 4481942? (preserve as in source: Observations562857529914481942). R2: 0.3856; 0.3897; 0.3802
- Interpretation: Effects of other monetary announcements on loan growth vary by bank interbank exposure and liquidity; notable negative coefficients for net interbank position around previous rate cuts, and positive liquidity associations around QE announcements.

### Table 6 — Negative policy rates and bank credit supply: firm heterogeneity (OLS, loan growth at bank-firm-month level)
- Window: ±3 months and ±6 months around June 2014 (June 2014 excluded)
- Panel A: Net interbank position interactions
  - Net interbank position x risky: 0.1609**; 0.1544*; 0.1815; 0.1721 (SEs: (0.0778), (0.0780), (0.1296), (0.1305))
  - Net interbank position x small: 0.1082**; 0.0975*; 0.1698**; 0.1532** (SEs: (0.0532), (0.0529), (0.0704), (0.0706))
  - Observations (columns 1–6): 439039; 441144; 439029; 440882; 442992; 440840 (preserve as source lists)
  - R2 range: 0.3629 to 0.3845
- Panel B: Liquidity interactions
  - Liquidity x risky: 0.0730***; 0.0690**; 0.1370***; 0.1352*** (SEs: (0.0272), (0.0275), (0.0393), (0.0399))
  - Liquidity x small: 0.0564***; 0.0511***; 0.0404; 0.0299 (SEs: (0.0184), (0.0185), (0.0271), (0.0276))
  - Observations identical to Panel A; R2 range: 0.3630 to 0.3847
- Definitions: Risky firms defined as those above the median of the Altman z-score (1 = best, 9 = worst); small firms defined below median of total assets. Standalone bank-level variables absorbed by bank fixed effects (time-invariant measures as of end-March 2014).

### Table 7 — Negative policy rates and lending rates (OLS, change in lending rates, 6-month window around June 2014)
- Dependent variable: change in lending rates (gross rates including commissions and fees)
- Net interbank position coefficients (columns 1–4): -0.0996***; -0.0842***; -0.0846***; -0.0806*** (SEs: (0.0282), (0.0159), (0.0169), (0.0188))
- Liquidity coefficients (columns 1–4): -0.0452***; -0.0427***; -0.0423***; -0.0402*** (SEs: (0.0132), (0.0118), (0.0139), (0.0135))
- Retail Deposits: 0.0007; -0.0069 (SEs: (0.0054), (0.0058)) — included in some specifications
- TLTRO: 0.0180** (SE: (0.0083)) — included in column reporting
- Observations: 188,690 (all columns). R2 range: 0.3959 to 0.3981
- Interpretation: Larger net interbank position and higher liquidity are associated with larger reductions in lending rates following NIRP.

### Table 8 — Negative policy rates, firm-level credit supply and real effects (model 2)
- Dependent variables: 1) loan growth at firm-month level (log difference over 6 months post vs pre NIRP); 2) net investment (growth rate of fixed assets between 2014 and 2013); 3) wage bill growth (2014 vs 2013).
- Firm-level bank controls: averages of bank characteristics weighted by share of credit to each firm (as of March 2014). Standard errors clustered at main bank level.
- Panel A: Manufacturing firms (columns 1–3)
  - Net interbank exposure: 0.3258***; 0.5228*; 0.3239*** (SEs: (0.0873), (0.2654), (0.1213))
  - Liquidity: 0.0903***; 0.3679***; 0.0256 (SEs: (0.0224), (0.0780), (0.0278))
  - Credit demand: 0.9833***; 0.4950***; 0.2655*** (SEs: (0.0051), (0.0328), (0.0092))
  - Observations: 4983; 1482; 574? (preserve as: Observations498314825747428). R2: 0.6231; 0.0298; 0.0551
- Panel B: All firms (columns 4–6)
  - Net interbank exposure: 0.2627***; 0.0433; 0.3157*** (SEs: (0.0408), (0.2692), (0.1150))
  - Liquidity: 0.1013***; 0.2193***; 0.0245 (SEs: (0.0199), (0.0681), (0.0324))
  - Credit demand: 1.0090***; 0.4985***; 0.3173*** (SEs: (0.0024), (0.0118), (0.0111))
  - Observations: 14230; 2134960; 127621? (preserve as in source: Observations142302134960127621). R2: 0.6080; 0.0242; 0.0554
- Interpretation: Higher bank net interbank exposure and liquidity at the firm’s lending banks are positively associated with firm loan growth and some positive associations with net investment (more pronounced in manufacturing).

### Appendix — selected figures and additional cross-sectional bank-level results
- Figure A1: Correlation plots of net interbank position (interbank loans minus interbank deposits up to one week, percent of total assets) and liquidity (securities over total assets, percent) for December 2013 and March 2014; variables winsorized at 1st and 99th percentiles. Sample source: Bank of Italy.
- Figure A2: Distributions (density plots) of net interbank position and liquidity (March 2014), winsorized at 1st and 99th percentiles. Sample includes 95 banks.
- Table A1: Fee income (bank-level cross-sectional OLS): change in income from fees over June–December 2014
  - Retail Deposits: 0.0031*** (column 1, no bank controls); 0.0023*** (column 2, with bank controls) (SEs: (0.0007), (0.0007))
  - Observations: 83; 83. R2: 0.2111; 0.4119
- Table A2: Changes between March and September 2014 (bank-level)
  - ∆net interbank position regressed on initial net interbank position: -0.2180** (SE: (0.0851))  
  - ∆interbank loans on net interbank position: -0.1644** (SE: (0.0741))  
  - ∆interbank deposits: 0.0536 (SE: (0.0545))  
  - ∆liquidity: 0.0530 (SE: (0.0501))  
  - Liquidity coefficients in separate regressions: e.g., ∆liquidity regressed on liquidity: -0.1016** (SE: (0.0394))
  - Observations: 95 in each column. R2 range: 0.0338 to 0.1985
- Table A3: Placebo (Sep 2013–Mar 2014) — changes and placebo regressions
  - Net interbank position coefficient: -0.1306* (SE: (0.0775))
  - Liquidity coefficient: 0.0433 (SE: (0.0372))
  - Observations: 95 per column. R2 range: 0.0714 to 0.1850
- Table A4: Cross-sectional correlates of March/June 2014 bank characteristics
  - Size coefficients: Net interbank position: -2.2878*** (SE: (0.4614)); Liquidity: -0.3338 (SE: (0.5773))
  - Capital: 0.2630 (Net interbank position, SE: (0.2178)); -0.8104*** (Liquidity, SE: (0.2526))
  - NPL: 0.0725 (Net interbank position, SE: (0.3030)); -1.8584*** (Liquidity, SE: (0.3580))
  - Observations: 95. R2: 0.3340 (Net interbank position), 0.2425 (Liquidity)
- Table A5: Balancing of observable firm characteristics by quartile of bank exposure to NIRP (net interbank position and liquidity). Reported quartile means and normalized differences for firm size, Z-score, Equity/Debt, Profitability (values preserved as in source).
- Table A6: Baseline regressions — 1-month window around May–June 2014 (loan growth, bank-firm-month)
  - Net interbank position: 0.1083**; 0.0992*; 0.0987*; 0.0973* (SEs: (0.0512), (0.0510), (0.0523), (0.0545))
  - Liquidity: 0.0356***; 0.0329***; 0.0343**; 0.0323** (SEs: (0.0112), (0.0110), (0.0133), (0.0132))
  - Observations: 487,882 across columns. R2: 0.3553–0.3554
- Table A7: Change in banks’ NPLs (March 2014–March 2015)
  - Net interbank position: -0.0133; -0.0101 (SEs: (0.0116), (0.0242))
  - Liquidity: -0.0213; 0.0102 (SEs: (0.0132), (0.0138))
  - Observations: 94; 94; 94 (columns with/without controls). R2: 0.0062; 0.0260; 0.4067
- Interpretation across appendix: Bank balance sheet structure (net interbank position, liquidity, retail deposits) shifted following NIRP; retail deposits positively associated with fee income growth; liquidity and interbank positions linked to changes in interbank lending/deposits.

*Source: wpiea2019044 (selected tables and notes as provided).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019044.pdf_