## _wp13118

## Source details

**Canonical URL:** [_wp13118](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13118.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13118.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13118.pdf.json)

---

### I. Introduction — purpose and key findings
- Research question: Evaluate heterogeneity in bank lending responses to target federal funds rate changes that are plausibly exogenous to expected output growth, inflation, unemployment, and capacity utilization; compare these responses to those estimated from realized federal funds rate changes.
- Conceptual motivation:
  - Lending channel: banks with greater access to non-deposit financing shield lending growth in tight monetary policy (Kashyap and Stein, 2000; Kishan and Opiela, 2000; Loutskina, 2011; Ashcraft, 2006; Cetorelli and Goldberg, 2012).
  - Identification challenge: realized federal funds rate changes reflect FOMC responses to expected output growth and inflation, potentially confounding loan supply and loan demand effects.
- Four highlighted empirical differences between responses to exogenous (UM) versus realized (FF) monetary policy:
  - One year after an exogenous monetary contraction, the reduction in lending growth at the average bank not part of a holding company is up to twice that from a rise in the realized federal funds rate.
  - The extent to which a bank can shield lending growth (via large asset base or funds from affiliates in a holding company) is up to four times larger when estimated in response to identified, exogenous monetary policy.
  - Securities-to-assets share mitigates the lending response to a realized federal funds rate increase, but amplifies the lending contraction to an exogenous federal funds rate increase. Cash-to-assets ratio shields lending growth from an exogenous monetary tightening but is only weakly correlated with lending responses to realized federal funds rate increases.
  - Post-1987 (source of strength doctrine), effects of asset composition on lending responses occur only among banks not part of a holding company; affiliated banks smooth lending via internal capital markets, rendering balance sheet composition unrelated to lending responses.
- Implication: forward-looking component of endogenous monetary policy matters; studies of bank lending behavior should account for policymakers’ forecasts and policy endogeneity.

### II. Bank lending and monetary policy — mechanisms of bias and interpretation
- Empirical specification (illustrative):
  - ΔL_{i,t} = α + β M_t + γ' S_{i,t} + δ' (S_{i,t} ∘ M_t) + φ' Z_t + ε_{i,t}
  - ΔL: percentage growth of total loans at current prices; M: monetary policy measure; S: vector of bank-specific characteristics; Z: control variables; interactions S ∘ M capture heterogeneity.
- Sources of bias when using realized federal funds rate (FF):
  - Endogenous policy: FOMC reacts to forecasts of output and inflation; such forecasts also affect loan demand and loan supply, producing omitted variable bias.
  - Loan demand channels: tighter policy implemented because of expected stronger output/inflation can coincide with rightward shifts in loan demand, attenuating observed lending reductions.
  - Loan supply channels: expected economic growth can increase access to non-reservable finance (equity issues, securitization) or reduce external finance premium, offsetting policy-induced supply contractions.
- Two examples of bias affecting heterogeneity estimates (δ coefficients):
  - (i) Expected macroeconomic improvements induce loan supply shifts that depend on bank characteristics (constrained banks expand more in expansions), attenuating the estimated effect of monetary tightening for constrained banks.
  - (ii) Expected macroeconomic changes induce bank-specific loan demand shifts correlated with bank characteristics (rational buffer stocking; cyclically sensitive customer mixes), also attenuating or biasing estimates of the shielding effect of liquidity or holding company status.

### III. Econometric methodology — identification of exogenous policy and regression framework
- Identification strategy: extend Romer and Romer (2004) two-step procedure to isolate exogenous variation in the federal funds target.
  - Step 1: Construct series of targeted federal funds rate changes at scheduled FOMC meetings using Romer and Romer (1969–96) narrative series plus appended FOMC announced target changes for 1997–2006.
  - Step 2: Regress targeted federal funds rate change on Federal Reserve Greenbook real-time forecasts for real output growth (Δy), inflation (π), unemployment (n), and manufacturing capacity utilization (û) across horizons up to two quarters; residual captures exogenous policy innovation (UM).
- Sample and key identification details:
  - Sample of 346 FOMC meetings from 1969–2006.
  - Inclusion of capacity utilization and additional unemployment terms raises R̄^2 to 31% as compared to 28% in original Romer and Romer (2004) specification.
  - Aggregation to quarterly frequency: cumulate post-meeting identified policy changes at daily frequency within a quarter to form UM; construct analogous quarterly changes in effective federal funds rate (FF) by daily cumulation.
  - Comparison of UM and FF (1969q2 to 2006q4):
    - Standard deviation of UM is 57 basis points.
    - Standard deviation of FF is 204 basis points.
    - Correlation(UM, FF) = 0.57.
- Identification assumptions:
  - Greenbook forecasts are not a function of the contemporaneous change in the federal funds rate target.
  - Greenbook forecasts account for factors driving policy that are also determinants of bank lending via expected economic conditions.
- Potential threat: Fed responses to banking sector conditions not captured by Greenbook forecasts could bias residuals; sample excludes post-2007 sub-prime crisis but includes episodes around Basel I (1988/1992) and LTCM rescue (1998).

### IV. Data and measurement notes
- Bank-level data:
  - Source: Reports of Condition and Income (“Call Reports”) filed with FDIC; end-of-quarter balance sheet reports.
  - Sample extended from 1968 to 2006; banks reported consistently exactly at the end of the quarter only from 1975.
  - Baseline dataset: unbalanced quarterly panel 1969q3 to 2006q4; maximum of 15,306 banks and minimum of 7,922 banks; average observations per bank = 112 quarters.
  - Outlier exclusion: from total 1,724,924 observations, sample reduced to 1,724,355 when UM is policy measure and 1,724,334 when FF is policy measure after DFITS-based trimming.
- Key variable definitions:
  - Dependent variable: total loans minus allowances for loan losses; includes loans under commitment and loans on flexible terms.
  - Total bank assets: reported net of loan loss reserves.
  - Securities: Total Investment Securities + Assets Held in Trading Accounts.
  - Total Equity Capital: book value of equity issued + cumulated retained earnings.
  - Cash: cash on the asset side of the balance sheet.
  - Holding company status: indicator set to unity post-1986 if a bank is part of a bank holding company (dated t rather than t−1).
- Macroeconomic data:
  - Income growth: seasonally adjusted real GDP (BEA via FRED).
  - Inflation: seasonally adjusted personal core consumption expenditure price index (PCE) (BEA via FRED).
- Aggregation rationale: use end-of-period bank balances to match Call Report timing and avoid intra-quarter averaging distortions.

### V. Representative dynamic lending responses (UM versus FF) — key statistics (Table 3)
- Dynamic responses are sums of contemporaneous and four lagged responses to a 100 basis point tightening. Standard errors clustered at bank level.
- Horizon 0 (QoQ)
  - UM: -0.3068, s.e. 0.0490
  - FF: -0.1360, s.e. 0.0177
  - YoY: UM -0.3240, s.e. 0.0727; FF -0.1125, s.e. 0.0328
- Horizon 1
  - UM: -0.9060, s.e. 0.0810
  - FF: -0.4259, s.e. 0.0396
  - YoY: UM -2.0864, s.e. 0.1133; FF -0.8077, s.e. 0.0697
- Horizon 2
  - UM: -1.9893, s.e. 0.1342
  - FF: -0.8124, s.e. 0.0690
  - YoY: UM -3.7095, s.e. 0.1607; FF -1.6786, s.e. 0.1070
- Horizon 3
  - UM: -2.5439, s.e. 0.1650
  - FF: -1.1207, s.e. 0.0916
  - YoY: UM -4.5173, s.e. 0.2164; FF -2.3476, s.e. 0.1357
- Horizon 4
  - UM: -3.2841, s.e. 0.1960
  - FF: -1.4813, s.e. 0.1068
  - YoY: UM -5.2595, s.e. 0.2686; FF -3.0291, s.e. 0.1521
- Findings:
  - At each horizon, lending reduction estimated from exogenous monetary policy (UM) exceeds that from realized federal funds rate (FF).
  - 95% confidence intervals for UM and FF estimates are non-overlapping at all horizons beyond the current quarter.
  - Autoregressive parameters are small (< 0.1) and virtually identical across UM and FF.

### VI. Heterogeneity in lending responses — size, affiliation, and balance sheet composition (Table 4 and Table 5 highlights)
- Representative bank marginal effect (sum through horizon 4)
  - UM (QoQ): -3.2841, s.e. 0.1960
  - FF (QoQ): -1.4813, s.e. 0.1068
  - YoY UM: -5.2595, s.e. 0.2686; YoY FF: -3.0291, s.e. 0.1521
- Holding Company Status interaction (effect when indicator = 1)
  - UM interaction: 5.5366, s.e. 0.4644
  - FF interaction: 2.7783, s.e. 0.2188
  - Marginal effect for affiliated bank (sum)
    - UM: 2.2526, s.e. 0.3971
    - FF: 1.2970, s.e. 0.1815
    - YoY marginal: UM 5.8324, s.e. 0.7730; FF 2.8215, s.e. 0.3237
- Assets interaction (1 standard deviation above mean)
  - UM interaction: 1.0616, s.e. 0.1702
  - FF interaction: 0.0954, s.e. 0.0891
  - Marginal effect
    - UM: -2.2225, s.e. 0.2225
    - FF: -1.3859, s.e. 0.1453
    - YoY marginal: UM -3.0612, s.e. 0.3944; FF -2.5587, s.e. 0.2168
- Securities/Assets interaction
  - UM interaction: -1.1357, s.e. 0.1453
  - FF interaction: -0.0443, s.e. 0.0853
  - Marginal effect
    - UM: -4.4197, s.e. 0.2578
    - FF: -1.5256, s.e. 0.1027
    - YoY marginal: UM -7.0822, s.e. 0.4589; FF -2.8567, s.e. 0.2265
- Cash/Assets interaction
  - UM interaction: 0.4139, s.e. 0.1656
  - FF interaction: 0.0249, s.e. 0.0906
  - Marginal effect
    - UM: -2.8701, s.e. 0.2462
    - FF: -1.4564, s.e. 0.1446
    - YoY marginal: UM -4.2783, s.e. 0.5116; FF -2.9304, s.e. 0.2997
- Equity Capital/Assets interaction
  - UM interaction: 0.0177, s.e. 0.4510
  - FF interaction: -0.2723, s.e. 0.2837
  - Marginal effect
    - UM: -3.2664, s.e. 0.5462
    - FF: -1.7536, s.e. 0.3511
    - YoY marginal: UM -5.5633, s.e. 1.1313; FF -3.7669, s.e. 0.7828
- Interpretation:
  - Total assets and holding company affiliation help banks shield lending growth from policy contractions, with effects much larger when using UM than FF.
  - Holding company banks appear to raise lending following an exogenous policy contraction (UM), while unaffiliated banks reduce lending.
  - Securities holdings amplify lending contractions under UM but appear to mute policy effects under FF.
  - Cash holdings shield banks from exogenous monetary contractions.
  - Equity capital shows little consistent shielding effect in baseline specifications.

### VII. Source of strength doctrine (post-1986) and bank holding company lending responses
- Federal Reserve Board statement (April 1987) on parent injections of liquidity into distressed subsidiaries altered dynamics of affiliated banks.
- Post-1986 findings:
  - Interaction terms for affiliated banks show post-1986 changes approximately equal in magnitude but opposite in sign to main interaction effects (one exception: FF with assets).
  - "Total effect of balance sheet related characteristics on lending responses to monetary policy, both exogenous and endogenous, is close to zero during the second half of the sample for affiliated banks."
  - Affiliation becomes the principal source of heterogeneity in the late 1980s and 1990s; affiliated banks smooth lending via internal capital markets.
- Selected coefficients from Table 5 (sums of contemporaneous and four lagged responses to a 100 b.p. tightening; s.e. in parenthesis):
  - Representative Bank marginal effect examples: -3.4154 (s.e. 0.2040); -1.5394 (s.e. 0.110); -5.0797 (s.e. 0.2734); -3.1292 (s.e. 0.166)
  - Holding Company Status interaction examples: 6.2261 (s.e. 0.6826); 3.018 (s.e. 0.3007); 38.525 (s.e. 2.6354); 55.398 (s.e. 0.5261)
  - Securities Ratio interaction examples: -1.5012 (s.e. 0.1791); -0.0551 (s.e. 0.1045); -2.2442 (s.e. 0.410); 0.0476 (s.e. 0.2398)
  - Cash Ratio interaction examples: 0.2837 (s.e. 0.1852); 0.0557 (s.e. 0.110); 1.3405 (s.e. 0.4265); 0.0968 (s.e. 0.2082)

### VIII. Robustness checks (section V highlights)
- Episodic dummies:
  - Dummy = 1 for 1990-1993 (Basel I related tightening) and for 1998-1999 (LTCM episode).
  - Interacting these with monetary policy terms produced only marginal increases in absolute size of estimated monetary policy effect on lending growth.
- Bank-level fixed effects: augmenting specification yields generally robust main findings.
- Predetermined characteristics (dating at t−1): preserves direct effect of policy and holding company status; UM-based size effects remain consistent.
- Use of Survey of Professional Forecasters as alternative forecasts: direct policy coefficients shrink somewhat, but difference between UM and FF remains meaningful (response three times as large in the quarterly model for UM vs FF; orders of magnitude larger in year-over-year quarterly model).
- Table 6 reported marginal effects and interactions across many specifications; UM responses consistently larger than FF responses in robustness checks.

### IX. Data summary (selected bank-level statistics, Q4 snapshots)
- Loan Growth
  - 1970 Mean 10.5589 Std. Dev. 23.2026
  - 1980 Mean 4.5732 Std. Dev. 16.6064
  - 1990 Mean 9.834 Std. Dev. 34.5314
  - 2000 Mean 15.6723 Std. Dev. 38.1374
- Assets
  - 1970 Mean 46,670 Std. Dev. 496,413
  - 1980 Mean 139,751 Std. Dev. 1,735,830
  - 1990 Mean 288,878 Std. Dev. 2,496,265
  - 2000 Mean 778,952 Std. Dev. 10,220,686
- Holding Company Status
  - 1970 Mean 0.1549 Std. Dev. 0.3618
  - 1980 Mean 0.3310 Std. Dev. 0.4706
  - 1990 Mean 0.6883 Std. Dev. 0.4632
  - 2000 Mean 0.7558 Std. Dev. 0.4296
- Loans/Assets
  - 1970 Mean 0.4809 Std. Dev. 0.1100
  - 1980 Mean 0.5407 Std. Dev. 0.1147
  - 1990 Mean 0.5475 Std. Dev. 0.1574
  - 2000 Mean 0.6286 Std. Dev. 0.1440
- Deposits/Assets
  - 1970 Mean 0.8892 Std. Dev. 0.0305
  - 1980 Mean 0.8839 Std. Dev. 0.0424
  - 1990 Mean 0.8828 Std. Dev. 0.0774
  - 2000 Mean 0.8321 Std. Dev. 0.0937
- Securities/Assets
  - 1970 Mean 0.3433 Std. Dev. 0.1173
  - 1980 Mean 0.2822 Std. Dev. 0.1163
  - 1990 Mean 0.2881 Std. Dev. 0.1546
  - 2000 Mean 0.2513 Std. Dev. 0.1360
- Equity Capital/Assets
  - 1970 Mean 0.0821 Std. Dev. 0.0244
  - 1980 Mean 0.0855 Std. Dev. 0.0246
  - 1990 Mean 0.0867 Std. Dev. 0.0434
  - 2000 Mean 0.1039 Std. Dev. 0.0462
- Cash/Assets
  - 1970 Mean 0.1231 Std. Dev. 0.0570
  - 1980 Mean 0.0923 Std. Dev. 0.0550
  - 1990 Mean 0.0715 Std. Dev. 0.0536
  - 2000 Mean 0.0490 Std. Dev. 0.0418
- Number of Banks (Q4)
  - 1970 13,195.0
  - 1980 14,351.00
  - 1990 12,365.0
  - 2000 8,309.0

### X. Conclusion and policy implications
- Main findings:
  - Using an explicitly identified exogenous monetary policy measure (UM) versus the realized federal funds rate (FF) reveals economically and statistically significant attenuation of estimated lending responses to monetary contractions when endogeneity is not accounted for.
  - Bank holding company affiliation provides a shielding effect on lending growth following policy tightenings; affiliation becomes the dominant source of heterogeneity in the late 1980s and 1990s.
  - Securities share in total assets amplifies transmission from exogenous interest rate changes but restricts transmission of realized interest rate changes.
- Research and policy recommendations:
  - Future studies on monetary transmission should consider exogenous policy measures alongside realized measures such as the federal funds rate.
  - Identification of exogenous monetary policy should incorporate forward-looking drivers of policy (growth and inflation forecasts), because these forward-looking variables likely impact lending markets.
  - Suggested avenues for future work:
    - Examine impact of exogenous monetary policy upon regional, industrial, and firm-specific lending and compare with realized federal funds rate effects.
    - Apply the methodology to study heterogeneity in lending rates as opposed to lending quantities.
    - Investigate effects of bank characteristics on responses to policy at a more disaggregated level, e.g., the particular mixture of securities held as bank assets.

*Source: _wp13118 (PDF chapter/section).*

### 1. Policy Identification Regression ....................................................................................

### 1. Policy Identification Regression

### Major Sections (chapter/section listing)
- 1. Policy Identification Regression ..........................................................................................12
- 2. Bank Level Summary Statistics ...........................................................................................23
- 3. Comparison of Dynamic Lending Response of a Representative Bank ..............................24
- 4. Heterogeneity in Lending Responses due to Bank Characteristics .....................................26
- 5. Bank Holding Company Lending Responses  .....................................................................30
- 6. Robustness Checks...............................................................................................................33

### Figures (figure listing)
- FIGURES
  - 1. Endogenous and Exogenous Monetary Policy Shocks ........................................................15
  - 2. Macroeconomic Controls During the Sample ......................................................................19
  - 3. Imperfectly Measured Variables Prior to 1975 ....................................................................20
  - 4. Assumptions Regarding Timing ..........................................................................................20
  - 5. Assumptions Regarding Interpolation .................................................................................21

*Source: _wp13118 - 1. Policy Identification Regression ....................................................................................*

### References .............................................................................................................

### _wp13118 - References

### I. Introduction — purpose and key findings
- Research question: Evaluate heterogeneity in bank lending responses to target federal funds rate changes that are plausibly exogenous to expected output growth, inflation, unemployment, and capacity utilization; compare these responses to those estimated from realized federal funds rate changes.
- Conceptual motivation:
  - Lending channel: banks with greater access to non-deposit financing shield lending growth in tight monetary policy (Kashyap and Stein, 2000; Kishan and Opiela, 2000; Loutskina, 2011; Ashcraft, 2006; Cetorelli and Goldberg, 2012).
  - Identification challenge: realized federal funds rate changes reflect FOMC responses to expected output growth and inflation, potentially confounding loan supply and loan demand effects.
- Four highlighted empirical differences between responses to exogenous versus realized monetary policy:
  - One year after an exogenous monetary contraction, the reduction in lending growth at the average bank not part of a holding company is up to twice that from a rise in the realized federal funds rate.
  - The extent to which a bank can shield lending growth (via large asset base or funds from affiliates in a holding company) is up to four times larger when estimated in response to identified, exogenous monetary policy.
  - Securities-to-assets share mitigates the lending response to a realized federal funds rate increase, but amplifies the lending contraction to an exogenous federal funds rate increase. Cash-to-assets ratio shields lending growth from an exogenous monetary tightening but is only weakly correlated with lending responses to realized federal funds rate increases.
  - Post-1987 (source of strength doctrine), effects of asset composition on lending responses occur only among banks not part of a holding company; affiliated banks smooth lending via internal capital markets, rendering balance sheet composition unrelated to lending responses.
- Implication: forward-looking component of endogenous monetary policy matters; studies of bank lending behavior should account for policymakers’ forecasts and policy endogeneity.

### II. Bank lending and monetary policy — mechanisms of bias and interpretation
- Typical empirical specification (illustrative):
  - ΔL_{i,t} = α + β M_t + γ' S_{i,t} + δ' (S_{i,t} ∘ M_t) + φ' Z_t + ε_{i,t}
  - ΔL: percentage growth of total loans at current prices; M: monetary policy measure; S: vector of bank-specific characteristics; Z: control variables; interactions S ∘ M capture heterogeneity.
- Sources of bias when using realized federal funds rate:
  - Endogenous policy: FOMC reacts to forecasts of output and inflation; such forecasts also affect loan demand and loan supply, producing omitted variable bias.
  - Loan demand channels: tighter policy implemented because of expected stronger output/inflation can coincide with rightward shifts in loan demand, attenuating observed lending reductions.
  - Loan supply channels: expected economic growth can increase access to non-reservable finance (equity issues, securitization) or reduce external finance premium, offsetting policy-induced supply contractions.
- Two examples of bias affecting heterogeneity estimates (δ coefficients):
  - (i) Expected macroeconomic improvements induce loan supply shifts that depend on bank characteristics (constrained banks expand more in expansions), attenuating the estimated effect of monetary tightening for constrained banks.
  - (ii) Expected macroeconomic changes induce bank-specific loan demand shifts correlated with bank characteristics (rational buffer stocking; cyclically sensitive customer mixes), also attenuating or biasing estimates of the shielding effect of liquidity or holding company status.
- Caveat: some bank characteristics (e.g., holding company affiliation) may be less correlated with loan demand composition and thus better proxies for supply-side access to funds (Ashcraft, 2006).

### III. Econometric methodology — identification of exogenous policy and regression framework
- Identification strategy: extend Romer and Romer (2004) two-step procedure to isolate exogenous variation in the federal funds target.
  - Step 1: Construct series of targeted federal funds rate changes at scheduled FOMC meetings using Romer and Romer (1969–96) narrative series plus appended FOMC announced target changes for 1997–2006.
    - Note: announced target series correlated essentially 1 with Romer and Romer series over 1994–96; one exception noted for September 28, 1994.
  - Step 2: Regress targeted federal funds rate change on Federal Reserve Greenbook real-time forecasts:
    - Forecast horizons: up to two quarters (forecasts/backcasts/nowcasts) for real output growth (Δy), inflation (π), unemployment (n), and manufacturing capacity utilization (û), plus additional unemployment terms.
    - Formal regression (symbolic): targeted change Δf_{n} regressed on sums of forecasted Δy, π, n, and capacity utilization across forecast horizons, with residual capturing exogenous policy innovation.
- Sample and estimation details:
  - Sample of 346 FOMC meetings from 1969–2006.
  - Inclusion of capacity utilization and additional unemployment terms raises R̄^2 to 31% as compared to 28% in original Romer and Romer (2004) specification.
  - Aggregation to quarterly frequency: cumulate post-meeting identified policy changes at daily frequency within a quarter to form UM; construct analogous quarterly changes in effective federal funds rate (FF) by daily cumulation.
  - Comparison of UM and FF (sample 1969q2 to 2006q4):
    - Standard deviation of UM is 57 basis points.
    - Standard deviation of FF is 204 basis points.
    - Correlation(UM, FF) = 0.57.
  - Identification assumptions:
    - (i) Greenbook forecasts are not a function of the contemporaneous change in the federal funds rate target (Greenbook typically assumes no policy change until at least the meeting after next).
    - (ii) Greenbook forecasts account for factors driving policy that are also determinants of bank lending via expected economic conditions.
  - Potential threats noted: Fed responses to banking sector conditions (e.g., liquidity concerns) that are not captured by Greenbook forecasts could bias residuals; sample excludes post-2007 sub-prime banking crisis but includes episodes around Basel I (1988/1992) and LTCM rescue (1998); robustness checks reported in Section 5 (not reproduced here).
- Regression specification for bank-level lending responses:
  - Quarterly panel regression of ΔL_{i,t} on lags and changes of a vector M (three macro variables), interactions with five bank characteristics S (details of variables described in paper), seasonal dummies, and dynamics (autoregressive terms and distributed lags).
  - Monetary policy variables constructed consistently with end-of-quarter bank balance sheet measurement.

### IV. Data and measurement notes (as described)
- Bank-level data: end-of-quarter balance sheet reports filed with the Federal Deposit Insurance Corporation (FDIC); reported for the final day of a quarter; banks have up to 30 days in the following quarter to confirm.
- Monetary policy measures:
  - UM: quarterly changes constructed by cumulating identified post-meeting policy changes at daily frequency within a quarter (exogenous component).
  - FF: quarterly changes in the effective federal funds rate constructed analogously from daily effective federal funds rate from FRED.
- Frequency and aggregation rationale: using end-of-period aggregation to align with end-of-quarter bank balance sheet reporting; avoids distortions that arise when bank lending responds fully within a month but policy is averaged across the quarter.

*Source: _wp13118 - References (excerpted content provided).*

### 1. a monetary policy measure, either UM or FF, as described in section 2;

### _wp13118 - 1. a monetary policy measure, either UM or FF, as described in section 2;

### Dependent and key macroeconomic variables
- 1. a monetary policy measure, either UM or FF, as described in section 2;
- 2. real GDP growth in percentage points;
- 16
- 3. growth in the PCE core price index in percentage points.

### Regression specifications
- Two versions of the regressions are presented:
  - (i) a less noisy version based on year-over-year percentage growth in lending (Δܮ
௜,௧
) and the non-policy macroeconomic controls.
  - (ii) a version based on annualized quarter-over-quarter percentage growth in lending (Δܮ
௜,௧
) and the non-policy macroeconomic controls.

### Vector of bank characteristics (ܤ)
- 1. the natural log of bank assets in millions of dollars, at current prices;
- 2. an indicator variable set to unity post-1986 if a bank is part of a bank holding company and zero otherwise (following Ashcraft (2006) this characteristic is dated ݐ rather than െ1ݐ);
- 15
- 3. the ratio of bank securities to assets;
- 4. the ratio of total equity capital to assets;

*Source: _wp13118 - 1. a monetary policy measure, either UM or FF, as described in section 2;*

### 5. the ratio of cash to assets.

### 5. the ratio of cash to assets.

### Regression specification and interactions
- The regression specification in equation (3) is closely related to those employed by Ashcraft (2006) and Loutskina (2011).
- Once-lagged bank characteristics are included as controls to allow for differences in lending growth conditional upon bank size, holding company affiliation, and balance sheet composition.
- The growth and inflation controls in the vector ܯ account for variations in nominal lending growth arising from contemporaneous changes in prices and economic activity.
- Interactions between the macroeconomic variables and bank characteristics capture heterogeneity in bank lending responses to monetary policy, income growth, and inflation.
- Key points:
  - Interacted bank characteristics are dated  െ1ݐ (pre-determined) except the bank holding company dummy which is dated ݐ; this mirrors Ashcraft (2006) and Loutskina (2011).
  - Each bank characteristic ratio (except the binary holding company variable) is demeaned by sample quarter and normalized by the standard deviation. Nominal assets are deflated by the GDP deflator.
  - The first component of the vector 
    ∑
    ߚ
    ℓ
    ସ
    ℓୀ଴
    measures the percentage change in lending a year after a 100 basis point (b.p.) monetary policy contraction for an unaffiliated bank at the sample mean of each characteristic (overlooking autoregressive contributions).
  - The ݇th component (݇൑ 5) of the vector 
    ∑
    ߜ
    ଵ,௞,ℓ
    ସ
    ℓୀ଴
    measures the increment to the marginal lending response to a monetary contraction (݉ൌ1) when the ݇th characteristic is 1 standard deviation above the sample mean (or a bank is affiliated with a holding company).
  - The regression includes a full set of interactions between real income growth, inflation, and bank characteristics to control for: (i) purely nominal effects on lending growth from inflation; and (ii) heterogeneity in the response of real lending growth to macro factors.
- Additional regression elements:
  - Seasonal dummies and a time trend are included (macro variables seasonally adjusted; bank-level variables are not).
  - Maximum lag order is 4. Lags in the dependent variable control for serial correlation.
  - All regression standard errors are calculated through clustering at the bank-level to handle residual heteroscedasticity and autocorrelation.
  - The first-stage uncertainty used to identify UM is not accounted for in the standard errors; Pagan (1984) implies this affects inference only for non-zero null hypotheses.

### Data: bank-level and macroeconomic
- Bank-level data:
  - Source: Reports of Condition and Income (“Call Reports”) usually submitted to the FDIC at the end of each quarter.
  - Sample extended from 1968 to 2006.
  - Banks have reported consistently exactly at the end of the quarter only from 1975; prior to 1975 some timing irregularities exist.
  - Assumptions for earlier data:
    - Timing irregularities treated as measurement error uncorrelated with other controls.
    - Interpolate bank-level variables for quarters with missing data; treated as measurement error that increases estimator variance if not systematically related to controls.
  - Variable definitions follow Ashcraft (2006); Call Report line numbers per Kashyap and Stein (2000).
  - Dependent variable: total loans minus allowances for loan losses; includes loans under commitment and loans on flexible terms.
  - Limitations: measure does not control for loans moved off balance sheets via securitization; distortions limited since sample ends in 2006 and mortgage securitization growth concentrated later.
  - Balance sheet composition measures:
    - Total bank assets: reported net of loan loss reserves.
    - Securities: sum of Total Investment Securities and Assets Held in Trading Accounts.
    - Total Equity Capital: book value of equity issued plus cumulated retained earnings.
    - Cash: cash on the asset side of the balance sheet.
  - Holding company status indicator from Ashcraft (2006) (same regulatory holder identification number).
  - Baseline dataset: unbalanced quarterly panel 1969q3 to 2006q4; maximum of 15,306 banks and minimum of 7,922 banks; average observations per bank = 112 quarters.
  - Exclusions: bank/quarter observations affected by mergers are excluded.
  - Outlier treatment:
    - Fit benchmark regression by OLS then eliminate observations with absolute DFITS > 2·
      ට
      ܭ
     ܰ
     ൗ
      (Welsch and Kuh, 1977), where K is total explanatory variables and N is overall sample size.
    - From total 1,724,924 observations, outlier exclusion reduces sample to 1,724,355 when UM is policy measure and 1,724,334 when FF is policy measure.
    - Results across UM and FF do not depend on outlier exclusion.
- Macroeconomic data:
  - Income growth: seasonally adjusted real GDP.
  - Inflation rate: seasonally adjusted personal core consumption expenditure price index (PCE).
  - Both series from Bureau of Economic Analysis (BEA) via FRED.
  - Income and price data are period average (flow) values; bank-level data are end-of-quarter (stock) values — potential measurement mismatch.
  - Two policy controls described in section 3.B: FF and UM.

- Table 2: Bank Level Summary Statistics (Q4 of each year: 1970, 1980, 1990, 2000). Selected entries:
  - Loan Growth
    - 1970 Mean 10.5589 Std. Dev. 23.2026
    - 1980 Mean 4.5732 Std. Dev. 16.6064
    - 1990 Mean 9.834 Std. Dev. 34.5314
    - 2000 Mean 15.6723 Std. Dev. 38.1374
  - Assets
    - 1970 Mean 46,670 Std. Dev. 496,413
    - 1980 Mean 139,751 Std. Dev. 1,735,830
    - 1990 Mean 288,878 Std. Dev. 2,496,265
    - 2000 Mean 778,952 Std. Dev. 10,220,686
  - Holding Company Status
    - 1970 Mean 0.1549 Std. Dev. 0.3618
    - 1980 Mean 0.3310 Std. Dev. 0.4706
    - 1990 Mean 0.6883 Std. Dev. 0.4632
    - 2000 Mean 0.7558 Std. Dev. 0.4296
  - Loans/Assets
    - 1970 Mean 0.4809 Std. Dev. 0.1100
    - 1980 Mean 0.5407 Std. Dev. 0.1147
    - 1990 Mean 0.5475 Std. Dev. 0.1574
    - 2000 Mean 0.6286 Std. Dev. 0.1440
  - Deposits/Assets
    - 1970 Mean 0.8892 Std. Dev. 0.0305
    - 1980 Mean 0.8839 Std. Dev. 0.0424
    - 1990 Mean 0.8828 Std. Dev. 0.0774
    - 2000 Mean 0.8321 Std. Dev. 0.0937
  - Securities/Assets
    - 1970 Mean 0.3433 Std. Dev. 0.1173
    - 1980 Mean 0.2822 Std. Dev. 0.1163
    - 1990 Mean 0.2881 Std. Dev. 0.1546
    - 2000 Mean 0.2513 Std. Dev. 0.1360
  - Equity Capital/Assets
    - 1970 Mean 0.0821 Std. Dev. 0.0244
    - 1980 Mean 0.0855 Std. Dev. 0.0246
    - 1990 Mean 0.0867 Std. Dev. 0.0434
    - 2000 Mean 0.1039 Std. Dev. 0.0462
  - Cash/Assets
    - 1970 Mean 0.1231 Std. Dev. 0.0570
    - 1980 Mean 0.0923 Std. Dev. 0.0550
    - 1990 Mean 0.0715 Std. Dev. 0.0536
    - 2000 Mean 0.0490 Std. Dev. 0.0418
  - Number of Banks (Q4)
    - 1970 13,195.0
    - 1980 14,351.00
    - 1990 12,365.0
    - 2000 8,309.0

### Empirical results — dynamic lending responses (UM versus FF)
- Table 3: Dynamic lending response of the representative bank to a 100 b.p. tightening (sums of contemporaneous and four lagged responses). Reported for UM and FF at horizons 0 to 4; standard errors clustered at bank level.
  - Horizon 0 (QoQ)
    - UM: -0.3068, s.e. 0.0490
    - FF: -0.1360, s.e. 0.0177
    - YoY: UM -0.3240, s.e. 0.0727; FF -0.1125, s.e. 0.0328
  - Horizon 1
    - UM: -0.9060, s.e. 0.0810
    - FF: -0.4259, s.e. 0.0396
    - YoY: UM -2.0864, s.e. 0.1133; FF -0.8077, s.e. 0.0697
  - Horizon 2
    - UM: -1.9893, s.e. 0.1342
    - FF: -0.8124, s.e. 0.0690
    - YoY: UM -3.7095, s.e. 0.1607; FF -1.6786, s.e. 0.1070
  - Horizon 3
    - UM: -2.5439, s.e. 0.1650
    - FF: -1.1207, s.e. 0.0916
    - YoY: UM -4.5173, s.e. 0.2164; FF -2.3476, s.e. 0.1357
  - Horizon 4
    - UM: -3.2841, s.e. 0.1960
    - FF: -1.4813, s.e. 0.1068
    - YoY: UM -5.2595, s.e. 0.2686; FF -3.0291, s.e. 0.1521
- Findings:
  - At each horizon, the lending reduction estimated from an exogenous monetary policy contraction (UM) exceeds that from a policy contraction measured by the realized federal funds rate (FF).
  - 95% confidence intervals for UM and FF estimates are non-overlapping at all horizons beyond the current quarter.
  - Autoregressive parameters are small (less than 0.1) and virtually identical across UM and FF; they do not affect inferences materially.
  - Interpretation: part of the sluggishness in lending behavior measured from FF may reflect endogeneity of policy to macro fundamentals; using exogenous UM reveals a quantitatively more important monetary transmission via credit markets.

### Heterogeneity: bank size, holding company status, and balance sheet composition
- Table 4: Heterogeneity in lending responses due to bank characteristics. Reported are sums of interaction terms through horizon 4 (interaction = effect when characteristic is 1 standard deviation above the quarter sample mean; holding company indicator set to unity). Representative bank marginal effect reproduced for context. Standard errors clustered at bank level.
- Representative bank marginal effect (YoY / QoQ)
  - UM: -3.2841 (s.e. 0.1960)
  - FF: -1.4813 (s.e. 0.1068)
  - YoY UM: -5.2595 (s.e. 0.2686); YoY FF: -3.0291 (s.e. 0.1521)
- Holding Company Status interaction
  - UM: 5.5366 (s.e. 0.4644)
  - FF: 2.7783 (s.e. 0.2188)
  - Marginal effect (affiliated bank)
    - UM: 2.2526 (s.e. 0.3971)
    - FF: 1.2970 (s.e. 0.1815)
    - YoY marginal: UM 5.8324 (s.e. 0.7730); FF 2.8215 (s.e. 0.3237)
- Assets interaction
  - UM: 1.0616 (s.e. 0.1702)
  - FF: 0.0954 (s.e. 0.0891)
  - Marginal effect
    - UM: -2.2225 (s.e. 0.2225)
    - FF: -1.3859 (s.e. 0.1453)
    - YoY marginal: UM -3.0612 (s.e. 0.3944); FF -2.5587 (s.e. 0.2168)
- Securities Ratio interaction
  - UM: -1.1357 (s.e. 0.1453)
  - FF: -0.0443 (s.e. 0.0853)
  - Marginal effect
    - UM: -4.4197 (s.e. 0.2578)
    - FF: -1.5256 (s.e. 0.1027)
    - YoY marginal: UM -7.0822 (s.e. 0.4589); FF -2.8567 (s.e. 0.2265)
- Cash Ratio interaction
  - UM: 0.4139 (s.e. 0.1656)
  - FF: 0.0249 (s.e. 0.0906)
  - Marginal effect
    - UM: -2.8701 (s.e. 0.2462)
    - FF: -1.4564 (s.e. 0.1446)
    - YoY marginal: UM -4.2783 (s.e. 0.5116); FF -2.9304 (s.e. 0.2997)
- Equity Capital Ratio interaction
  - UM: 0.0177 (s.e. 0.4510)
  - FF: -0.2723 (s.e. 0.2837)
  - Marginal effect
    - UM: -3.2664 (s.e. 0.5462)
    - FF: -1.7536 (s.e. 0.3511)
    - YoY marginal: UM -5.5633 (s.e. 1.1313); FF -3.7669 (s.e. 0.7828)
- Model fit and sample
  - R2: UM 0.4603; FF 0.4627 (QoQ). YoY R2: UM 0.0066; FF 0.0067.
  - Observations (reported): UM 1,737,581 and 1,762,369; FF 1,750,726 and 1,775,773 (YoY / QoQ noted in table context).
  - Note: Interaction effects calculated for a bank one standard deviation above the quarter sample average; standard errors clustered at bank level. ** and * indicate significance at the 1 and 10 percent levels respectively.

- Key heterogeneity findings:
  - Total assets and holding company affiliation: interaction sums are positive → these characteristics help banks shield lending growth from policy contractions, with effects much larger when policy is measured using UM versus FF.
  - Holding company banks appear to raise lending following an exogenous policy contraction, while unaffiliated banks reduce lending. Baseline regression estimates:
    - Exogenous policy contraction (UM): representative unaffiliated bank reduces lending 3.28 percentage points in the first year; representative affiliated bank raises lending 2.25 percentage points over same period.
  - Aggregate effect when excluding bank characteristics from vector ܤ:
    - Lending reduction from UM: 1.85 percentage points (0.14)
    - Lending reduction from FF: 0.19 percentage points (0.06)
    - Conclusion: an aggregate transmission mechanism exists and is roughly twice as strong when estimated from UM.
  - Securities-to-assets ratio:
    - Banks with securities one standard deviation above the mean reduce lending an additional 1.14 percentage points (YoY) and 1.82 percentage points (QoQ) after a 100 b.p. increase in UM.
    - Under FF, securities provide no meaningful shielding YoY and only 0.17 percentage points shielding QoQ (not significant at 10%).
    - Interpretation: securities may suffer capital losses when interest rates rise, prompting lending contraction; market liquidity of securities can be low in such episodes.
  - Cash-to-assets ratio:
    - Cash holdings do shield banks from monetary contraction: shielding estimates range from 0.4139 (YoY interaction) to 0.9812 (QoQ marginal effect noted in text).
  - Equity capital: estimated to neither shield nor amplify lending responses to monetary shocks in these specifications—heterogeneity due to capitalization may be captured by interactions with macro proxies.

- Interpretation and implications:
  - Controlling for endogeneity of monetary policy (using UM) not only shows stronger lending responses at the representative bank but also greater dispersion in responses across banks.
  - Heterogeneity suggests lending by small banks and non-holding-company banks is more responsive to expected economic growth, which can attenuate observed lending responses when using endogenous policy measures (FF).
  - The larger cross effects estimated from exogenous policy may help explain:
    - Different effects of policy across regions and industries.
    - Possible trend toward weaker propagation of monetary policy in recent decades attributable to banking sector consolidation and growth of holding companies.
  - Caveats:
    - Some interaction effects (e.g., assets) may reflect heterogeneity in loan demand as well as loan supply.
    - Affiliations and other bank characteristics may change jointly; interpreting affiliation effects as purely supply-driven requires caution.

### Stability and robustness notes
- Temporal stability of results is an important concern; sample may include structural changes (discussion continues beyond supplied content).
- Robustness checks discussed elsewhere include dating interacted characteristics at െݐℓെ1 to ensure pre-determination; results change very little due to low within-quarter temporal variation relative to cross-sectional variation.
- Outlier exclusion and the handling of earlier reporting irregularities are used to mitigate measurement issues; results are robust to using full or trimmed samples.

*Italic: Source: _wp13118 - 5. the ratio of cash to assets. (PDF chapter/section).*

### introduction of the source of strength doctrine (Ashcraft, 2006).

### introduction of the source of strength doctrine (Ashcraft, 2006).

### Effect of the source of strength doctrine on bank lending responses
- The Federal Reserve Board issued a formal statement in April 1987 indicating that failure by a parent bank to inject liquidity into a financially distressed subsidiary when funds are available would be considered an unsafe banking practice.
- Post-1986 (post-1987 implementation), interaction terms for banks that belong to a holding company (holding company banks) show post-1986 changes to the interaction coefficients that are of approximately equal magnitude but opposite sign to the main interaction effects (the one exception is the interaction of FF with bank assets).
- As a result, "the total effect of balance sheet related characteristics on lending responses to monetary policy, both exogenous and endogenous, is close to zero during the second half of the sample for affiliated banks" (and affiliated banks represent over two thirds of all banks in this period).
- During the late 1980s and the 1990s, the principal source of heterogeneity in lending responses to monetary policy is affiliation with a holding company, not balance sheet composition.
- Mechanisms:
  - When affiliated banks face write-downs in securities prices or loan values following policy tightening, they are able to tap loanable funds within the network, thus shielding their lending growth.
  - The roles of security holdings in amplifying and cash in mitigating the effects of exogenous policy on lending growth are quantitatively smaller from the late 1980s onwards because they are observed only amongst banks that cannot access the financing networks provided by holding companies.

### Key empirical estimates (Table 5 — Bank Holding Company Lending Responses)
- Representative Bank marginal effect (sum of contemporaneous and four lagged responses to a 100 basis points policy tightening):
  - - 3.4154, - 1.5394, -5.0797, - 3.1292
  - s.e.: 0.2040, 0.110, 0.2734, 0.166
- Holding Company Status interaction with policy:
  - 6.2261, 3.018, 38.525, 55.398
  - s.e.: 0.6826, 0.3007, 2.6354, 0.5261
- Assets interaction with policy:
  - 1.4369, 0.1382, 2.6568, 0.3424
  - s.e.: 0.1764, 0.1193, 0.2989, 0.1762
- Assets interaction with policy and holding company:
  - - 0.9142, 0.0089, -3.1439, - 0.7028
  - s.e.: 0.5205, 0.2135, 0.8637, 0.3737
- Securities Ratio interaction with policy:
  - - 1.5012, - 0.0551, -2.2442, 0.0476
  - s.e.: 0.1791, 0.1045, 0.410, 0.2398
- Securities Ratio interaction with policy and holding company:
  - 1.3514, - 0.0949, 2.049, - 0.0101
  - s.e.: 0.4818, 0.2185, 1.3317, 0.571
- Cash Ratio interaction with policy:
  - 0.2837, 0.0557, 1.3405, 0.0968
  - s.e.: 0.1852, 0.110, 0.4265, 0.2082
- Cash Ratio interaction with policy and holding company:
  - 0.2215, 0.0266, -1.0476, - 1.2427
  - s.e.: 0.5619, 0.3043, 1.7275, 1.0053
- Equity Capital Ratio interaction with policy:
  - 0.0073, - 0.4491, 0.3627, - 1.3477
  - s.e.: 0.4502, 0.2711, 1.045, 0.5221
- Equity Capital Ratio interaction with policy and holding company:
  - 2.1446, 1.1781, -14.5082, - 1.5695
  - s.e.: 2.2481, 1.0711, 12.886, 1.9497
- Goodness of fit and sample:
  - R2: 0.4435, 0.452, 0.0061, 0.0063 (reported per specification)
  - Observations: 1,737,141; 1,750,290; 1,757,740; 1,772,101 (across reported columns)
- Note: Interaction effects are calculated for a bank one standard deviation above the quarter sample average for a characteristic. Standard errors obtained after clustering at the bank level are reported in parenthesis. The reported lending responses are the sum of contemporaneous and four lagged responses to a 100 basis points policy tightening.

### Robustness exercises and principal results (section V, Table 6 summary)
- Two episodic dummies were introduced to test potential endogeneity of the UM policy measure:
  - Dummy equal to unity for all quarters in the period 1990-1993 (Basel I related tightening).
  - Dummy equal to unity for all quarters in 1998-1999 (the LTCM rescue occurred in 1998).
- Interacting these dummies with monetary policy terms produced only marginal increases in the absolute size of the estimated monetary policy effect on lending growth, indicating little evidence that the estimated effects of monetary policy were attenuated during the two episodes considered.
- Bank-level fixed effects:
  - Augmenting the specification with bank-level fixed effects shows main findings are generally robust to this extension.
- Predetermined characteristics:
  - Re-dating characteristics so that they are pre-determined with respect to the policy variable (characteristics dated െݐℓെ1) preserves the direct effect of policy and bank holding company status.
  - For UM measure, effects of bank size remain consistent; for FF the policy interaction with bank size is not significant and reverses sign.
  - For securities and cash ratio, there is some evidence of a shielding function. Capitalization clearly shields bank level lending growth based on the UM measure.
- Use of forecasts (Survey of Professional Forecasters) for non-policy macro controls:
  - The magnitude of coefficients on the direct policy response shrinks somewhat, but the difference between UM and FF measures remains meaningful: the response is three times as large in the quarterly model (UM vs FF) and orders of magnitude larger in the year-over-year quarterly model.
  - Holding company status and bank size have a shielding effect consistent with other estimates.
  - Securities act as policy amplifiers in the UM case and mute policy in the FF case. Cash shows similar patterns to securities in this exercise. Capitalization interactions are either insignificant (year-over-year quarterly model) or have an amplifying role.

- Selected marginal effects and interaction coefficients from Table 6 (robustness checks; marginal effects and holding company status interaction reported across multiple specifications):
  - Marginal Effect examples (sum of contemporaneous and four lagged responses to a 100 basis points policy tightening):
    - UM: - 3.3248, -5.3221, -3.8808, - 5.3008, -4.3402, -5.2191, - 1.4789, -1.9937
    - FF: - 1.4573, -3.1338, - 1.7400, - 3.5769, -1.0774, - 2.3703, - 0.0335, -0.7334
    - Corresponding s.e. examples: 0.2066, 0.1051, 0.2785, 0.1688, 0.2355, 0.1129, 0.2817, 0.1755, 0.1872, 0.0630, 0.2451, 0.1524, 0.1994, 0.0641, 0.3276, 0.1559
  - Holding Company Status interaction with policy (examples across specifications):
    - UM: 5.1537, 7.9903, 5.7015, 10.5771, 7.0534, 11.0793, 4.0981, 10.9462
    - FF: 2.4256, 4.8200, 2.9052, 5.4951, 2.6544, 5.3545, 1.0215, 3.3575
    - s.e. examples: 0.5000, 0.2102, 0.9235, 0.3819, 0.6801, 0.3076, 1.3095, 0.5308, 0.4112, 0.1660, 0.7587, 0.3121, 0.5790, 0.1643, 1.0716, 0.3570
  - Assets interaction with policy (selected values and s.e.):
    - Examples: 1.1023, 2.0355, 1.3907, 2.4594, 10.9937, - 19.4118, 14.8279, - 7.9247, 0.7792, 0.4612, 2.0750, 0.8524
    - s.e. examples: 0.1657, 0.0873, 0.2879, 0.1510, 0.1807, 0.1196, 0.3053, 0.1995, 8.5891, 2.0432, 6.1151, 1.7760, 0.1683, 0.0614, 0.3104, 0.1116
  - Observations across columns in Table 6 (examples):
    - 1,737,400; 1,750,566; 1,762,256; 1,775,695; 1,737,141; 1,750,290; 1,757,740; 1,772,101; 1,734,838; 1,748,146; 1,760,314; 1,773,614; 1,737,594; 1,750,739; 1,760,701; 1,774,230
  - Note: The reported lending responses are the sum of contemporaneous and four lagged responses to a 100 basis points policy tightening. Interaction effects are calculated for a bank one standard deviation above the quarter sample average for a characteristic. Standard errors obtained after clustering at the bank level are reported in parenthesis. ** and * indicate significance at the 1 and 10 percent levels respectively.

### Conclusion and policy implications
- Main summary of findings:
  - Using an explicitly identified exogenous monetary policy measure (UM) versus the realized federal funds rate (FF) reveals economically and statistically significant attenuation of estimated lending responses to monetary contractions when endogeneity is not accounted for.
  - Bank holding company affiliation provides a shielding effect on lending growth following policy tightenings; affiliation becomes the dominant source of heterogeneity in the late 1980s and 1990s.
  - Securities share in total assets amplifies transmission from exogenous interest rate changes but restricts transmission of realized interest rate changes; one explanation is adverse valuation effects on securities following exogenous monetary contractions.
- Research and policy recommendations:
  - Future studies on monetary transmission should consider exogenous policy measures alongside realized measures such as the federal funds rate.
  - Identification of exogenous monetary policy should incorporate forward-looking drivers of policy (growth and inflation forecasts), because these forward-looking variables likely impact lending markets.
  - Suggested avenues for future work:
    - Examine impact of exogenous monetary policy upon regional, industrial, and firm-specific lending and compare with realized federal funds rate effects.
    - Apply the methodology to study heterogeneity in lending rates as opposed to lending quantities.
    - Investigate effects of bank characteristics on responses to policy at a more disaggregated level, e.g., the particular mixture of securities held as bank assets.

*Source: introduction of the source of strength doctrine (Ashcraft, 2006).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13118.pdf_