## wpiea2023055-print-pdf

## Source details

**Canonical URL:** [wpiea2023055-print-pdf](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023055-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023055-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023055-print-pdf.pdf.json)

---

### Stylized facts and sections overview
- Section 3 presents the preferred measure of monetary policy shocks.
- Section 4 analyzes the impact of monetary policy shocks on the size of the non-bank sector.
- Section 5 illustrates the impact on flows and returns of long-term mutual funds.
- Section 6 presents the MS-VAR analysis to show how the impact of monetary policy on non-banks has evolved over time.
- Section 7 concludes.

### Supervision and regulation of the non-bank sector
- Post-GFC changes in U.S. supervision and regulation potentially underlie structural changes in the non-bank sector.
- Key regulators: Securities and Exchange Commission; Commodity Futures Trading Commission; Financial Stability Oversight Council (FSOC) — established as part of the 2010 Dodd-Frank Wall Street Reform and Consumer Protection Act.
- FSOC authorities and tools:
  - Authority to subject a systemically risky non-bank financial institution to consolidated supervision and enhanced regulatory safeguards.
  - Designation triggers stronger consolidated federal oversight by the Fed and enhanced financial stability rules, including capital requirements, liquidity rules and stress tests.
- Dodd-Frank outcomes and tools:
  - New registration requirements for hedge funds and private equity firms.
  - Federal office to monitor the insurance industry and to negotiate international insurance agreements.
  - Executive compensation restrictions for financial firms.
  - Authority for regulators to wind down systemically important nonbanks in an orderly fashion.
  - Consumer Financial Protection Bureau authority to use traditional law enforcement to stop non-banks from engaging in conduct that pose risks to consumers.
- Banks vs. non-banks regulatory intensity:
  - Dodd-Frank tightened traditional banking sector oversight, including tighter than Basel III capital and liquidity requirements, heightened prudential standards for the largest banking firms, and regular stress tests.
  - Heightened prudential standards require the largest and most interconnected banks to meet capital surcharges and stricter risk-management standards.
  - The Volcker Rule imposes broad prohibitions and restrictions on proprietary trading and investing in hedge funds or private equity funds by banking organizations and their affiliates.
- Evidence and ongoing work-streams:
  - Some empirical evidence links increased bank capital requirements to growth in non-bank finance (e.g., Irani et al. (2021)).
  - FSOC prioritized evaluating risks posed by hedge funds, open-end funds, and money market funds (MMFs) in 2021.
  - President’s Working Group on Financial Markets outlined potential reform options for MMFs in December 2020; work on hedge funds and open-ended funds is ongoing.

### Measuring the non-bank sector — definitions and approaches
- Definitions:
  - Non-banks reliant on short-term funding: perform credit, liquidity and maturity transformation, funded on a short-term basis, lacking access to deposit insurance and Federal Reserve liquidity facilities; subject to run-risk.
  - Non-banks reliant on long-term funding: perform similar bank-like activities but have more stable funding less subject to run risk.
- Institutional approach:
  - Sum inflation-adjusted, real assets of particular institutions or particular real liabilities of a short-term nature.
  - For short-term funded non-banks follow Pozsar et al. (2010): sum liabilities of total outstanding open market paper, total repo liabilities, net securities loaned by broker-dealers, total GSE liabilities and agency and GSE mortgage pool securities, total liabilities of ABS issuers, and total shares outstanding of money market mutual funds.
  - Alternative Adrian et al. (2010) measure: assets of agency and GSE backed mortgage pools, ABS issuers, finance companies and funding corporations.
  - Data: Federal Reserve’s quarterly ‘Financial Accounts of the United States (Z.1)’ up to 2022 Q1.
- Functional approach:
  - Consolidates chains of intermediation to avoid double counting (methodology following Gallin (2015)).
  - Short-term funders: money market mutual funds, unregistered liquidity funds, local government investment pools, cash-collateral reinvestment pools.
  - Intermediate funders: broker-dealers, GSEs, finance companies and private ABS issuers.
  - Long-term funders: mutual funds other than money market funds, pension funds, insurance companies.
  - Functional measures deflated by the GDP deflator; analogous bank sector size computed by substituting banks and credit unions.
- Comparative dynamics and interpretation:
  - Institutional and functional measures of short-term funded non-banks evolve similarly but functional measures have a lower level (avoid double counting).
  - Pre-GFC: non-bank sector grew rapidly; then declined.
  - Non-bank sector reliant on short-term funding began growing again around 2014 and accelerated before the pandemic, with GSEs accounting for a rising share of intermediate funding.
  - Since the pandemic, institutional measures continue to grow while functional measure declines — interpretation: cross-holdings between financial institutions, netted out by the functional approach, drove growth during the pandemic rather than expansion of credit to the non-financial sector.
  - Functional measure of non-banks reliant on long-term funding grew since the GFC and continued during the pandemic, although it has begun to decline more recently.
- Decomposition of asset growth:
  - Asset growth decomposed into changes in flows and a residual (assumed to comprise valuation effects); valuation effects were significant during the pandemic, particularly for mutual funds.
  - Monthly ICI data on long-term mutual funds used to analyze flows and net assets.

### ICI long-term mutual fund facts (2000–2021)
- Focus: U.S. domiciled open-ended mutual funds investing in domestic and international markets; disaggregated by fund type (equity domestic and international; investment grade; high yield; world; government; multisector; municipal; hybrid bond).
- Key figures:
  - Sector quadrupled over last twenty years to what is now a 22 trillion industry.
  - Equity mutual funds make up more than half of total net assets.
  - ICI reports individual investors hold about 90 percent of open-ended mutual fund assets.
  - Bond mutual funds show continued inflows during the pandemic while equity funds experienced persistent outflows.

### Identifying monetary policy shocks
- Main measure: Jarociński and Karadi (2020) shocks — interest rate surprises in the three-month fed funds future.
  - The three-month fed funds future exchanges a constant interest rate for the average federal funds rate over the course of the third calendar month in the contract.
  - Because regular FOMC meetings are six weeks apart, the three-month future reflects the shift in the expected federal funds rate after the following policy meeting, not the immediate next meeting.
  - Shocks do not capture surprises to the balance sheet, implicitly assuming such changes are orthogonal to surprises to the policy rate.
  - Shocks can be aggregated to monthly or quarterly frequency.
- Separation from Fed information shocks:
  - Jarociński and Karadi (2020) separate pure monetary policy shocks from signaling shocks (“Fed information” shocks).
  - Fed information shocks capture agents interpreting Fed actions as signals about the state of the economy; effects go in the opposite direction to monetary policy shocks.
  - Distinguishing assumption: correlation between changes in interest rates and stock prices following an information shock is positive; following a monetary policy shock it is negative.
  - Monetary policy shocks exhibit both significant tightening and loosening periods over the sample period.

### Empirical strategy and data
- Empirical method:
  - Local projections following Jorda (2005) at quarterly frequency.
  - Outcome variables y_t include functional measures based on Gallin (2015) for short- and long-term non-bank funding (preferred approach).
- Data:
  - Institutional measures from Federal Reserve flow-of-funds up to 2022 Q1.
  - Monthly ICI data on long-term mutual funds cover 2000–2021 for flows and net assets.

### Regression specification and interpretation
- Regression (OLS with Newey and West (1987) standard errors):
  - y_{t+h} − y_{t} = α_{h} + β_{h} ε_{t} + γ_{h} y_{t−1} + u_{h t}  for quarter h = 1,2,....,12.
  - ε_{t} = ε^{MP}_{t} (monetary policy shock) or ε^{Info}_{t} (Fed information shock).
  - Sample: 1990 Q1 through 2019 Q2.
  - Interpretation: estimated β_{h} is the percentage point change in the outcome after quarter h in response to a 100 basis point contractionary monetary policy shock or a 100 basis point positive Fed information shock.

### Main empirical findings (functional measures)
- Dislocation effect: contractionary monetary policy boosts short-term funded non-bank sector while shrinking long-term funded non-bank sector.
- Quantified impacts of a 100 basis point contractionary monetary policy shock:
  - Expands inflation-adjusted non-bank assets reliant on short-term funding by around 1/4 percentage point after five quarters (statistically significant at the 10 percent level).
  - Reduces inflation-adjusted size of non-bank sector reliant on long-term funding by 1/4 percentage point after 12 quarters (statistically significant at the 10 percent level), following an initial increase.
- Impact on bank assets (functional measure) is not statistically significant in the baseline regression without controls.
- Fed information shocks (functional measures):
  - Positive Fed information shock reduces size of non-bank sector reliant on long-term funding by around 1/2 percentage points after 12 quarters (Gallin (2015) measure).
  - Initially reduces size of sector reliant on short-term funding by around 1/4 percentage points after five quarters, then eventually turns positive.

### Institutional measures and robustness
- Using institutional measures with preferred shock identification:
  - Contractionary monetary policy is not found to have a statistically significant impact on institutional measures of non-bank size generally.
  - Contractionary shock found to reduce the size of mutual funds (excluding money market funds), though this effect is not necessarily robust to adding covariates.
- Robustness to controls:
  - Expansion of short-term non-bank sector (functional) following contractionary policy is robust to including lags of shocks and controls X_{t} (controls: year-over-year real GDP growth; Fed Funds Rate (midpoint of target range at end of quarter); term spread (10-year minus 3-month); year-over-year change in S&P 500; year-over-year change in CoreLogic Case-Shiller house price index; year-over-year change in real assets of broker dealers).
  - Contractionary policy effect on long-term funded non-banks loses statistical significance when all controls are included.

### Comparisons to earlier literature
- Magnitudes in this paper are smaller than in some earlier studies focusing on pre-GFC samples and institutional measures:
  - Nelson et al. (2017): 100 basis point contractionary shock increased inflation-adjusted U.S. non-bank sector by a peak of 1.5 percent after one year (sample 1966-2007; institutional measure).
  - Den Haan and Sterk (2011): 100 basis point contractionary shock increased inflation-adjusted non-bank mortgages and consumer credit by 1 and 1.5 percent respectively after 18 and 36 months (sample 1984-2008).
- Differences likely reflect sample period, shock identification methods, and measures of non-bank size.

### Mutual funds: flows and returns (monthly VAR, ICI data)
- Data and method:
  - Monthly ICI data, 2000-2019.
  - Two-variable VAR per fund category: ratio of net flows to total net assets f_{i,t} = c_{i,t} / a_{i,t−1} and monthly price return r_{i,t}.
  - Exogenous monetary policy surprise shocks included; vector of controls X (core PCE, ADS business conditions index, BBB corporate-Treasury spreads, term spread, VIX, real effective exchange rates as robustness).
  - Two lags based on Akaike criterion.
- Aggregate bond mutual funds (2000-2019) results:
  - A 100 basis points contractionary monetary policy surprise shock → 2 percent fall in bond mutual fund net flows and a 4 percent fall in returns (effects significant at the 5 percent level over 20 months).
- Aggregate equity mutual funds:
  - A 100 basis points contractionary shock → decline of net flows by about 1 percentage point on impact, and a larger impact on returns of about 12 percentage points on impact.
- Heterogeneity:
  - Greater impact on high yield bond funds (relative to other bond funds) and international equity funds (relative to domestic equity funds).
- Robustness:
  - Results robust to different shock measures (e.g., Nakamura and Steinsson (2018)) and inclusion of additional explanatory variables.
  - Jarociński and Karadi (2020) information shocks produce opposite signs: positive information shocks → inflows and higher returns.

### Time-varying impacts (MS-VAR results)
- MS-VAR monthly setup for flows and returns (2000-2019):
  - Five endogenous variables: flows and returns for mutual funds, monthly unemployment rate, core PCE inflation (month over month), Fed Funds Rate (midpoint of target range at end of month); two lags; three regimes for constants/AR coefficients and three regimes for variance-covariance matrix.
  - Identification: sign restrictions—contractionary monetary policy shock must make core PCE inflation negative for five months, Fed Funds Rate and unemployment rate positive for five months.
  - Estimation: Bayesian with flat priors; posterior mode used to compute regime probabilities.
- MS-VAR monthly findings:
  - Three regimes correspond to distinct time periods; regime after 2013 shows statistically significant reduction in bond fund flows by almost −0.2 percentage points on impact.
  - Contractionary monetary policy reduced flows into bond mutual funds more recently (post-2013).
  - For equity mutual funds, contractionary policy reduced flows in pre-2007 regime; not statistically significant in more recent regime.
- Quarterly MS-VARs on bank and non-bank size (1973 Q3 through 2019 Q4):
  - Functional measures:
    - Non-banks reliant on long-term funding: contractionary shock expanded sector prior to 1980s but shrank it by up to 0.2 percentage points during the 2000s.
    - Non-banks reliant on short-term funding: strong dislocation effect after 2001—contractionary shocks boost the size of the sector.
  - Institutional measures:
    - Mutual funds (excluding money market funds): contractionary shocks shrink size beyond early 2000s.
    - Short-term funded non-bank institutional measures less clear; some pre-2001 dislocation effects found but results differ by institutional measure (Adrian et al. (2010) vs Pozsar et al. (2010)).
  - Banks:
    - Evidence that contractionary monetary policy caused a reduction in the size of traditional banking more recently in some MS-VAR specifications (functional definitions), after an initial expansion on impact.
    - Institutional MS-VAR results for banks are mixed; some specifications show a small reduction in bank size after 2001 (Adrian and Shin (2010) institutional definitions).

### Conclusions and policy implications
- Empirical conclusions:
  - Contractionary monetary policy significantly reduces balance sheets of non-banks reliant on long-term funding and leads to significant outflows from long-term mutual funds.
  - Higher risk bond funds and international equity funds are most impacted, consistent with the risk-taking channel.
  - Evidence of time-varying transmission: contractionary impact on long-term funded non-banks has generally strengthened in recent years; contractionary policy increases size of short-term funded non-banks in recent years (dislocation effect) when measured functionally.
  - Monetary tightening may induce flows out of mutual funds, reduce returns, shrink the overall long-term funded non-bank sector, but could expand short-term funded non-banks and thus increase susceptibility to run risks.
- Policy implications and research avenues:
  - Monetary policy conduct needs to adapt as transmission mechanisms change with the growth of nonbank intermediation.
  - Stronger effects on the real economy and financial soundness via the risk-taking channel imply greater vigilance by prudential and regulatory authorities.
  - Further research recommended to explore underlying drivers of changing transmission to non-banks and consequences for monetary policy transmission to the real economy.

*Source — wpiea2023055-print-pdf*

### Section 3 presents our preferred measure of monetary policy shocks.  Section 4 analyzes the

### wpiea2023055-print-pdf - Section 3 presents our preferred measure of monetary policy shocks.  Section 4 analyzes the

### Stylized Facts: Non-Bank Finance in the U.S.
- Sections overview:
  - Section 3 presents the preferred measure of monetary policy shocks.
  - Section 4 analyzes the impact of monetary policy shocks on the size of the non-bank sector.
  - Section 5 illustrates the impact on flows and returns of long-term mutual funds.
  - Section 6 presents the MS-VAR analysis to show how the impact of monetary policy on non-banks has evolved over time.
  - Section 7 concludes.

### Supervision and regulation of the non-bank sector
- Post-GFC changes in U.S. supervision and regulation potentially underlie structural changes in the non-bank sector.
- Key regulators involved in non-bank financial intermediation:
  - Securities and Exchange Commission.
  - Commodity Futures Trading Commission.
  - Financial Stability Oversight Council (FSOC) — established as part of the 2010 Dodd-Frank Wall Street Reform and Consumer Protection Act.
- FSOC authorities and tools:
  - Authority to subject a systemically risky non-bank financial institution to consolidated supervision and enhanced regulatory safeguards.
  - Designation of systemically important non-bank financial institutions triggers stronger consolidated federal oversight by the Fed and enhanced financial stability rules, including capital requirements, liquidity rules and stress tests.
- Dodd-Frank outcomes and tools:
  - New registration requirements for hedge funds and private equity firms.
  - Federal office to monitor the insurance industry and to negotiate international insurance agreements.
  - Executive compensation restrictions for financial firms.
  - Authority for regulators to wind down systemically important nonbanks in an orderly fashion.
  - Consumer Financial Protection Bureau authority to use traditional law enforcement to stop non-banks from engaging in conduct that pose risks to consumers.
- Banks vs. non-banks regulatory intensity:
  - Dodd-Frank tightened traditional banking sector oversight, including tighter than Basel III capital and liquidity requirements, heightened prudential standards for the largest banking firms, and regular stress tests.
  - Heightened prudential standards require the largest and most interconnected banks to meet capital surcharges and stricter risk-management standards.
  - The Volcker Rule imposes broad prohibitions and restrictions on proprietary trading and investing in hedge funds or private equity funds by banking organizations and their affiliates.
- Evidence and ongoing work-streams:
  - Some empirical evidence links increased bank capital requirements to growth in non-bank finance (e.g., Irani et al. (2021)).
  - FSOC prioritized evaluating risks posed by hedge funds, open-end funds, and money market funds (MMFs) in 2021.
  - President’s Working Group on Financial Markets outlined potential reform options for MMFs in December 2020; work on hedge funds and open-ended funds is ongoing.

### Measuring the non-bank sector
- Definitions:
  - Non-banks reliant on short-term funding: perform credit, liquidity and maturity transformation, funded on a short-term basis, lacking access to deposit insurance and Federal Reserve liquidity facilities; subject to run-risk.
  - Non-banks reliant on long-term funding: perform similar bank-like activities but have more stable funding less subject to run risk.
- Two measurement approaches:
  - Institutional approach:
    - Sum inflation-adjusted, real assets of particular institutions or particular real liabilities of a short-term nature.
    - For non-banks reliant on short-term funding, follow Pozsar et al. (2010): sum liabilities of total outstanding open market paper, total repo liabilities, net securities loaned by broker-dealers, total GSE liabilities and agency and GSE mortgage pool securities, total liabilities of ABS issuers, and total shares outstanding of money market mutual funds.
    - Consider alternative Adrian et al. (2010) measure: assets of agency and GSE backed mortgage pools, ABS issuers, finance companies and funding corporations.
    - Data source: Federal Reserve’s quarterly ‘Financial Accounts of the United States (Z.1)’ (flow of funds), up to 2022 Q1.
    - Comparable to Federal Reserve’s measure of ’runnable’ assets and the ’narrow’ FSB measure of size of non-banks.
    - Analogous measure for traditional banking system: real assets of commercial banks, credit unions and savings institutions from flow of funds.
  - Functional approach:
    - Consolidates chains of intermediation to avoid double counting (methodology following Gallin (2015)).
    - Short-term funders defined as money market mutual funds (money funds), unregistered liquidity funds, local government investment pools, and cash-collateral reinvestment pools from securities lending programs.
    - Intermediate funders: broker-dealers, GSEs, finance companies and private ABS issuers.
    - Size of non-banks reliant on long-term funding: substitute long-term funders (mutual funds other than money market funds, pension funds, insurance companies).
    - Bank sector size computed by substituting banks and credit unions for short- or long-term funders.
    - All functional measures deflated by the GDP deflator to adjust for inflation, same as institutional measures.
- Comparative dynamics and interpretation:
  - Institutional and functional measures of non-banks reliant on short-term funding evolve similarly but functional measures have a lower level (avoid double counting).
  - Pre-GFC: non-bank sector grew rapidly; then declined.
  - Share of private ABS issuers in intermediate funding rose before the GFC and declined.
  - Non-bank sector reliant on short-term funding began growing again around 2014 and accelerated before the pandemic, with GSEs accounting for a rising share of intermediate funding.
  - Since the pandemic began, institutional and functional measures diverge:
    - Institutional measures continue to grow.
    - Functional measure declines.
    - Interpretation: cross-holdings between financial institutions, netted out by the functional approach, drove growth of non-banks during the pandemic rather than expansion of credit to the non-financial sector.
  - Functional measure of non-banks reliant on long-term funding grew since the GFC and continued during the pandemic, although it has begun to decline more recently.
- Decomposition of asset growth:
  - Overall size of non-banks affected by flows into institutions and valuation effects through growth of assets.
  - Simple decomposition of asset growth into changes in flows and a residual (assumed to comprise valuation effects) suggests valuation effects were significant during the pandemic, particularly for mutual funds.
  - Importance of considering flows and returns of non-banks; monthly ICI data on long-term mutual funds used to analyze flows and net assets.
- ICI long-term mutual fund data (2000–2021):
  - Focus: U.S. domiciled open-ended mutual funds investing in domestic and international markets.
  - Data allow disaggregation by fund type: equity mutual funds (domestic and international), investment grade, high yield, world, government, multisector, municipal, hybrid bond mutual funds.
  - ICI net assets confirm Federal Reserve flow-of-funds data:
    - Sector quadrupled over last twenty years to what is now a 22 trillion industry.
    - Equity mutual funds make up more than half of total net assets.
    - ICI reports individual investors hold about 90 percent of open-ended mutual fund assets.
    - Bond mutual funds show continued inflows during the pandemic while equity funds experienced persistent outflows.

### Identifying Monetary Policy Shocks
- Main measure used: Jarociński and Karadi (2020) shocks.
  - Focus on interest rate surprises in the three-month fed funds future.
  - The three-month fed funds future exchanges a constant interest rate for the average federal funds rate over the course of the third calendar month in the contract.
  - Because regular FOMC meetings are six weeks apart, the three-month future reflects the shift in the expected federal funds rate after the following policy meeting, not the immediate next meeting.
  - These shocks do not capture surprises to the balance sheet, implicitly assuming such changes are orthogonal to surprises to the policy rate (and that balance sheet measures would not affect 3-month futures).
  - Shocks can be aggregated to monthly or quarterly frequency.
- Separation of monetary policy shocks from Fed information shocks:
  - Jarociński and Karadi (2020) separate pure monetary policy shocks from signaling shocks related to the state of the economy (“Fed information” shocks).
  - Fed information shocks capture agents interpreting Fed actions as signals about the state of the economy; a surprise monetary loosening can be taken as a sign that the economy is performing poorly.
  - Effect of Fed information shocks goes in the opposite direction to monetary policy shocks; mixing them can bias results and confound channels.
  - Distinguishing assumption: correlation between changes in interest rates and stock prices following an information shock is positive; following a monetary policy shock it is negative.
  - Figure A.1 presents the time-series of Jarociński and Karadi (2020) shocks, including both monetary policy and Fed information shocks. Monetary policy shocks exhibit both significant tightening and loosening periods over the sample period.

### The Impact of Monetary Policy on the Size of the Non-Bank Sector
- Empirical strategy:
  - Quantify impact of monetary shocks on the size of the non-bank sector using local projections following Jorda (2005) at quarterly frequency.
  - Outcome variables y_t include the functional measures based on Gallin (2015) for both short- and longer-term non-bank funding (preferred approach).
- Data and coverage:
  - Institutional measures drawn from Federal Reserve flow-of-funds up to 2022 Q1.
  - Monthly ICI data on long-term mutual funds cover 2000–2021 for flows and net assets (used to analyze flows and returns; abstracts from valuation issues).

*Italic: Source — wpiea2023055-print-pdf*

### Section 2. We also provide results from regressions using institutional measures for the outcome

### Section 2. We also provide results from regressions using institutional measures for the outcome

### Methodology
- Outcome variables:
  - Institutional measures of Pozsar et al. (2010) and Adrian et al. (2010) for non-banks reliant on short-term funding.
  - Real mutual funds assets (excluding money market funds) from the flow of funds for non-banks reliant on longer-term funding (institutional).
  - Corresponding functional and institutional measures for traditional banks.
- Regression specification (OLS with Newey and West (1987) standard errors):
  - y_{t+h} − y_{t} = α_{h} + β_{h} ε_{t} + γ_{h} y_{t−1} + u_{h t}  for quarter h = 1,2,....,12.
  - ε_{t} = ε^{MP}_{t} (monetary policy shock) or ε^{Info}_{t} (Fed information shock identified per Jarociński and Karadi (2020)).
  - Sample: 1990 Q1 through 2019 Q2.
  - Interpretation: estimated β_{h} is the percentage point change in the outcome after quarter h in response to a 100 basis point contractionary monetary policy shock or a 100 basis point positive Fed information shock.

### Main empirical findings (functional measures)
- Dislocation effect: contractionary monetary policy boosts short-term funded non-bank sector while shrinking long-term funded non-bank sector.
  - A 100 basis point contractionary monetary policy shock:
    - Expands inflation-adjusted non-bank assets reliant on short-term funding by around 1/4 percentage point after five quarters (statistically significant at the 10 percent level).
    - Reduces inflation-adjusted size of non-bank sector reliant on long-term funding by 1/4 percentage point after 12 quarters (statistically significant at the 10 percent level), following an initial increase.
  - Impact on bank assets (functional measure) is not found to be statistically significant in the baseline regression without controls.
- Fed information shocks (functional measures):
  - Positive Fed information shock reduces size of non-bank sector reliant on long-term funding by around 1/2 percentage points after 12 quarters (Gallin (2015) measure).
  - Initially reduces size of sector reliant on short-term funding by around 1/4 percentage points after five quarters, then eventually turns positive.

### Institutional measures and robustness
- Using institutional measures over the more recent sample and the preferred shock identification:
  - Contractionary monetary policy is not found to have a statistically significant impact on institutional measures of non-bank size generally.
  - The contractionary shock is found to reduce the size of mutual funds (excluding money market funds), though this effect is not necessarily robust to adding covariates.
- Robustness to controls:
  - The expansion of short-term non-bank sector (functional) following contractionary policy is robust to including lags of shocks and controls X_{t}:
    - Controls include year-over-year real GDP growth, Fed Funds Rate (midpoint of target range at end of quarter), term spread (10-year minus 3-month), year-over-year change in S&P 500, year-over-year change in CoreLogic Case-Shiller house price index, and year-over-year change in real assets of broker dealers.
  - The contractionary policy effect on long-term funded non-banks loses statistical significance when all controls are included.

### Comparisons to earlier literature
- Magnitudes found in this paper are smaller than in some earlier studies that focused on pre-GFC samples and institutional measures:
  - Nelson et al. (2017): a 100 basis point contractionary shock increased inflation-adjusted U.S. non-bank sector by a peak of 1.5 percent after one year (sample 1966-2007; institutional measure).
  - Den Haan and Sterk (2011): 100 basis point contractionary shock increased inflation-adjusted non-bank mortgages and consumer credit by 1 and 1.5 percent respectively after 18 and 36 months (sample 1984-2008).
  - Differences likely reflect sample period, shock identification methods, and measures of non-bank size.

### Mutual funds: flows and returns (monthly VAR, ICI data)
- Data and method:
  - Monthly ICI data, 2000-2019.
  - Two-variable VAR for each fund category: ratio of net flows to total net assets (f_{i,t} = c_{i,t} / a_{i,t−1}) and monthly price return r_{i,t} (defined per equations (2) and (3)).
  - Exogenous monetary policy surprise shocks (preferred measure) included; vector of controls X (core PCE, ADS business conditions index, BBB corporate-Treasury spreads, term spread, VIX, real effective exchange rates as robustness).
  - Two lags based on Akaike criterion.
  - Compute dynamic impact multipliers of exogenous monetary policy surprise shocks on flows and returns.
- Aggregate bond mutual funds (2000-2019):
  - A 100 basis points contractionary monetary policy surprise shock → 2 percent fall in bond mutual fund net flows and a 4 percent fall in returns (effects significant at the 5 percent level over 20 months).
- Aggregate equity mutual funds:
  - A 100 basis points contractionary shock → decline of net flows by about 1 percentage point on impact, and a larger impact on returns of about 12 percentage points on impact.
- Heterogeneity across fund types:
  - Greater impact of monetary policy on high yield bond funds (relative to other bond funds) and international equity funds (relative to domestic equity funds).
- Robustness and alternative shocks:
  - Results robust to different shock measures (e.g., Nakamura and Steinsson (2018)) and inclusion of additional explanatory variables.
  - Using Jarociński and Karadi (2020) information shocks produces opposite signs: positive information shocks → inflows and higher returns.

### Time-varying impacts (MS-VAR results)
- MS-VAR setup:
  - Monthly MS-VAR for flows and returns with five endogenous variables: flows and returns for mutual funds, monthly unemployment rate, core PCE inflation (month over month), Fed Funds Rate (midpoint of target range at end of month); two lags; three regimes for constants/AR coefficients and three regimes for variance-covariance matrix.
  - Identification: sign restrictions—contractionary monetary policy shock must make core PCE inflation negative for five months, Fed Funds Rate and unemployment rate positive for five months.
  - Estimation: Bayesian with flat priors; posterior mode used to compute regime probabilities.
- MS-VAR monthly mutual funds (2000-2019) findings:
  - Three regimes correspond to distinct time periods; regime after 2013 shows statistically significant reduction in bond fund flows by almost −0.2 percentage points on impact.
  - Contractionary monetary policy reduced flows into bond mutual funds more recently (post-2013).
  - For equity mutual funds, contractionary policy reduced flows in pre-2007 regime; not statistically significant in more recent regime.
- Quarterly MS-VARs on bank and non-bank size:
  - Sample: 1973 Q3 through 2019 Q4.
  - Endogenous variables include bank/non-bank size (functional or institutional), Fed Funds Rate, year-over-year real GDP growth, year-over-year GDP deflator growth.
  - Functional measures:
    - Non-banks reliant on long-term funding: contractionary shock expanded sector prior to 1980s but shrank it by up to 0.2 percentage points during the 2000s.
    - Non-banks reliant on short-term funding: strong dislocation effect after 2001—contractionary shocks boost the size of the sector.
  - Institutional measures:
    - Mutual funds (excluding money market funds): contractionary shocks shrink size beyond early 2000s.
    - Short-term funded non-bank institutional measures less clear; some pre-2001 dislocation effects found but results differ by institutional measure (Adrian et al. (2010) vs Pozsar et al. (2010)).
  - Banks:
    - Evidence that contractionary monetary policy caused a reduction in the size of traditional banking more recently in some MS-VAR specifications (functional definitions), after an initial expansion on impact.
    - Institutional MS-VAR results for banks are mixed; some specifications show a small reduction in bank size after 2001 (Adrian and Shin (2010) institutional definitions).

### Conclusions and policy implications
- Summary of empirical conclusions:
  - Contractionary monetary policy significantly reduces balance sheets of non-banks reliant on long-term funding and leads to significant outflows from long-term mutual funds.
  - Higher risk bond funds and international equity funds are most impacted, consistent with the risk-taking channel.
  - Evidence of a time-varying transmission: contractionary impact on long-term funded non-banks has generally strengthened in recent years; contractionary policy increases size of short-term funded non-banks in recent years (dislocation effect) when measured functionally.
  - Monetary tightening may induce flows out of mutual funds, reduce returns, shrink the overall long-term funded non-bank sector, but could expand short-term funded non-banks and thus increase susceptibility to run risks.
- Policy implications and avenues for research:
  - Monetary policy conduct needs to adapt as transmission mechanisms change with the growth of nonbank intermediation.
  - Stronger effects on the real economy and financial soundness via the risk-taking channel imply greater vigilance by prudential and regulatory authorities.
  - Further research recommended to explore underlying drivers of changing transmission to non-banks and consequences for monetary policy transmission to the real economy.

*Source: wpiea2023055-print-pdf - Section 2. Canonical URL: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023055-print-pdf.pdf*

### References

### References

### Contents overview
- Section comprises the bibliography cited in the report (references to academic papers, working papers, staff reports, and technical reports).
- Includes an Appendix A titled "Additional Tables and Figures" containing tables and figures referenced in the main text.
- Report title and identifier preserved exactly: The Heterogeneous Effects of U.S. Monetary Policy on Non-Bank Finance — Working Paper No. WP/2023/055.

### Appendix A — Additional Tables and Figures (enumerated items and notes preserved)
- Figure A.1: Monetary Policy Shocks (JK 2020)
  - Sources: Jarociński and Karadi (2020).
- Table A.1: Long-term Mutual Fund Assets Over Time
  - Heading: Total net assets(billions of US$, year-end)20002005201020152021
  - Note: This table summarizes net assets of various mutual fund classes based on ICI data. Sources: ICI and authors’ calculations.
- Figure A.2: Bond and Equity Mutual Funds Net Flows
  - Sources: ICI, authors’ calculations.
- Figure A.3: Bond Mutual Fund Flows by Different Type of Fund ($ million)
  - Sources: ICI, authors’ calculations.
- Figure A.4: Monetary Policy Shocks Versus Information Shocks
  - Sources: ICI, Jarociński and Karadi (2020), authors’ calculations.
- Figure A.5: Impact of 100 bps Monetary Shock on Non-Banks Assets (ppts)
  - Notes: Local projections based on regressions including the contemporaneous value and first lag of the shocks, and first lag of controls: year-over-year real GDP growth, the Federal Funds Rate (midpoint of target range at end of quarter), the spread between the yield on ten-year and three month US treasuries (term spread), the year-over-year change in the S&P 500 index, the year-over-year change in the CoreLogic Case-Shiller house price index, and the year-over-year change in the real assets of broker dealers.
  - Source: Authors’ calculations.
- Figure A.6: Impact of 100 bps Fed Information Shock on Non-Banks Assets (ppts)
  - Motes: Local projections based on regressions including the contemporaneous value and first lag of the shocks, and first lag of controls (same set as Figure A.5).
  - Source: Authors’ calculations.
- Figure A.7: Impact of 100 bps Monetary Shock on Bank Assets (ppts)
  - Notes: The ’institutional’ measure comprises real assets of commercial banks, credit unions and savings institutions, from the flow of funds. The ’functional’ measure is based on chains of intermediation beginning with commercial banks, credit unions and thrifts that reside in the U.S.
  - Sources: Federal Reserve and authors’ calculations.
- Figure A.8: Impact of 100 bps Fed Information Shock on Bank Assets (ppts)
  - Notes: The ’institutional’ measure comprises real assets of commercial banks, credit unions and savings institutions, from the flow of funds. The ’functional’ measure is based on chains of intermediation beginning with commercial banks, credit unions and thrifts that reside in the U.S.
  - Sources: Federal Reserve and authors’ calculations.
- Table A.2: Total Bond Flows and Returns Vector-Autoregression
  - Notes: This table summarizes VARs with different sets of explanatory variables.
  - Sources: ICI and authors’ calculations.
- Table A.3: Total Equity Flows and Returns Vector-Autoregression
  - Notes: This table summarizes VARs with different sets of explanatory variables.
  - Sources: ICI and authors’ calculations.
- Figure A.9: Time-Varying Impact of 1 Std. Dev. Monetary Shock on Equity Mutual Fund Flows (ppts)
  - Sources: ICI and authors’ calculations.
- Figure A.10: Probability of Regimes Within-Sample for MS-VAR with Bond Mutual Fund Flows
  - Notes: Probabilities evaluated at the posterior mode. The three possible regimes allowed for VAR coefficients and the three possible regimes allowed for the Variance-Covariance Matrix (VCOV) are assumed to be independent.
  - Sources: ICI and authors’ calculations.
- Figure A.11: Probability of Regimes Within-Sample for MS-VAR with Equity Mutual Fund Flows
  - Notes: Probabilities evaluated at the posterior mode. The three possible regimes allowed for VAR coefficients and the three possible regimes allowed for the Variance-Covariance Matrix (VCOV) are assumed to be independent.
  - Sources: ICI and authors’ calculations.
- Figure A.12: Probability of Regimes Within-Sample for MS-VAR with Assets of Long-Term Funded Non-Banks (Functional Definition)
  - Notes: Probabilities evaluated at the posterior mode. The three possible regimes allowed for VAR coefficients and the three possible regimes allowed for the Variance-Covariance Matrix (VCOV) are assumed to be independent.
  - Sources: Federal Reserve and authors’ calculations.
- Figure A.13: Probability of Regimes Within-Sample for MS-VAR with Assets of Short-Term Funded Non-Banks (Functional Definition)
  - Notes: Probabilities evaluated at the posterior mode. The three possible regimes allowed for VAR coefficients and the three possible regimes allowed for the Variance-Covariance Matrix (VCOV) are assumed to be independent.
  - Sources: Federal Reserve and authors’ calculations.
- Figure A.14: Probability of Regimes Within-Sample for MS-VAR with Assets of Long-Term Funded Non-Banks (Institutional Definition)
  - Notes: The institutional definition of the size of long-term funded non-banks comprises real mutual funds assets (excluding money market funds) from the flow of funds. Regime probabilities evaluated at the posterior mode. The three possible regimes allowed for VAR coefficients and the three possible regimes allowed for the Variance-Covariance Matrix (VCOV) are assumed to be independent.
  - Sources: Federal Reserve and authors’ calculations.
- Figure A.15: Probability of Regimes Within-Sample for MS-VAR with Assets of Short-Term Funded Non-Banks (Institutional Definition)
  - Notes: The institutional definition of the size of short-term funded non-banks is based on Adrian et al. (2010). Regime probabilities evaluated at the posterior mode. The three possible regimes allowed for VAR coefficients and the three possible regimes allowed for the Variance-Covariance Matrix (VCOV) are assumed to be independent.
  - Sources: Federal Reserve and authors’ calculations.
- Figure A.16: Time-Varying Impact of 1 Std. Dev. Monetary Shock on Bank Assets: Functional Defn. (ppts)
  - Source: Authors’ calculations.
- Figure A.17: Time-Varying Impact of 1 Std. Dev. Monetary Shock on Bank Assets: Institutional Defn. (ppts)
  - Source: Authors’ calculations.

*The Heterogeneous Effects of U.S. Monetary Policy on Non-Bank Finance — Working Paper No. WP/2023/055*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023055-print-pdf.pdf_