## _wp1676

## Source details

**Canonical URL:** [_wp1676](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1676.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1676.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1676.pdf.json)

---

### Methodology (Section III)
- Baseline identification blends the traditional recursive VAR (CEE identification) with a proxy-VAR using monetary policy surprise series.
- Sample period: starts in 1970 and ends in the second quarter of 2007.
- Monetary surprise series used: Coibion’s series (extends back to 1969).
- Monetary policy shock identified as the shock to the monetary policy surprise series rescaled to raise the fed funds rate by 100 basis points at peak.
- Baseline macro variables: oil dummy, log-differences of real GDP, log-differences of the PCE deflator, the series of monetary policy surprises, and the fed funds rate.
- Additional controls to capture financial transmission: 10-year term spread and log-difference of real house prices.
- VAR specification: multiple VARs estimated by adding flow of fund (FA) variables one at a time; credit variables ordered last and thus free to react contemporaneously (with robustness checks where they do not react contemporaneously).
- VAR estimated in both differences and levels; differences used primarily but levels produce broadly similar results.

### Data (Section IV)
- Main source: Financial Accounts (FAs) of the United States (Flow of Funds).
- Sectors: banks (private depository institutions), Government Sponsored Enterprises (GSEs), Agency- and GSE-backed mortgage pools, issuers of asset-backed securities (ABS), Finance Companies, and Life Insurance Companies.
- Key series collected: total assets, total mortgages, and subcategories (including “Other Loans and Advances” from the GSE balance sheet).
- Private-label securitization activity measure: change in mortgages held by ABS issuers (table L.216 in the FAs).
- Non-private securitization measure: total mortgages held by agency- and GSE-backed mortgage pools (table L.125 in the FAs).
- Deposit volumes: private domestic deposits of US-chartered depository institutions and credit unions (Table L.204 in the FAs).
- CD rates: average dealer bid rates for CDs actively traded in the secondary market (high-denomination, typically $1,000,000 or greater).

### Empirical results — How do interest rates affect financial intermediaries? (Section V)
- Monetary policy tightening shock of a 100bp increase in the federal funds rate:
  - Reduces output growth by about 0.9 percent after three quarters.
  - Reduces inflation by about 0.4 percent after three quarters.
- House prices and total mortgages decline persistently, consistent with reduced household credit demand.
- Term spread and lending slowdown suggest banks’ interest margin decline.
- Banks’ response:
  - Banks cut back lending and shrink balance sheets; growth decline starts in the second quarter and peaks 1 year after the shock.
  - Bank-level decline in the level of total assets is about 0.8 percent after 3 years.
  - Mortgage lending growth falls by almost 0.5 percentage points after 3 quarters.
- Other financial intermediaries:
  - Mortgages held by finance companies decline after an interest rate shock.
  - Mortgages held by life insurers increase slightly after the shock (life insurers account for about 5 percent of total mortgages).
  - Private-label securitization expands: the increase in ABS activity lasts around two years and adds around 1 percentage point to growth in the second quarter after the shock.
- Deposit and liability spreads:
  - Rates on bank liabilities (Eurodollar deposits and high-denomination CDs) rise one-to-one and almost immediately with the fed funds after a shock.
  - Volume of reservable banks’ checkable deposits falls.
  - Figure note: responses illustrated for a 50 bps tightening surprise in a specific spread/deposit exercise.

### Empirical results — GSE behavior (Section V)
- GSE mortgage pools’ assets decrease for around 4 quarters and by around 1 percentage point after a monetary tightening.
- Aggregate effect: taken together, private and GSE securitization activity declines in response to interest rate shocks (GSE mortgage pools historically larger than ABS issuers).
- Counterintuitive finding: GSE total assets growth rate rises for around 1½ years after an interest rate shock, converging to a 1.5 percent higher level than pre-shock.
- Decomposition of GSE assets:
  - Bank mortgages and GSE mortgages (conforming mortgages) decline as expected.
  - GSE holdings of agency MBS increase but are not precisely estimated.
  - ‘Other loans and advances’ category grows markedly: grows by almost 2 percentage points faster at impact and in the next quarter.
    - This category represented about 30 percent of total GSEs’ assets in 2007 and mostly consists of loans extended by GSEs to Federal Home Loan Banks (FHLBs).
    - Interpretation: tighter funding conditions shift mortgage demand to FHLB providers and GSEs react by financing FHLB banks; GSEs indirectly increase their role as a funding source for a variety of conforming and nonconforming mortgages of FHLB member banks.
- Table 1 (GSEs and FHLB Balance Sheets, 2007) (values reproduced exactly as in source):
  - GSEs Balance Sheet (2007) Share of Total Assets Bil$
    - Other loans and advances 29.7% 943
    - of which: to FHLB 27.3% 867
    - Agency- and GSE-backed securities 22.1% 703
    - Mortgages 20.3% 643
    - Miscel. 27.9% 887
    - total assets 3176
  - FHLB Balance Sheet (2007) % Bil$
    - tot assets 1,016,469
    - of which: advances 63.0% 640681
    - tot liabilities 1,016,469
    - of which: consolid. debt obligations 91.9% 934,214

### Robustness, alternative specifications, and appendix findings
- Robustness checks:
  - Credit variables sometimes treated as predetermined; results broadly align with baseline but the overall fall in total mortgages, bank mortgages, and bank assets can be larger.
  - Alternative CEE identification yields qualitatively similar results, though CEE suffers from the price puzzle even with the oil dummy.
  - Estimation in (log) levels broadly consistent with baseline; levels estimation can pose instability issues for confidence bands.
- Estimation details:
  - Proxy-VAR uses the policy surprise series as an external instrument for the fed funds shock.
  - Oil dummy included to alleviate the price puzzle (captures refiner acquisition cost of crude oil before 1986 and is zero thereafter).
  - NBR targeting period (1979Q3 to 1982Q2) excluded in some specifications by zeroing the surprise series.
  - Confidence bands computed via bootstrap Monte Carlo with 1000 random drawings; reported percentiles: 15th and 33rd.

### Conclusions and policy-relevant implications (Section VI)
- Heterogeneous intermediation response:
  - Different financial intermediaries react differently to policy rate shocks; higher rates can induce migration of financial activity from banks to less-monitored non-bank sectors via securitization.
  - As a result, higher interest rates may unintentionally raise systemic risk even while overall credit aggregate growth slows.
- GSE–FHLB interaction:
  - Loans of GSEs to other government-sponsored entities (notably FHLB) may increase after an interest rate hike, reflecting mandates and potential implicit guarantees.
- Current context caveats:
  - Post-crisis structural changes (private-label securitization has not recovered) may reduce the past channel of risk migration from banks to non-banks while rates remain historically low.
  - New nonbank channels have emerged after the GFC; the impact of monetary policy on these new areas requires careful analysis.
  - The bank lending channel of monetary policy may operate differently today due to:
    - Substantial amount of excess reserves that may dampen the effects of higher fed funds rates on bank liquidity and lending.
    - Different nature of substitution through nonbank channels with uncertain implications for systemic risk.
- Policy takeaway:
  - When evaluating monetary policy responses to financial instability, it is critical to account for these heterogeneous effects and measure their relative magnitudes.

*Source: _wp1676 - Section III presents our methodology to identify a monetary policy shock, Section IV*

### Section III presents our methodology to identify a monetary policy shock, Section IV

### _wp1676 - Section III presents our methodology to identify a monetary policy shock, Section IV

### Methodology (Section III)
- Presents the methodology to identify a monetary policy shock.

### Data (Section IV)
- Describes the financial data used in the empirical analysis.

### Results (Section V)
- Presents the results.

*Source: _wp1676 - Section III presents our methodology to identify a monetary policy shock, Section IV*

### Section VI concludes.

### _wp1676 - Section VI concludes.

### A brief overview of mortgage securitization in the U.S.
- Securitization market grew at very rapid pace in the decades preceding the Great Financial Crisis (GFC).
- Total loans typically represent about 2/3 of a bank total asset, making securitization a tool to increase bank balance-sheet liquidity.
- Home mortgages are the easiest loans to securitize; even after the GFC about 80 percent of new home mortgages are being securitized.
- The overall value of outstanding securitized mortgages was almost US$8 trillion in 2014, exceeding the size of the corporate bond market.
- Private-label ABS issuers’ share rose to a peak of 40 percent of total market in 2006, but the industry unwound after the crisis and is virtually non-existent in the current financial structure.
- Two periods of high growth in ABS issuance coincided with monetary policy tightening cycles.

### Identification strategy of interest rate shocks
- Baseline identification blends the traditional recursive VAR (CEE identification) with a proxy-VAR using monetary policy surprise series.
- Sample period: starts in 1970 and ends in the second quarter of 2007.
- Monetary surprise series used: Coibion’s series (extends back to 1969).
- Monetary policy shock identified as the shock to the monetary policy surprise series rescaled to raise the fed funds rate by 100 basis points at peak.
- Baseline macro variables: oil dummy, log-differences of real GDP, log-differences of the PCE deflator, the series of monetary policy surprises, and the fed funds rate.
- Additional controls to capture financial transmission: 10-year term spread and log-difference of real house prices.
- VAR specification: multiple VARs estimated by adding flow of fund (FA) variables one at a time; credit variables ordered last and thus free to react contemporaneously (with robustness checks where they do not react contemporaneously).
- VAR estimated in both differences and levels; differences used primarily but levels produce broadly similar results.

### Description of the data
- Main source: Financial Accounts (FAs) of the United States (Flow of Funds).
- Sectors: banks (private depository institutions), Government Sponsored Enterprises (GSEs), Agency- and GSE-backed mortgage pools, issuers of asset-backed securities (ABS), Finance Companies, and Life Insurance Companies.
- Key series collected: total assets, total mortgages, and subcategories (including “Other Loans and Advances” from the GSE balance sheet).
- Private-label securitization activity measure: change in mortgages held by ABS issuers (table L.216 in the FAs).
- Non-private securitization measure: total mortgages held by agency- and GSE-backed mortgage pools (table L.125 in the FAs).
- Deposit volumes: private domestic deposits of US-chartered depository institutions and credit unions (Table L.204 in the FAs).
- CD rates: average dealer bid rates for CDs actively traded in the secondary market (high-denomination, typically $1,000,000 or greater).

### Empirical results — How do interest rates affect financial intermediaries?
- Monetary policy tightening shock of a 100bp increase in the federal funds rate:
  - Reduces output growth by about 0.9 percent after three quarters.
  - Reduces inflation by about 0.4 percent after three quarters.
- House prices and total mortgages decline persistently, consistent with reduced household credit demand.
- Term spread and lending slowdown suggest banks’ interest margin decline.
- Banks’ response:
  - Banks cut back lending and shrink balance sheets; growth decline starts in the second quarter and peaks 1 year after the shock.
  - Bank-level decline in the level of total assets is about 0.8 percent after 3 years.
  - Mortgage lending growth falls by almost 0.5 percentage points after 3 quarters.
- Other financial intermediaries:
  - Mortgages held by finance companies decline after an interest rate shock.
  - Mortgages held by life insurers increase slightly after the shock (life insurers account for about 5 percent of total mortgages).
  - Private-label securitization expands: the increase in ABS activity lasts around two years and adds around 1 percentage point to growth in the second quarter after the shock.
- Deposit and liability spreads:
  - Rates on bank liabilities (Eurodollar deposits and high-denomination CDs) rise one-to-one and almost immediately with the fed funds after a shock.
  - Volume of reservable banks’ checkable deposits falls.
  - Figure note: responses illustrated for a 50 bps tightening surprise in a specific spread/deposit exercise.

### Empirical results — GSE behavior
- GSE mortgage pools’ assets decrease for around 4 quarters and by around 1 percentage point after a monetary tightening.
- Aggregate effect: taken together, private and GSE securitization activity declines in response to interest rate shocks (GSE mortgage pools historically larger than ABS issuers).
- Counterintuitive finding: GSE total assets growth rate rises for around 1½ years after an interest rate shock, converging to a 1.5 percent higher level than pre-shock.
- Decomposition of GSE assets:
  - Bank mortgages and GSE mortgages (conforming mortgages) decline as expected.
  - GSE holdings of agency MBS increase but are not precisely estimated.
  - ‘Other loans and advances’ category grows markedly: grows by almost 2 percentage points faster at impact and in the next quarter.
    - This category represented about 30 percent of total GSEs’ assets in 2007 and mostly consists of loans extended by GSEs to Federal Home Loan Banks (FHLBs).
    - Interpretation: tighter funding conditions shift mortgage demand to FHLB providers and GSEs react by financing FHLB banks; GSEs indirectly increase their role as a funding source for a variety of conforming and nonconforming mortgages of FHLB member banks.
- Table 1 (GSEs and FHLB Balance Sheets, 2007) (values reproduced exactly as in source):
  - GSEs Balance Sheet (2007) Share of Total Assets Bil$
    - Other loans and advances 29.7% 943
    - of which: to FHLB 27.3% 867
    - Agency- and GSE-backed securities 22.1% 703
    - Mortgages 20.3% 643
    - Miscel. 27.9% 887
    - total assets 3176
  - FHLB Balance Sheet (2007) % Bil$
    - tot assets 1,016,469
    - of which: advances 63.0% 640681
    - tot liabilities 1,016,469
    - of which: consolid. debt obligations 91.9% 934,214

### Robustness, alternative specifications, and appendix findings
- Robustness checks:
  - Credit variables sometimes treated as predetermined; results broadly align with baseline but the overall fall in total mortgages, bank mortgages, and bank assets can be larger.
  - Alternative CEE identification yields qualitatively similar results, though CEE suffers from the price puzzle even with the oil dummy.
  - Estimation in (log) levels broadly consistent with baseline; levels estimation can pose instability issues for confidence bands.
- Estimation details:
  - Proxy-VAR uses the policy surprise series as an external instrument for the fed funds shock.
  - Oil dummy included to alleviate the price puzzle (captures refiner acquisition cost of crude oil before 1986 and is zero thereafter).
  - NBR targeting period (1979Q3 to 1982Q2) excluded in some specifications by zeroing the surprise series.
  - Confidence bands computed via bootstrap Monte Carlo with 1000 random drawings; reported percentiles: 15th and 33rd.

### Conclusions and policy-relevant implications
- Heterogeneous intermediation response:
  - Different financial intermediaries react differently to policy rate shocks; higher rates can induce migration of financial activity from banks to less-monitored non-bank sectors via securitization.
  - As a result, higher interest rates may unintentionally raise systemic risk even while overall credit aggregate growth slows.
- GSE–FHLB interaction:
  - Loans of GSEs to other government-sponsored entities (notably FHLB) may increase after an interest rate hike, reflecting mandates and potential implicit guarantees.
- Current context caveats:
  - Post-crisis structural changes (private-label securitization has not recovered) may reduce the past channel of risk migration from banks to non-banks while rates remain historically low.
  - New nonbank channels have emerged after the GFC; the impact of monetary policy on these new areas requires careful analysis.
  - The bank lending channel of monetary policy may operate differently today due to:
    - Substantial amount of excess reserves that may dampen the effects of higher fed funds rates on bank liquidity and lending.
    - Different nature of substitution through nonbank channels with uncertain implications for systemic risk.
- When evaluating monetary policy responses to financial instability, it is critical to account for these heterogeneous effects and measure their relative magnitudes.

*Source: _wp1676 - Section VI concludes.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1676.pdf_