## 1. Estimating Nonbank Funding to U.S. Banks (2008–16)

## Source details

**Canonical URL:** [1. Estimating Nonbank Funding to U.S. Banks (2008–16)](https://www.imf.org/-/media/files/publications/wp/2018/wp1862.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp1862.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp1862.pdf.json)

---

### Introduction: motivation and core argument
- Traditional leverage metrics use on-balance sheet bank data and omit substantial off-balance sheet interconnections between banks and nonbanks.
- Pledged collateral is proposed as a useful measure of nonbank funding to banks; omission of off-balance sheet items implies a substantial underestimation of bank leverage.
- Approach:
  - Use the theoretical framework described in Shin (2010) to estimate aggregate nonbank funding (wholesale and household deposits) to banks across large, interconnected financial systems.
  - Collect and analyze data on pledged collateral funding for the largest 15 global systemically important banks (G-SIBs).
  - Estimate the impact of fully accounting for pledged collateral on bank leverage at the G-SIB level.
- Key takeaway:
  - Accounting for pledged collateral transactions could increase bank leverage values by about a third at the G-SIB level.
- Policy implication preview:
  - Systematic capture of off-balance sheet items, especially pledged collateral transactions, is needed in bank leverage metrics to improve global financial stability monitoring.

### Literature review: data gaps and existing metrics
- Common data sources and metrics cited:
  - National statistics (flow of funds), SNL Financial, BIS consolidated international banking statistics, Fitch’s Bankscope, IMF Financial Soundness Indicators (FSI).
- Limitations highlighted:
  - Flow-of-funds and typical databases tear up holding-company balance sheets and aggregate by entity type, losing the holding-company level picture.
  - BIS consolidated banking statistics and FSI generally do not capture large off-balance sheet transactions such as pledged collateral.
  - Basel III does not require a modified leverage metric that spans all off-balance sheet transactions.
- Relevant prior findings summarized:
  - Adrian and Shin (2009): M2 is a good proxy for liquid claims; M2 “has been slow-moving or stable over time, expanding ‘by a factor of 2.4 since 1994.’”
  - Schularick and Taylor (2012): use credit-to-GDP as proxy for leverage and argue off-balance sheet activities should be included in the “entire balance sheet.”
  - Other cited works (Borio and Drehmann (2009); Lee et al. (2017); Claessens and Ratnovski (2014); Adrian et al. (2013); Fischer (2015)) underscore nonbank funding and off-balance sheet activities complicate leverage measurement and crisis prediction.

### Financial system leverage: mechanics and quantitative context
- Conceptual framework (Box 1 / Shin (2010)):
  - yi = total lending to ultimate borrowers by bank i
  - ei = equity of bank i
  - λi = leverage of bank i (defined as asset over equity)
  - zi = fraction of nonbank funding bank i receives
  - Nonbank funding decomposed: Zi = Zh + Zw (Zh = household/core funding; Zw = wholesale market funding)
- Non-M2 funding and pledged collateral:
  - Non-M2 funding is neither “sticky” nor confined to household deposits; asset managers’ short-term investments are primarily non-M2 liabilities.
  - Pledged collateral circulates as a “cash equivalent” in cross-border secured funding, margining, repo, securities lending, and OTC derivatives; it has regulatory and settlement value.
- Quantitative magnitudes:
  - The gross volume of wholesale funding intermediated by banks in the U.S. may have been as high as US$25 trillion at end-2007 (Pozsar and Singh 2011).
  - The banking system estimate at that time was US$13 trillion.
- Relative scale of pledged collateral versus money:
  - Annual savings in major advanced economies are described as in the range of 20–25 percent of GDP.
  - “Savings comprise about 25 percent of M2 in advanced economies” and only this fraction of money interfaces with pledged collateral markets.
  - From this perspective, pledged collateral used in financial transactions is probably at par with money when considering the portion of M2 that functions as savings interfacing with market transactions.

### Empirical evidence and findings
- Data sources and empirical focus:
  - Aggregate estimates of nonbank funding use Shin (2010) theoretical framework.
  - Detailed dataset on pledged collateral funding for the largest 15 G-SIBs to assess magnitude and trends relative to balance sheet assets.
- Aggregate and trend findings:
  - Aggregate nonbank funding to banks has been rising among economies with large and interconnected financial systems.
  - Pledged collateral funding is large and has been rising relative to banks’ balance sheet assets for major global banks.
  - Fully accounting for pledged collateral transactions could increase measured bank leverage by about a third at the G-SIB level.
- Market characteristics and magnitudes:
  - Pledged collateral market volume has remained at $6 trillion since 2009.
  - QE-related increase in deposits makes the leverage ratio more binding.
  - Reuse rate ranged from a high of 3 before the Lehman crisis to below 2 in recent years (see Box 3 discussion).
  - Sizable dispersion: about 40 percent to 70 percent of off-balance-sheet funding may not have appeared on-balance sheet pre-crisis.
- Limitations acknowledged:
  - Pledged collateral accounting does not capture all interconnectedness nor fully resolve cross-border identification issues in national reporting.
  - Many pledged collateral transactions are not identifiable by jurisdiction and therefore fall through national balance sheet reporting cracks.

### Empirical estimates and illustrative bank-level examples
- Table 1: Estimating Nonbank Funding to U.S. Banks (2008–16)
  - 2008:
    - Bank Capital e: 1.6 (In trillions of U.S. dollars)
    - Bank Credit y: 8.4 (In trillions of U.S. dollars)
    - Bank Leverage λ: 9.2
    - Average Nonbank Funding to Banks (fraction) z: 0.52
  - 2013:
    - Bank Capital e: 1.7 (In trillions of U.S. dollars)
    - Bank Credit y: 8.1 (In trillions of U.S. dollars)
    - Bank Leverage λ: 5.9
    - Average Nonbank Funding z: 0.74
  - 2016:
    - Bank Capital e: 2.0 (In trillions of U.S. dollars)
    - Bank Credit y: 9.7 (In trillions of U.S. dollars)
    - Bank Leverage λ: 5.6
    - Average Nonbank Funding z: 0.86
  - Sources: FSI (IMF), OECD, and BIS.
- Bank-level pledged collateral example (Barclays, 2016):
  - Fair value of securities accepted as collateral: 466.2 (In billions of British pounds)
  - Off-Balance Sheet Collateral Received: 466.2 (In billions of British pounds)
  - Collateral Received of which On-Pledged: 405.5 (In billions of British pounds)
  - Readily Available Assets: 59.7 (In billions of British pounds)
  - Of Barclays’ total pledged collateral in 2016—£466 billion—only £34 billion made it onto the balance sheet; £432 billion did not make it onto the balance sheet.
  - Barclays’ reported total assets in 2016: £1.2 trillion.
  - Including pledged collateral not on the balance sheet, Barclays’ true total assets in 2016 were 35 percent greater than assets reported on the balance sheet.
  - Barclays’ leverage (assets/equity) rises from about 20.7 (based on balance sheet data alone) to 28.1 if pledged collateral transactions are fully recorded.
- General patterns across major dealer-banks:
  - Balance-sheet sizes declined for many banks aside from JPMorgan, Bank of America, and Nomura (growth partly due to acquisitions).
  - Off-balance-sheet pledged collateral volumes did not decline after the initial post-Lehman years.
  - Some banks (e.g., JPMorgan and Barclays) increased their market share in pledged collateral.

### International accounting standards, reporting challenges, and leverage measurement
- Two central challenges for international comparison:
  - (i) removing effects of accounting standard differences (GAAP vs. IFRS);
  - (ii) capturing transactions recorded off-balance sheet.
- Basel QIS participation is voluntary; latest QIS report suggests nearly all banks meet a 3 percent leverage requirement, yet leverage measured as capital-asset ratios differ across countries:
  - about 11.5 percent in the U.S.
  - 6 percent in Germany
  - 5.5 percent in Japan
  - about 7 percent in U.K. and Switzerland.
- Basel III principle: “banks must not take account of physical or financial collateral, guarantees or other credit risk mitigation techniques to reduce the [leverage ratio] exposure measure” (BIS, 2014).
- New accounting standards have forced many secured financing assets on-balance sheet; exception: borrow-vs-pledge (BVP) transactions can remain off-balance sheet because no cash changes hands.
- Differences in netting policies under IFRS and GAAP complicate cross-country comparison of off-balance-sheet funding.

### Policy recommendations and data priorities
- Primary recommendations:
  - Augment bank leverage metrics by systematically capturing off-balance sheet items, especially pledged collateral transactions, to better reflect total financial system leverage.
  - Improve international and cross-border statistical collection to identify and measure pledged collateral flows that are not attributed to a single jurisdiction in national balance sheet reporting.
- Rationale and expected benefits:
  - Enhanced measurement would reduce underestimation of financial leverage and improve monitoring of risks related to credit intermediation and market liquidity.
  - Better data would enrich macroprudential surveillance and academic research on leverage, credit supply, and financial stability.
- Operational suggestions:
  - Collect more granular data at the holding-company level and on off-balance-sheet collateral intermediation across dealer banks, custodians, asset managers, hedge funds, CCPs, and other nonbank actors.
  - Harmonize reporting and disclosure requirements that better capture pledged collateral and off-balance-sheet funding to assess true leverage and systemic risk.
  - Use international fora (e.g., BIS QIS) to request information on the extent of pledged collateral and other off-balance-sheet transactions to augment national leverage data with a “cross-border leverage” metric.

### Box 4 — Econometric implications of using a broader leverage metric
- Measurement error in leverage and econometric bias:
  - Measurement error in explanatory variables biases regression coefficients toward zero (attenuation bias), especially in panel regressions with fixed effects.
  - The expected value of the measurement error for leverage may be nonzero because actual leverage is argued to be much higher due to unaccounted nonbank funding.
  - Attenuation persists because it is driven by the variance of the measurement error; given the collateral market multiplier (ranged from 3 pre-Lehman to below 2 in recent years), the variance of measurement error may be comparable to recorded variance.
  - Result: estimated beta coefficients on leverage in current research are suspected to be much smaller than the “true” beta.
- Implications for empirical research and inference:
  - Research using leverage from existing databases suffers from measurement bias; significant or insignificant estimates of leverage effects could change materially with a more complete leverage measure.
  - Accounting differences can be differenced away with panel fixed effects, but measurement bias from incomplete leverage measurement persists.
- Policy and data recommendations (Box 4):
  - Policymakers and researchers need a deeper understanding of links between globally active banks and their off-balance-sheet funding from nonbank sources.
  - Accurately accounting for nonbank funding (households and wholesale) would provide a more complete picture of the financial system’s leverage and improve early detection of financial cycles.
  - Develop databases to fill the information gap on off-balance sheet transactions; properly account for pledged collateral transactions, wealth management accounts, and other shadow banking areas.
  - BIS QIS and related work can help bridge GAAP and IFRS to conform to a standard template for leverage and should request information on pledged collateral not on the balance sheet.

### Concluding assessment
- Off-balance sheet pledged collateral is a significant component of nonbank funding to banks and a key element of financial plumbing that affects market liquidity and leverage.
- Standard on-balance-sheet leverage metrics understate true leverage when they omit pledged collateral and broader non-M2 funding channels.
- Systematic data collection and revised leverage metrics that incorporate pledged collateral would materially improve understanding and monitoring of global financial leverage.

*Source: https://www.imf.org/-/media/files/publications/wp/2018/wp1862.pdf*

### 1. Estimating Nonbank Funding to U.S. Banks (2008–16) .....................................................13

### 1. Estimating Nonbank Funding to U.S. Banks (2008–16) .....................................................13

### Introduction: motivation and core argument
- Traditional leverage metrics rely on on-balance sheet bank data (e.g., total assets to capital, risk-weighted assets to Tier 1 capital) and therefore omit substantial off-balance sheet interconnections between banks and nonbanks.
- The paper argues that pledged collateral is a useful measure of nonbank funding to banks, and that omission of off-balance sheet items implies a substantial underestimation of bank leverage.
- Approach:
  - Use the theoretical framework described in Shin (2010) to estimate aggregate nonbank funding (wholesale and household deposits) to banks across large, interconnected financial systems.
  - Collect and analyze data on pledged collateral funding for the largest 15 global systemically important banks (G-SIBs).
  - Estimate the impact of fully accounting for pledged collateral on bank leverage at the G-SIB level.
- Key takeaway:
  - Accounting for pledged collateral transactions could increase bank leverage values by about a third at the G-SIB level.
- Policy implication preview:
  - Further efforts are needed to systematically capture off-balance sheet items, especially pledged collateral transactions, in bank leverage metrics to improve global financial stability monitoring.

### Literature review: data gaps and existing metrics
- Common data sources and metrics:
  - National statistics (flow of funds), SNL Financial, BIS consolidated international banking statistics, Fitch’s Bankscope.
  - IMF Financial Soundness Indicators (FSI) measure of leverage (capital to total assets) appears to include on-balance sheet measures only.
- Limitations highlighted:
  - Flow-of-funds and typical databases tear up holding-company balance sheets and aggregate by entity type, losing the holding-company level picture.
  - BIS consolidated banking statistics and FSI generally do not capture large off-balance sheet transactions such as pledged collateral.
  - Recent regulations (Basel III) do not require a modified leverage metric that spans all off-balance sheet transactions (see Box 2 in source).
- Relevant findings from prior literature cited in the text:
  - Adrian and Shin (2009) suggest M2 is a good proxy for liquid claims held by ultimate creditors and note M2 “has been slow-moving or stable over time, expanding ‘by a factor of 2.4 since 1994.’”
  - Schularick and Taylor (2012) use credit-to-GDP as a proxy for leverage and argue off-balance sheet activities should be included in the “entire balance sheet.”
  - Other works (Borio and Drehmann (2009); Lee et al. (2017); Claessens and Ratnovski (2014); Adrian et al. (2013); Fischer (2015)) underscore that nonbank funding and off-balance sheet activities complicate leverage measurement and crisis prediction.

### Financial system leverage: mechanics and quantitative context
- Conceptual framework (Box 1 / Shin (2010)):
  - yi = total lending to ultimate borrowers by bank i
  - ei = equity of bank i
  - λi = leverage of bank i (defined as asset over equity)
  - zi = fraction of nonbank funding bank i receives
  - Aggregate lending relates to banks’ equity and the total funding from nonbanks; nonbank funding can be decomposed as Zi = Zh + Zw (Zh = household/core funding; Zw = wholesale market funding).
- Non-M2 funding and the pledged collateral market:
  - Non-M2 funding is neither “sticky” nor confined to household deposits; asset managers’ short-term investments are primarily non-M2 liabilities.
  - Pledged collateral circulates as a form of “cash equivalent” in cross-border secured funding, margining, repo, securities lending, and OTC derivatives, and has regulatory as well as settlement value.
- Quantitative magnitudes given in the text:
  - The gross volume of wholesale funding intermediated by banks in the U.S. may have been as high as US$25 trillion at end-2007 (Pozsar and Singh 2011).
  - The banking system estimate at that time was US$13 trillion.
- Relative scale of pledged collateral versus money:
  - Annual savings in major advanced economies are described as in the range of 20–25 percent of GDP.
  - The paper notes that “savings comprise about 25 percent of M2 in advanced economies” and that only this fraction of money interfaces with pledged collateral markets (Figure 2b discussion).
  - From this perspective, pledged collateral used in financial transactions is probably at par with money when considering the portion of M2 that functions as savings interfacing with market transactions.

### Empirical evidence and findings (methods summarized)
- Data sources and empirical focus:
  - Aggregate estimates of nonbank funding use the Shin (2010) theoretical framework.
  - Detailed dataset collected on pledged collateral funding for the largest 15 G-SIBs to assess magnitude and trends relative to balance sheet assets.
- Findings:
  - Aggregate nonbank funding to banks has been rising among economies with large and interconnected financial systems.
  - Pledged collateral funding is large and has been rising relative to banks’ balance sheet assets for major global banks.
  - Fully accounting for pledged collateral transactions could increase measured bank leverage by about a third at the G-SIB level.
- Limitations acknowledged:
  - Pledged collateral accounting does not capture all interconnectedness within the financial system, nor does it fully resolve cross-border identification issues inherent in national balance sheet reporting.
  - Many pledged collateral transactions are not identifiable by jurisdiction and therefore fall through national balance sheet reporting cracks.

### Policy recommendations and data priorities
- Primary recommendations:
  - Augment bank leverage metrics by systematically capturing off-balance sheet items, especially pledged collateral transactions, to better reflect total financial system leverage.
  - Improve international and cross-border statistical collection to identify and measure pledged collateral flows that are not attributed to a single jurisdiction in national balance sheet reporting.
- Rationale and expected benefits:
  - Enhanced measurement would reduce underestimation of financial leverage and improve monitoring of risks related to credit intermediation and market liquidity.
  - Better data would enrich both macroprudential surveillance and academic research on leverage, credit supply, and financial stability.
- Operational suggestion implied by text:
  - Collect more granular data at the holding-company level and on off-balance-sheet collateral intermediation across dealer banks, custodians, asset managers, hedge funds, CCPs, and other nonbank actors.

### Concluding assessment
- Off-balance sheet pledged collateral is a significant component of nonbank funding to banks and a key element of financial plumbing that affects market liquidity and leverage.
- Standard on-balance-sheet leverage metrics understate true leverage when they omit pledged collateral and broader non-M2 funding channels.
- Systematic data collection and revised leverage metrics that incorporate pledged collateral would materially improve understanding and monitoring of global financial leverage.

*Source: https://www.imf.org/-/media/files/publications/wp/2018/wp1862.pdf*

### 16.1 percent 14.5 percent, 23.3 percent, and 22.7 percent, respectively. Source: World Bank.

### wp1862 - 16.1 percent 14.5 percent, 23.3 percent, and 22.7 percent, respectively. Source: World Bank.

### International accounting standards, reporting challenges, and leverage measurement
- Recent changes to Basel’s framework attempt to consolidate G-SIB financial data but national reporting remains constrained by domestic laws and accounting standards (GAAP vs. IFRS).
- Two central challenges for international comparison:
  - (i) removing effects of accounting standard differences (GAAP vs. IFRS);
  - (ii) capturing transactions recorded off-balance sheet.
- Basel QIS participation is voluntary; latest QIS report suggests nearly all banks meet a 3 percent leverage requirement, yet leverage measured as capital-asset ratios differ across countries:
  - about 11.5 percent in the U.S.
  - 6 percent in Germany
  - 5.5 percent in Japan
  - about 7 percent in U.K. and Switzerland.
- Basel III principle: “banks must not take account of physical or financial collateral, guarantees or other credit risk mitigation techniques to reduce the [leverage ratio] exposure measure” (BIS, 2014).
- New accounting standards have forced many secured financing assets on-balance sheet; exception: borrow-vs-pledge (BVP) transactions can remain off-balance sheet because no cash changes hands.
- Differences in netting policies under IFRS and GAAP complicate cross-country comparison of off-balance-sheet funding.

### A broader view of leverage: importance of off-balance-sheet pledged collateral
- Bank credit determinants include equity, leverage, and the share of off-balance-sheet pledged collateral funding; traditional leverage metrics need augmentation to capture nonbank funding.
- Key empirical findings:
  - Nonbank funding (from balance sheet data) in the U.S. rose by about 50 percent since the crisis (see Table 1, last column).
  - Up to half (with wide variance) of pledged collateral activity may not be on-balance sheet, per market discussions.
  - Pledged collateral is the largest off-balance-sheet component for the 15 G-SIBs active in the market.
  - Pledged collateral volumes have not declined since the initial years after the Lehman crisis; in relative terms, off-balance-sheet funding is higher now than in 2007.
- Pledged collateral market characteristics:
  - The pledged collateral market volume has remained at $6 trillion since 2009.
  - QE-related increase in deposits makes the leverage ratio more binding.
  - Reuse rate (the average number of times pledged collateral is reused) is lower now compared to pre-Lehman levels.
  - There is sizable dispersion: about 40 percent to 70 percent of off-balance-sheet funding may not have appeared on-balance sheet pre-crisis.

### Empirical estimates and illustrative bank-level examples
- Table 1: Estimating Nonbank Funding to U.S. Banks (2008–16)
  - 2008:
    - Bank Capital (In trillions of U.S. dollars) e: 1.6
    - Bank Credit (In trillions of U.S. dollars) y: 8.4
    - Bank Leverage λ: 9.2
    - Average Nonbank Funding to Banks (fraction) z: 0.52
  - 2013:
    - Bank Capital e: 1.7
    - Bank Credit y: 8.1
    - Bank Leverage λ: 5.9
    - Average Nonbank Funding z: 0.74
  - 2016:
    - Bank Capital e: 2.0
    - Bank Credit y: 9.7
    - Bank Leverage λ: 5.6
    - Average Nonbank Funding z: 0.86
  - Sources: FSI (IMF), OECD, and BIS.
  - Note: European banks’ leverage ratios are being harmonized under Basel’s leverage rules; Table 1 shows U.S. data as representative of the leverage trend under Basel.
- Bank-level pledged collateral evidence (examples from Figures and Boxes):
  - Barclays (2016):
    - Fair value of securities accepted as collateral: 466.2 (In billions of British pounds)
    - Off-Balance Sheet Collateral Received: 466.2 (In billions of British pounds)
    - Collateral Received of which On-Pledged: 405.5 (In billions of British pounds)
    - Readily Available Assets: 59.7 (In billions of British pounds)
    - Of Barclays’ total pledged collateral in 2016—£466 billion—only £34 billion made it onto the balance sheet; £432 billion did not make it onto the balance sheet.
    - Barclays’ reported total assets in 2016: £1.2 trillion.
    - Including pledged collateral not on the balance sheet, Barclays’ true total assets in 2016 were 35 percent greater than assets reported on the balance sheet.
    - Barclays’ leverage (assets/equity) rises from about 20.7 (based on balance sheet data alone) to 28.1 if pledged collateral transactions are fully recorded.
  - Figures 3a and 3b panels show pledged collateral and total assets for major dealer-banks (Morgan Stanley, Goldman Sachs, Bank of America, JPMorgan Chase, Citigroup, Lehman Brothers (2004–07), Societe Generale, BNP Paribas, Deutsche Bank, Credit Suisse, UBS, Barclays, Royal Bank of Scotland, Nomura, HSBC). General patterns:
    - Balance-sheet sizes (blue lines) declined for many banks aside from JPMorgan, Bank of America, and Nomura (growth partly due to acquisitions).
    - Off-balance-sheet pledged collateral volumes did not decline after the initial post-Lehman years.
    - Some banks (e.g., JPMorgan and Barclays) increased their market share in pledged collateral.
- Implications for measurement and econometrics:
  - True leverage is systematically higher than captured in common databases; leverage data imported from these databases are measured with error.
  - Measurement error in explanatory variables biases regression estimates toward zero; t-statistics are biased downward.
  - Some smaller and/or insignificant effects of leverage in research could stem from leverage measurement error.

### Policy and regulatory implications
- Single-country and cross-country empirical work that relies on balance-sheet data will understate brokers’ true dependence on volatile off-balance-sheet financing.
- Leverage ratio frameworks that ignore off-balance-sheet pledged collateral will understate bank leverage and interconnectedness among G-SIBs.
- For global financial stability, sizable off-balance-sheet transactions—particularly pledged collateral—cannot be ignored because they provide a basis for interconnectedness among G-SIBs.
- Disclosure and harmonization:
  - Harmonized reporting and disclosure requirements that better capture pledged collateral and off-balance-sheet funding are needed to assess true leverage and systemic risk.
  - Monitoring efforts such as Basel QIS help but are limited by voluntary participation and national reporting constraints.
- Research and early warning:
  - Aggregation of data can obscure the left tail of distribution where early warning signals of financial instability may reside; granular disclosure of off-balance-sheet activity in G-SIBs would improve early warning capacity.

*Source: Excerpt from wp1862 (IMF working paper content provided).*

### Box 4. Econometric Implications of Using a Broader Leverage Metric

### Box 4. Econometric Implications of Using a Broader Leverage Metric

### Measurement error in leverage and econometric bias
- Presence of measurement error in explanatory variables may cause estimates of their regression coefficients to be biased towards zero.
- This attenuation bias will typically be more pronounced in panel regressions with fixed effects.
- Only under certain assumptions will the size of the bias in Ordinary Least Squares and panel with fixed effects regressions be of the same size.
- The assumption that the expected value of the measurement error is zero may not be entirely accurate for leverage:
  - The paper argues actual leverage is much higher in advanced economies (due to unaccounted nonbank funding).
  - Observed leverage in academic literature is either lower than the banks described in this paper that have sizeable transactions recorded off-balance sheet, or at the very most it would be equal to actual leverage (for those small banks that do not have an off-balance sheet).
  - Therefore the expected value of the measurement error is nonzero.
- Attenuation in coefficient estimates resulting from measurement error still holds under this structure of the measurement error, since the attenuation is driven by the variance of the measurement error.
- Given the size and multiplier effect of the collateral market (it has ranged from a high of 3 before the Lehman crisis to below 2 in recent years (Box 3)), it is suspected that the variance of the measurement error would be comparable in magnitude to the recorded variance.
- It is therefore reasonable to suspect that estimated beta coefficients on leverage in the current library of research are much smaller than the “true” beta.

### Implications for empirical research and inference
- Current empirical research that uses leverage from existing databases suffers from measurement bias.
- Studies that find significant (insignificant) estimates of the impact of leverage (on some dependent variable) might find their impacts to be much greater (significant and higher in magnitude) if they were to have a more complete measure of leverage.
- Measurement bias arises when leverage is measured using only on-balance sheet data, or even incomplete off-balance sheet information.
- While accounting differences (GAAP vs. IFRS) can be differenced away by using panel fixed effects, measurement bias from incomplete leverage measurement persists.
- Note that despite Basel’s “uniformity” via the QIS—see Box 2—the database that researchers use will still incorporate data filed via GAAP or IFRS to the local regulators.

### Policy and data recommendations
- Policymakers and researchers interested in monitoring leverage and fully appreciating its impact on financial cycles and vulnerabilities would benefit from a deeper understanding of the links between banks with a global footprint, and their off-balance sheet funding from non-bank sources.
- Accurately accounting for nonbank funding (households and wholesale), which has been on the rise since the Lehman-crisis and constitutes a major source of bank credit to the economy, would provide a more complete picture of the financial system’s leverage.
- Having a broader measure of leverage in the economy could enable policymakers to perceive financial cycles earlier and more accurately, allowing them to better apply policy options such as monetary and macroprudential policies in a timely fashion.
- There is a need to develop databases that can fill the information gap on off-balance sheet transactions; information on off-balance sheet transactions is currently not used in empirical research, and is not readily available.
- A key step is to properly account for pledged collateral transactions and other transactions such as wealth management accounts, or other areas that generally fall under the rubric of “shadow banking.”
- The BIS’s Quantitative Impact Study and related work bridge international accounting standards such as GAAP and IFRS, to conform to a standard template for leverage.
- The paper suggests such international fora also request information on the extent of pledged collateral (and other off-balance sheet transactions) that are not accounted for on the balance sheet.
  - This information can augment the leverage data at a national level with an additional “cross-border leverage” metric.

*Source: wp1862 - Box 4. Econometric Implications of Using a Broader Leverage Metric*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp1862.pdf_