## _wp14239

## Source details

**Canonical URL:** [_wp14239](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14239.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14239.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14239.pdf.json)

---

### Research objective and scope
- Study whether a common assumption in models used to estimate default probabilities from CDS spreads is appropriate when tested against post-default data.
- Compare risk-neutral default probabilities calculated from CDS spreads using standard valuation models with default probabilities calculated using recovery rates established by final bond prices at CDS settlement auctions.
- Analyze whether the cost of credit risk protection quoted by the CDS market prior to default reflected appropriately the probability of default as estimated using post-default data.
- Investigate determinants of the “excess spread”: the difference between the cost of protection CDS buyers should have paid given final bond prices and the actual CDS cost charged in the market.
- Briefly discuss auction participants’ payoffs and possible strategic behaviors affecting auction outcomes.

### Key findings
- Adjusting implied default probability estimations for actual recovery rates yields estimates that are significantly different than those estimated using standard assumptions.
- If quoted CDS spreads are computed using standard expected values on recoveries, market participants may be paying too little or too much for default protection compared with probabilities estimated using actual recovery values received at auctions.
- Observed “excess spreads” across the cross-section of default experiences where a CDS auction was held.
- Large mismatches between demand and supply of bonds at CDS settlement auctions may bias final recovery prices of the bonds and thus generate excess spreads.

### Methodology highlights
- Use of market CDS spread data and standard valuation models to compute risk-neutral implied default probabilities.
- Use of post-default auction-determined bond prices to compute realized recovery rates and corresponding ex-post default probabilities.
- Econometric analysis to study determinants of excess spreads (Tables 1–4: baseline regressions, additional controls, five-year ahead implied probabilities, excess spreads at different dates).
- Presentation of recovery outcomes and relationships between CDS spreads and recoveries (Figures 1–4).

### Suggested analytical focus for practitioners and policymakers
- When using CDS-implied default probabilities for decision-making, explicitly account for sensitivity to assumed recovery rates and consider scenario analysis with actual or auction-based recoveries.
- Monitor CDS settlement auction conditions—particularly potential mismatches between demand and supply of bonds—that may bias recovery prices.
- Consider further analysis of auction design and participant incentives to mitigate strategic behaviors that could distort recovery outcomes and pricing signals.

### Conclusions — Summary of paper and purpose
- Examines CDS settlement auctions (ISDA/ Creditex) and the relationship between market-implied default probabilities (from CDS spreads) and ex-post recovery values established at auctions.
- Uses auction data between 2005 and June 2014 to assess whether standard CDS pricing assumptions (notably a recovery rate of 0.4) led to biased market signals about default probabilities.

### Data and descriptive statistics
- Auctions with a credit event: 116 reference entities; 156 final auction results.
- Final recovery-price observations: 93 unsecured notes (senior and subordinated bonds) and 24 first lien loan obligations.
- Event-of-default breakdown: 73 bankruptcies, 35 failure-to-pay, 32 restructurings.
- Average recovery prices (expressed in percent of par): senior bonds: 42.74; loans: 51.47.
- Common market recovery assumption used in analysis: 0.4.
- Moody’s long-term average loss rate on senior unsecured debt: around 60 percent.
- Net Open Interest (NOI) is highly correlated with total volume of physical settlement requests: correlation = 0.99.
- Sample period for auctions and ISDA/ Creditex data: 2005–June 2014.

### Key empirical methods
- CDS-pricing framework: discrete-time reduced-form valuation using parity between protection payments (spread) and expected payout (Equations (1)–(3)), with discrete-time default probabilities and recovery rate R.
- Two approaches to infer implied default probabilities:
  - Bootstrapped (market term-structure) implied probabilities using standard practice (one-, three-, five-year maturities).
  - Parametric reduced-form (Nelson–Siegel-type functional form, Equation (4)) estimating parameters α and β.
- Main comparison: market-implied default probabilities computed assuming standard recovery (0.4) versus probabilities implied using actual auction recovery rates (RA).
- Defined excess spread ΔS as the relative difference between the protection payments that would prevail using RA and the actual market-quoted spread prior to the credit event.

### Main empirical findings (selected)
- Market signals (CDS spreads) often under- or over-estimated ex-post recovery outcomes; many cases where very low auction recoveries were associated with very low pre-default CDS spreads (market “surprises”).
- The distribution of recovery rates displays significant skewness and kurtosis; not close to normal.
- NOIindex (NOI/(bid+ask)) is strongly and negatively associated with the excess spread:
  - Baseline regression coefficient examples: NOIindex coefficients reported include -0.410***, -0.352***, -0.354***, -0.333***, -0.195**, -0.300*** across specifications (Table 1).
  - Interpretation provided: an increase in the size of the NOIindex by 1 would reduce the size of the excess spread by around 30 percent (qualitative example in text).
- Market volatility (VIX) at the time of the credit event is positively associated with the excess spread:
  - Example coefficient: vix 0.0178** (see Tables 1–4).
- Controls for CDS-market liquidity (number of CDS contracts outstanding; NetNotional share) and macro variables (3-month Libor; 5-year zero coupon rates) show limited or no systematic influence on the excess spread in most specifications.
- Results are robust across maturities (one-, three-, five-year CDS) and across estimation methods (reduced-form vs. bootstrapped probabilities).

### Interpretation and mechanism
- Auction structure and participant incentives generate “one-sided” auctions:
  - CDS sellers submit buy orders when they expect final liquidation value to exceed the auction price; otherwise they prefer cash settlement.
  - CDS buyers submit sell orders when they can source bonds or expect auction price below final payout or their sourcing cost.
  - Auctions often attract a minority of participants with divergent views, generating large NOIs and biased final prices in favor of the minority side.
- Negative relationship between NOIindex and excess spread implies auctions with significant sell interest (negative NOI) produce higher excess spreads (final bond prices below market expectations), while auctions with buy interest produce lower (or negative) excess spreads (final bond prices above expectations).
- Illiquidity at auctions (large mismatch of demand and supply) can bias final recovery prices, consistent with theoretical predictions (Chernov et al.; Du and Zhu) and prior empirical evidence (Gupta and Sundaram; Sundaram and Gupta).

### Policy implications and recommendations for future research
- Systematic bias between ex-ante market-implied spreads (using standard recovery assumptions) and ex-post auction recoveries suggests potential mispricing of CDS protection when standard recovery assumptions are used without accounting for auction dynamics.
- Potential policy considerations:
  - Re-assess auction design and settlement protocols to mitigate one-sidedness and participation constraints.
  - Consider transparency or incentives that broaden participation on both sides of auctions to reduce susceptibility to minority-driven price bias.
- Suggested avenues for future research:
  - Develop a full structural model of auction behavior incorporating agents’ expectations about final payout values (extending Chernov et al., Du–Zhu).
  - Use dealer-level actual CDS transaction spreads and valuations (rather than publicly quoted spreads) to check robustness and further understand dealer incentives and market microstructure effects on settlement outcomes.

### Appendix I — Definitions of Variables (selected)
- y1cdsce: 1-year maturity CDS spread observed on the day of the credit event
- y1cds1y: 1-year maturity CDS spread observed 1 year before the credit event
- NOI: Net Open Interest: bid - ask
- NOIindex: Normalized index of NOI : NOI/(bid+ask)
- VIX: Volatility index of the US stock market on the day of the credit event
- VIX _1y: Volatility index of the US stock market 1 year before the credit event
- default year: Number of defaulted firms in a year when the auction is held
- default prob. 5 year ahead: Estimated probability of default in five years (see footnote 34)
- 3 month libor: 3-month libor observed on the day of the credit event
- NCDS contract: Number of outstanding CDS contracts in the week of the credit event
- NetNotional Index: Outstanding amount of net-notional values divided by gross-notional values in the week of the credit event
- Zero coupon 5 year rate: “Risk-free” interest rate on the 5-year maturity US Treasury Bill on the day of the credit event
- Avg. Pr. Bnds: Average price of quoted bonds on the day of the credit event
- Pr. 1y Bnds: Price of 1 year residual maturity bonds on the day of the credit event
- CDS region: Dummy variable, =1 if the region is associated with the reference entity
- Auction Year: Dummy variable, =1 if auction is held that year
- Industry: Dummy variable, =1 if the industry is associated with the reference entity

### Appendix II — Summary Statistics and Correlations (selected)
- Summary statistics (selected rows as presented)
  - 90 -0.030 0.119 0.014 -3.875 20.667
  - 90 -0.028 0.099 0.010 -3.047 14.332
  - 90 -0.028 0.093 0.009 -2.586 11.112
  - NOI 156 -362.122 212.948 9710.0 -8.224 72.027
  - NOIindex "142 -0.198 0.714 0.510 0.530 1.889
  - 3 month libor 160 1.148 1.307 1.707 1.794 5.199
  - VIX 160 28.093 13.058 170.518 0.823 2.953
  - No CDS contracts (DTCC) 79 1873.214 66.221 4976.01 1.567 5.721
  - NetNotional Index 79 0.083 0.037 0.001 1.545 6.442

- Selected correlations (matrix excerpts)
  - Physical sttlm req. correlations:
    - with NOI: 1
    - with NOIindex: -0.1117
    - with NetNotional Index: -0.1469
    - with Net Notional: 0.1159
    - with Gross Notional: 0.1569
    - with No CDS contracts (DTCC): 0.2494
    - with VIX: -0.0516
  - NOI correlations:
    - with Physical sttlm req.: -0.99761
    - with NOIindex: 0.13191
    - with NetNotional Index: 0.1429
    - with Net Notional: -0.0859
    - with Gross Notional: -0.1309
    - with No CDS contracts (DTCC): -0.2305
    - with VIX: 0.0495
  - NOIindex correlations:
    - with Physical sttlm req.: -0.11170
    - with NOI: -0.11170
    - with NetNotional Index: -0.17811
    - with Net Notional: -0.14110
    - with Gross Notional: -0.0989
    - with No CDS contracts (DTCC): -0.1261
    - with VIX: -0.2629
  - NetNotional Index correlations:
    - with Physical sttlm req.: -0.14690
    - with NOI: 0.1429
    - with NOIindex: -0.17811
    - with Net Notional: 0.01161
    - with Gross Notional: -0.2826
    - with No CDS contracts (DTCC): -0.3032
    - with VIX: 0.190
  - Net Notional correlations:
    - with Gross Notional: 0.8672
    - with No CDS contracts (DTCC): 0.8833
    - with VIX: 0.1096
  - Gross Notional correlations:
    - with No CDS contracts (DTCC): 0.8921
    - with VIX: 0.0096
  - No CDS contracts (DTCC) correlations:
    - with VIX: 0.0411

*Source: _wp14239 - References (PDF).*

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

### _wp14239 - References

### Content inventory
- Tables:
  - 1. Determinants of Excess Spreads 1 - Baseline Regressions .................................................18
  - 2. Determinants of Excess Spreads 2 - Additional Controls ....................................................19
  - 3. Estimations Using Five-year Ahead Implied Probabilities of Default ...............................19
  - 4. Estimations Using Excess Spreads at Different Dates ........................................................21
- Figures:
  - 1. Recovery Rates at CDS Settlement Auctions ........................................................................8
  - 2. CDS Spread and Recovery Rates ...........................................................................................9
  - 3. Implied Default Probabilities One-year Ahead ....................................................................13
  - 4. Implied Default Probabilities Estimated Using Average and Actual Recoveries ................13
- Appendices:
  - I. Definitions of Variables........................................................................................................28
  - II. Summary Statistics and Selected Correlations ....................................................................29

### Research objective and scope
- Study whether a common assumption in models used to estimate default probabilities from CDS spreads is appropriate when tested against post-default data.
- Compare risk-neutral default probabilities calculated from CDS spreads using standard valuation models with default probabilities calculated using recovery rates established by final bond prices at CDS settlement auctions.
- Analyze whether the cost of credit risk protection quoted by the CDS market prior to default reflected appropriately the probability of default as estimated using post-default data.
- Investigate determinants of the “excess spread”: the difference between the cost of protection CDS buyers should have paid given final bond prices and the actual CDS cost charged in the market.
- Briefly discuss auction participants’ payoffs and possible strategic behaviors affecting auction outcomes.

### Key findings
- Adjusting implied default probability estimations for actual recovery rates yields estimates that are significantly different than those estimated using standard assumptions.
- Implication: if quoted CDS spreads are computed using standard expected values on recoveries, market participants may be paying too little or too much for default protection compared with probabilities estimated using actual recovery values received at auctions.
- Observed “excess spreads” across the cross-section of default experiences where a CDS auction was held.
- Conclusion: large mismatches between demand and supply of bonds at CDS settlement auctions may bias final recovery prices of the bonds and thus generate excess spreads.

### Methodology highlights
- Use of market CDS spread data and standard valuation models to compute risk-neutral implied default probabilities.
- Use of post-default auction-determined bond prices to compute realized recovery rates and corresponding ex-post default probabilities.
- Econometric analysis to study determinants of excess spreads (see Tables 1–4 for baseline regressions, additional controls, five-year ahead implied probabilities, and excess spreads at different dates).
- Presentation of recovery outcomes and relationships between CDS spreads and recoveries (see Figures 1–4).

### Interpretations and implications
- Market quotes of CDS spreads are widely used to signal default risk and extract market-implied default probabilities, liquidity conditions, and systemic risk.
- Mis-specification of recovery assumptions in standard models can materially change implied default probabilities and therefore affect valuations and hedging decisions.
- Settlement auction mechanics and demand–supply imbalances can influence realized recoveries and lead to systematic divergences (excess spreads) between pre-default market pricing and ex-post outcomes.
- Strategic behavior and payoffs of auction participants are relevant to understanding auction outcomes and potential biases in recovery rates.

### Suggested analytical focus for practitioners and policymakers
- When using CDS-implied default probabilities for decision-making, explicitly account for sensitivity to assumed recovery rates and consider scenario analysis with actual or auction-based recoveries.
- Monitor CDS settlement auction conditions—particularly potential mismatches between demand and supply of bonds—that may bias recovery prices.
- Consider further analysis of auction design and participant incentives to mitigate strategic behaviors that could distort recovery outcomes and pricing signals.

*Source: _wp14239 - References*

### conclusions.

### Conclusions

### Summary of paper and purpose
- Examines CDS settlement auctions (ISDA/ Creditex) and the relationship between market-implied default probabilities (from CDS spreads) and ex-post recovery values established at auctions.
- Uses auction data between 2005 and June 2014 to assess whether standard CDS pricing assumptions (notably a recovery rate of 0.4) led to biased market signals about default probabilities.

### Data and descriptive statistics
- Auctions with a credit event: 116 reference entities; 156 final auction results.
- Final recovery-price observations: 93 unsecured notes (senior and subordinated bonds) and 24 first lien loan obligations.
- Event-of-default breakdown: 73 bankruptcies, 35 failure-to-pay, 32 restructurings.
- Average recovery prices (expressed in percent of par): senior bonds: 42.74; loans: 51.47.
- Common market recovery assumption used in analysis: 0.4.
- Moody’s long-term average loss rate on senior unsecured debt: around 60 percent.
- Net Open Interest (NOI) is highly correlated with total volume of physical settlement requests: correlation = 0.99.
- Sample period for auctions and ISDA/ Creditex data: 2005–June 2014.

### Key empirical methods
- CDS-pricing framework: discrete-time reduced-form valuation using parity between protection payments (spread) and expected payout (Equations (1)–(3)), with discrete-time default probabilities and recovery rate R.
- Two approaches to infer implied default probabilities:
  - Bootstrapped (market term-structure) implied probabilities using standard practice (one-, three-, five-year maturities).
  - Parametric reduced-form (Nelson–Siegel-type functional form, Equation (4)) estimating parameters α and β.
- Main comparison: market-implied default probabilities computed assuming standard recovery (0.4) versus probabilities implied using actual auction recovery rates (RA).
- Defined excess spread ΔS as the relative difference between the protection payments that would prevail using RA and the actual market-quoted spread prior to the credit event.

### Main empirical findings
- Market signals (CDS spreads) often under- or over-estimated ex-post recovery outcomes; many cases where very low auction recoveries were associated with very low pre-default CDS spreads (market “surprises”).
- The distribution of recovery rates displays significant skewness and kurtosis; not close to normal, implying systematic differences between protection-leg valuation and payment-leg valuation.
- The NOIindex (NOI as a share of total bid+offer, constrained between -1 and +1) is strongly and negatively associated with the excess spread:
  - Baseline regression coefficient examples: NOIindex coefficients reported include -0.410***, -0.352***, -0.354***, -0.333***, -0.195**, -0.300*** across specifications (Table 1).
  - Interpretation: an increase in the size of the NOIindex by 1 would reduce the size of the excess spread by around 30 percent (qualitative example given in text).
- Market volatility (VIX) at the time of the credit event is positively associated with the excess spread (coefficients reported, e.g., vix 0.0178**; see Tables 1–4), meaning higher volatility is associated with lower auction recoveries relative to expectations.
- Controls for CDS-market liquidity (number of CDS contracts outstanding; NetNotional share) and macro variables (3-month Libor; 5-year zero coupon rates) show limited or no systematic influence on the excess spread in most specifications.
- Results are robust across maturities (one-, three-, five-year CDS) and across estimation methods (reduced-form vs. bootstrapped probabilities).

### Interpretation and mechanism
- Auction structure and participant incentives generate “one-sided” auctions:
  - CDS sellers participate in auctions (submit buy orders) only when they expect final liquidation value (post-auction payout) to exceed the auction price; otherwise they prefer cash settlement.
  - CDS buyers participate (submit sell orders) when they can source bonds or expect auction price below final payout or their sourcing cost.
  - As a result, auctions often attract a minority of participants with divergent views, generating large NOIs and biased final prices in favor of the minority side.
- The observed negative relationship between NOIindex and excess spread implies that auctions with significant sell interest (negative NOI) produce higher excess spreads (final bond prices below market expectations), while auctions with buy interest produce lower (or negative) excess spreads (final bond prices above expectations).
- Illiquidity at auctions (large mismatch of demand and supply) can bias final recovery prices, consistent with theoretical predictions (Chernov et al.; Du and Zhu) and prior empirical evidence (Gupta and Sundaram; Sundaram and Gupta).

### Policy implications and recommendations for future research
- The systematic bias between ex-ante market-implied spreads (using standard recovery assumptions) and ex-post auction recoveries suggests potential mispricing of CDS protection when standard recovery assumptions are used without accounting for auction dynamics.
- Potential policy considerations:
  - Re-assess auction design and settlement protocols to mitigate one-sidedness and participation constraints (the paper suggests the auction mechanism might need re-thinking if confirmed).
  - Consider transparency or incentives that broaden participation on both sides of auctions to reduce susceptibility to minority-driven price bias.
- Suggested avenues for future research:
  - Develop a full structural model of auction behavior incorporating agents’ expectations about final payout values (extending Chernov et al., Du–Zhu).
  - Use dealer-level actual CDS transaction spreads and valuations (rather than publicly quoted spreads) to check robustness and further understand dealer incentives and market microstructure effects on settlement outcomes.

*Source: _wp14239 - conclusions.*

### References

### _wp14239 - References

### References
- Acharya, V.V., S.T. Bharath, and A. Srinivasan (2007) “Does Industry-Wide Distress Affect Defaulted Firms? Evidence from Creditor Recoveries” Journal of Financial Economics, 85, pp:787-821.
- Bloomberg (2009) “The Bloomberg CDS model”. Technical Report, Bloomberg.
- Díaz, A., J. Groba and P. Serrano (2013). "What drives corporate default risk premia? Evidence from the CDS market." Journal of international money and finance, 37, pp. 529-563.
- Chan-Lau, J.A. (2006) “Market-Based Estimation of Default Probabilities and Its Application to Financial Market Surveillance” IMF WP/06/104.
- Chernov, M. A.S, Gorbenko, and I. Makarov (2013) “CDS Auctions” Review of Financial Studies, 26 (3), 768-805.
- Creditex and Markit (2010) “Credit Event Auction Primer”. Technical Report, Creditex and Markit.
- Deliannedis, G. and R. Lagnado (2002) “Recovery Assumptions in the Valuation of Credit Derivatives” The Journal of Fixed Income; March 2002, pp 20-30.
- Duffie, D. (1999) “Credit Swap Valuation” Financial Analysts Journal, 52 (1), pp. 73-85.
- Duffie, D. and K. Singleton (1999) “Modeling the Term Structure of Defaultable Bonds” Review of Financial Studies, Vol 12, N.4; pp-687-720.
- Du, S. and H. Zhu (2013) “Are CDS Auctions Biased and Inefficient?” memo, Graduate School of Business, Stanford University.
- Gupta, S. and R.K. Sundaram (2012) “CDS Auctions and in Informative Bias in CDS Recovery Rates”, mimeo New York University.
- Helwege, J; Maurer, S.; Sarkar, A.; and Y. Wang (2009) “Credit Default Swap Auctions” Staff Report, Federal Reserve Bank of New York. N. 372.
- Hull J. and A. White (2000) “Valuing Credit Default Swaps I: No Counterparty Default Risk” The Journal of Derivatives, 8;pp. 29-49.
- Longstaff, F. A., S. Mithal, and E. Neis (2005) “Corporate Yield Spreads: Default Risk or Liquidity? New Evidence from the Credit Default Swap Market”. The Journal of Finance, 60; pp:2213–2253.
- Jarrow, R. and S. Turnbull (1995) “Pricing Options on Financial Securities Subject to Default Risk” Journal of Finance, 50; pp.53-86.
- International Swaps and Derivatives Association (2009) “2009 ISDA Credit Derivatives Determination Committee and Auction Settlement CDS Protocol” Technical Report, International Swaps and Derivatives Association.
- Merrick, J. Jr. (2006) “Evaluating Pricing Signals from the Bond Markets” in Managing Economic Volatility and Crisis: A Practitioner Guide” J. Aizenman and B. Pinto eds. Cambridge University Press.
- Moody’s Investor Services (2014) “Annual Default Study: Corporate Default and Recovery Rates, 1920-2013. February 28, 2014.
- O’Kane, D. and S. Turnbull (2003) “ Valuation of Credit Default Swaps” Fixed Income Quantitative Credit Research, Lehman Brothers.
- Shaefer, T. and M. Uhrig-Homburg (2014) “Is recovery risk priced?” Journal of Banking and Finance, 40; pp:257-270.
- Sundaram, Rangarajan K. (2013) “CDS Auctions and Recovery Rates: An Appraisal” mimeo, New York University.
- Vrugt, Evert (2011) “Estimating Implied Default Probabilities and Recovery Values from Sovereign Bond Prices” Journal of Fixed Income. Fall2011, Vol. 21 Issue 2, pp:5-14.
- Zhan, Frank Xiaoling (2008) “Market Expectations and Default Risk Premium in Credit Default Swaps: A Study of the Argentine Default” The Journal of Fixed Income; Vol. 18 Issue 1, pp:37-55.

### Appendix I — Definitions of Variables
- y1cdsce: 1-year maturity CDS spread observed on the day of the credit event
- y1cds1y: 1-year maturity CDS spread observed 1 year before the credit event
- NOI: Net Open Interest: bid - ask
- NOIindex: Normalized index of NOI : NOI/(bid+ask)
- VIX: Volatility index of the US stock market on the day of the credit event
- VIX _1y: Volatility index of the US stock market 1 year before the credit event
- default year: Number of defaulted firms in a year when the auction is held
- default prob. 5 year ahead: Estimated probability of default in five years (see footnote 34)
- 3 month libor: 3-month libor observed on the day of the credit event
- NCDS contract: Number of outstanding CDS contracts in the week of the credit event
- NetNotional Index: Outstanding amount of net-notional values divided by gross-notional values in the week of the credit event
- Zero coupon 5 year rate: “Risk-free” interest rate on the 5-year maturity US Treasury Bill on the day of the credit event
- Avg. Pr. Bnds: Average price of quoted bonds on the day of the credit event
- Pr. 1y Bnds: Price of 1 year residual maturity bonds on the day of the credit event
- CDS region: Dummy variable, =1 if the region is associated with the reference entity
- Auction Year: Dummy variable, =1 if auction is held that year
- Industry: Dummy variable, =1 if the industry is associated with the reference entity

### Appendix II — Summary Statistics and Correlations

- Table 1. Summary Statistics (selected rows as presented)
  - Obs Mean Std. Dev. Va r i a n c e S k e w n e s s Kurtosis
  - 90 -0.030 0.119 0.014 -3.875 20.667
  - 90 -0.028 0.099 0.010 -3.047 14.332
  - 90 -0.028 0.093 0.009 -2.586 11.112
  - NOI 156 -362.122 212.948 9710.0 -8.224 72.027
  - NOIindex "142 -0.198 0.714 0.510 0.530 1.889
  - 3 month libor 160 1.148 1.307 1.707 1.794 5.199
  - VIX 160 28.093 13.058 170.518 0.823 2.953
  - No CDS contracts (DTCC) 79 1873.214 66.221 4976.01 1.567 5.721
  - NetNotional Index 79 0.083 0.037 0.001 1.545 6.442

- Table 2: Selected Correlations (matrix excerpts as presented)
  - Variables listed: Physical sttlm req., NOI, NOIindex, NetNotional Indx, Net Notional, Gross Notional, No CDS contracts, VIX
  - Physical sttlm req. correlations:
    - with NOI: 1
    - with NOIindex: -0.1117
    - with NetNotional Index: -0.1469
    - with Net Notional: 0.1159
    - with Gross Notional: 0.1569
    - with No CDS contracts (DTCC): 0.2494
    - with VIX: -0.0516
  - NOI correlations:
    - with Physical sttlm req.: -0.99761 (table alignment note)
    - with NOIindex: 0.13191
    - with NetNotional Index: 0.1429
    - with Net Notional: -0.0859
    - with Gross Notional: -0.1309
    - with No CDS contracts (DTCC): -0.2305
    - with VIX: 0.0495
  - NOIindex correlations:
    - with Physical sttlm req.: -0.11170
    - with NOI: -0.11170 (table alignment note)
    - with NetNotional Index: -0.17811
    - with Net Notional: -0.14110
    - with Gross Notional: -0.0989
    - with No CDS contracts (DTCC): -0.1261
    - with VIX: -0.2629
  - NetNotional Index correlations:
    - with Physical sttlm req.: -0.14690
    - with NOI: 0.1429
    - with NOIindex: -0.17811
    - with Net Notional: 0.01161
    - with Gross Notional: -0.2826
    - with No CDS contracts (DTCC): -0.3032
    - with VIX: 0.190
  - Net Notional correlations:
    - with Gross Notional: 0.8672
    - with No CDS contracts (DTCC): 0.8833
    - with VIX: 0.1096
  - Gross Notional correlations:
    - with No CDS contracts (DTCC): 0.8921
    - with VIX: 0.0096
  - No CDS contracts (DTCC) correlations:
    - with VIX: 0.0411

*Source: _wp14239 - References (PDF).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14239.pdf_