Estimating a Missing Long-Maturity Credit Spread from Sparse Data

Summary

The question asks how to estimate one bank’s missing 20-year credit spread using the other bank’s 20-year spread and the observed relationship between their 5- and 10-year spreads. The response first challenges the strength of the evidence: correlation cannot be assessed reliably from only two observations, and a spread difference cannot reasonably be characterized as white noise on that basis. It then suggests simple provisional estimates, such as using the other bank’s long-maturity spread or an average of the shorter-maturity spreads.

The exchange offers no calibrated model, uncertainty estimate, or validation against known 20-year values. Its suggestions are rough fallbacks, not a method justified by the shorter-tenor correlations, and averaging shorter maturities may fail to capture the credit curve’s shape. A defensible estimate would require clarifying the amount and quality of the available data and how the spreads are measured. The key lesson is that sparse observations do not support confident statistical inference, even when the reported shorter-term relationship appears strong.

Key ideas

• Two observations are insufficient to support a reliable correlation estimate or a white-noise claim.
• Using the other bank’s long-maturity spread or averaging shorter spreads are only rough fallback estimates.
• Shorter-tenor correlation does not establish that the banks’ 20-year spreads move together.
• The exchange provides no model validation or uncertainty range for its proposed estimates.

Tags

# Estimating a Missing 20-Year Credit Spread Using Correlated Data


# Estimating a Missing 20-Year Credit Spread Using Correlated Data












I have credit spreads for two banks:





The 5-year and 10-year spreads for both banks are highly correlated. I want to estimate Bank A’s 20-year spread using Bank B’s 20-year spread and the relationships at shorter maturities.

What I’ve Tried:

I modeled the differences in 5-year and 10-year spreads between the two banks, but the differences themselves appear to be mostly white noise.

Question:

How can I estimate Bank A’s 20-year spread using this data? Are there specific methods or models suited to this problem?

## Answer by MaybeMaybeNot (score 1)

https://quant.stackexchange.com/a/81605

I want to first make sure that I fully understand what you mean. I assume you have calculated their spread over the same benchmark. How did you estimate the correlation between their spreads? (Estimating the correlation with only two data points is not very reasonable). With only two datapoints it is not feasible to assume white noise regarding their spread difference.

The simplest solution I can come up with, by only having two sets of data, is to estimate the 20Y spread for bank A to be average of 5Y and 10Y spread or set it equal to the 20y spread for the other bank.

One rule that I have learned is "Keep it simple, keep it clean". It works most of the times.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.