Choosing CDS Spread Quotes for Historical Regression Analysis
Summary
The document compares Markit’s par spread, conventional spread, and upfront payment as inputs for analyzing relationships between sovereign and banking credit default swaps. It argues that par spreads persist mainly for compatibility with older conventions and that conventional spread and upfront payment encode equivalent pricing information when the running spread, interest rate curve, and recovery assumption are held consistent.
For historical observations spanning the 2009 ISDA Big Bang, some names may have par spreads where conventional quotes are unavailable. The proposed approach is to convert those observations to conventional spreads using the earliest available running spread and recovery assumption for the name, along with the historical interest rate curve. This takes more work than joining par and conventional spread series directly, but is presented as more accurate. The advice is specific to quote conversion and the assumptions needed for it; the answer also cautions that financial institution CDS data can be unusual, so regression findings should be interpreted carefully.
Key ideas
- Par spreads may appear in legacy CDS data, while conventional spreads reflect later quoting standards.
- Conventional spread and upfront payment can be converted into one another given consistent running spread, curve, and recovery assumptions.
- Pre-2009 par spread observations can be converted using name-specific assumptions and historical interest rate curves.
- Combining unconverted legacy par spreads with later conventional spreads may reduce comparability in a time series.
- CDS market peculiarities can limit the reliability of conclusions drawn from financial institution data.
Tags
Full text
# Which spread to use to analyse CDS data from Markit # Which spread to use to analyse CDS data from Markit So I am currently doing some analysis on CDS Data and I am using Markit to extract this data. However, I am a bit confused regarding the quotation standards here. I want to investigate the relationship between Sovereign and Banking CDS using some standard regressions. Through Markit, I can get the Par Spread, the ConvSpread and the Upfront Payment. I am wondering which of these to use for my analysis as for example, some sovereign bonds do not exhibit any ConvSpread for the years 2005-2007 but for me the Par Spread wouldnt make too much sense. Or should I be using the Upfront Payment to analyse this relationship? Thanks ## Answer by Dimitri Vulis (score 2) https://quant.stackexchange.com/a/70903 The par spread is retained just for compatibility with legacy quoting conventions - pre April 2009 ISDA Big Bang. In my opinion, it's best ignored. Some people may disagree. it's close to the market standard quote spread anyway. The msq spread and the upfront really convey exactly the same information. If you use the same standard running spread (usually 100 bps, but sometimes 500 bps or some other running spread), and the same interest rate curve, and the same recovery assumption, then you can convert from either one to the other. A problem arises if you are looking at pre-Big Bang historical data, and some name has only par spreads, and no msq spread. My siggestion is to convert to msq spread using the earliest available running spread and recovery assumption for that name; and the the historical IR curve, although that is less material. It's more work than just concatenating time series of par spread before big bang and msq spread after, but a little more accurate. Both before the big bang and today, financial institutions CDS is a very weird market and I'm skeptical of any information extracted from it.
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.