2
$\begingroup$

I'm new to this kind of matters hence probably this is a stupid question. I would like to build a return time series for backtesting purposes and I was wondering how to handle pv changes when contracts roll, in particular in case of vanilla IMM swaps and vanilla CDSs.

In case of a regular 5Y swap, I would roll it every day and with the curves built at day $t$ I would price the swap I sold in $t-1$ as of $t$, and get the PV difference. It is not clear to me if for standardized contracts such a 5Y-IMM swaps or, even more standardized, a 5Y-CDS, this method can still work.

If someone more experienced on this could share any feedback it would be deeply appreciated.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

Yes, it can get a bit confusing.

The CDS quotes for maturities, say, March 2030, September 2030, March 2031, don't jump.

But the meaning of the "on the run" tenors changes on a roll date. E.g., one day, March 2030 is the on the run 4.5 years, September 2030 is on the run 5 years, and March 2031 is on the run 5.5 years. But the next day, March 2030 is now the on the run 4.0 years, September 2030 is on the run 4.5 years, and March 2031 is on the run 5.0 years. (Moreover tenors other than 5Y might be not liquid and stop being quoted soon after roll date.)

So, a time series of "on the run 5 year" CDS quotes, will have a jump whenever it begins to refer to a different CDS maturity, i.e. to September 2030 up to the roll date, and to March 2031 afterwards.

Exchange traded futures might be a helpful analogy. Suppose you look at a series of "next out" futures quotes. Whenever the "near" one expires, the "next out" one becomes "near", and the following one becomes "next out", so you have a jump because "next out" again refers to a different IMM date.

Improve this answer
$\endgroup$
6
  • 1
    $\begingroup$ Thanks a lot for your reply Dimitri, and all your contributions in SE. I think I understood the point but just to fix the ideas: now, the on-the-run 5Y CDS expires on 20-06-2030. From September 20th, the on-the-run 5Y CDS will expire on 20-12-2030. So, in order to build my time series, I would: register daily PV differences holding the current 5Y on-the-run from today till September roll. After rolling, I should still be able to price the old one with credit curves built with the current liquid CDS tenors in principle, and get my PV difference. Is that right or completely off? Many thanks! $\endgroup$ Commented Apr 26 at 16:18
  • 1
    $\begingroup$ You're v welcome, my pleasure. If you have a CDS contract maturing June 20, 2030, then every day it's has 1 day less left to maturity. You mark to market using a CDS quote for this maturity date, which doesn't change. Sometimes this swap's maturity date happens to coincide with the on-the-run 5 year , later with on-the-run 4.5 year, later with on-the-run 4 year. Once the maturity date no longer coincide with the on-the-run 5 year , the bid-ask spread may get wider, and you may have difficulty obtaining the quote for mark to market, may need to interpolate from 5 years. $\endgroup$ Commented Apr 26 at 17:07
  • 1
    $\begingroup$ So, there's no change in the CDS's mark to market, no P&L just from the roll occurring. However: I've seen Taylor expansion CDS P&L attributions set up inconveniently, so that instead of multiplying the sensitivity to "June 2030" by the change in "June 2030" quote, they multiplying the sensitivity to "on the run 5Y" by the change in "on the run 5Y" quote. Whenever the meaning of "on the run 5Y" changes, bogus P&L appears in the attribution, so they introduce "P&L from the roll" to offset the bogus P&L, since there's no reason for the mark to market to change. $\endgroup$ Commented Apr 26 at 18:01
  • 1
    $\begingroup$ Thanks a lot Dimitri, that is very valuable info, what I said was a bit nonsense. I follow what you are saying but regarding the last part: isn't that what people want when building a PV time series for the 5Y CDS? From my ignorant stand point, I would want the series to reflect that the "on the run 5Y" changes meaning, so I would want the jump. You seems to hint, that instead people don't want this. I definitely trust more you than me, but I don't understand this point:) $\endgroup$ Commented Apr 27 at 7:31
  • 1
    $\begingroup$ Consider a daily time series of "shopping days left till Christmas". On most days, the change is due to 1 less day being left - referencing the same Christmas. However on some days, there is a jump because two consecutive days reference Christmases from different years. $\endgroup$ Commented Apr 27 at 12:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.