Using firstPeriodStartDate and IRD example 32

[scolebourne]

I am currently reviewing the example ird-ex32-zero-coupon-swap.xml in v5.8 and have found a question with firstPeriodStartDate.

In the fixed leg, the example has:
– effectiveDate = 2005-02-22
– terminationDate = 2035-02-22
– firstPeriodDate = 2005-02-20
– calculationPeriodFrequency = 1Y
– rollConvention = NONE

Since there is no “firstRegularPeriodStartDate” tag, our code assumed that there is no initial stub. It therefore assumed that “firstPeriodDate” indicates the start of the first regular period (with effectiveDate having no meaning for schedule generation in this case). However, since 2005-02-20 to 2035-02-22 does not divide evenly into 1 year periods there is a problem with our interpretation.

Is the example wrong? ie. is there a missing “firstRegularPeriodStartDate” tag or incorrect dates?

Or is there a two day stub implied here from 2005-02-20 to 2005-02-22?

Note that this post, http://www.fpml.org/dev/modules/newbbex/viewtopic.php?topic_id=80&forum=4#forumpost178, almost answers the question, but not quite.

Also, my understanding was that if the payment frequency does not equal the calculation frequency then compounding applies. If so, why is there no “compoundingMethod” tag in the fixed leg?

[roma]

Hi,

i believe this configuration means that first period is 20.02.2005-22.05.2005 (for float leg) and 20.02.2005-31.12.2005 (and then apply adj.) (for fix leg). No “firstRegularPeriodStartDate” = no initial stub.

Also, i think there is no “compoundingMethod” on fix, because calculation period = payment period = 1Y.

this can be useful – http://www.fpml.org/_wgmail/_im-custodianmail/msg00030.html

[scolebourne]

Thanks for replying.

For me, a float leg of 20.02.2005-22.05.2005 is a stub, as the dates do not both match the implied roll convention of 22. If your reply is correct, then I’d have to assume that the intended rule is to calculate the schedule of periods using the effective date first, then replace the effective date with the firstPeriodStartDate extending the first period as necessary. (I’d also have to assume that the fact that the roll convention is “NONE” rather than “22” is significant.)

I don’t see how to get 20.02.2005-31.12.2005 for the fixed leg. Nothing in the example indicates periods based on calendar years. IIUC, the yearly frequency is just that, periods of one year rolling just like periods of 3 months would. Thus, if we adopt the rule outlined above, I’d calculate 20.02.2005-22.02.2005 for the fixed leg.

On compounding, the example has calculation period of 1Y and payment period of 1T (term), hence my belief that a compounding method is needed. For a 30 year swap, accruing yearly but paying once at the end, compounding would be quite significant!

[roma]

About fix leg – you are absolutely right. It’s my mistake – i used example from 5.6 version which has some errors (wrong roll convention and payment freq.).

There is post about this – http://www.fpml.org/issues/view.php?id=1200

About compounding, i’m still think that there is no compounding on fixed leg (why not?), but i think best practice to use compoundingMethod = NONE in this cases.

[Author]

Posts