Re: COUP_ bases.

["Andreas J. Guelzow" <aguelzow taliesin ca>]

Neil Booth wrote:

> Andreas J. Guelzow wrote:-
>
>
> > Currently the bases are implemented as follows. Likely the names used 
> > for them are not completely correct. bases 0 to 4 have to be like that 
> > to match what Excel appears to be doing:
>
> Yup.
>
>
> > -----------------------------------------------------------------
> > base 0: BASIS_30Ep360
> >
> > date adjustment:
> > to_date is changed from last of month (whether 28, 29, 30, or 31) to 30th
> > from_date is changed from last of month (whether 28, 29, 30, or 31) to 30th
> >
> > date difference:
> > each month 30 days, within a month serial delta
>
> No, both for name and description.  Called either "PSA 30/360" or,
> only by Excel, "NASD 30/360".  Try (28Feb01,28Feb01) to see why your
> description is wrong.  Answer is -2, not zero as your description
> would get.


Pardon? Considering that you haven't seen the code yet, you would neeed 
to trust me that that is in fact what my current code has to do to Match 
XL.In XL when you calculate COUPDAYBS ("28-Feb-2001","28-Feb-2002",1,0) 
you get: 0 _not_ -2.

Which name should we use? Remember _this_ _is_ the default!


> This convention is not used in *any* markets to the best of my
> knowledge.  Bloomberg doesn't implement it either, and they cover damn
> near everything (albeit with a fair number of bugs, too, but far less
> than MS 8^)).  Seems to be at least 5 years old, or a figment of MS's
> imagination.


If this is what XL does the only question is what is an appropriate name.


> > -----------------------------------------------------------------
> > base 1: BASIS_ACTACT
> >
> > date adjustment:
> > no change
> >
> > date difference:
> > serial delta
> >
> > currently serial delta also for length of coupon period but that
> > is not Excel comp.
>
> I don't grok what you're saying here.  There are 2 meanings of
> Act/Act, one for bonds, one for swaps.  For bonds its actual days /
> actual days (doh! 8^)).
>
> Used for accrued int and yields.


Well XL seems to calculate this simply as the difference of the date 
serial numbers, except when it tries to calculate COUPDAYS.


> > -----------------------------------------------------------------
> > base 2: BASIS_ACT360
> >
> > date adjustment:
> > no change
> >
> > date difference:
> > serial delta, even for times of 1 year between settlement and
> > coupon date
> >
> > 360/freq for length of coupon period
>
> I don't understand what you're saying here either.  Act/360 is as easy
> as could be: number of days / 360.  Used only for accrued int, not
> yields.  CoupDays seems to period-ize it, but that's a simple issue.


What do you expect COUPNCD ("28-Feb-2003","29-Feb-2004",1,2) or
COUPNCD ("28-Feb-2003","29-Feb-2004",1,2) to be in XL ?  Well neither is 
360! Therefore what do you mean with number of days / 360 ? Remember 
these distance functions usually return integers.



> > -----------------------------------------------------------------
> > base 3: BASIS_ACT365
> >
> > date adjustment:
> > no change
> >
> > date difference:
> > serial delta, even for times of 1 year between settlement and
> > coupon date
> >
> > 365/freq for length of coupon period, (with decimal answer)
>
> Number of days / 365.  Easy.  Used for accrued int and yields.
>
>
> > -----------------------------------------------------------------
> > base 4: BASIS_30E360
> >
> > date adjustment:
> > from_date is changed from 31st to 30th
> > to_date is changed from 31st to 30th
> >
> > date difference:
> > each month 30 days, within a month serial delta
>
> Yup.  Used for accrued int and yields.
>
>
> > base 5: BASIS_30_360
> >
> > date adjustment:
> > from_date is changed from 31st to 30th
> > if from_date is 30 or 31, to_date is changed from 31st to 30thdate 
> > difference:
>
> Yup.  Used for accrued int and yields.
>
> Then there's basis 6:
>
> The real meaning of 30E+360 (30Ep360 as an identifier, say):
>
> d1:  31->30 (as for all 30/360 conventions)
> d2:  31->1, m2++ if d1 >= 30.
>
> This is used for calculating yields according to the "EuroBond"
> convention.  It's not used for accrued interest.  FWIW.
>
> Neil.

Andreas




--
Prof. Dr. Andreas J. Guelzow
http://www.math.concordia.ab.ca/aguelzow