Re: COUPDAY* functions

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

Neil Booth wrote:

> Andreas J. Guelzow wrote:-
>
>
> > > I didn't read his message properly then.
> > >
> > > The implementations of Act/Act and 30/360 (US) are wrong.
> > Pardon me?
> >
> > COUPNCD is independent from the bases chosen.
>
> I know, after all, it was me who told _you_ that!  But have you read
> the subject of the message, or the content of the original?
>
> Now, I didn't download the file that was linked to, but the original
> message and the subject don't make it sound like the only problem is
> coupncd.  From glancing at the code, I think you've got ncd / pcd
> correct, which is great.


It wasn't quite correct (ie. it didn't match XL because they fell into a 
trap. Well, we can do that too. I committed appropriate change s a few 
hours ago.

>  If Excel has it wrong, that's really sad; I
> know you want to introduce the compatibility option.


They ar edefinitely wrong... because a 11/29/2001 maturity date should 
not have a quarterly coupon at 11/28/2000, I think


> If you want to be compatible, what exactly is the Excel implementation
> doing?  I'm curious, because it appears awfully broken from the
> example you posted; 


They simply backtrack by 3 or 6 months as appropriate. Whene they hit 
February 29th becomes 28th... they never go back to 29th.

> I can't imagine how you could get the answers it
> gives.  (After all, the obvious implementation is to take the day of
> the maturity date and shove it into a new create-serial-date function,
> with the appropriate month and year, and with the day adjusted for EOM
> if necessary).  I guess they must be counting days to get coupon dates
> or something daft like that.


It turns out that using serial dates makes things awfully complicated. 
That's why I decided to stay with GDate. couppcd and coupncd now mathces 
XL. (We will have a correct implementation with a fifth argument eom/oem.


> But anyway, my message was about the day count stuff like DAYS360 and
> COUPDAYBS and COUPDAYSNC; I _think_ the latter are broken for Act/Act
> and and 30/360.


THe basis=0 30/360 convention wasn't quite right, but I nearly got it 
now. (Febraury 28th sometimes becomes February 30th in the calscs)

 As always, I could be wrong - the more you find out
> about bond conventions, the less willing you are to ever be certain
> 8-) Some conventions are mind numbingly weird, and no two countries
> have the same conventions.
>
> Neil.


Why can't they use real dates???

(Don't answer that, it's rhetorical)

Andreas

PS: Getting close to finishing coup...





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