Re: [Quantlib-users] Inconsistency in ZSpread calculation?

[jian Xu <jia...@gm...>]

Aha, Thanks for pointing that out, Peter.  After I pass the same
compounding and bond coupon frequency to the constructor of the
ZeroSpreadedTermStructure, the clean price comes back to be the same
as the new_clean_prce.

>>> ts_spreaded1 = ql.ZeroSpreadedTermStructure(ts_handle, spread_handle1, ql.Compounded, fixed_rate_bond.frequency())
>>> ql.BondFunctions.cleanPrice(fixed_rate_bond, ts_spreaded1, settle_date)

101.00000001031341

As a side note, I think a better practice is to always specify
compounding and frequency, no matter constructing the baseline
"ts_yield" or the spread curve "ts_spreaded1".  I don't understand why
there isn't inspectors like compounding()/frequency() for the
FlatForward and the ZeroSpreadedTermStructure class.

On Sun, Nov 22, 2020 at 12:37 PM Peter Caspers <pca...@gm...> wrote:
>
> Hi Jian, this might be due to
>
> - different compounding conventions for the spread in the z-spread
> calculation (compounded in your code) vs. the zero-spreaded term
> structure (defaults to continuous, you can change it but I am not sure
> if this is this possible in Python)
>
> - different day count conventions for the z-spread (given by day_count
> in your code) and for the spread in the zero-spreaded term structure,
> which is always the day counter of the underlying rate curve (your
> ts_handle) - again you can overwrite this in the zero-spreaded term
> structure constructor, but this is unfortunately ignored (a known bug,
> or an open to do as we often like to call it)
>
> Best regards,
> Peter
>
> On Sun, 22 Nov 2020 at 19:00, jian Xu <jia...@gm...> wrote:
> >
> > Hi,
> > I encountered some inconsistency in ZSpread calculation.  Basically, I
> > calculate the zspread using the BondFunctions.zSpread.  But when I add
> > this zspread on top of the yield curve, I cannot recover the clean
> > price.  Following is what I did:
> >
> > I used the official notebook example at:
> > http://gouthamanbalaraman.com/blog/bonds-with-spreads-quantlib-python.html,
> > everything the same up to the part which sets the bond_engine.
> >
> > Then instead of using the engine, I used BondFunctions:
> > >>> settle_date = fixed_rate_bond.settlementDate(calc_date)
> > >>> ql.BondFunctions.cleanPrice(fixed_rate_bond, ts_yield, settle_date)
> >
> > This returns the clean price of 114.09294985282332, which translates
> > to the same NPV =114.18461651948999 on the official site (by adding
> > the accrued interest).  So far so good.
> >
> > Then I ask if the clean price is 101, what's the zspread:
> > >>> new_clean_price = 101
> > >>> zspread = ql.BondFunctions.zSpread(fixed_rate_bond, new_clean_price, ts_yield, day_count, ql.Compounded, fixed_rate_bond.frequency(), settle_date)
> > print(zspread)
> >
> > This shows the zspread=0.02592546532075354.
> >
> > Then I add this zspread on top of the original yield curve to obtain a
> > new yield curve.  My understanding is that by using this new yield
> > curve to price the bond, I should recover the new_clean_price.  But it
> > does not:
> >
> > >>> spread1 = ql.SimpleQuote(zspread)
> > >>> spread_handle1 = ql.QuoteHandle(spread1)
> > >>> ts_spreaded1 = ql.ZeroSpreadedTermStructure(ts_handle, spread_handle1)
> > >>> ql.BondFunctions.cleanPrice(fixed_rate_bond, ts_spreaded1, settle_date)
> >
> > This resulted clean price is 100.91139815763174, which is about 0.0886
> > different from the new_clean_price.  I tried to increase the
> > "accuracy" parameter in the ql.BondFunctions.zSpread, but doesn't seem
> > to help.
> >
> > Does anyone know why?  Did I misunderstood ZSpread?  Or did I pass
> > wrong parameters in the zSpread function call?  Thanks.
> >
> > Jian
> >
> >
> > _______________________________________________
> > QuantLib-users mailing list
> > Qua...@li...
> > https://lists.sourceforge.net/lists/listinfo/quantlib-users