Re: [Quantlib-users] Swap floating cash flows with negative fwd rates seem wrong

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

Hi Philippe,

The discounting price engine for the swap calculates the amount() of each Coupon using a class method called Cashflows::npvbps(….) that computes the absolute values for cashflows depending on their “type” (Ibor coupons, CPI coupons, etc….)

Once these are calculated, then they get multiplied by the swap::argument::payer[] flag which is set to -1 or 1 depending on the type payer or receiver. I suspect what you are calling is the amount() method on the coupons. The correct flag of the leg is handled after calculation of the absolute amount.

Hope it helps.
Amine

> On 13 Jul 2020, at 11:11, philippe hatstadt <pha...@ma...> wrote:
> 
> It’s the underlying swap of a swaption. US curve. Swap starts in 2y and ends in 12y. Semi annual fixed act/act versus USDLibor 3M floating a/360.
> The Hull White path is harder to share other than by providing daily discount factors.
> Defining dfs and dfe as the discount factors at 2y and 12y, I was able to get the correct fairRate manually by stating that floating leg NPV is equal to dfs - dfe. swap.fixedLegNPV() is correct but not swap.flioatingLeg.NPV().
> I used today as evaluation date.
> As you can see below, the floating leg cash flow amounts never change in sign even when the forward rate does. I realize my graph is the instantaneous forward, but the 3M can easily be calculated from the discount factors.
> 
> Regards
> 
> Philippe Hatstadt
> +1-203-252-0408
> pha...@ma...
> https://www.linkedin.com/in/philippe-hatstadt
> 
> 
>> On Jul 13, 2020, at 5:57 AM, Amine Ifri <ami...@gm...> wrote:
>> 
>> I use the C++ library and never had an issue with swap.fairRate(). It gives me positive as well as negative values depending on the term structures.
>> 
>> As far as checking the CPP code, it looks correct to me. Do you mind sharing the swap details you are trying to price/simulate?
>> 
>> Amine
>> 
>>> On 12 Jul 2020, at 03:02, Philippe Hatstadt <phi...@ex... <mailto:phi...@ex...>> wrote:
>>> 
>>> I'm using a Hull-White model to generate forward paths of the short rate, from which I then build a discounting curve, from which I compute forward swap values towards Monte-carlo integration of certain path dependent swap rate linked derivatives. Here, I am looking at a 2y into 10y swaption, and for each path, I want to calculate the forward swap rate and the swap NPV.
>>> My problem is that I have a curve that has negative rates on a certain path, the 2y into 10y forward swap.fairRate() function returns like 94bp. Since all rates are negative, it's pretty clear that fairRate() should be negative, and if I use the curve's discount factors and use the good old approximation for the swap rate with the formula C/2 * Sum_i(FixedLegDiscountFactor(i))/(1-DiscountFactor(TMat)) then I get the correct -250bp forward swap rate (where i=1 to 20 corresponding to 20 forward semi-annual fixed payments).
>>> So I am wondering if the swap cash flows do not work with negative rates or else, but as it stands right now, the swap.fairRate() function doesn't work at all. I also checked that by solving for the fixed rate that would make the swap net cash-flows to be equal to zero, then I get the value returned by swap.fairRate(). This would further indicate that the latter function is correct, but that the cash flows are not. lastly, whether I select a path with positive or negative forward rates, the cf.amount() is always positive.
>>> This is the code I use to generate the cash-flows:
>>> def swap_cash_flows(swap: ql.Swap,
>>>                     crvh,
>>>                     ):
>>>     output_list = []
>>>     for i, cf in enumerate(swap.leg(0)):
>>>         dt = cf.date()
>>>         dfact = crvh.discount(dt)
>>>         output_list.append(['leg0', cf.date(), dfact, cf.amount()])
>>>     for i, cf in enumerate(swap.leg(1)):
>>>         dt = cf.date()
>>>         dfact = crvh.discount(dt)
>>>         output_list.append(['leg1', cf.date(), dfact, cf.amount()])
>>>     df1 = pd.DataFrame(columns=['leg', 'cf_date', 'disc_fact', 'cf_amount'], data=output_list)
>>>     return df1
>>> The short-rate path corresponding to the curve handle crvh is as follows:
>>> <image.png>
>>> and lastly, this is the table of cash flows, dates and discount factors:
>>> leg	cf_date	disc_fact	cf_amount
>>> leg0	2023-01-11	1.031727084	0.004717496
>>> leg0	2023-07-11	1.044326706	0.004640581
>>> leg0	2024-01-11	1.056475309	0.004717496
>>> leg0	2024-07-11	1.067323978	0.004666219
>>> leg0	2025-01-13	1.080208917	0.004768773
>>> leg0	2025-07-11	1.095773345	0.004589303
>>> leg0	2026-01-12	1.107358267	0.004743135
>>> leg0	2026-07-13	1.119706713	0.004666219
>>> leg0	2027-01-11	1.134937064	0.004666219
>>> leg0	2027-07-12	1.148588556	0.004666219
>>> leg0	2028-01-11	1.164662739	0.004691858
>>> leg0	2028-07-11	1.184811281	0.004666219
>>> leg0	2029-01-11	1.202933688	0.004717496
>>> leg0	2029-07-11	1.22143445	0.004640581
>>> leg0	2030-01-11	1.244698418	0.004717496
>>> leg0	2030-07-11	1.268437607	0.004640581
>>> leg0	2031-01-13	1.289269217	0.004768773
>>> leg0	2031-07-11	1.310335034	0.004589303
>>> leg0	2032-01-12	1.330460609	0.004743135
>>> leg0	2032-07-12	1.351119197	0.004666219
>>> leg1	2022-10-11	1.025077227	0.000595151
>>> leg1	2023-01-11	1.031727084	0.000595151
>>> leg1	2023-04-11	1.037740592	0.000582209
>>> leg1	2023-07-11	1.044326706	0.00058868
>>> leg1	2023-10-11	1.050712784	0.001151809
>>> leg1	2024-01-11	1.056475309	0.001190658
>>> leg1	2024-04-11	1.062040699	0.001177708
>>> leg1	2024-07-11	1.067323978	0.001177708
>>> leg1	2024-10-11	1.073341181	0.001190658
>>> leg1	2025-01-13	1.080208917	0.001216557
>>> leg1	2025-04-11	1.087963402	0.00113886
>>> leg1	2025-07-11	1.095773345	0.001322081
>>> leg1	2025-10-14	1.102080743	0.002477114
>>> leg1	2026-01-12	1.107358267	0.002346587
>>> leg1	2026-04-13	1.112978456	0.002372691
>>> leg1	2026-07-13	1.119706713	0.002372691
>>> leg1	2026-10-13	1.127732208	0.002398796
>>> leg1	2027-01-11	1.134937064	0.002346587
>>> leg1	2027-04-12	1.141923849	0.002372691
>>> leg1	2027-07-12	1.148588556	0.002390753
>>> leg1	2027-10-12	1.155929127	0.002537285
>>> leg1	2028-01-11	1.164662739	0.002509671
>>> leg1	2028-04-11	1.174741361	0.002509671
>>> leg1	2028-07-11	1.184811281	0.002509671
>>> leg1	2028-10-11	1.194083437	0.002537285
>>> leg1	2029-01-11	1.202933688	0.002537285
>>> leg1	2029-04-11	1.211815927	0.002482058
>>> leg1	2029-07-11	1.22143445	0.002509671
>>> leg1	2029-10-11	1.232822712	0.002537285
>>> leg1	2030-01-11	1.244698418	0.002537285
>>> leg1	2030-04-11	1.256926582	0.002482058
>>> leg1	2030-07-11	1.268437607	0.003472659
>>> leg1	2030-10-11	1.278474854	0.004092082
>>> leg1	2031-01-13	1.289269217	0.004181226
>>> leg1	2031-04-11	1.299839855	0.003914165
>>> leg1	2031-07-11	1.310335034	0.004047153
>>> leg1	2031-10-14	1.320633529	0.004225801
>>> leg1	2032-01-12	1.330460609	0.004002946
>>> leg1	2032-04-12	1.340898895	0.004047513
>>> leg1	2032-07-12	1.351119197	0.004047513
>>> Regards
>>> 
>>> Philippe
>>> 
>>> 
>>> 
>>> Brokerage services offered through Exos Securities LLC, member of SIPC <http://www.sipc.org/> / FINRA <http://www.finra.org/>.  For important disclosures, click here <https://www.exosfinancial.com/disclosures>. 
>>> _______________________________________________
>>> QuantLib-users mailing list
>>> Qua...@li... <mailto:Qua...@li...>
>>> https://lists.sourceforge.net/lists/listinfo/quantlib-users
>> 
>> _______________________________________________
>> QuantLib-users mailing list
>> Qua...@li...
>> https://lists.sourceforge.net/lists/listinfo/quantlib-users