Cross-currency swaps

At this time, there’s no instrument class in the library modeling cross-currency swaps. However, it’s possible to calculate their value by working with cashflows. Here is a short example.

import QuantLib as ql
import pandas as pd

today = ql.Date(27, ql.October, 2021)
ql.Settings.instance().evaluationDate = today

bps = 1e-4

Sample data

For the purposes of this notebook, we’ll use mock rates. The data frame below holds made-up zero rates from a few different curves at a number of nodes…

sample_rates = pd.DataFrame(
    [
        (ql.Date(27, 10, 2021), 0.0229, 0.0893, -0.5490, -0.4869),
        (ql.Date(27, 1, 2022), 0.0645, 0.1059, -0.5584, -0.5057),
        (ql.Date(27, 4, 2022), 0.0414, 0.1602, -0.5480, -0.5236),
        (ql.Date(27, 10, 2022), 0.1630, 0.2601, -0.5656, -0.5030),
        (ql.Date(27, 10, 2023), 0.4639, 0.6281, -0.4365, -0.3468),
        (ql.Date(27, 10, 2024), 0.7187, 0.9270, -0.3500, -0.2490),
        (ql.Date(27, 10, 2025), 0.9056, 1.1257, -0.3041, -0.1590),
        (ql.Date(27, 10, 2026), 1.0673, 1.2821, -0.2340, -0.0732),
        (ql.Date(27, 10, 2027), 1.1615, 1.3978, -0.1690, -0.0331),
        (ql.Date(27, 10, 2028), 1.2326, 1.4643, -0.1041, 0.0346),
        (ql.Date(27, 10, 2029), 1.3050, 1.5589, -0.0070, 0.1263),
        (ql.Date(27, 10, 2030), 1.3584, 1.5986, 0.0272, 0.1832),
        (ql.Date(27, 10, 2031), 1.4023, 1.6488, 0.0744, 0.2599),
        (ql.Date(27, 10, 2036), 1.5657, 1.8136, 0.3011, 0.4406),
        (ql.Date(27, 10, 2041), 1.6191, 1.8749, 0.3882, 0.5331),
        (ql.Date(27, 10, 2046), 1.6199, 1.8701, 0.3762, 0.5225),
        (ql.Date(27, 10, 2051), 1.6208, 1.8496, 0.3401, 0.4926),
    ],
    columns=[
        "date",
        "SOFR",
        "USDLibor3M",
        "EUR-USD-discount",
        "Euribor3M",
    ],
)

…and this helper function uses them to create an interest-rate curve. Depending on the context, curves will be used for either forecasting or discounting.

def sample_curve(tag):
    curve = ql.ZeroCurve(
        sample_rates["date"], sample_rates[tag] / 100, ql.Actual365Fixed()
    )
    return ql.YieldTermStructureHandle(curve)

Const-notional cross-currency swaps

The first kind of cross-currency swaps we’ll model is less common but simpler. The notionals are exchanged at the beginning (where they have the same value, given the exchange rate at that time) and again at the end. There is no rebalancing during the life of the swap. All the coupons in either leg have the same notional in the leg’s payment currency.

In this example we’ll model a 5-years swap paying quarterly coupons, based on 3M Euribor on one leg and 3M USD Libor on the other. We start by creating the indexes and their forecasting curves.

euribor3M_curve = sample_curve("Euribor3M")
euribor3M = ql.Euribor(ql.Period(3, ql.Months), euribor3M_curve)

usdlibor3M_curve = sample_curve("USDLibor3M")
usdlibor3M = ql.USDLibor(ql.Period(3, ql.Months), usdlibor3M_curve)

For each of the currencies, we create a corresponding sequence of cashflows, including the notional exchanges (which, unlike for vanilla swaps, don’t cancel out—at least at maturity). The reference notional will be in dollars, converted in EUR at the present exchange rate (also made up). Just for kicks, we’ll also add a spread to the USD leg.

notional = 1_000_000
fx_0 = 0.85
spread = 50.0 * bps

As I mentioned, the swaps make quarterly payments. The corresponding schedule starts spot and ends in five years.

calendar = ql.UnitedStates(ql.UnitedStates.FederalReserve)
start_date = calendar.advance(today, ql.Period(2, ql.Days))
end_date = calendar.advance(start_date, ql.Period(5, ql.Years))
tenor = ql.Period(3, ql.Months)
rule = ql.DateGeneration.Forward
convention = ql.Following
end_of_month = False

schedule = ql.Schedule(
    start_date,
    end_date,
    tenor,
    calendar,
    convention,
    convention,
    rule,
    end_of_month,
)

Since the notionals don’t change, the two legs are similar: an initial lending of notional, the interest payments received according to the schedule and the index fixings, and the final payment of the notional.

usd_leg = (
    (ql.SimpleCashFlow(-notional, schedule[0]),)
    + ql.IborLeg(
        nominals=[notional],
        schedule=schedule,
        index=usdlibor3M,
        spreads=[spread],
    )
    + (ql.SimpleCashFlow(notional, schedule[-1]),)
)

For the EUR leg, of course, we’ll have to convert the notional in the proper currency.

eur_leg = (
    (ql.SimpleCashFlow(-notional * fx_0, schedule[0]),)
    + ql.IborLeg(
        nominals=[notional * fx_0], schedule=schedule, index=euribor3M
    )
    + (ql.SimpleCashFlow(notional * fx_0, schedule[-1]),)
)

Now, we can get the NPV of each leg in its own currency by discounting them with the corresponding curve. For the USD leg, that would be the SOFR curve. For the EUR leg, ideally, a discount curve bootstrapped on cross-currency instruments so that it can capture the basis.

sofr_curve = sample_curve("SOFR")
usd_npv = ql.CashFlows.npv(usd_leg, sofr_curve, True)
usd_npv

35427.6553257405

eurusd_curve = sample_curve("EUR-USD-discount")
eur_npv = ql.CashFlows.npv(eur_leg, eurusd_curve, True)
eur_npv

6908.723028828779

And of course, we can convert the NPV of the EUR leg in USD:

ql.CashFlows.npv(eur_leg, eurusd_curve, True) / fx_0

8127.909445680917

We can also look at each cashflow by means of a short(ish) helper function:

def cashflow_data(leg):
    data = []
    for cf in sorted(leg, key=lambda c: c.date()):
        coupon = ql.as_floating_rate_coupon(cf)
        if coupon is None:
            data.append((cf.date(), None, None, cf.amount()))
        else:
            data.append(
                (
                    coupon.date(),
                    coupon.nominal(),
                    coupon.rate(),
                    coupon.amount(),
                )
            )
    return pd.DataFrame(
        data, columns=["date", "nominal", "rate", "amount"]
    ).style.format(
        {"amount": "{:.2f}", "nominal": "{:.2f}", "rate": "{:.2%}"}
    )

Here are the cashflows in USD…

cashflow_data(usd_leg)

	date	nominal	rate	amount
0	October 29th, 2021	nan	nan%	-1000000.00
1	January 31st, 2022	1000000.00	0.61%	1585.56
2	April 29th, 2022	1000000.00	0.72%	1750.57
3	July 29th, 2022	1000000.00	0.81%	2040.59
4	October 31st, 2022	1000000.00	0.91%	2386.92
5	January 30th, 2023	1000000.00	1.22%	3080.34
6	May 1st, 2023	1000000.00	1.40%	3541.30
7	July 31st, 2023	1000000.00	1.58%	3999.86
8	October 30th, 2023	1000000.00	1.76%	4444.64
9	January 29th, 2024	1000000.00	1.79%	4518.96
10	April 29th, 2024	1000000.00	1.93%	4890.80
11	July 29th, 2024	1000000.00	2.08%	5262.77
12	October 29th, 2024	1000000.00	2.22%	5682.53
13	January 29th, 2025	1000000.00	2.06%	5257.82
14	April 29th, 2025	1000000.00	2.16%	5388.64
15	July 29th, 2025	1000000.00	2.25%	5695.32
16	October 29th, 2025	1000000.00	2.35%	6000.94
17	January 29th, 2026	1000000.00	2.27%	5807.00
18	April 29th, 2026	1000000.00	2.35%	5873.71
19	July 29th, 2026	1000000.00	2.43%	6133.38
20	October 29th, 2026	1000000.00	2.50%	6388.32
21	October 29th, 2026	nan	nan%	1000000.00

…and here are those in EUR.

cashflow_data(eur_leg)

	date	nominal	rate	amount
0	October 29th, 2021	nan	nan%	-850000.00
1	January 31st, 2022	850000.00	-0.50%	-1108.91
2	April 29th, 2022	850000.00	-0.53%	-1109.57
3	July 29th, 2022	850000.00	-0.49%	-1042.88
4	October 31st, 2022	850000.00	-0.46%	-1020.88
5	January 30th, 2023	850000.00	-0.30%	-644.90
6	May 1st, 2023	850000.00	-0.22%	-479.05
7	July 31st, 2023	850000.00	-0.15%	-314.08
8	October 30th, 2023	850000.00	-0.07%	-158.20
9	January 29th, 2024	850000.00	-0.12%	-266.58
10	April 29th, 2024	850000.00	-0.08%	-163.55
11	July 29th, 2024	850000.00	-0.03%	-60.50
12	October 29th, 2024	850000.00	0.02%	42.55
13	January 29th, 2025	850000.00	0.04%	96.24
14	April 29th, 2025	850000.00	0.09%	188.22
15	July 29th, 2025	850000.00	0.13%	284.92
16	October 29th, 2025	850000.00	0.18%	383.98
17	January 29th, 2026	850000.00	0.20%	443.61
18	April 29th, 2026	850000.00	0.25%	523.68
19	July 29th, 2026	850000.00	0.29%	619.73
20	October 29th, 2026	850000.00	0.33%	708.11
21	October 29th, 2026	nan	nan%	850000.00

To get the NPV of the swap, we add those of the two legs after converting the EUR leg back to dollars:

NPV = usd_npv - eur_npv / fx_0
NPV

27299.745880059585

Mark-to-market cross-currency swaps

In this more common kind of cross-currency swaps, the notionals are rebalanced at each coupon date so that their value remain the same (according, of course, to the value of the FX rate at each coupon start).

In order to model this rebalancing feature we will need to estimate the FX rates in the future, the corresponding notionals in Euro for the floating cashflows, and the amounts exchanged due to rebalancing.

The future FX rates can be forecast from the discount curves, since they model the cost of money. We’ll write a convenience function to extract it at any given date:

def FX(date):
    return fx_0 * sofr_curve.discount(date) / eurusd_curve.discount(date)

The notionals at the start of each coupon can now be calculated from the FX rates:

start_dates = list(schedule)[:-1]

notionals = [notional * FX(d) for d in start_dates]

Given the notionals, we can also calculate the rebalancing cashflows:

rebalancing_cashflows = []
for i in range(len(notionals) - 1):
    rebalancing_cashflows.append(
        ql.SimpleCashFlow(notionals[i] - notionals[i + 1], schedule[i + 1])
    )

Finally, we can create the two legs and price them. The USD leg is as before, since its notional doesn’t change:

usd_leg = (
    (ql.SimpleCashFlow(-notional, schedule[0]),)
    + ql.IborLeg(
        nominals=[notional],
        schedule=schedule,
        index=usdlibor3M,
        spreads=[spread],
    )
    + (ql.SimpleCashFlow(notional, schedule[-1]),)
)

cashflow_data(usd_leg)

	date	nominal	rate	amount
0	October 29th, 2021	nan	nan%	-1000000.00
1	January 31st, 2022	1000000.00	0.61%	1585.56
2	April 29th, 2022	1000000.00	0.72%	1750.57
3	July 29th, 2022	1000000.00	0.81%	2040.59
4	October 31st, 2022	1000000.00	0.91%	2386.92
5	January 30th, 2023	1000000.00	1.22%	3080.34
6	May 1st, 2023	1000000.00	1.40%	3541.30
7	July 31st, 2023	1000000.00	1.58%	3999.86
8	October 30th, 2023	1000000.00	1.76%	4444.64
9	January 29th, 2024	1000000.00	1.79%	4518.96
10	April 29th, 2024	1000000.00	1.93%	4890.80
11	July 29th, 2024	1000000.00	2.08%	5262.77
12	October 29th, 2024	1000000.00	2.22%	5682.53
13	January 29th, 2025	1000000.00	2.06%	5257.82
14	April 29th, 2025	1000000.00	2.16%	5388.64
15	July 29th, 2025	1000000.00	2.25%	5695.32
16	October 29th, 2025	1000000.00	2.35%	6000.94
17	January 29th, 2026	1000000.00	2.27%	5807.00
18	April 29th, 2026	1000000.00	2.35%	5873.71
19	July 29th, 2026	1000000.00	2.43%	6133.38
20	October 29th, 2026	1000000.00	2.50%	6388.32
21	October 29th, 2026	nan	nan%	1000000.00

The EUR leg, instead, changes notionals and thus it includes the rebalancing payments:

eur_leg = (
    [ql.SimpleCashFlow(-notionals[0], schedule[0])]
    + list(
        ql.IborLeg(nominals=notionals, schedule=schedule, index=euribor3M)
    )
    + rebalancing_cashflows
    + [ql.SimpleCashFlow(notionals[-1], schedule[-1])]
)

cashflow_data(eur_leg)

	date	nominal	rate	amount
0	October 29th, 2021	nan	nan%	-849973.31
1	January 31st, 2022	849973.31	-0.50%	-1108.87
2	January 31st, 2022	nan	nan%	1361.41
3	April 29th, 2022	848611.90	-0.53%	-1107.76
4	April 29th, 2022	nan	nan%	1140.19
5	July 29th, 2022	847471.71	-0.49%	-1039.78
6	July 29th, 2022	nan	nan%	1688.83
7	October 31st, 2022	845782.88	-0.46%	-1015.82
8	October 31st, 2022	nan	nan%	2036.91
9	January 30th, 2023	843745.97	-0.30%	-640.16
10	January 30th, 2023	nan	nan%	1989.74
11	May 1st, 2023	841756.23	-0.22%	-474.40
12	May 1st, 2023	nan	nan%	2164.38
13	July 31st, 2023	839591.85	-0.15%	-310.23
14	July 31st, 2023	nan	nan%	2337.65
15	October 30th, 2023	837254.19	-0.07%	-155.83
16	October 30th, 2023	nan	nan%	2508.90
17	January 29th, 2024	834745.29	-0.12%	-261.80
18	January 29th, 2024	nan	nan%	2661.04
19	April 29th, 2024	832084.25	-0.08%	-160.10
20	April 29th, 2024	nan	nan%	2825.60
21	July 29th, 2024	829258.65	-0.03%	-59.02
22	July 29th, 2024	nan	nan%	2988.43
23	October 29th, 2024	826270.22	0.02%	41.36
24	October 29th, 2024	nan	nan%	3181.28
25	January 29th, 2025	823088.95	0.04%	93.19
26	January 29th, 2025	nan	nan%	3166.37
27	April 29th, 2025	819922.58	0.09%	181.56
28	April 29th, 2025	nan	nan%	3227.34
29	July 29th, 2025	816695.24	0.13%	273.75
30	July 29th, 2025	nan	nan%	3392.07
31	October 29th, 2025	813303.17	0.18%	367.40
32	October 29th, 2025	nan	nan%	3550.54
33	January 29th, 2026	809752.63	0.20%	422.60
34	January 29th, 2026	nan	nan%	3259.96
35	April 29th, 2026	806492.67	0.25%	496.88
36	April 29th, 2026	nan	nan%	3266.85
37	July 29th, 2026	803225.82	0.29%	585.63
38	July 29th, 2026	nan	nan%	3380.28
39	October 29th, 2026	799845.53	0.33%	666.33
40	October 29th, 2026	nan	nan%	799845.53

Finally, as in the previous case, we can calculate the NPV of each leg and of the swap:

usd_npv = ql.CashFlows.npv(usd_leg, sofr_curve, True)
usd_npv

35427.6553257405

eur_npv = ql.CashFlows.npv(eur_leg, eurusd_curve, True)
eur_npv

6678.756064412276

NPV = usd_npv - eur_npv / fx_0
NPV

27570.295249961353

Future developments

The above calculations work but are not very well integrated with the rest of the library; if you’re familiar with QuantLib, you know that usually instruments can monitor market quotes and recalculate automatically when they change. This is not possible with the above; it will be up to you to recreate the cashflows when the FX rate or the curves are updated. A cross-currency swap modeled as an instrument would be way more convenient.

The reason we don’t have such an instrument yet is that, unfortunately, the calculations above are a bit awkward to encapsulate in an instrument class. The constant-notional case is easy enough, but the mark-to-market case requires changing the notionals of the coupons when market quotes change, which is something our current classes were not prepared for. We’ll have to think about it and try to add the required functionality in future releases.