Mechanics of Interest-Rate Swaps

A mid-cap manufacturer borrowed $50 million three years ago on a floating-rate term loan pegged to 6-month SOFR. Rates have been climbing all year and the CFO is losing sleep over every central-bank meeting. Refinancing the loan itself would mean lawyers, covenants, and a prepayment penalty. There's a cleaner fix: leave the loan exactly as it is, and separately enter an agreement with a bank to receive floating and pay fixed on the same notional. The floating payments received from the swap now offset the floating interest owed on the loan, and what's left is a fixed rate, locked in for years, without touching the original debt at all.

That agreement is an interest-rate swap — the single most heavily traded derivative in the world, with hundreds of trillions of dollars of notional outstanding. A swap is really nothing exotic: it's a string of forward rate agreements, one per reset date, packaged into a single contract. Once you understand how one FRA settles, you already understand how every payment on a swap works — a swap just repeats that same settlement over and over, on a schedule, until maturity.

The plain-vanilla structure

The most common interest-rate swap — the plain-vanilla fixed-for-floating swap — has two counterparties who agree, at initiation, on four things: a notional principal N, a fixed rate R_{\text{fix}}, a floating-rate index (historically LIBOR, today typically SOFR-based), and a schedule of reset dates running to maturity. On each reset date, the floating rate for the next period is observed, and at the end of that period two payments are calculated on the notional:

\text{fixed leg: } N \cdot R_{\text{fix}} \cdot \tau, \qquad \text{floating leg: } N \cdot L \cdot \tau,

where L is the floating index rate set at the start of the period and \tau is the accrual fraction of a year covered (for a semiannual swap, \tau \approx 0.5, adjusted for the exact day-count convention). In practice only the net of the two legs actually changes hands — if the fixed-rate payer owes more than they're owed, they simply write one check for the difference.

Worked example: three years of cash flows

Take our manufacturer's swap directly. Notional N = \$50{,}000{,}000, fixed rate R_{\text{fix}} = 4.20\%, semiannual payments (\tau = 0.5), 6-month SOFR as the floating index, running three years — six settlement dates. The manufacturer is the fixed-rate payer: it pays fixed, receives floating. Each period's fixed leg is the same:

N \cdot R_{\text{fix}} \cdot \tau = 50{,}000{,}000 \times 0.042 \times 0.5 = \$1{,}050{,}000.

The floating leg moves with the market. Suppose SOFR is set as follows at the start of each period:

Period SOFR set Floating leg Fixed leg Net to fixed-rate payer
1 3.80% $950,000 $1,050,000 −$100,000
2 4.10% $1,025,000 $1,050,000 −$25,000
3 4.60% $1,150,000 $1,050,000 +$100,000
4 4.90% $1,225,000 $1,050,000 +$175,000
5 5.10% $1,275,000 $1,050,000 +$225,000
6 4.70% $1,175,000 $1,050,000 +$125,000

Read the last column as the story of the hedge: in periods 1 and 2, SOFR sits below the 4.20% fixed rate, so the manufacturer's swap actually costs it money net — but that's fine, because its floating-rate loan is also cheap in those periods. From period 3 onward SOFR climbs past 4.20%, the swap starts paying out net, and those payments exactly offset the extra interest now due on the floating loan. Add the six net figures and the swap paid the company a net $500,000 over three years — but the real point isn't that total, it's that in every single period, floating-loan interest plus swap net settles to a flat 4.20% on $50 million. That's the whole design: convert an unpredictable stream into a predictable one.

See the two legs move

Each payment date is just this one relationship: a flat fixed payment against a floating payment that scales linearly with wherever the index rate lands. Drag the fixed rate and watch the crossing point — the rate at which neither side would owe anything net — move with it.

To the right of the crossing point the floating-rate line lies above the fixed one: the fixed-rate payer profits on that period. To the left, it's the mirror image. A swap doesn't eliminate risk — it relocates it, from "will my loan rate rise?" to nothing at all for the fixed-rate payer, and onto the floating-rate receiver on the other side of the trade, who is betting the opposite way.

This convention — reset in advance, pay in arrears — isn't an accident, and it's the same logic that makes an FRA settle the way it does. At the start of the period, 6-month SOFR is observed and both sides now know exactly what the floating payment will be six months later: there's no more uncertainty left for that period, only a fixed number waiting to be paid. Structuring it any other way — say, using the rate observed at the end of the period — would leave both sides pricing an unknown quantity right up until the last moment, which defeats the purpose of a rate-fixing mechanism. Reset-in-advance turns each period of a swap into, in effect, a tiny FRA that's already been cash-settled in everyone's head the moment the rate is set — only the paperwork waits six months.

The single most common beginner error is treating the "$50 million" in a $50-million swap as money that moves between the two parties. It never does — not at initiation, not at maturity. The notional exists purely as an input to a multiplication; only the tiny difference between two interest rates, applied to that notional for one accrual period, ever changes hands. This is precisely what makes swaps so capital-efficient: two firms can transform $50 million of interest-rate exposure while only a few hundred thousand dollars a year actually crosses accounts.

Don't let this lull you into thinking the notional is irrelevant, though — it still determines every payment's size, and (as a later lesson shows) it is exactly what does get exchanged in a currency swap, which is why currency swaps carry a kind of risk that interest-rate swaps simply don't.