Valuing Interest-Rate Swaps
A pension fund entered a $20 million, three-year
interest-rate swap
exactly one year ago, receiving a fixed 3.50% and paying floating. Rates have risen since
then. The fund's risk officer needs a number, today, for the swap's current market
value — not what it was worth at initiation, not what it will be worth at maturity, but its
value right now, for the quarterly mark-to-market report. Unlike a stock, a swap has
no ticker to look up. This lesson builds the number from the two pieces the swap is made of.
A swap is the difference of two bonds
The cleanest way to value a swap leans on a trick: separate the two legs and value each one as
if it were a stand-alone bond. From the point of view of the party who receives
fixed, pays floating, holding the swap is equivalent to
- being long a fixed-rate bond — receiving each fixed coupon, plus the notional at maturity, and
- being short a floating-rate bond — paying out each floating coupon, plus the notional at maturity.
Neither bond's principal is real (no notional ever changes hands, as we established in the
first lesson of this module) — but the value of the swap is still exactly the
difference of what those two imaginary bonds would be worth if the principal were real, because
the imaginary principal amounts are identical and cancel:
V_{\text{receive-fixed}} = B_{\text{fix}} - B_{\text{fl}}, \qquad V_{\text{pay-fixed}} = B_{\text{fl}} - B_{\text{fix}}.
B_{\text{fix}} is priced exactly like any coupon bond — discount
every remaining coupon and the final "principal" back to today. The genuinely useful fact is
what happens to B_{\text{fl}}.
-
Immediately after a reset date, the floating-rate note is about to pay exactly
the current market rate on its notional for the next period — precisely what any newly
issued floating-rate note at that moment would pay.
-
A floating-rate note that always pays the going market rate is, by definition, worth
exactly its par value (the notional) right after each reset — buyers
would neither pay a premium nor demand a discount for a bond guaranteed to yield the
market rate.
-
So B_{\text{fl}} = N (the notional) at every reset date,
regardless of what happens to interest rates in between resets or what the general level
of the yield curve is. This is what makes swap valuation so much simpler than it looks.
Worked example: revaluing the pension fund's swap
The fund's swap: notional N = \$20{,}000{,}000, fixed rate
3.50\%, semiannual, three-year original term. One year (exactly two
resets) has now passed, so we are valuing it right at the third reset date — two
years and four semiannual coupons remain. Today's zero-coupon discount curve (continuously
compounded) reads:
| Maturity |
Zero rate |
Discount factor |
Fixed coupon |
PV of coupon |
| 0.5y |
4.00% |
0.9802 |
$350,000 |
$343,070 |
| 1.0y |
4.20% |
0.9589 |
$350,000 |
$335,615 |
| 1.5y |
4.40% |
0.9362 |
$350,000 |
$327,670 |
| 2.0y |
4.50% |
0.9139 |
$350,000 |
$319,865 |
The fixed leg's coupons total \$1{,}326{,}220 in present value; add
the "notional" repaid at 2.0 years, discounted at the same 0.9139:
B_{\text{fix}} = 1{,}326{,}220 + 20{,}000{,}000 \times 0.9139 = 1{,}326{,}220 + 18{,}278{,}000 = \$19{,}604{,}220.
Because today happens to be a reset date, the floating leg is worth par exactly:
B_{\text{fl}} = \$20{,}000{,}000. The swap's value to the
fund (which receives fixed) is:
V_{\text{receive-fixed}} = B_{\text{fix}} - B_{\text{fl}} = 19{,}604{,}220 - 20{,}000{,}000 = -\$395{,}780.
A negative number for the fund — sensible, since market rates have risen well above the
3.50% the fund locked in as a receiver of fixed; it is now stuck receiving a
below-market rate. Its swap counterparty, who pays fixed and receives floating, holds the
mirror-image position worth exactly +\$395{,}780. Swap valuation is
always zero-sum this way: one side's mark-to-market gain is exactly the other side's loss.
See the value curve
Because B_{\text{fl}} always collapses to par right at a reset, the
entire value of a receive-fixed swap right after a reset is just a fixed-coupon bond's price
minus its own par value — and a bond priced at its own coupon rate is worth par. That's why
the curve below always passes almost exactly through zero at the swap's own original fixed
rate, however the current market rate wanders elsewhere.
Slide the original fixed rate and watch the whole curve — and its zero crossing — shift with
it. To the right of the crossing (current rates above the swap's fixed rate) the receive-fixed
position is under water; to the left, it's in the money. This is exactly the same shape as a
bond's price-yield curve, because that is precisely what it is.
A $20 million swap losing $395,780 sounds dramatic; a $2 billion swap book losing the
proportional amount would be a genuinely different headline. Trading desks instead track a
swap's DV01 (or PV01) — the change in value for a one-basis-point
(0.01%) parallel shift in the discount curve — which behaves almost exactly like a bond's
duration
because, as this lesson shows, a swap effectively is two bonds. A desk with many
swaps on its book sums the DV01s across the whole portfolio to get one aggregate risk number
instead of re-deriving a discount curve calculation for each trade individually — the same
instinct that turns a warehouse full of individual positions into one manageable risk report.
It is tempting to conclude that a floating-rate leg is always worth its notional,
full stop — but the theorem only holds exactly at a reset date. Between resets, the
floating leg's next payment has already been locked in (reset in advance, remember) at a
rate that may no longer match the market's current view, so its value can drift slightly
away from par until the next reset resets it back. For most practical valuation this drift is
small relative to the swap's overall size, and desks often value mid-period as if sitting
exactly at the next reset for simplicity — but a risk manager marking a large book on a date
that lands mid-period should account for the (usually small) known next coupon separately
rather than silently assuming exact par.
There's a second, sharper subtlety worth flagging even here: which discount curve you use
for B_{\text{fix}} and B_{\text{fl}}
in the first place turns out to matter more than most students expect — enough that it
reshaped the entire swap market after 2008, which is exactly where this module ends up.