Skip to content

New to Omni?·Get 20% off your first purchase with code WELCOME20 at checkout.See plans

FRM® CMT swaps: why must the tree match each reset date?

A constant maturity Treasury swap needs a fresh long-maturity yield at every reset. See how time steps, node discounting and payment dates shape its value.

ConceptOmni Curriculum Team6 min read

Wooden markers aligned above an unlabelled branching curve on white paper on a navy desk.
On this page

A constant maturity Treasury swap resets one leg to a Treasury yield of a specified maturity at each agreed reset date. A short-rate tree used to value it must represent those dates and generate the relevant yield curve at every reset node; a single short rate at a node cannot stand in for the longer-maturity reset yield. The time grid matters because a missed reset date forces an extra approximation before the cash flow can even be valued.

The rate being reset is a curve observation

The word “constant” refers to the maturity of the yield reference, not to an unchanging payment. The Federal Reserve explains that Treasury constant-maturity yields are read from a curve at fixed maturities, so a chosen maturity remains comparable even when no outstanding security has precisely that remaining life. A swap linked to such a yield can therefore have a new rate at every observation. The referenced maturity and the time between observations are two separate contract features. Confusing them is the first source of a wrong valuation.

Suppose a contract references a ten-year Treasury constant-maturity yield but settles every six months. At each reset, the question is what the ten-year yield is on that date. It is not the yield of a bond whose ten-year original term has shortened since trade inception. Nor is it the six-month rate used to discount the next payment. The former is the floating reference; the latter is one input into valuation. The Federal Reserve’s H.15 description makes the fixed-maturity convention clear.

In a tree, future market curves are unknown today. The model represents possible states at future dates. At a reset node, the analyst takes the model-implied discount factors for the reference maturity beyond that node and derives the corresponding par yield under the chosen convention. Each node at the same calendar date can imply a different par yield. That is why one deterministic forward-rate list is insufficient to describe all possible future swap payments when the problem asks for a tree valuation.

Separate the reference yield, payment and discount rate

A useful discipline is to write three labels beside every reset date: reference yield, accrual fraction, and discount factor. The reference yield determines the variable leg. The accrual fraction turns an annualised rate difference into the payment for that period. The discount factor brings the state-contingent payment back through the tree. The fixed rate is a contractual comparison rate, not a discount rate. Even if numerical values happen to match in a toy example, their jobs remain different.

CFt=N αt (yCMT,t−K)CF_t=N\,\alpha_t\,(y_{CMT,t}-K)
Illustrative receive-CMT net settlement: N is notional, alpha is the accrual fraction, y is the reset yield, and K is the fixed rate.

The sign convention in that expression is for receiving the CMT-linked leg and paying fixed. Reverse the position and every net cash flow changes sign. Real contracts also specify observation, payment, day-count and fallback conventions. An exam calculation usually supplies enough of these to remove ambiguity. If it does not, state the assumed accrual convention before inserting a number; silently treating a half-year as a full year doubles the illustrative payment.

A deliberately small worked scenario

Consider invented figures for a receive-CMT swap with a notional of 1,000,000 currency units, a fixed rate of 4%, and half-year accruals. At the next reset there are two model states, each assigned risk-neutral probability one half. The model-implied reference yields are 5% in the upper state and 3% in the lower state. Assume for this teaching example that the payment is made at that reset and that the current one-period discount factor is 0.98. None of these numbers is a market quote or a proposed contract.

The upper-state net payment is 1,000,000 times one half times the difference between 5% and 4%, giving positive 5,000 to the receive-CMT side. The lower-state payment is 1,000,000 times one half times the difference between 3% and 4%, giving negative 5,000. Their equal-probability expected net payment is zero, so this single-settlement illustration has a present value of zero after multiplying by 0.98. This result follows from deliberately symmetric yields; it says nothing about the value of a real CMT swap or the fair fixed rate.

Now change only the upper-state reference yield to 6%. Its payment becomes positive 10,000 while the lower-state payment remains negative 5,000. The risk-neutral expected net payment is 2,500 and its value at the stated discount factor is 2,450. The point of the arithmetic is the order: determine a long-maturity yield for each state, calculate that state’s payment, average under the model’s pricing probabilities, then discount. Averaging the yields first happens to give the same answer in this linear one-payment setup, but that shortcut is unreliable once reset dates and state-dependent discounting multiply.

Why the spacing of the tree changes the answer

If payments are scheduled at half-year intervals, a tree with only annual nodes omits the first reset. An analyst could interpolate a state at the missing date, but that is an additional modelling choice. It can alter both the reset yield and the discount applied to the resulting payment. A grid containing every contractual observation and settlement date makes the bookkeeping explicit. A finer grid may also reduce numerical discretisation error, though finer does not automatically mean accurate: calibration, yield-curve construction and model assumptions still matter.

The tree must also extend far enough beyond a reset node to infer the reference yield’s maturity, or otherwise supply a consistent curve construction there. A ten-year reference yield at a future reset cannot be inferred from that node’s next six-month short rate alone. This is the deeper reason that CMT valuation is a term-structure exercise rather than a simple one-period floating-rate calculation. The Federal Reserve’s nominal yield-curve explanation distinguishes the curve across maturities from any single observed security.

Work backward only after the future-node cash flows have been defined. At the final relevant payment node, record the net amount under the contract’s sign convention. At the preceding node, take the pricing-probability weighted continuation values, discount using the rate and period appropriate to that node, then add any payment occurring there. Repeat toward today. That is the same general tree discipline used elsewhere in market-risk valuation; the CMT-specific extra work is rebuilding the referenced long-maturity yield at every reset state.

What to check in an exam answer

Start with a timeline. Mark observation dates separately from payment dates if the question distinguishes them. Mark the reset reference maturity separately from the interval between resets. At each reset node, ask whether the quoted number is a par or constant-maturity yield, a short rate, or a discount factor. Only the first belongs in the floating coupon. Use the stated day-count fraction, and keep the receiver’s sign convention visible throughout the rollback.

A common error is to look at the short rate printed inside a node and treat it as the CMT reset. That may look plausible because both are interest rates, but they describe different horizons. Another error is to carry today’s reference yield forward unchanged. “Constant maturity” keeps the tenor of the observation constant; the level of the future observation is allowed to move. Finally, coarse time steps can hide a missing cash flow, so check grid alignment before trusting any elegantly computed terminal value.

This topic sits in the market-risk portion of GARP’s FRM learning objectives. GARP’s exam FAQ also distinguishes the foundational tools in Part I from their applications in Part II. For a wider route through the course, use the FRM programme page and the methodology page. The useful study habit here is to draw the reset timeline before doing any tree arithmetic.

Sources

  1. FRM Exam Learning Objectives, Syllabus & more | GARP · garp.org · Retrieved
  2. FRM Exam FAQs | GARP · garp.org · Retrieved
  3. Selected Interest Rates H.15 | Federal Reserve · federalreserve.gov · Retrieved
  4. Nominal Yield Curve | Federal Reserve · federalreserve.gov · Retrieved

Explore FRM study resources

Educational content, not investment advice. This article is written to help candidates prepare for professional exams. It is not a recommendation to buy, sell or hold any security, and it does not take your personal circumstances into account.

Omni Finance Academy is an independent exam-preparation provider and is not affiliated with or endorsed by CFA Institute, GARP, CAIA Association or CFP Board. CFA®, FRM®, CAIA® and CFP® are trademarks of their respective owners. Full disclaimer.