On this page
A risk-neutral short-rate tree is a way to value an interest-rate claim using rates that can change at future nodes. Start with the claim's payoff at the end, weight the possible next values using the tree's pricing probabilities, and discount at the short rate at each earlier node. The weights are for valuation, not a forecast of how often rates will actually rise. For an FRM® Part II candidate, that distinction is the point of the method.
Why the tree exists
A single discount rate cannot describe a claim whose cash flow depends on the path of interest rates. A fixed bond may have known promised payments, but an option on a bond, a cap or an instrument with an exercise choice responds differently after rates rise and after they fall. A tree preserves those alternative states. Each column represents a point in time, and each node gives the one-period rate that applies from that node to the next column. The model turns a time path into a set of small, checkable calculations rather than one opaque present-value number.
GARP identifies market risk as a Part II area and distinguishes the application of risk tools in Part II from their introduction in Part I. Its FRM study materials are the current source for the formal curriculum. The FRM programme page is a useful orientation page for readers deciding where this concept fits in their preparation. In an exam setting, the tree often arrives already specified. The candidate's task is then to read the node rates and payoffs carefully, rather than to infer an entire model from a yield curve.
What risk-neutral means here
The up and down weights used in a valuation tree are pricing weights. They are selected with the rates so that values are consistent with observable prices under the model. A forecast probability asks a different question: how likely is a rate move in the real world? The two need not coincide. If an analyst expects an up move more often than the pricing weight suggests, substituting that expectation into a valuation calculation changes the question being answered. It produces a forecast-weighted amount, not the price implied by the calibrated tree.
Some short-rate trees use equal up and down pricing weights and put the calibration work into the node rates. That is a model convention, not a general rule about the likelihood of market moves. Other constructions can use different weights. Read the information supplied with the tree before calculating. The important consistency test is whether the same rates, weights and timing conventions reproduce the reference instruments the tree was built to price. An arbitrary collection of plausible future rates is not automatically a pricing model.
Work backwards from the payoff
At maturity, set out the amount delivered in each terminal state. Move one column left. For each earlier node, take the weighted average of its possible next-node values and discount for one period at the rate attached to the earlier node. Repeat until the initial node is reached. This sequence is backward induction. It is easy to audit because every value can be traced to its two successors, and the short rate used for a step is visible at the node from which that step begins.
The timing of a coupon or exercise payment matters. If a cash flow is received at a node, add it at that node; if it is received one period later, include it in the next value before discounting. An exam calculation can be numerically tidy yet economically wrong if a payment is shifted by one column. The safest approach is to mark every cash flow at the date it is received before doing any arithmetic. That prevents a familiar formula from disguising a timing error.
An illustrative two-period calculation
Consider an invented tree with annual periods. Today's short rate is 5%. At the next date, the up-node rate is 8% and the down-node rate is 2%. The stipulated risk-neutral weights are one half for each move. A zero-coupon claim pays 100 at the end of the second period in every state. These are deliberately round teaching numbers, not observed rates, market prices or a calibrated curve. There is no claim that this tree describes a traded instrument.
At the up node, the value one period before the payment is 100 divided by 1.08, or 92.5926. At the down node, it is 100 divided by 1.02, or 98.0392. Back at today's node, average those two values using the stipulated weights: the average is 95.3159. Discount it at today's 5% short rate. The result is approximately 90.7771. Retaining the unrounded intermediate values matters; prematurely rounding each branch can create a small difference in the final answer.
Now imagine the claim were an option that paid only when a specified underlying value exceeded a strike at the next date. The terminal entries would no longer all be 100. You would write the relevant payoff in each state and then apply the same rollback. This is why learning the procedure is more valuable than memorising one zero-coupon example. The cash-flow map changes from instrument to instrument, while the movement from right to left through a specified tree remains the organising method.
Calibration comes before extension to another claim
The illustrative price above is a model output only. Before using a tree to price another traded claim, a modeller checks it against the reference prices that are supposed to anchor the model. If the tree misses a reference price, it must be adjusted or the mismatch explained. Calling a tree risk-neutral does not itself perform this calibration. In practice, there can also be modelling choices about volatility, curve inputs and conventions. A candidate should distinguish an exercise where calibration is already given from one asking how a tree would be made consistent with reference prices.
This distinction helps interpret a common exam prompt. If the question supplies a fully calibrated tree and asks for a contingent claim's value, use the provided node rates and pricing weights. If it supplies a market price and an unknown node rate, the unknown may need to be solved so the tree reprices that instrument. Those are different operations. The first is valuation within a model; the second is fitting a model to a price. Confusing them can send a correct calculation toward the wrong target.
Read a node before you calculate
Write down four facts for each node: the date, the state, the local short rate and the claim value after any payment or exercise decision at that date. This compact notation exposes the common mistake of discounting from a future node with the rate that applies after, rather than before, the transition. It also helps when the tree recombines. Two paths may lead to one economic state, so one value belongs at that shared node. Count states, not route labels, when recording its value.
For an instrument with an exercise right, continuation value and exercise value are separate objects. First calculate the continuation value from later nodes. Then apply the contract's decision rule at the current node. The issuer and the holder have opposite incentives in a callable or putable structure, so a mechanically applied maximum is not always the right operation. State who controls exercise before comparing numbers. That reasoning is more robust than a memorised instruction to choose the larger branch.
The main conceptual traps
The first trap is treating a pricing weight as a real-world forecast. The second is treating a rate at a node as if it discounted all future cash flows in one jump. The third is taking a simple average of final payoffs and discounting once, which ignores that later discount rates depend on the intervening state. The fourth is using a tree without checking what prices it was fitted to. Each error removes one reason the tree was introduced in the first place: state dependence and price consistency.
A useful study check is to solve the invented zero-coupon tree twice: once by writing a value at every node, and once by tracing each terminal payoff through its path and verifying the same initial value. If the answers disagree, look first at period timing and the location of each discount rate. Then compare the Omni methodology page with GARP's official programme and study-material descriptions for the wider preparation context. The exam lesson is to name the pricing question before choosing probabilities or pressing a calculator key.
Sources
- FRM Study Materials | GARP · garp.org · Retrieved
- FRM Exam and Certification | GARP · garp.org · Retrieved



