What it means
Interest-rate models translate uncertain future rates into scenarios that can be used for valuation and risk analysis. CIR specifies one random factor, the instantaneous short rate, and longer-term bond prices are then linked to the expected path of that rate under the chosen valuation framework.
A one-factor model compresses many real-world forces into a single state variable, and because central-bank policy, inflation, credit spreads and demand for safe assets can all move yields, CIR is a controlled lens for a specified question, not a complete macroeconomic forecast. Mean reversion means a rate above its assumed long-run level tends to drift downward, while a rate below tends upward, and the speed parameter controls how strongly it is pulled back.
This tendency can be overwhelmed temporarily by random moves. Consider a rate of 2% and an assumed long-run level of 4%: the drift term points upward, but the next observed rate could still fall because of a negative random shock, so mean reversion describes conditional tendency, not a one-way path.
The square-root term gives a distinctive volatility pattern: when the current rate is low, shocks become smaller in absolute size, and when it is higher, the same volatility parameter can produce larger absolute moves. That helps keep the standard process from crossing below zero.
New York University Stern teaching material presents CIR as a mean-reverting square-root process and contrasts it with the constant-volatility Vasicek model, which can generate negative model rates while standard CIR keeps the short rate nonnegative. Nonnegative model rates were appealing in many historical settings, but real markets have experienced negative yields and policy rates, so a standard CIR model may fit some environments poorly.
One should not force the model onto a period its central assumption cannot represent. Parameters are estimated from data or calibrated to market prices, and the long-run level, reversion speed and volatility can change the value of a bond or derivative, so a neat output can be misleading if estimates are unstable or the curve used for calibration is stale.
In pricing a zero-coupon bond, future cash paid at maturity is discounted using the model's rate dynamics. Different maturities respond differently to the assumed process, and a mismatch between the model-implied and observed yield curve can reveal calibration problems.
Scenario analysis should ask what happens when estimated parameters are wrong, so a risk manager may compare CIR outputs with market-implied curves and alternative models, and stress periods when rates jump or central-bank policy shifts abruptly. When a report quotes a CIR price, request the observation date, input curve, parameters and instrument terms.
Those inputs are necessary to interpret the model value and reproduce a result.
In practice
Real-world examples.
Example
An analyst models a short rate at 2% with an assumed 4% long-run level. CIR gives upward drift on average but still permits a short-term downward realisation, so the analyst presents a range of outcomes rather than one path.
Example
A bank values a five-year interest-rate option under several reversion speeds. The price changes materially, so management sees a range rather than one unquestioned model number, and the risk team records which speed was used in the official valuation.
Example
A historical period includes negative market yields. A team flags that standard CIR cannot generate a negative modelled short rate and compares an alternative framework before using any output in a report.
Formula
Calculation
The standard continuous-time form is dr = k(theta - r)dt + sigma sqrt(r)dW. Here r is the short rate, theta the long-run level, k the reversion speed, sigma the volatility scale, and dW a random shock. This is a model equation, not an instruction to calculate a future rate by simple arithmetic.
Worked illustration of the two parts, using a rough one-year step. Let r = 2%, theta = 4%, k = 0.5 and sigma = 0.1.
- Drift: k(theta - r) = 0.5 x (0.04 - 0.02) = 0.5 x 0.02 = 0.01, so the expected pull is upward by 1 percentage point, from 2% towards 3%, before any random shock.
- Shock size at r = 2%: sigma x sqrt(r) = 0.1 x sqrt(0.02) = 0.1 x 0.1414 = about 0.0141, or roughly 1.4 percentage points.
- Shock size at r = 8%: sigma x sqrt(r) = 0.1 x sqrt(0.08) = 0.1 x 0.2828 = about 0.0283, or roughly 2.8 percentage points.
The same sigma therefore produces larger moves when the rate is higher, which is the square-root effect. The 1 percentage point drift is only an average pull; a random shock of the size above could easily outweigh it in any single year.Case study
Seen in the real world.
Fictional case: A treasury group uses CIR to assess the value of a callable bond. It calibrates inputs to a dated market curve and models the exercise feature. The first output suggests a gain from refinancing, but sensitivity tests show the result disappears when volatility rises. The team documents the model assumptions, compares another rate model and avoids presenting the theoretical price as an executable market offer. It also checks whether observed negative rates in a reference market make standard CIR inappropriate for that use.
Watch out
Common mistakes.
- Calling mean reversion a guarantee that rates will move toward the assumed average next month.
- Applying a nonnegative-rate model to negative-rate observations without explaining the mismatch.
- Treating a single calibrated model value as an executable market price.
Questions
People also ask.
Why does CIR use a square root of the rate?
It makes shock size depend on the current rate and supports a nonnegative short-rate process under standard assumptions.
Can it predict the exact next policy rate?
No. It generates a distribution under assumptions rather than an exact forecast.
What changes model results most?
The input curve, reversion level and speed, volatility, and the security's terms can all matter.
From the founder's library
Take it further with the book.
Build your financial confidence beyond this definition. Shihan's full-length guide, Accounting Fundamentals, takes the same plain-English approach and turns it into a complete, practical playbook for non-finance managers, business owners and students - with chapter-end quiz answers and presentation slides included.
25% off with code MMHQ25, applied at checkout. Priced in USD - checkout may show the equivalent in your local currency.
View the book and save 25%Related