How Vasicek and CIR Models Bound Interest Rates

In financial mathematics, modeling the future path of interest rates requires balancing random market fluctuations with realistic economic boundaries. This article provides an overview of how two foundational frameworks—the Vasicek model and the Cox-Ingersoll-Ross (CIR) model—prevent interest rates from exploding to infinity using a mechanism called mean reversion, and how the CIR model specifically modifies the volatility structure of the Vasicek approach to prevent interest rates from turning negative.

Preventing Exploding Rates: Mean Reversion

Unlike stock prices, which can theoretically grow indefinitely over time, interest rates cannot rise to infinity without collapsing the economy. Both the Vasicek and CIR models prevent interest rates from exploding by incorporating a mathematical concept known as mean reversion.

In both models, the change in the interest rate (\(dr_t\)) over a tiny fraction of time (\(dt\)) is governed by two parts: a deterministic "drift" term and a random "volatility" term. The drift term is identical in both models and is written as:

\[a(b - r_t)dt\]

  • \(r_t\) is the current interest rate.
  • \(b\) is the long-term mean (average) interest rate.
  • \(a\) is the speed of mean reversion (a positive constant).

This drift term acts like a gravitational pull toward the long-term average, \(b\).

If the current interest rate \(r_t\) shoots up and becomes much higher than the long-term average \(b\), the term \((b - r_t)\) becomes negative. Consequently, the drift pulls the interest rate back down. Conversely, if the current rate falls below \(b\), the term \((b - r_t)\) becomes positive, pushing the rate back up. The parameter \(a\) dictates how quickly and forcefully the model corrects these deviations. Because of this elastic tether to the average, the rate is mathematically prevented from exploding to infinity.

The Vasicek Model and the Negative Rate Problem

While the Vasicek model successfully stops rates from exploding, it fails to prevent them from becoming negative. The full Vasicek equation is:

\[dr_t = a(b - r_t)dt + \sigma dW_t\]

Here, \(\sigma\) represents constant volatility, and \(dW_t\) represents random market shocks (Brownian motion).

Because the volatility \(\sigma\) is constant, the size of the random market shock does not care where the current interest rate sits. If the interest rate \(r_t\) is hovering just above zero, a large negative random shock (\(dW_t\)) can easily overpower the upward pull of the mean reversion drift. This pushes the interest rate into negative territory. While negative rates have occurred in modern central banking, they were long considered a theoretical flaw in the standard Vasicek model.

Preventing Negative Rates: The CIR Model's Solution

To fix the negative rate flaw of the Vasicek model, John Cox, Jonathan Ingersoll, and Stephen Ross introduced the CIR model. They kept the exact same mean-reverting drift to prevent exploding rates, but altered the volatility term by introducing the square root of the current interest rate:

\[dr_t = a(b - r_t)dt + \sigma \sqrt{r_t} dW_t\]

This addition of \(\sqrt{r_t}\) makes the volatility conditional on the current level of the interest rate, which directly prevents negative rates through the following mechanics:

  1. Dampening Volatility Near Zero: As the interest rate \(r_t\) drops and gets closer to zero, \(\sqrt{r_t}\) also gets closer to zero. Because this term is multiplied by the random shock (\(\sigma dW_t\)), the overall randomness of the model shrinks as rates fall.
  2. Shutting Off Randomness at Zero: If the interest rate actually hits zero (\(r_t = 0\)), the volatility term \(\sigma \sqrt{0} dW_t\) equals zero. At this exact point, the random shock is completely turned off, making it impossible for a negative shock to push the rate below zero.
  3. The Upward Bounce: When the rate is at zero and volatility is turned off, the only thing left in the equation is the drift term: \(a(b - 0)dt\), which simplifies to \(ab \cdot dt\). Since both \(a\) and \(b\) are positive numbers, the drift is strictly positive. The interest rate is immediately forced to bounce back up into positive territory.

The Feller Condition

In the CIR model, to guarantee that the interest rate never even touches zero, mathematicians apply a rule known as the Feller condition, expressed as:

\[2ab \ge \sigma^2\]

This inequality ensures that the upward pull of the mean reversion (\(a\) and \(b\)) is strong enough relative to the maximum possible variance (\(\sigma^2\)) that the rate will always be repelled upward before it can actually reach the zero boundary.