When designing continuous-time financial models for interest rates or asset volatility, mathematical convenience often guides the choice of stochastic differential equations (SDEs). Models like the Heath-Jarrow-Morton (HJM) framework or the SABR volatility model are widely celebrated because they can capture complex market behaviors, such as the volatility smile.
However, a critical hazard lurks within the feedback loops of these mathematical formulations: finite-time explosion. This phenomenon occurs when a modeled financial variable—such as a short-term interest rate or an asset's volatility—escapes to infinity in a finite amount of time with non-zero probability. When a model permits an explosion, the financial consequences are catastrophic, yielding absurd pricing results like infinite call option prices or bonds that collapse to a value of zero.
The Intuition Behind Mathematical Explosion
To understand why a financial model explodes, it is helpful to look at a simple deterministic analogy. Consider the ordinary differential equation:
$$\frac{dx}{dt} = x^2, \quad x(0) = x_0 > 0$$
By separating variables and integrating, we find the solution:
$$x(t) = \frac{1}{\frac{1}{x_0} - t}$$
As $t$ approaches $1/x_0$, the denominator shrinks to zero, and $x(t)$ flies off to infinity. The growth rate of $x$ is superlinear ($\propto x^2$), meaning that as $x$ gets larger, it accelerates even faster. It moves so quickly that it reaches infinity in a finite window of time.
In stochastic models, a similar mechanism can be triggered by either the drift coefficient or the diffusion (volatility) coefficient. If the variance of a process grows too rapidly as a function of the state variable itself, the random trajectories can be driven to infinity before reaching the contract's maturity date.
Instability in Interest Rate Frameworks
In interest rate modeling, the quasi-Gaussian HJM model is a popular choice for capturing the dynamics of the yield curve. However, certain volatility specifications within this framework introduce severe instabilities.
In Explosion in the quasi-Gaussian HJM model, Dan Pirjol investigates what happens when these models utilize Constant Elasticity of Variance (CEV)-type volatility specifications. The author rigorously proves that under specific parameter regimes, the short-term interest rate can explode to infinity in finite time. When the short rate explodes, the discount factor $e^{-\int_0^t r_s ds}$ collapses to zero, causing theoretical bond prices to drop to zero and long-term derivative prices to grow without bound.
This vulnerability is not limited to exotic CEV specifications. In a companion paper, Small-noise limit of the quasi-Gaussian log-normal HJM model, Dan Pirjol evaluates the popular log-normal variant of the one-factor quasi-Gaussian HJM model. By analyzing the model under a "small-noise" limit, the author demonstrates that interest rates can explode to infinity in finite time with a probability of one.
The physical consequence of this log-normal explosion is that standard financial instruments, such as Eurodollar futures, attain infinite prices within the model. This reveals that the potential for unbounded prices is an inherent, structural flaw of log-normal volatility specifications within this class of HJM models, severely limiting their valid application to long-dated contracts.
Volatility Blow-Up in the SABR Model
The phenomenon of finite-time explosion is not unique to interest rate models; it also compromises stochastic volatility models when applied to underlying volatility indices like the VIX.
The standard SABR model describes an asset $X_t$ and its volatility $\alpha_t$ using coupled SDEs:
$$dX_t = \alpha_t X_t^\beta dW_{1,t}$$ $$d\alpha_t = \nu \alpha_t dW_{2,t}$$
where $W_1$ and $W_2$ have a correlation $\rho$.
In VIX options in the SABR model, Dan Pirjol uncovers a fundamental limitation when using this framework to price VIX derivatives. When the model parameters are configured such that $0 < \beta < 1$ and the correlation is negative ($\rho < 0$), the mathematical formulation forces the volatility process into a regime where it explodes in finite time.
Intuitively, the negative correlation means that as the asset price drops, the volatility spikes. Because the CEV exponent $\beta$ alters the feedback loop between the asset price and its volatility, the acceleration becomes unmanageable. The volatility process blows up, which mathematically translates into infinite prices for VIX futures and VIX call options, alongside zero prices for VIX put options. This pathology renders the standard SABR model completely unusable for these parameter combinations.
Mitigating the Explosion Risk
Recognizing the conditions that cause finite-time explosion allows researchers and practitioners to develop targeted remedies. Two primary strategies exist to handle these model limitations:
- Horizon Restrictions: As shown in the HJM studies, practitioners must restrict the model's application strictly to shorter time horizons, ensuring that the contract maturity occurs well before the probability of explosion becomes significant.
- Volatility Capping: To restore the practical utility of these models over longer horizons, the underlying SDEs can be modified. For instance, in VIX options in the SABR model, the author proposes a "capped volatility" modification. By placing an artificial ceiling on the volatility coefficient, the feedback loop is broken, preventing the trajectory from escaping to infinity. This capped framework eliminates infinite pricing anomalies and allows for the valid derivation of short-maturity asymptotics.
Key Foundational Work
Explosion in the quasi-Gaussian HJM model (2018) — contributed 8 concept elements to this direction.
Small-noise limit of the quasi-Gaussian log-normal HJM model (2017) — contributed 5 concept elements to this direction.