Explosion in the quasi-Gaussian HJM model

Dan Pirjol, Lingjiong ZhuFinance Stochastics 2018 · DOI

TL;DR

This paper investigates the "explosion" phenomenon in quasi-Gaussian Heath-Jarrow-Morton (HJM) models, a popular framework for modeling interest rate curves. The authors rigorously prove that under certain conditions, the short rate in these models can explode to infinity in finite time, with implications for pricing financial derivatives.

What Problem Does This Paper Address?

The Heath-Jarrow-Morton (HJM) model is a cornerstone for modeling the term structure of interest rates. However, its full generality can be computationally challenging. Quasi-Gaussian HJM models offer a simplification by assuming a separable volatility structure, leading to a Markovian representation that aids simulation. Despite their practical appeal, a crucial question arises: do these simplified models exhibit undesirable behaviors, such as the "explosion" of interest rates to infinity? Previous work has shown this phenomenon in log-normal HJM models. This paper extends that investigation to a broader class of quasi-Gaussian HJM models, specifically those with Constant Elasticity of Variance (CEV)-type volatility, and crucially, accounts for the impact of Brownian noise. The goal is to understand the conditions under which these models can lead to exploding short rates and what this means for financial modeling.

Methods

The core of this paper lies in the rigorous mathematical analysis of stochastic differential equations (SDEs) that describe the dynamics of the interest rate models. The authors employ techniques from the theory of stochastic processes, specifically focusing on the conditions for explosion of multidimensional diffusions.

Key methods include:

  • Lyapunov Functions: A central tool is the construction of carefully chosen Lyapunov functions. These functions are used to analyze the behavior of the system and determine whether it can escape to infinity.
  • Khasminskii Criterion and Related Criteria: The paper draws upon established criteria for explosion in multidimensional diffusions, such as those by Khasminskii and others, which often rely on the existence of a Lyapunov function satisfying certain properties.
  • Comparison Arguments: For certain cases, comparison arguments are used to bound the solutions of the SDEs and relate their behavior to simpler, known processes.
  • Analysis of SDEs: The paper involves detailed analytical work on two-dimensional SDEs, examining the behavior of the coefficients and the drift and diffusion terms.

Datasets

Not discussed in the paper. The paper focuses on theoretical analysis and numerical illustrations rather than empirical data analysis.

Key Insights and Ideas

  • Explosion is a Real Risk: The paper provides rigorous mathematical proof that the short rate in quasi-Gaussian HJM models with CEV-type volatility can indeed explode to infinity in finite time, with positive probability, and even with probability one under stronger assumptions. This is not just a theoretical curiosity but a potential pitfall for practical model usage.
  • CEV Volatility Matters: The CEV-type volatility specification, which includes the log-normal model as a special case, is shown to be susceptible to this explosion phenomenon, particularly for exponents $\gamma \in (\frac{1}{2}, 1]$.
  • Brownian Noise is Crucial: Unlike some prior deterministic analyses, this paper explicitly incorporates the impact of Brownian noise, making the results more relevant to realistic stochastic models.
  • Implications for Pricing: The explosion of the short rate directly translates to the collapse of zero-coupon bond prices to zero and the explosion of prices for interest rate derivatives like Eurodollar futures. This highlights a fundamental limitation of these models for long maturities.
  • Model Calibration Challenges: The possibility of explosion implies that model parameters must be chosen carefully, and that the model's applicability might be restricted to shorter time horizons.

Key Results

  • Explosion with Positive Probability: For CEV-type quasi-Gaussian HJM models with volatility exponent $\gamma \in (\frac{1}{2}, 1]$, the short rate explodes in finite time with positive probability under specific conditions on model parameters.
  • Explosion with Probability One: Under additional, stronger assumptions on model parameters, the short rate is shown to explode with probability one.
  • Numerical Illustration: Figure 1 provides a visual demonstration of the explosion phenomenon for the log-normal case ($\gamma=1$) with different parameter values, showing sample paths of the short rate diverging.
  • Conditions for Explosion: The paper derives explicit conditions on the model parameters (volatility $\sigma$, mean-reversion parameter $\beta$, and elasticity parameter $\gamma$) that govern the likelihood and certainty of explosion.

Limitations and Future Work

  • Model Specificity: The results are derived for a specific class of quasi-Gaussian HJM models with CEV-type volatility. Extensions to other volatility specifications or more complex HJM models might be necessary.
  • Practical Avoidance: While the paper highlights the problem, it briefly mentions that capping the short rate volatility can prevent explosion. Further research could explore optimal capping strategies or alternative model modifications.
  • Parameter Space Exploration: The paper provides conditions for explosion. A more extensive numerical study exploring the parameter space and the transition between explosive and non-explosive regimes would be valuable for practitioners.
  • Impact on Other Derivatives: While the paper discusses zero-coupon bonds and Eurodollar futures, a more detailed analysis of the impact on other complex interest rate derivatives would be beneficial.

Summary updated

BibTeX
@article{pirjol2018explosion,
  author = {Dan Pirjol and Lingjiong Zhu},
  title = {{Explosion in the quasi-Gaussian HJM model}},
  journal = {Finance Stochastics},
  year = {2018},
  doi = {10.1007/s00780-018-0367-5},
}
Figure