TL;DR
This paper investigates how the price of an option behaves when its expiration date is very close (small-time asymptotics). The authors discover that the leading term in this price behavior, particularly for "at-the-money" options, is primarily determined by the initial volatility of the underlying stock price and whether the stock's price movements are "smooth" (finite variation) or "jumpy" (infinite variation). This work provides a deeper understanding of option pricing in complex financial models.
What Problem Does This Paper Address?
In financial mathematics, option pricing models often rely on complex stochastic processes to describe the evolution of underlying asset prices. A key challenge is understanding how these models behave in extreme scenarios, such as when an option is about to expire. Specifically, this paper focuses on the "small-time asymptotics" of option prices, which describes how the price of an option approaches its intrinsic value as the time to expiration shrinks to zero. This is crucial for calibrating financial models to market data and for understanding the sensitivity of option prices to various parameters. The authors aim to provide a unified framework for analyzing these asymptotics for a broad class of martingales, going beyond simpler models.
Methods
The core of the paper's methodology involves approximating a general stock price process $S$ with a simpler Lévy process $Z$. This approximation allows the authors to leverage the well-developed theory of Lévy processes. The key steps include:
- Approximation by Lévy Processes: The authors show that a general martingale $S$ can be approximated by a Lévy process $Z$ (Equation 1.3) with a small error, particularly for small times. This approximation is achieved by "freezing" the coefficients of the stochastic differential equation defining $S$ at their initial values.
- Analysis of Lévy Processes: The bulk of the work then focuses on analyzing the small-time asymptotics of option prices for Lévy processes. This involves detailed calculations and the use of probabilistic inequalities, such as the Burkholder-Davis-Gundy inequality.
- Decomposition and Case Analysis: The analysis is broken down into different cases based on the properties of the Lévy process, specifically whether it has a continuous component (driven by Brownian motion) or is purely jump-driven, and whether it has finite or infinite variation.
- Asymptotic Expansions: The authors derive asymptotic expansions for the option prices, focusing on the leading order terms. This involves identifying the dominant factors influencing the price as time approaches zero.
Datasets
Not discussed in the paper.
Key Insights and Ideas
- Initial Volatility Dominance: For options with a continuous component in the underlying asset's price process, the leading term in the small-time asymptotics is solely determined by the initial volatility ($\sigma_0$) and is proportional to $\sqrt{T}$, regardless of any jumps.
- Finite Variation Matters: If the stock price process has finite variation (meaning its price movements are "smooth" and don't involve sudden jumps), the leading term in the option price asymptotics is linear in time ($O(T)$). This is a significantly faster convergence rate than for processes with infinite variation.
- Infinite Variation Spectrum: For processes with infinite variation, the convergence rate lies between $\sqrt{T}$ and $T$. The exact form depends on the nature of the small jumps, with stable-like jumps leading to rates of $T^{1/\alpha}$ or $T|\log T|$.
- Approximation is Key: The strategy of approximating general martingales with Lévy processes is a powerful tool that simplifies the analysis considerably, allowing for a unified treatment of various models.
- Absolute Moment Connection: The small-time asymptotics of at-the-money call option prices are directly related to the first centered absolute moment of the stock price process.
Key Results
The paper establishes precise formulas for the leading order term of at-the-money call option prices as time $T \to 0$.
- Continuous Component Present ($\sigma_0 \neq 0$): The leading term is $|\sigma_0|/\sqrt{2\pi} \sqrt{T}$.
- Pure Jump, Finite Variation: The leading term is $\frac{1}{2}CT + o(T)$, where $C = \int_{\mathbb{R}} |\kappa_0(x)| F(dx) + |\int_{\mathbb{R}} \kappa_0(x) F(dx)|$.
- Pure Jump, Infinite Variation (Stable-like small jumps):
- For $\alpha \in (1,2)$, the leading term is $\frac{1}{2}C(\alpha, \alpha, f_+, f_-) T^{1/\alpha} + o(T^{1/\alpha})$.
- For $\alpha = 1$ and $f_+ = f_-$, the leading term is $\frac{1}{2}(f_+ + f_-) T|\log T| + o(T|\log T|)$.
The paper also shows that the implied volatility converges to the initial spot volatility of the continuous martingale part of the log price.
Limitations and Future Work
- Assumptions on Lévy Measure: The analysis for infinite variation processes relies on specific assumptions about the "small jumps" resembling those of an $\alpha$-stable Lévy process. Extending these results to more general jump structures might be challenging.
- Model Specificity: While the framework is general, obtaining explicit constants for the leading terms requires knowledge of the specific parameters of the Lévy measure (e.g., $f_+, f_-, \alpha$).
- Future Work: The authors suggest exploring the case where the stock price process is not a martingale, which is a common scenario in practice. Further investigation into more general jump processes and their impact on option pricing asymptotics is also a natural direction. The paper also hints at the possibility of extending the results to higher-order terms in the asymptotic expansion.
Summary updated