TL;DR
This paper establishes a rigorous mathematical bridge between financial utility theory and information theory. It proves that the additional profit (logarithmic utility) an insider earns is exactly equal to the "Shannon information" of their superior knowledge, generalized across complex, evolving market conditions.
What Problem Does This Paper Address?
In the world of finance, information is literally money. If you know something the rest of the market doesn't—an "insider" advantage—you can achieve higher expected returns. Mathematically, this is often modeled by giving an insider a larger "filtration" (a more detailed history and future of events) than a regular trader.
While previous research had linked specific types of insider information to utility gains, a general, unified theory was missing. Specifically, how do we quantify the value of information that isn't just a single secret revealed at the start, but a continuous stream of extra data over time? The authors seek to define the "information drift"—the subtle shift in how an insider perceives price movements—and link it directly to the fundamental units of information theory: entropy and mutual information.
Methods
The authors employ a sophisticated toolkit from stochastic calculus and information theory:
- Enlargement of Filtrations: They model the insider's advantage by "enlarging" the standard market information flow ($\mathcal{F}_t$) into a larger flow ($\mathcal{G}_t$).
- Information Drift: They identify a process called "information drift." From a regular trader's perspective, a price might look like a random walk (a martingale). To an insider, that same price has a predictable "drift" based on their extra knowledge.
- Radon-Nikodym Densities: The core proof involves representing the information drift as a density of the insider's conditional probability relative to the market's probability.
- Discretization and Limits: To handle general information flows, they approximate continuous information as a sequence of "discrete bits" revealed over time, then use monotone convergence theorems to prove the relationship holds in the continuous limit.
- Logarithmic Utility: The study focuses on logarithmic utility, a standard measure in finance that represents a trader's satisfaction with their wealth.
Datasets
As a theoretical paper in mathematical finance and probability, no empirical datasets were used. The authors provide analytical proofs and illustrative examples based on standard mathematical models, such as the Wiener process (Brownian motion) and Gaussian random variables.
Key Insights and Ideas
- Utility = Information: The most profound "aha moment" is the confirmation that the expected logarithmic utility increment is identical to the Shannon information difference between the two agents' knowledge horizons.
- The Drift Bridge: The "information drift" acts as the physical bridge between the two worlds. It is the mathematical translation of "knowing more" into "trading better."
- Universality of Entropy: The authors show that this relationship is independent of the specific price dynamics of the assets. Whether the market is volatile or calm, the value of the information remains tied to its entropy.
- Gaussian Maximization: In the final section, they leverage a classic information theory result: Gaussian distributions maximize entropy for a given covariance. This allows them to set upper bounds on how much an insider can possibly gain if their information is "noisy" or perturbed by Gaussian error.
Key Results
- Theorem 5.11: This is the paper's crown jewel. It states that the additional utility ($u_{\mathcal{G}} - u_{\mathcal{F}}$) is equal to the "information difference" ($A(\mathcal{G}^0, \mathcal{F}^0)$).
- Initial Enlargement: For the specific case where an insider gets a "tip" at the very beginning, the utility gain is exactly the Shannon mutual information between that tip and the future state of the market (Theorem 5.13).
- Quantitative Estimates: Using the eigenvalues of covariance matrices, the authors provide concrete formulas (Corollary 6.3) to estimate the maximum utility gain an insider can expect when their information is a noisy version of a market variable.
Limitations and Future Work
- Logarithmic Constraint: The results are specific to logarithmic utility. The authors acknowledge that different utility functions (like power or exponential utility) would likely correspond to different information-theoretic measures, such as $f$-divergences or Renyi entropy.
- Small Trader Assumption: The model assumes "small traders" who do not influence the market price. In reality, large-scale insider trading moves the needle, which would require a much more complex feedback model.
- Standard Borel Spaces: The proofs rely on the existence of regular conditional probabilities, which requires the underlying probability space to be "standard." While this covers most practical applications, it is a theoretical constraint.
- Future Directions: The authors suggest that exploring "progressive enlargements"—where information is revealed at random times (like a surprise earnings call)—is a promising area for further refining these estimates.
Summary updated