Abstract
The background for the general mathematical link between utility and information theory investigated in this paper is a simple financial market model with two kinds of small traders: less informed traders and insiders, whose extra information is represented by an enlargement of the other agents' filtration. The expected logarithmic utility increment—the difference between the insider's and the less informed trader's expected logarithmic utility—is described in terms of the information drift. This is the drift one must eliminate to perceive price dynamics as a martingale from the insider's perspective. We describe the information drift in a general setting using natural quantities expressing the probabilistic better-informed view of the world, allowing us to identify the additional utility through entropy-related quantities known from information theory. In particular, in a complete market where the insider has fixed additional information, the utility increment is represented by the Shannon information of the extra knowledge.
Introduction
In financial mathematics, the value of information is often quantified by the advantage it provides an agent in maximizing their expected utility. While it has long been understood intuitively that "more information leads to better returns," providing a rigorous, universal bridge between financial utility and information theory has remained a complex challenge. This paper addresses this by examining the additional logarithmic utility gained by an insider—an agent with access to a larger filtration (information set) than the general market.
The central question is: can we quantify the "price" of knowledge using the fundamental units of information theory, such as entropy and Shannon information?
Methodology and Mathematical Framework
Enlargement of Filtrations
The study models the market using two distinct flows of information. The standard market information is represented by the filtration $\mathcal{F} = (\mathcal{F}_t)_{t \in [0,T]}$. The insider possesses a larger filtration, $\mathcal{G} = (\mathcal{G}_t)_{t \in [0,T]}$, such that $\mathcal{F}_t \subseteq \mathcal{G}_t$. This "enlargement of filtrations" allows the insider to observe events or price trajectories with greater precision or foresight.
Information Drift
A key concept introduced is the information drift. To a standard trader, a price process might appear as a martingale (a fair game with no predictable direction). However, to an insider with superior knowledge, that same process contains a predictable component—the drift. The authors identify this drift as the mathematical translation of "knowing more" into "trading better." By identifying the drift that must be removed for the insider to see the process as a martingale, the researchers can isolate the specific advantage provided by the extra information.
Logarithmic Utility
The research focuses on logarithmic utility, $U(x) = \ln(x)$. In finance, this utility function is significant because it relates to the growth rate of wealth and is additive over time. The "utility increment" is defined as the difference between the maximum expected logarithmic utility of the insider ($u_{\mathcal{G}}$) and that of the less-informed trader ($u_{\mathcal{F}}$).
Key Insights: The Bridge to Information Theory
The paper's most significant contribution is the proof that the financial advantage of an insider is exactly equal to the information-theoretic measure of their knowledge.
Utility as Shannon Information
The authors demonstrate that the expected logarithmic utility increment is identical to the Shannon information difference between the two agents' knowledge horizons. This relationship holds regardless of the specific price dynamics of the assets, suggesting a universal law: the value of information is tied to its entropy.
Theorem 5.11 and 5.13
- Theorem 5.11: This core result establishes 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 (Theorem 5.13): In scenarios where an insider receives a "tip" at the start of the trading interval, the utility gain is shown to be exactly the Shannon mutual information between that initial tip and the future state of the market.
Gaussian Maximization
By applying the principle that Gaussian distributions maximize entropy for a given covariance, the authors provide upper bounds for utility gains. Using the eigenvalues of covariance matrices, they derive concrete formulas (Corollary 6.3) to estimate the maximum utility an insider can expect when their information is "noisy" or perturbed by Gaussian error.
Discussion and Limitations
While the paper provides a robust link between finance and information theory, several constraints apply:
- Utility Function: The results are specific to logarithmic utility. Other functions (e.g., power utility) would likely correspond to different information measures, such as Renyi entropy.
- Small Trader Assumption: The model assumes agents do not influence market prices. In reality, significant insider trading can move prices, creating a feedback loop not captured here.
- Mathematical Constraints: The proofs require the probability space to be "standard" (Standard Borel Spaces) to ensure the existence of regular conditional probabilities.
Conclusion
This research provides a rigorous foundation for the "value of information" in financial markets. By proving that the additional utility of an insider is equivalent to the Shannon information of their filtration, the authors unify two previously distinct fields. This framework allows for quantitative estimates of the maximum possible gain from superior information, providing a theoretical limit to the profitability of insider knowledge.
Summary updated