Utility theory front to back - inferring utility from agents' choices

Alexander M. G. Cox, David Hobson, Jan Obłój

Abstract

We pursue an inverse approach to utility theory and consumption-investment problems. Rather than specifying a utility function to derive optimal actions, we observe an agent's consumption and investment strategies to determine if a utility function exists that renders this behavior optimal. In a deterministic setting, we find that infinitely many utility functions can generate a given consumption pattern. However, in a stochastic Black-Scholes complete market, consumption and investment strategies must satisfy a specific consistency condition—a partial differential equation (PDE)—to be consistent with classical utility maximization. Furthermore, we demonstrate that an agent's risk attitudes, such as Decreasing Absolute Risk Aversion (DARA), and their beliefs about market volatility can be deduced directly from their observed choices.

Introduction

In classical financial economics, the standard paradigm—often exemplified by the Merton problem—begins with a predefined utility function representing an agent's risk appetite and time preference. From this mathematical abstraction, researchers derive optimal consumption and investment trajectories. However, this "forward" approach faces a significant empirical hurdle: an agent's internal utility is rarely observable, and self-reported risk preferences are often unreliable.

In "Utility theory front to back – inferring utility from agents' choices," Cox, Hobson, and Obłój propose an inverse approach. By treating consumption ($c$) and investment ($\pi$) as the primary observables, they investigate the conditions under which these behaviors are "rational"—meaning they can be mapped back to a well-behaved utility function.

Methodology and Theoretical Framework

The authors analyze the problem across two distinct environments in continuous time:

1. The Deterministic Setting

In a risk-free world where an agent decides how to consume wealth over an infinite horizon, the problem is under-specified. The authors demonstrate that for any given consumption strategy, there exists an infinite family of utility functions that could have produced it. Without the "stress test" of market volatility, the agent's choices do not provide enough information to uniquely identify their underlying preferences.

2. The Stochastic Setting (Black-Scholes Market)

The complexity increases in a stochastic market where the agent allocates wealth between a risk-free bond and a risky asset following geometric Brownian motion. Here, the agent must balance immediate consumption against the risk of future wealth fluctuations. This environment imposes a much stricter structure on behavior.

The core of the analysis involves deriving a consistency condition known as Black’s PDE (first suggested by Fischer Black in 1968). Using both primal (martingale) and dual (state-price density) methods, the authors establish a rigorous link between the observed strategies $(c, \pi)$ and the latent utility $u$.

Key Results: Theorem 3.4 and Black's PDE

The paper’s central contribution is Theorem 3.4, which provides an "if and only if" condition for optimality. For a pair of strategies to be consistent with the maximization of a standard von Neumann-Morgenstern utility function, they must satisfy the integrated form of Black’s PDE.

If the observed strategies do not satisfy this relationship, the agent is not acting as a classical utility maximizer. This makes the PDE a powerful diagnostic tool for economic rationality. This result provides a unified, mathematically rigorous treatment of a problem that was previously only explored in informal or finite-horizon contexts.

Decoding the Investor’s Mind

Beyond verifying rationality, the inverse approach allows for the extraction of specific psychological and cognitive traits:

  • Risk Aversion Profiles: The authors show that characteristics like Decreasing Absolute Risk Aversion (DARA) or Decreasing Relative Risk Aversion (DRRA) are encoded in the slopes of the consumption and investment functions. For instance, DARA can be identified if the investment-to-consumption ratio satisfies specific inequalities relative to wealth derivatives.
  • Market Beliefs: A striking finding is that an agent’s actions reveal their subjective belief regarding market volatility ($\sigma$). Even if the observer does not know the agent's utility, the agent's optimal response to risk "leaks" their perception of that risk. Interestingly, the Sharpe ratio remains hidden, as the agent's utility can be adjusted to compensate for different beliefs about expected returns.
  • Time Discounting: The paper proves that linear strategies (where an agent consumes and invests a fixed percentage of wealth) necessarily imply exponential time discounting.

Discussion and Limitations

While the paper provides a robust mathematical foundation for the inverse utility problem, it operates under the assumption of a complete market. In the Black-Scholes model, all risks can be perfectly hedged. In incomplete markets—where stochastic volatility or jumps exist—the mapping between actions and utility becomes significantly more complex.

Additionally, the results assume "regular" utility functions (smooth and strictly concave). Future research may explore how "behavioral" traits, such as loss aversion or satiation points, manifest in the inverse problem, or how the theory adapts to "Knightian uncertainty" where the agent is unsure of the underlying market dynamics.

Conclusion

By flipping the perspective on utility theory, Cox, Hobson, and Obłój transform consumption and investment data from mere outcomes into diagnostic tools. This "front to back" approach provides a rigorous framework for revealed preference theory, allowing economists to decode the risk preferences and market beliefs of agents simply by observing their financial footprints.

Summary updated

Abstract

We pursue an inverse approach to utility theory and consumption & investment problems. Instead of specifying an agent's utility function and deriving her actions, we assume we observe her actions (i.e. her consumption and investment strategies) and ask if it is possible to derive a utility function for which the observed behaviour is optimal. We work in continuous time both in a deterministic and stochastic setting. In the deterministic setup, we find that there are infinitely many utility functions generating a given consumption pattern. In the stochastic setting of the Black-Scholes complete market it turns out that the consumption and investment strategies have to satisfy a consistency condition (PDE) if they are to come from a classical utility maximisation problem. We show further that important characteristics of the agent such as her attitude towards risk (e.g. DARA) can be deduced directly from her consumption/investment choices.
BibTeX
@article{cox2014utility,
  author = {Alexander M. G. Cox and David Hobson and Jan Obłój},
  title = {{Utility theory front to back - inferring utility from agents' choices}},
  journal = {International Journal of Theoretical and Applied Finance},
  year = {2014},
  doi = {10.1142/s0219024914500186},
}
Figure