Revenue Maximization with an Uncertainty-Averse Buyer

Shuchi Chawla, Kira Goldner, J. Benjamin Miller, Emmanouil Pountourakis

TL;DR

This paper explores how to maximize seller revenue when buyers have "uncertainty aversion," a behavioral trait where they undervalue uncertain outcomes beyond what standard economic models predict. The authors show that while optimal mechanisms can be complex, simple strategies like posted prices can still achieve good revenue, especially when buyer risk aversion is bounded. However, in dynamic settings, this uncertainty aversion can make robust revenue extraction challenging.

What Problem Does This Paper Address?

Traditional economic models often assume buyers are "risk-neutral" or follow "expected utility theory" (EUT). However, real-world behavior, as highlighted by phenomena like the Allais paradox, deviates from EUT. Prospect Theory offers a more nuanced model, suggesting that agents not only have a non-linear utility for money but also fundamentally misvalue random events – they are "uncertainty-averse." This paper tackles the crucial question of how mechanism design, specifically revenue maximization, changes when we incorporate this more realistic, non-EUT buyer behavior. The core challenge is understanding if and how sellers can design mechanisms that effectively extract value from buyers who discount uncertain outcomes, and whether these mechanisms can be robust to variations in the buyer's specific risk preferences.

Methods

The authors adapt the framework of mechanism design to incorporate a specific model of uncertainty aversion derived from Prospect Theory. Instead of standard expected utility, they use a "weighted expected utility" where probabilities are transformed by a "weighting function" y(x). This function captures how much an agent discounts an event with probability x. They focus on convex y(x) to model risk aversion.

Key methodological contributions include:

  • Characterizing Optimal Mechanisms: For single-item sales, they show that optimal mechanisms are composed of binary lotteries (offering the item with a certain probability at a certain price).
  • Risk-Robustness: A significant focus is on designing mechanisms that work well across a range of possible buyer weighting functions, not just a single, precisely known one. This involves analyzing how revenue guarantees change when the buyer's risk attitude is bounded.
  • Dynamic Settings: They extend the analysis to two-stage sales, where a buyer purchases two items sequentially. Here, the challenge is how the seller can commit to future prices and how buyer uncertainty about future values interacts with their risk aversion.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Uncertainty Aversion is Different: Unlike standard risk aversion (which can be captured by a concave utility function), uncertainty aversion (modeled by a convex weighting function) fundamentally changes how buyers evaluate lotteries. This can lead to sellers extracting more revenue than in the risk-neutral case by exploiting this aversion.
  • Binary Lotteries are Key: For single-item sales, optimal mechanisms boil down to offering menus of simple binary lotteries. This simplifies the structure of optimal mechanisms considerably.
  • Posted Prices Can Be Surprisingly Robust: Even with uncertainty aversion, simple posted-price mechanisms can achieve a constant-factor approximation of optimal revenue, provided the buyer's risk aversion is "bounded." This is a powerful result for practical mechanism design.
  • Dynamic Settings are Tricky: While single-item sales can be handled robustly, the dynamic, multi-item setting presents a major challenge. The paper shows that it's impossible to guarantee a constant-factor risk-robust approximation in these dynamic scenarios, even with bounded risk aversion. This implies that precise knowledge of buyer preferences becomes critical for maximizing revenue in sequential sales.

Key Results

  • Single-Shot Sales: Optimal mechanisms are menus of binary lotteries. Myerson's posted-price mechanism achieves a constant-fraction approximation of optimal revenue under a "β-bounded" risk aversion assumption.
  • Dynamic Sales: A 2-approximation to the revenue bound is achieved using posted-price mechanisms. However, a key result (Theorem 7) demonstrates that no posted-price mechanism can achieve a constant-factor risk-robust approximation in the two-stage setting, even with bounded risk aversion.

Limitations and Future Work

  • Unrealistic Extreme Aversion: Theorem 2, which shows near-full revenue extraction, requires extreme risk aversion that is likely unrealistic. The paper addresses this by focusing on bounded risk aversion for robustness.
  • Computational Complexity: While Theorem 3 provides an existential result for a risk-robust approximation in dynamic settings, constructing such a mechanism is not computationally efficient.
  • Future Work: The paper suggests exploring more general classes of buyer behavior beyond the specific Prospect Theory model used, as well as investigating mechanisms that are robust to uncertainty about the buyer's value distribution itself, not just their risk attitude. The challenges in dynamic settings also point to a need for new approaches to robust mechanism design in sequential environments.

Summary updated

Abstract

Most work in mechanism design assumes that buyers are risk neutral; some considers risk aversion arising due to a non-linear utility for money. Yet behavioral studies have established that real agents exhibit risk attitudes which cannot be captured by any expected utility model. We initiate the study of revenue-optimal mechanisms under buyer behavioral models beyond expected utility theory. We adopt a model from prospect theory which arose to explain these discrepancies and incorporates agents under-weighting uncertain outcomes. In our model, an event occurring with probability $x < 1$ is worth strictly less to the agent than $x$ times the value of the event when it occurs with certainty. In contrast to the risk-neutral setting, the optimal mechanism may be randomized and appears challenging to find, even for a single buyer and a single item for sale. Nevertheless, we give a characterization of the optimal mechanism which enables positive approximation results. In particular, we show that under a reasonable bounded-risk-aversion assumption, posted pricing obtains a constant approximation. Notably, this result is "risk-robust" in that it does not depend on the details of the buyer's risk attitude. Finally, we examine a dynamic setting in which the buyer is uncertain about his future value. In contrast to positive results for a risk-neutral buyer, we show that the buyer's risk aversion may prevent the seller from approximating the optimal revenue in a risk-robust manner.
BibTeX
@inproceedings{chawla2018revenue,
  author = {Shuchi Chawla and Kira Goldner and J. Benjamin Miller and Emmanouil Pountourakis},
  title = {{Revenue Maximization with an Uncertainty-Averse Buyer}},
  booktitle = {SODA},
  year = {2018},
  url = {https://openreview.net/forum?id=Z9Ixzq8Gel},
}
Figure