Abstract
We study revenue variance in the sale of $k$ homogeneous items to risk-neutral, unit-demand bidders with independent private values. Although the Revenue Equivalence Theorem implies that standard auctions generate the same expected revenue, the distribution of revenue differs across mechanisms. Prior work shows that, in single-item environments with ex-post individual rationality (IR), the first-price auction minimizes revenue variance. We show that this result is fragile. Under interim IR, the optimality of the first-price auction breaks down in asymmetric single-item settings, and we characterize the variance-minimizing mechanisms for any implementable allocation rule in this environment. In multi-item symmetric regular environments with interim IR, we construct a mechanism that implements the efficient allocation and guarantees constant revenue while maintaining non-negative payments. Under ex-post IR, we show that revenue variance can be reduced relative to winner-pays-bid formats by introducing negative correlations in payments. Nevertheless, we show that the variance ranking between the winner-pays-bid auction and the uniform $(k+1)$-st price auction is maintained in multi-unit settings.
Introduction: The Problem of Revenue Volatility
In auction theory, the Revenue Equivalence Theorem is a cornerstone: it suggests that as long as bidders have independent private values, the seller's average take is the same whether they use a first-price, second-price, or all-pay auction. However, real-world sellers—like governments funding public works or platforms managing ad slots—care deeply about revenue variance. High volatility creates budget uncertainty and financial risk.
As illustrated in the figure above, while the mean revenue remains constant at 1/3 across different formats, the "spread" or density of outcomes varies significantly. Previous research established that for selling a single item where losers pay nothing (ex-post individual rationality), the first-price auction is the "gold standard" for minimizing variance. This paper investigates whether this optimality holds when participation constraints are relaxed or when selling multiple items.
Methodology
The authors employ a mix of rigorous mathematical proofs and numerical simulations using synthetic distributions (e.g., Uniform $[0, 1]$ and asymmetric power-law distributions) to explore variance-minimizing mechanisms:
- Revenue Decomposition: A novel technique to break down the revenue function into two parts: one tied to a specific bidder’s type and another that satisfies a "complementary slackness" property.
- Fixed-Point Theory: To prove that variance-minimizing payment rules exist in complex asymmetric settings, the authors utilize the Schauder fixed-point theorem on the space of probability measures.
- Differential Equations: For multi-item auctions, the optimal payment rule is characterized as the solution to a first-order linear ordinary differential equation.
- Convex Optimization: In cases where analytical solutions are elusive, quadratic programming is used to numerically find optimal payment functions for multi-item ex-post IR settings.
Key Insights: The Fragility of Traditional Optimality
The core finding is that the "first-price is best" rule is surprisingly fragile. It breaks down the moment bidders are asymmetric (i.e., they don't draw their values from the same distribution). In these cases, a novel auction format that spreads payments across more states of the world is superior.
Furthermore, the paper highlights Interim Individual Rationality (IR) as a powerful risk management tool. By requiring bidders to be happy on average rather than in every possible outcome (ex-post IR), the seller gains the flexibility to "smooth" revenue. This is realistic in settings like government procurement or contests where participation itself has a cost.
In multi-item auctions, the seller can reduce variance by introducing negative correlations between payments. For example, making the highest bidder pay more when the second-highest bidder pays less can stabilize the total sum.
Key Results
- Asymmetric Single-Item: The authors characterize a new class of variance-minimizing mechanisms for asymmetric bidders, proving they always exist and outperform first-price auctions.
- The Multi-Item "Holy Grail": For $k$ items and $n$ bidders, they construct a specific payment rule that yields zero revenue variance (constant revenue) while maintaining non-negative payments and interim IR.
- Ranking Persistence: Even though the winner-pays-bid (discriminatory) auction isn't perfectly optimal in multi-unit settings, it still results in lower revenue variance than the uniform $(k+1)$-st price auction.
- Numerical Validation: For a 3-bidder, 2-item uniform distribution, they show the second moment of revenue can be reduced from $\approx 0.264$ (first-price) to $\approx 0.255$ via their optimized payment rule.
Discussion and Future Work
While the zero-variance mechanism is a significant theoretical breakthrough, its reliance on interim IR assumes bidders won't "rage quit" if they see an unfavorable ex-post outcome. In many digital markets, ex-post IR remains a practical necessity.
Future research should extend these variance-minimizing rules to heterogeneous items (e.g., different positions in a search result) and explore whether these rankings hold for other convex risk measures, such as Value-at-Risk (VaR).
Summary updated