Selecting Weight Functions to Spur Competition in Quasi-Proportional Auctions.

Zheng Wen, Eric Bax, James Li

Abstract

In quasi-proportional auctions, each bidder receives a fraction of the allocation equal to the weight of his bid divided by the sum of the weights of all bids, where a bid's weight is determined by a chosen weight function. We study the relationship between the weight function, bidders' private values, and the auctioneer's revenue in equilibrium. It has been shown that if one bidder has a much higher private value than the others, then the revenue-optimizing weight function is nearly flat. Essentially, threatening the bidder with the highest valuation to share the allocation of the rewards maximizes the auctioneer's revenue. We show that as bidder private values approach parity, the revenue-maximizing weight function becomes steeper, making the quasi-proportional auction more like a winner-take-all auction. For flatter weight functions, there is known to be a unique pure-strategy Nash equilibrium. We show that this also holds for steeper weight functions, at least in the two-bidder case. We also give some analysis and intuition about why this may hold for more bidders.

Introduction

Standard auctions, such as the second-price auction, are efficient but often require the seller to possess prior knowledge of bidder valuations to set optimal reserve prices. In dynamic or opaque markets, such "priors" are often unavailable. Quasi-proportional auctions, or Tullock auctions, provide a robust, prior-free alternative. In these mechanisms, the allocation is shared: a bidder's share is proportional to the weight of their bid relative to the total weighted volume of all bids. The central design question for the auctioneer is the selection of the weight function, $f(x)$, to maximize competitive tension and, consequently, revenue.

Equilibrium Analysis

This study focuses on winners-pay quasi-proportional auctions utilizing power-law weight functions of the form $f(x) = x^p$. A critical challenge in analyzing these auctions is ensuring the existence of a pure-strategy Nash equilibrium. While equilibria are well-defined for $p \le 1$ due to the concavity of the utility functions, the situation becomes more complex for $p > 1$.

Utility function $u(x)$ and its derivatives for different exponents $p$.

As illustrated in the figure above, as $p$ increases, the second derivative of the utility function, $d^2u(x)$ (green line), is no longer strictly negative, indicating a loss of concavity. To address this, the authors employ Brouwer’s fixed-point theorem to prove that a pure-strategy Nash equilibrium exists for any $p > 0$. This theoretical foundation allows for the exploration of "steep" weight functions that drive higher competition.

Revenue Optimization in the OLOS Case

To analyze revenue dynamics, the authors consider the "One Larger, Others Symmetric" (OLOS) case, featuring one high-valuation "whale" bidder $(\alpha)$ and $n-1$ smaller, identical bidders. Through numerical optimization and line search, the study identifies the revenue-maximizing exponent $p^*$ across various competitive landscapes.

The "Threat" of Sharing vs. Parity

A key insight is the strategic use of the weight function's steepness. When a single bidder dominates the market (high $\alpha$), a flatter weight function (low $p$) is optimal. This "threatens" the dominant bidder with a significant loss of share to "bottom-feeders" unless they bid aggressively. Conversely, as bidders approach parity (low $\alpha$), a steeper weight function (high $p$) is required. This shifts the auction toward a winner-take-all dynamic, which is necessary to spark competition among equals.

The simulations reveal several critical trends:

  • Optimal $p^*$ Dynamics: The optimal exponent $p^*$ is a strictly decreasing function of the valuation ratio $\alpha$. In highly competitive scenarios (small $\alpha$), $p^*$ can exceed 10, whereas it drops toward 1 or lower as dominance increases.
  • Revenue Superiority: With an adaptively chosen $p$, quasi-proportional auctions can outperform second-price auctions. Equilibrium revenue increases with both the number of bidders $n$ and the valuation ratio $\alpha$, provided the weight function is correctly tuned.
  • Lower Bounds: As $p \to \infty$, the lower bound for bids at equilibrium approaches the minimum private valuation among the participants.

Conclusion and Future Work

This research confirms the viability of using steep weight functions in quasi-proportional auctions. While existence is proven, future work remains to establish the uniqueness of equilibria in non-symmetric cases and to derive closed-form solutions for general bid distributions. Additionally, exploring alternative weight function shapes, such as exponential functions, and developing adaptive mechanisms that learn bidder valuations over time represent promising avenues for maximizing auctioneer revenue in real-world applications.

Summary updated

Abstract

In quasi-proportional auctions, each bidder receives a fraction of the allocation equal to the weight of his bid divided by the sum of the weights of all bids, where a bid's weight is determined by a chosen weight function. We study the relationship between the weight function, bidders' private values, and the auctioneer's revenue in equilibrium. It has been shown that if one bidder has a much higher private value than the others, then the revenue-optimizing weight function is nearly flat. Essentially, threatening the bidder with the highest valuation to share the allocation of the rewards maximizes the auctioneer's revenue. We show that as bidder private values approach parity, the revenue-maximizing weight function becomes steeper, making the quasi-proportional auction more like a winner-take-all auction. For flatter weight functions, there is known to be a unique pure-strategy Nash equilibrium. We show that this also holds for steeper weight functions, at least in the two-bidder case. We also give some analysis and intuition about why this may hold for more bidders.
BibTeX
@misc{wen2015selecting,
  author = {Zheng Wen and Eric Bax and James Li},
  title = {{Selecting Weight Functions to Spur Competition in Quasi-Proportional Auctions.}},
  year = {2015},
  eprint = {1504.08333},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1504.08333},
}
Figure