Exponential Weight Functions for Quasi-Proportional Auctions

Eric Bax, James Li, Zheng Wen

Abstract

In quasi-proportional auctions, the allocation is shared among bidders in proportion to their weighted bids. The auctioneer selects a bid weight function, and bidders know the weight function when they bid. In this note, we analyze how weight functions that are exponential in the bid affect bidder behavior. We show that exponential weight functions have a pure-strategy Nash equilibrium, we characterize bids at an equilibrium, and we compare it to an equilibrium for power weight functions.

Introduction

Traditional auction mechanisms, such as the second-price or first-price auction, are typically "winner-take-all." While these mechanisms are theoretically optimal under specific conditions, they often face practical challenges. They require the auctioneer to possess accurate statistical priors to set reserve prices effectively, and they may discourage participation from lower-value bidders who perceive a near-zero probability of winning.

Quasi-proportional auctions offer an alternative by sharing the allocation among all participants. In this model, every bidder receives a share of the "prize" proportional to their weighted bid. This structure ensures that every participant receives a "slice of the pie," incentivizing broader participation and ensuring a sale occurs even without complex reserve price settings. The critical design choice for the auctioneer is the weight function $f(x)$, which determines how raw bids are translated into shares. While power functions ($f(x) = x^p$) have been the subject of prior study, this research investigates the properties and performance of exponential weight functions: $f(x) = e^{cx} - 1$.

Theoretical Framework and Equilibrium Analysis

The core of the paper's theoretical contribution is the proof of existence for a pure-strategy Nash equilibrium (PSNE) when using exponential weights. By applying Brouwer’s fixed-point theorem, the authors demonstrate that in a field of rational bidders, a stable set of bids will emerge where no bidder can improve their utility by unilaterally changing their bid.

Characterizing the Equilibrium

The authors derive a mathematical characterization of the equilibrium bid $b_i^*$ for a bidder with private value $v_i$:

$$b_i^* = v_i - \frac{1}{c} \left( 1 - \frac{1}{e^{cb_i^*}} \right) \left( \frac{1}{1 - a_i} \right)$$

In this equation, $a_i$ represents the bidder's resulting share of the allocation. This formula reveals a fundamental relationship: as the "steepness" parameter $c$ increases, the "discount" a bidder takes from their private value decreases. This suggests that steeper functions can effectively compel bidders to bid closer to their true valuation, potentially increasing total revenue.

Comparative Performance: Exponential vs. Power Weights

To evaluate the efficacy of exponential weights, the authors conducted numerical simulations using a best-response iteration. This process allows bidders to repeatedly update their bids to maximize utility until the system converges to an equilibrium. Two primary scenarios were tested:

  1. Two-Bidder Scenario ($n=2$): A direct competition between two bidders with varying value ratios ($\alpha = v_1/v_2$).
  2. Ten-Bidder "OLOS" Scenario ($n=10$): A "One Large and Others the Same" configuration, featuring one high-value bidder and nine identical lower-value competitors.

Key Insights

  • The Steepness Trade-off: Increasing the parameter $c$ makes the weight function steeper. While this forces high-value bidders to bid higher to protect their share, there is a limit. If the function becomes too steep, the high-value bidder captures such a dominant share that the competitive threat from others diminishes, eventually allowing the leader to lower their bid.
  • Revenue Dominance: The simulations indicate that exponential functions are generally more effective at revenue extraction than power functions in competitive environments. In the two-bidder case, exponential weights generated higher revenue as long as the high-value bidder's value was less than 70 times that of the low-value bidder ($\alpha < 70$).
  • Scalability of Competition: In the ten-bidder scenario, the advantage of exponential weights was even more robust, outperforming power functions even when the value ratio $\alpha$ reached 100. This suggests that exponential weights are particularly effective at leveraging the presence of multiple competitors to drive up bids.

Conclusion and Future Directions

Exponential weight functions provide a powerful tool for auction design, offering a stable equilibrium and superior revenue potential compared to traditional power-law weights in many scenarios. Their "smooth" response to bid changes makes them less sensitive to small errors in parameter tuning than winner-take-all mechanisms.

However, the research also highlights areas for further exploration. Selecting the optimal steepness $c$ remains a challenge, as it requires some knowledge of the distribution of bidder values. Future work may investigate hybrid weight functions—perhaps those that are steep for low bids to encourage entry but flatter for high bids to maintain competition—or the use of "phantom bidders" as a mechanism to simulate reserve prices within a proportional framework.

Summary updated

Abstract

In quasi-proportional auctions, the allocation is shared among bidders in proportion to their weighted bids. The auctioneer selects a bid weight function, and bidders know the weight function when they bid. In this note, we analyze how weight functions that are exponential in the bid affect bidder behavior. We show that exponential weight functions have a pure-strategy Nash equilibrium, we characterize bids at an equilibrium, and we compare it to an equilibrium for power weight functions.
BibTeX
@misc{bax2015exponential,
  author = {Eric Bax and James Li and Zheng Wen},
  title = {{Exponential Weight Functions for Quasi-Proportional Auctions}},
  year = {2015},
  eprint = {1509.01314},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1509.01314},
}
Figure