Revenue Maximization for Correlated Multi-Item Auctions

This direction is merged into another direction: interdependent valuations. Its last version stays readable and citable here.

Version 1 of 114 papers23 conceptsfrozen 2026-10-04

In the design of auctions, there is a constant tension between optimality and simplicity. For a single item, the problem is well-understood: a simple posted price or a standard auction with a reserve price is optimal. However, when a seller has multiple items, the landscape shifts dramatically.

For years, researchers focused on cases where a buyer’s valuations for different items were independent. In those cases, "simple" mechanisms—like selling each item separately or bundling them all together—often capture a significant fraction of the maximum possible revenue. But real-world valuations are rarely independent. This research direction explores a troubling discovery: when item valuations are correlated, simple mechanisms can fail catastrophically, and the complexity required to extract revenue grows exponentially.

The Infinite Gap

The central challenge in this field is the "revenue gap." This is the difference between the revenue generated by an optimal, potentially infinite-sized menu of options and the revenue generated by simple strategies like selling items individually.

Sergiu Hart demonstrated in Selling Multiple Correlated Goods: Revenue Maximization and Menu-Size Complexity that this gap is not just a minor inefficiency. Even with only two goods, if their valuations are correlated, simple mechanisms can capture essentially 0% of the potential revenue. This means that the "optimal" revenue can be infinitely larger than what you could get by just setting a price for each item.

The intuition here is that correlation allows for "pathological" distributions. If the value of Item A tells you something very specific about the value of Item B, a sophisticated seller can design a complex "menu" of lotteries—offers like "pay $10 for a 20% chance of Item A and a 50% chance of Item B"—that extract far more surplus than a simple price tag ever could.

If simple mechanisms fail, how complex does an auction need to be? Researchers use menu size—the number of distinct options or lotteries offered to the buyer—as the primary metric for auction complexity.

Sergiu Hart argues that menu size is the fundamental driver of revenue. In the independent case, a small menu is often "good enough." But as Aviad Rubinstein notes in Simple Mechanisms for a Combinatorial Buyer and Applications to Revenue Monotonicity, correlation renders these low-complexity models ineffective. To avoid leaving money on the table, the seller must offer an increasingly large and nuanced menu of choices, tailored to the specific statistical relationship between the items.

Bridging Pathologies and Practice

The existence of "infinite gaps" suggests that multi-item auction design might be impossible in practice. If the optimal mechanism requires an infinite menu to outperform a simple one, then simple mechanisms are useless, but optimal ones are too complex to implement.

To resolve this, researchers like Alexandros Psomas, Christos-Alexandros Psomas, Ariel Schvartzman, and S. Matthew Weinberg turned to smoothed analysis. In Smoothed Analysis of Multi-Item Auctions with Correlated Values, they ask: are these infinite gaps a robust feature of markets, or are they mathematical artifacts that disappear with a little bit of real-world noise?

Their findings provide a middle ground. They discovered that while some gaps are robust, natural perturbations (like the "Square-Shift" model) collapse the infinite gap into a finite one. By adding a small amount of random noise to the buyer's valuations, the advantage of an infinitely complex menu is curtailed. This suggests that while correlation makes auctions difficult, the most extreme "pathological" failures of simple mechanisms may be less common in noisy, real-world environments than pure theory suggests.

Papers in this version

History & citationVersion 1 · 2026-10-04

Merged into another direction · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_revenue_maximization_for_correlated_mult_2cc95088,
  title = {Revenue Maximization for Correlated Multi-Item Auctions},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/revenue-maximization-for-correlated-multi-item-auctions/rdr_2cc95088fe5f433a927816fea9b47f00},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_revenue_maximization_for_correlated_mult_2cc95088_v1,
  title = {Revenue Maximization for Correlated Multi-Item Auctions},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/revenue-maximization-for-correlated-multi-item-auctions/rdr_2cc95088fe5f433a927816fea9b47f00/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}