← News

ResearchSources: arXiv · 20 Nov 2026

Economists propose data-sharing contracts that keep competitors' market shares stable

A new theoretical paper models when rival firms should pool data for predictive models, identifying contracts that Pareto-improve profits while preventing firms from gaining by misreporting their data holdings.

In plain words

Companies that compete in the same market often hesitate to share data, fearing rivals will benefit more. The paper's authors design sharing contracts where, mathematically, every firm's profit rises compared with not sharing, and no firm can gain by lying about how much data it holds. One contract in their family meets both conditions at once.

What was announced

The authors, writing in an arXiv abstract, model firms whose profits depend on two factors: prediction precision, driven by each firm's own number of datapoints, and market share, which depends on the data endowments of all firms. Within this model, they characterize Pareto-optimal contracts as those where at least one firm receives all of the other firms' datapoints.

They then introduce a parametric family of contracts under which relative market shares stay constant, which they say removes the downside of handing data to competitors and equalizes gains relative to a no-sharing baseline. One contract in this family is both Pareto-optimal and strategyproof against firms under-declaring their initial endowment. Using a real-world dataset, the authors report that profit gains from sharing are largest when the machine learning task is harder or when firms' data endowments are more similar.

What it does not show

  • The evidence is the paper's abstract only; the full article, proofs and dataset details were not available for review.

Sources

Source-support review completed 21 Nov 2026. This story was drafted with AI assistance and checked against retained source passages. Automated checking is not independent verification.

Coverage: arXiv: abstract.

Updates and corrections

No corrections so far.

First published 21 Nov 2026

New to the subject? AI from the Ground Up starts from the beginning.