A Patreon dataset finds creator income concentrated like capital

The test

A preprint titled 'Rich-Get-Richer'? Platform Attention and Earnings Inequality using Patreon Earnings Data, first posted to arXiv on 30 September 2025 by Ilan Strauss, Jangho Yang and Mariana Mazzucato, examines how concentrated creator earnings are across platforms. The authors use Patreon earnings data as a proxy for the attention creators capture on external platforms, reasoning that Patreon offers little native audience discovery of its own, so a creator's Patreon income mostly reflects the following they built elsewhere.

What the evidence says

The paper's methodology section states that the authors purchased official monthly Patreon earnings data directly from Patreon for March of 2018, 2021 and 2024, covering 104,719 creators, restricted to those earning over $10 a month and identified as active on a single external platform, YouTube, Instagram, Twitter/X, Twitch or Facebook, or on Patreon alone. Fitting a power-law model to each platform-year, the authors report a Pareto tail exponent near 2, which they describe as closer to the pattern typically seen in concentrated capital income than in labor income, and consistent with a compounding dynamic in which top earners pull further ahead. The paper is a preprint; arXiv's own site states that 'materials on this site are not peer-reviewed by arXiv', and no peer-reviewed publication is confirmed as of the version examined.

The sample and the variance

The authors state their own central limitation plainly: Patreon-sourced earnings represent a small subset of total creator earnings on the external platforms studied, since it captures only the portion of a following that converts to paid Patreon membership, not ad revenue, brand deals or platform payout-program income. Missing earnings, especially among larger creators who the authors say disclose less often, were imputed using a model built on subscriber counts and content type, a choice that shapes the resulting inequality estimate. The finding that algorithmic attention gains are drawn disproportionately from the creator 'middle class' describes this specific dataset and time frame, three snapshot months across six years, not a real-time or universal measurement of platform payouts.

What to try next

A researcher or publisher citing this paper should describe it as a preprint using a Patreon-earnings proxy for cross-platform attention, name the three years sampled, and avoid restating the Pareto exponent as a measure of any platform's actual payout distribution. This is an editorial framing suggestion, not an endorsement of the paper's causal claims.

• Is the earnings figure being cited Patreon-specific income, or is it being generalized to a creator's total earnings across platforms?
• Does the source distinguish a measured Pareto exponent from a causal claim about what algorithms do to the creator 'middle class'?
• Has this preprint been revised or published in a peer-reviewed venue since the version examined here?

Used carefully, a Patreon-earnings proxy offers a rare window into cross-platform attention concentration that platform-disclosed figures alone do not provide, provided the proxy's own limits travel with every figure drawn from it.