The Distribution Nobody Quotes When They Quote the Average

Why the Average Is the Wrong Number

Every few months a new creator-economy report arrives with a headline figure that makes the sector sound like a reliable path to a professional income. Adobe's 2022 survey put average creator earnings at $29,000 annually. Linktree's figures from the same period were cited widely in coverage that cast the creator economy as a mass employer of meaningful scale. These numbers travel fast, get stripped of context, and end up in pitch decks, op-eds, and programme marketing.

The problem is methodological before it is political. Both the Adobe and Linktree surveys were commissioned by companies with a commercial interest in a large, prosperous creator economy. More fundamentally, neither figure is a median. Both are means — arithmetic averages that add up all reported earnings and divide by the number of respondents. When a distribution has a long right tail, as income distributions almost always do, the mean is pulled sharply upward by the earners at the top. The median — the value that sits in the middle of the ranked distribution, with half of earners above and half below — is not affected by this distortion in the same way. It is, in almost every case, substantially lower than the mean, and it is almost always the figure nobody leads with.

This is not a subtle distinction. The Bureau of Labor Statistics' Current Population Survey, which tracks multiple jobholders — people the BLS counts as holding more than one job in the reference week — reported roughly 8.4 million such workers as of 2023, representing about 5.1 percent of employed people. That figure already suggests that secondary income from gig or platform work is far less universal than aggregate survey counts imply. But the BLS data says nothing about how much those workers earn from each job. For that, the Federal Reserve's Survey of Household Economics and Decisionmaking is more informative, and its picture is considerably more modest.

What the Published Data Actually Shows

The Federal Reserve's SHED survey asks respondents directly about income from gig and platform work. Its consistent finding is that for the large majority of people reporting such income, it is supplemental rather than primary — and often small in absolute terms. The 2022 SHED found that among adults who earned money from a gig platform in the prior month, most reported that income as less than half of their total household earnings. The share reporting gig income as their sole or dominant source was small, and the typical amounts involved were modest.

Patreon's own published figures illustrate the distributional problem more concretely. The platform has disclosed, at various points in press releases and impact reports, that a small fraction of its creators — the top tier — account for a disproportionate share of total creator earnings on the platform. Patreon itself, at its 2017 scale, reported that roughly two million patrons were supporting creators, but the income flowing from those supporters was concentrated among a relatively small number of high-follower accounts. By the time Patreon's published figures reached several billion dollars in cumulative creator payouts, the platform had approximately 250,000 active creators. Dividing the annual payout figures by the number of active creators produces an average that tells very little about the experience of the median account, because the distribution is steep: the top few percent of creators generate earnings that are orders of magnitude above what most accounts see.

Substack's disclosures follow the same shape. The platform has named its highest-earning publications and cited figures — in some cases, publications earning more than a million dollars annually from subscriber revenue — in its own promotional material. But Substack's top-earner disclosures are precisely that: disclosures about the top. The company has not published a median figure for active publications. What can be inferred from the structure of its business is that the majority of Substack newsletters carry free subscriber lists and earn nothing, while a modest middle tier earns enough to be meaningful supplemental income, and a small number earn at a level that would constitute a professional wage. Taking the cumulative payout figure and dividing by the number of active publications would produce a mean that looks flattering and a median that would not.

This is the structure of essentially every platform in the creator economy. YouTube's published 55 percent revenue share sounds generous until the denominator is considered: the overwhelming majority of YouTube channels never reach the monetisation threshold — currently 1,000 subscribers and 4,000 watch-hours — required to join the Partner Programme at all. Channels below that threshold earn nothing from AdSense regardless of how many videos they upload.

The Median Is Structural, Not Incidental

It would be convenient if the gap between mean and median were simply a data-quality problem — if better surveys with more respondents would produce a tighter, more accurate average. But the gap is structural. Platform economics produce winner-take-most distributions by design. Algorithmic recommendation systems on YouTube, TikTok, Spotify, and Twitch direct a disproportionate share of total attention to a small number of accounts, which in turn generates a disproportionate share of total monetisable activity. The advertising CPM model compounds this: because programmatic advertising values audience scale, a channel with ten times the audience earns far more than ten times the revenue of a smaller one, because it clears quality and scale thresholds that open more valuable inventory.

Spotify's royalty pool operates on a pro-rata basis, meaning each track's payment is proportional to its share of total streams. The published per-stream rate of $0.003 to $0.005 is itself an average across a distribution, and a track with a hundred streams earns a figure that rounds, at the platform level, to almost nothing. The minimum-streams threshold Spotify introduced in 2024 — below which tracks receive no royalty payment at all — was an explicit acknowledgment that the long tail of the distribution was generating payments so small they were administratively meaningless.

The FTC's Business Opportunity Rule exists, in part, because of how systematically earnings claims exploit the mean-versus-median confusion. When enforcement actions have been brought against schemes promoting online income — including MOBE and Digital Altitude, the high-ticket coaching operations the Commission pursued — the central charge in each case was that income representations cited the earnings of a small number of top participants and implied those figures were representative of typical outcomes. The Commission's 2021 Notice of Penalty Offenses on money-making claims told companies explicitly that atypical earnings require a "clear and conspicuous disclaimer" — which is a regulatory way of acknowledging that the average, absent context, is a misleading representation of the distribution.

The honest version of any creator-income figure comes with three numbers: the mean, the median, and the percentage of participants earning above a liveable threshold. Published surveys rarely provide all three. The ones produced by platforms or platform-adjacent companies — which have an interest in making the opportunity look attractive — almost never do. What they publish is the figure that generates headlines. The figure that describes the experience of the typical participant sits somewhere further down the spreadsheet, uncited, and usually quite a lot lower than the average that led the press release.