An artist posts a screenshot of a streaming statement, the number is absurdly small, and the replies divide into blaming the platform or blaming the label. Working out how music royalties work means noticing that the argument is usually about the wrong copyright, and that one of the two prices involved is not negotiated by anyone.
Every recorded song contains two separate copyrights: the composition, meaning the melody and lyrics, and the sound recording, meaning that particular performance captured on tape. They can be owned by entirely different people, and they are licensed under entirely different rules.
One of them is set by federal regulation. The other is whatever the parties can agree.
Every song is two copyrights
A cover version makes the split obvious. When one artist records another’s song, the composition belongs to the original writer and their publisher while the new recording belongs to the performer and their label.
The same division exists in every original recording. It is simply invisible when the writer and the performer happen to be the same person, which is why so many people never learn that the two rights exist at all.
Money flows along both channels at once, and they rarely carry the same amount. This is why a songwriter with no recording career and a performer who wrote nothing can both be paid for the same play, in different sums, from different organisations, on different schedules.
Almost every confused argument about streaming income comes from treating these as one thing. They are not one thing, and they were never priced by the same method.
One of the two cannot refuse
The composition side operates under a compulsory licence. Once a musical work has been distributed to the public, anyone else may record and distribute their own version of it without asking permission, provided they pay the statutory rate and follow the procedure.
The owner cannot say no. They also cannot name their price.
This is section 115 of the Copyright Act, and its origin is older than recorded music as an industry. Congress created it in 1909, when player piano rolls were the disruptive technology and legislators feared that a single firm would corner the rights to popular songs.
The remedy was to remove the negotiation entirely: anybody may use the work, at a price the state fixes. That structure survived the gramophone, the cassette, the CD and the download, and it now governs streaming — a law written about piano rolls setting the price of half of modern recorded music.
What the government sets the rate at
The rates live in the Code of Federal Regulations, updated through proceedings before the Copyright Royalty Board. For interactive streaming the calculation is not a per-play figure at all.
The regulation defines an “all-in” royalty for each offering as the greater of two calculations: a set percentage of the service provider’s revenue, or a figure derived from what the service pays the record labels. The higher of the two applies.
The revenue percentage is not a single figure either. It is a schedule, fixed years ahead so that services and publishers both know what is coming.
| Royalty year | Percent of service provider revenue |
|---|---|
| 2023 | 15.1% |
| 2024 | 15.2% |
| 2025 | 15.25% |
| 2026 | 15.3% |
| 2027 | 15.35% |
Two things follow from that table, and both are routinely missed.
The first is that composition royalties are a share of a pool, not a price per play. The service works out what it owes in total, and each rights holder takes a slice proportional to their share of that month’s listening. A songwriter’s income therefore depends on what everybody else was playing.
The second is that nobody in the music business agreed to 15.3%. It was determined by a government body after an adversarial proceeding, and it applies whether or not any party thinks it is fair.
The other copyright is a normal negotiation
No compulsory licence covers the sound recording for interactive streaming. A platform that wants to offer a particular recording has to agree terms with whoever controls it, which for commercially significant music usually means a major label.
That negotiation happens privately, at scale, between a small number of large companies, and the terms are not published. Catalogue is the leverage on one side and distribution is the leverage on the other, which is why the deals tend to include advances and minimum guarantees rather than a simple rate.
The asymmetry explains most of the disparity people notice. One side of the song is priced by regulation with the outcome printed in the Federal Register; the other is priced by companies with considerable bargaining power on both sides of the table.
It is the difference between a regulated ceiling and a negotiated deal, and it runs in the opposite direction to the pattern in how airline overbooking works, where the regulated figure is a cap the parties settle beneath rather than a floor they cannot go under.
A single stream sets off more than one right
Copyright law grants a bundle of separate rights, and streaming a song exercises several at once.
Delivering audio to a listener reproduces the work, because a copy is made in the course of transmission. It also performs the work publicly. Those are distinct rights, and they generate distinct royalties on the composition — the mechanical and the performance — collected by different organisations.
The sound recording follows its own path, with performance income for digital transmissions administered separately from the label’s negotiated share.
Even the performance side of the composition is not a free market. The two largest American performing rights organisations have operated for decades under antitrust consent decrees agreed with the Department of Justice, which oblige them to license anyone who asks and send disputed rates to a federal rate court — so a second layer of government price-setting sits underneath what looks like ordinary licensing.
So one play can trigger four separate payment streams, arriving from four organisations on four different timetables. When an artist says the statement is incomprehensible, that is an accurate description rather than a complaint.
Where the money stops before it arrives
None of the above describes what an artist receives, because the royalty owed for a recording is paid to whoever owns that recording, and under a traditional deal that is the label rather than the performer.
The artist’s contractual percentage is then applied, and applied only after recoupment. Advances, recording costs and often a share of marketing are treated as loans against future income, so an artist earns nothing personally until the recording has repaid what was spent on it — while the label collects from the first play.
This is why two artists with identical stream counts can report wildly different income, and why a screenshot of a statement proves very little on its own. It reflects a contract, not a rate.
American radio pays for only half the song
Here is the quirk that most surprises people who assume the system is at least internally consistent, and it has survived every attempt to remove it.
When a song plays on AM or FM radio in the United States, the station pays the composition side — songwriters and publishers collect through performing rights organisations. It pays the owner of the sound recording nothing.
The performance right in a sound recording exists in American law only for digital audio transmissions. Terrestrial broadcast sits outside it, on the historical argument that airplay promotes record sales and the exposure is payment enough.
Most comparable countries do not draw the line there, which means a recording earning broadcast royalties across Europe can earn none at home. The position survives because it is entrenched, not because anyone designed it this way.
What the Music Modernization Act changed
Until recently the compulsory licence was administered song by song, and the paperwork was the problem. A service had to identify and notify the owner of every composition it reproduced, which at streaming scale was not achievable.
Unmatched money piled up. Recordings were streamed, royalties were owed, and nobody knew who to pay.
The Music Modernization Act restructured this by replacing song-by-song licensing with a blanket licence, administered by a body created for the purpose — the Mechanical Licensing Collective, which maintains a public database and distributes what it can match.
It is an administrative fix rather than a change in the price. The rate is still set by regulation, and the unmatched money problem is reduced rather than solved, because a database is only as good as the metadata submitted to it.
What to check when someone quotes a per-stream rate
Per-stream figures circulate constantly and are almost always misleading, because the underlying system does not work in per-stream terms.
- Which copyright is being described — a composition figure and a recording figure are not comparable
- Whether it is gross or the share reaching the artist after the label’s contractual split
- Whether the platform is interactive; the rules for services where a listener picks the track differ from radio-style streaming
- What the denominator was, since a revenue-share pool divided by a month’s total plays produces a different number every month
The system’s real weakness is not that anyone is stealing. It is that a mechanism designed in 1909 to stop one company monopolising sheet music now sets the price of half of every song, while the other half is decided by private negotiation between a handful of firms.
Nobody would design it this way from scratch. It persists for the same reason the shuffling of shows between streaming services persists — the parties who would have to agree on a replacement are the parties currently doing well out of the arrangement.
