Exactly. To really be fair you need to compare the audience for each play, and for Pandora/Spotify it's just one person per play versus X (hundred/thousand) for regular radio.
Looks like the answer is probably "thousands" [1] of listeners per radio station. The back-of-the-envelope is that the top station will have 1% of its metro area listening at a time (metro areas as defined here [2]).
Even the 100th largest metro area would typically have 5k people listening to a play of a track.
If you assume 5000 people listening per play at 18797 plays, and then calculate a cost PER LISTENER (at $1373.78 royalties paid) you end up with $0.014617 royalty per 1000 "performances" of the song (counting every listener's radio as a "performance"). AND, when you multiply that back out with the number of "performances" on Pandora, you get....$16.94.
Considering WAY more people probably heard the song on the radio, possibly by a factor of 5 or more, this is a HIGH estimate; actually doing the math would likely come out with a much lower equivalence, but it's the right order of magnitude.
I am all about protecting artists rights to earn, and for buying music that I'm listening to. As the parallel comment by smack_fu mentions, though, this is really about the insignificance of the Pandora market size more than anything else. That and the inability of a particular blogging artist to do enough basic math to realize the royalty per performance is higher on Pandora than it is on the radio.
If you follow the links he provides they have actual market data on how many listeners any given radio station has, so his numbers have that behind it. Also, it's not that 1% of people listen to a popular radio station 24 hours a day, it's that at any given point in time, an average of 1% of the population are tuned in to that station.
I'd say the Pandora 1 track per customer is a safe assumption. Sure, I've known cafes and dinner parties where Pandora was streaming to multiple people at once, but I've also known lots of friends who have Pandora playing on their headphones even when their headphones aren't on their heads. Nobody I know who uses Pandora actively pauses it every time they leave their computers, so I expect it washes out in the end.
You can generalize my point: Pandora should be judged the same way as radio or the comparison is unfair.
The methodology of arbitron says that if radio can be identified within audio range, you are deemed to be 'listening' to it.
Can we still claim, by this methodology, that every Pandora device has an average of one listener? Only if you claim every radio has an average of one listener by this methodology as well (which is highly doubtful) or if radio and Pandora are used significantly differently (which I don't strongly believe: as you say, it's played in cafes in place of radio).
> if radio and Pandora are used significantly differently (which I don't strongly believe: as you say, it's played in cafes in place of radio).
I have to disagree: I think the vast majority of people listening to Pandora are doing it on their laptop on headphones, while many radios (today) are not being listened to by a single listener.
Regardless, the estimates I used intentionally underestimated the Arbitron radio audience numbers by an order of magnitude. Some of the larger markets go up to 150k listeners on the top station. If you're right and significantly more than 1.0 people listen to a particular Pandora stream on average, then it still have to be more than about 5-10 listeners per instance before it hits the numbers that radio is claiming. And I would be very, very surprised if the average were anywhere near that high. I think if Pandora had evidence of that, they would be able to demand more money from advertisers, for one thing.
As to the methodology of Arbitron: Not really relevant for my purposes, because I think that it's accepted by advertisers as reasonable. At some point, if the numbers were complete garbage, the advertisers would have noticed and stopped relying on them; advertisers certainly track the results of their campaigns, after all.
My argument is trying to judge Pandora in the same way as radio, in that the number of plays as counted on Pandora reaches far less people than the number of plays as counted on radio. What the artist wants is to be paid for every person that listens to their song right?
So when you say "Only if you claim every radio has an average of one listener by this methodology as well", I think you misunderstand the core problem, which is that the article compares stats that aren't equivalent.
The issue is in how these listens are counted. So yes, you can pretty much claim every Pandora device has an average of one listener. Users of Pandora have their own personalized streams, and every time a song is played on a person's stream, it's counted as 1 play, and is heard by 1 person.
In contrast, radio stations have one stream that many many people listen to. When a radio station broadcasts a song once it's counted as 1 play even though that song is received by X radio-receivers and heard by X people (where X is the reach that station has, for example, 150k).
The issue is the article compares plays on Pandora (where each play as counted reaches 1 listener on average) to plays on radio (where each play as counted reaches upwards of 150k listeners).
