I checked my calculations again and did find a couple of errors, but I don't think they affect my argument.
The mistakes: I realized that just prorating the album price for single sales is inaccurate, since albums sell for $9.90 and songs are $0.99 each. And although I wrote that artist & producer royalties are calculated on 130% of the wholesale price, somehow I entered 140% in my spreadsheet, which incorrectly added 10 cents per album to the label's costs. I've fixed the relevant numbers in the post.
In response to your points:
You said: Traditionally you pay more per track for a single than an album, there is no reason for this to be different when selling mp3's online.
The price for MP3 singles is already well established: It's 99 cents on iTunes (by far the dominant service), lower on most competing services. 99 cents per song compared to $9.90 per album is not a very large premium, unless there are lots of songs on the album.
Labels have traditionally worked hard to sell singles in order to a) promote albums, b) get a higher margin, c) sell something rather than nothing to people who don't want the album.
Labels in the US have been abandoning the physical singles market for a long time, and that market is now very small. For 2004, The RIAA reports US CD album unit sales of 766.9 million, and CD single unit sales of 3.1 million (see http://www.riaa.com/news/newsletter/032105.asp).
In your calculations you neglected to consider that the number of people who will be at least 1 single is greater than the number of people who will buy an album.
That seems quite possible, especially if marketing changes to be focused on selling singles rather than albums. In my scenarios I assumed that interest in an album marketed at this level would be roughly equal to interest in the single(s) from that album, since most marketing is currently focused on selling albums, not singles. (Labels market singles to radio in order to promote album sales, but they don't make money from radio; publishers and songwriters do). One obstacle that I think will have to be overcome if there's a shift towards singles is that a similar marketing investment probably has to be made to draw attention to an artist's new release whether it's an album or a single, but with quite a lot less money to be made from the single. Also, the explosion of content made possible by the lowered cost of entry seems likely to dilute the market for any individual single. So even if more people are willing to buy a 99 cent single than a $9.90 album, it might still be hard to make good money. I hope it works out to be a good business, but I don't think we know enough yet to be sure that it will.
Also you are forgetting that online sales are often in addition to traditional sales and somebody may buy only 1 or 2 tracks online (or even god forbid hear them for free or get an illict copy) before buying the complete album on CD.
I'm not sure how we know that that's actually happening often enough to offset declining CD sales--can you point me to a reference? US CD album sales have been declining every year since 2000 except for an uptick in 2004 (I believe the trend is down again this year), and meanwhile digital sales are rising fast. It seems likely to me that digital sales are taking over from CD sales, similarly to how CD sales took over from vinyl sales.
It looks to me that you had already decided what result you wanted, and fudged the figures and maths until you got it.
I really want downloading to work out to be a good business, especially if it helps artists get a better deal. But I'm also trying to distinguish what I can actually see so far from what I hope turns out to be true.