September 4, 2009

The "market for lemons" among song and video game compilations

People who criticize the idea of selling entire albums as bundles, rather than allowing you to buy each song individually, complain that the record company is forcing you to buy all those other bad songs when all you want are the good songs. But this is wrong, and the reason is simple: before they release the album to the public, the record company has zero clue which songs are good and which are bad, according to the consumers' tastes.

If we measure song quality by how many downloads it sells on iTunes, we will probably see a distribution of songs that has a sharp peak near 0 where the innumerable also-ran songs are, and a long, fat tail where the superstars are. I'm assuming that songs are enough like movies, and Arthur De Vany showed in Hollywood Economics (featured in the Amazon box above) that if we look at a superstar movie, the expected revenues given a current level of revenues is even greater -- that is, we expect it to keep going and going (although we know it eventually must end). So, movie studios may have a guess about how well a movie will do (that is, what the average is), but the variance is infinite, which makes the guess worthless.

The same must go for songs. (Hopefully some economist has already looked into this.) The record studios may have a guess about how many downloads a song will get (i.e., what its quality level is), but that guess has an infinitely large error bar around it, so their guess is as good as anyone else's -- namely, not very good at all. Therefore, they don't know beforehand which songs on the album are good and bad, and so the charge about them ripping us off by bundling good and bad songs is wrong.

The record studios only find out which songs are good and bad once they've been out in the market long enough for consumers to vote with their dollars. Of course, the studios can't travel back in time to use this information before the album is released, but they can use it when they put together compilation albums. They have two options: bundle a few good with a majority bad songs, as the complainers fear, or put only good songs on (again, good as judged by sales -- there may be one or two that you don't like, but that everyone else does, and most of them everyone would agree are good).

It's striking how many greatest hits albums there are, and how few garbage compilations there are. Here, the studios actually do have the ability to bundle a few good ones with a bunch of crap -- anyone who has lost or sold their original albums, or who never heard the originals in the first place, would be forced to buy bad songs in order to get the good ones. And yet that rarely happens. My guess is that most of the target customers are savvy about the quality of the artist's various songs and would instantly spot a "bad bundle" compilation, get offended, and not buy it out of spite. Studios recognize that the target audience is this savvy and potentially spiteful, so they don't bother trying to pull a fast one on them. Instead, they're forced to use up all or most of the artist's good songs on a single greatest hits album, rather than boost profits even more by spreading them out across several bad bundles.

However, there are a few exceptions, and they stem from the target audience being fairly ignorant of song quality, while again the studio already knows which are good and bad by this time. We have a problem of asymmetric information, and as with the hypothetical example of the market for lemons, we get something close to a market failure. It isn't that bad, but it's close. Consider all of those compilations based on a decade, period, or really any concept that would pool together songs from a wide variety of artists (as opposed to hits from a single artist). There are tons of these, and if a savvy person thumbs through them, they'll notice that they're all bad bundles (lemon bundles?). In all my searching, I've never come across the equivalent of a greatest hits album. I've only bothered buying one, a new wave compilation with four or so good songs and a bunch of shit.

I believe most people buying such compilations weren't die-hard fans of the music when it was popular, just as I was only a toddler when new wave was big. It's also harder to know everything about an entire period that includes many artists, as opposed to knowing everything about a particular artist, even if you were alive during the period and vaguely recall some of the good songs. Most customers will therefore be pretty ignorant of song quality, while the studios will know a lot.

They can therefore sell bad bundles, and the target consumer won't be able to tell until it's too late. If a studio wanted to release a true "greatest hits of the '80s" album, it wouldn't sell too well because customers will have become jaded by all the other bad bundles flooding the market. So the studio has little incentive to put out a good period compilation. Why doesn't the market collapse? Because the jaded customers are replaced each year with a new round of ignorant customers, as the population ages or as new parts of the population become aware of the period and become curious about it. The bad compilations never gain much traction over time within a given group, but there are new naive buyers who enter the market each year, ensuring that the market doesn't totally collapse.

You see the same thing with video game compilations. If the bundle is based on a single franchise -- say, those featuring Nintendo's Mario -- then it's like the greatest hits album of a single artist. If you're part of the target audience for that compilation, you're pretty savvy about which games are good and which are bad. No one in the target market would buy a compilation that had the arcade Mario Bros. game, Super Mario Land, and Dr. Mario, but didn't come with Mario 3, Super Mario World, Mario Kart 64, and so on. So, just as with music, these compilations abound and are great (e.g., those for Mario, Mega Man, Sonic the Hedgehog, etc.).

However, there are counterparts to the decade or period compilations too -- namely, arcade game compilations. These bundle together anywhere from 4 to 20 or more arcade games. As with music, a few people who lived their lives in arcades may know enough about the games' quality levels just by looking at the list, but most people will not. I spent plenty of time in arcades and yet only recognize a handful of the titles. Customers may know which Mortal Kombat games are better than which others, but they will be pretty ignorant about the entire bundle. The video game companies have a decent idea of how well each game did in the arcade way back when, so they have much clearer information about each one's quality.

As a result, they bundle together a few gotta-have arcade games with a bunch of throw-aways. The customers won't find out they've over-paid until it's too late. They'll become jaded, although they'll be replaced with new naive customers each year, so that the market doesn't entirely collapse. And as before, a video game company has little incentive to offer a true greatest hits of arcade games. Most people would respond, "Yeah right, just like all those other alleged greatest hits compilations -- not falling for that again." So most of the arcade game compilations are "lemons." Just like music, I've got one or two of those too, but only because I'm one of the few who will bite the bullet and pay for the good ones rather than turn the whole thing down out of spite.

By the way, these bad bundles are also the bulk of lot sales on Ebay or some other second-hand market. Wow, 20 Sega Genesis games for one price -- too bad 18 of them are retarded sports games.

These facts are pretty interesting, given that finding real-world examples of a "market for lemons" wasn't as easy as some had thought. (See Market Failure or Success, also featured in the Amazon box above.) It looks like a market for lemons, or something close enough, is more likely to occur when the information required covers too broad of an area for the typical buyer to be savvy. If it's narrow enough, then those in the target market can just tell whether the product is good or bad. There's also a neat twist, in that the tiny handful of models I've seen don't look at the replenishment of suckers over time, whether due to one age group maturing to the point where they are now in the market for the product, or because some part of the population that used to not care about the product now finds itself in the market for it. This keeps the market from collapsing, although it probably does remain at a very low level of activity.


  1. It's somewhat true that studios and labels have no clue which songs are good; if you ask them, for example, what song you should make a video to promote they'll just tell you "Do whatever you want" or "They're your songs - you should know better than us".

    And that's true - a band that's been playing its material live for a while will have a pretty good idea of what gets the best audience reactions. They have some idea about quality, so if anyone's guilty of bundling bad songs with good ones on albums it's not record companies but bands.

  2. I liked Super Mario Land. And the sequel was great. But World is clearly the masterpiece, with Bros. III some distance behind.

  3. I'm more skeptical of the "labels don't know what's going to be popular" theory than you. That they release singles in advance of the album implies that they *think* they know. Looking at long-term sales of those particular songs (versus the rest of the album) might be a good indication of how accurate they are.

  4. Interesting theory. I wonder if the same goes for literature -- if anthologies of poems, short stories, essays, etc., tend to be of a higher caliber when they are drawn from a single author rather than from a time period or school.

    I don't really know from my own experience because I tend not to buy multi-author anthologies -- perhaps, I now realize, because I subconsciously associate them with crappy "greatest hits of the 80s" type compilations.

  5. You forgot to mention another kind of high quality compilation. Smaller record companies put out compilations with one good song from several different bands on their label. These compilations generally sell for less than the cost of a regular album. They cover the costs but use the compilation more as a marketing tool.

    Living no where near Texas, I never would have been exposed to At the Drive-In in high school if I hadn't purchased Fearless Record's Serial Killer compilation.

  6. Looking at long-term sales of those particular songs (versus the rest of the album) might be a good indication of how accurate they are.

    They may have a better sense of the quality ranking among songs on the album, but what I mean is whether the quality level for some song will exceed the consumer's threshold of "good." That is very uncertain.

    I wonder if the same goes for literature

    Yeah, in principle this idea applies to any intellectual property that would have any demand for compilations. (Another example is movie boxed sets -- say, movies starring Audrey Hepburn vs. movies from the film noir period.)

    Maybe I'll look into Amazon sales figures or customer ratings for author-specific anthologies vs. broader ones.

    Smaller record companies put out compilations with one good song from several different bands on their label.

    Right, if the point of these compilations is not to generate lots of revenue directly, we might not see a bunch of lemons -- like, 5 records with those good songs spread out among them.

    So good call, whether the compilation is for promotional purposes or not is key.

  7. i don't think this applies to literature anthologies if they are chosen by one person for others. for example, the Norton Anthologies of Literature don't suck because the professors presumably know all the pieces (or at least could read them before selecting them as course material). or alternatively, the pieces are selected because they make up part of the "canon" of literature.

    would it be useful to think about more informal anthologies, like a particular person's favorite links or favorite blogs? would readers of one blog generally enjoy other blogs on that author's blogroll?

  8. the Norton Anthologies of Literature don't suck because the professors presumably know all the pieces

    Good point. Many literature anthologies are designed for classroom use and are therefore catering to the savviest of the savvy -- professors who have devoted their careers to studying the product. Agnostic's reasoning would probably be more applicable to detective stories, sci-fi, etc., than to canonical literature.

    would it be useful to think about more informal anthologies, like a particular person's favorite links or favorite blogs?

    No. It doesn't cost you anything to add a link to your list of favorites, nor do you make any money if a reader likes your list, so the economic incentives in question don't apply.

  9. Generally speaking, music industry professionals have better than random predictive powers about songs' popularity. Irving Berlin immediately said "White Christmas" was the best song he ever wrote, and Paul McCartney and George Martin knew "Yesterday" was going to be a big deal. There is a market among singers for songs written by others, and good decisions make or break a career.

    On the other hand, there are lots of surprises as well.

  10. I agree though with music it comes down to genres. When I was younger I bought so many albums where only 2 or 3 songs were really great and 2 or 3 songs just weren't good at all. When my music tastes changed I found myself liking complete albums and realizing what the problem was.

    With Pop Music they are trying to appeal to a VERY large crowd. They'll throw in some fast upbeat songs and then they'll hit you with a boring ass ballad. I have a fairly broad taste in music but for the same reason I dislike the shuffle feature of the iPod I hate albums that throw every genre under the sun at you.

    I listen to an album or band because I want a certain feeling, groove, or emotion. If I want to relax I'll listen to something laid back, if I want to rock out I'll listen to that. With pop music it was typical to throw everything at you on one album and you would have to skip the songs you either hated or weren't in the mood for.

    While far from my first choice pop music has gotten better with consistency over the past few years. If you listen to Britney Spears latest album, Kanye West latest album, Lady Gaga or numerous others they are all quite consistent. All the songs have a similar feel which is good for those that like it. Still with numerous songwriters per album(since pop musicians don't usually actually write anything) it's not the easiest thing to accomplish.


You MUST enter a nickname with the "Name/URL" option if you're not signed in. We can't follow who is saying what if everyone is "Anonymous."