How do songs become popular? When and why do people pay for play? On July 26, I (virtually) sat down with Gabriel Rossman, a sociologist at the University of California, Los Angeles, and the author of the brilliant "Climbing the Charts," an analysis of radio airplay, and what it tells us about the diffusion of innovation. We spent an hour-and-a-half on instant messenger, hashing out what we know -- and what we don’t -- about the way songs climb the charts, and what that might tell us about insider trading and supermarket samples. The resulting interview was lightly edited for readability, and is now presented here for your reading pleasure.
Megan McArdle: Start out by telling me about yourself, and how you got interested in the music industry.
Gabriel Rossman: I grew up in LA and so I've always had some exposure to the entertainment industry, but when I started graduate school I was really interested in news. The trick is, though, that news is really hard to study because you need people to code it. So most studies of news end up devolving into fights about whether you did your content analysis correctly.
So fairly early on in grad school I started looking for aspects of media content that you could code objectively. I first hit on movie reviews (specifically the 4-star summaries) but then moved on to pop music as songs are very nice atoms of content. If two stations play the same song, they are in a very meaningful sense doing the same thing, no argument about it.
In my dissertation I applied this to questions of corporate ownership affecting bias and creativity, but in my work as an assistant professor I moved on to using it as a sandbox for diffusion of innovation. It's very well suited for this since there are lots of songs, they spread very quickly, and it's all very well documented. So I decided to make pop songs my own little fruit fly.
McArdle: Do the pop songs end up with red eyes, or white eyes?
Rossman: My equivalent to red eyes, or curly wings, or what have you, is to look at the pattern by which the song spreads across radio. When something grows fast, then slow, this reflects a process of something outside the system pushing it onto the system but constrained by saturation. Like Zeno's paradox of Achilles and the tortoise.
McArdle: Tell the readers what that paradox is, please.
Rossman: Sure. Zeno's paradox says that fleet-footed Achilles races a turtle but gives it a head start. If every second he closes half the remaining distance, then you can ask “when will he lap the animal?” The answer is “never.” He'll just get closer and closer.
This type of diffusion pattern tends to characterize when people learn about things, especially if they're broadcast in the mass media or the subject of an extensive marketing campaign. As such you can call it "external influence" or "exogenous" because (according to theory) it results from a new idea or practice being pushed onto the system from outside the peer group.
The alternative pattern is that something spreads internally to a system, and this gives you slow, then fast, then slow growth. You get this when you combine Zeno's paradox with exponential growth. There's also a little fable to explain this.
An emperor once wished to reward a sage, and the sage asked for a grain of wheat on the first square of a chessboard, doubling with each square. This doesn't sound like a lot of wheat but compounding the doubling means that you very quickly run out of all the wheat in the world.
McArdle: I remember an example of this from statistics. Some unfortunate graduate student had written that "the number of abused children doubles every year." He wrote that in 1957 .
Of course, in real-life, exponential growth is always constrained in some way, which is why you get an S-curve (that tapers off eventually), not a J-curve (that just continuously shoots into heaven). You see this not just with the sorts of fads that [Malcolm] Gladwell talks about, but also things like bacteria on a petri dish.
McArdle: Implying that by the mid-1980s, everyone on Earth had been abused as children. My parents must not have gotten the memo.
Rossman: Good thing, too.
McArdle: So, there is no song that will literally reach every person on Earth? Not even "Happy Birthday" or "Solsbury Hill"?
Rossman: There very well could be a song that reaches every person on Earth, but the point is that eventually you run out of people. Once every human being on Earth has learned to sing "Happy Birthday" it's not as if logic requires that Martians start singing it.
McArdle: "Gangnam Style" seems close to reaching that point.
Rossman: Yes, I haven't checked, but "Gangnam Style" is probably a good example of S-curve diffusion.
McArdle: So are these the only two diffusion patterns we see in music, or just the most common?
Rossman: Those are the two basic patterns, yes, and most songs follow one or the other fairly closely.
McArdle: I assume there's a third pattern, where it just dies and no one ever listens to it. That was the pattern my college band showed. "The Crib Death Curve."
Rossman: True, the literature calls this "success bias" and I have to admit that my work falls prey to it.
McArdle: I'm not sure there's enough storage space in the world to contain all the terrible songs produced by college bands. But I'm dragging us off topic.
Rossman: Well, briefly on the subject of failed songs, as you know, [Princeton University's Matthew] Salganik et al did a great study showing that to a large extent failure is random. Anyway, when I started the work, I expected to see a lot of S-curves in radio because sociology is fascinated with anything "social," including social networks but also just general imitation.
McArdle: Hence the name "sociology."
Rossman: Exactly. If you read the sociology literature there are tons of S-curves and emphasis on "imitation." This includes cases where it doesn't really exist, most famously a study of when doctors started prescribing the antibiotic tetracycline.
The investigators, Coleman, Katz, and Menzel collected really cool social network data and argued that the doctors learned about it from their mentors. But secondary analysis by Van Den Bulte and Lilien a generation later showed pretty convincingly that it was Zeno's paradox, not S-curves -- the doctors were more attuned to Pfizer marketing than to each other's advice.
So in the case of pop songs, it turns out that there are some special circumstances under which you see S-curves, but typically a pop song looks exactly like Zeno's paradox. This strongly indicates that radio stations are NOT imitating each other, but rather are all imitating some central source.
McArdle: So what's the central source?
Rossman: Well, first let me tell you what it isn't. Just like sociologists would expect diffusion to be structured by S-curves/imitation, radio experts would expect it to be structured by corporate ownership. I once had a grant reviewer tell me there was no point in studying radio stations because all the behavior is at the chain level and so there'd be no variance by station.
You see this especially in criticisms of Clear Channel, which until being displaced by Spotify was the favorite agent of theodicy for musicians.
Turns out though that chain ownership is an extremely bad explanation for a couple of reasons.
1. Within each chain, songs get picked up in a staggered fashion, whereas if it were truly decided at corporate headquarters, you'd expect the songs to appear on all corporate properties simultaneously.
2. When you graph the chains together, they look very similar to each other. In technical terms, the variance in adoption time is entirely within chain, not between chain.
McArdle: So Clear Channel's station in Nashville might be more different from Clear Channel's station in New York than some other station in Nashville?
Rossman: Yes, exactly. Another way to put is that if station A and B are owned by Clear Channel and station C by some other company, then knowing what A did gives exactly the same predictive power about what B will do as it does about what C will do.
McArdle: So it's not corporate ownership, and it's not social contagion.
What is it?
Rossman: There's one more thing it isn't, which is unsated demand. You can tell this by looking at multiple singles from the same album.
Anyway, but to answer your question, as best as I can tell it's aggressive record label promotional efforts. That's a broad concept that encompasses a lot of legitimate activity (mailing sample CDs and press kits, sending sales people to talk about what a great song it is, etc…) but in the extreme case it includes payola. That is, record labels or their proxies sometimes bribe radio stations.
I have a few reasons for saying this. One is just that when you've eliminated radio chains and unsated demand, it's the only thing left. But there's more direct evidence based on court records. As you know well, before he was candidate for comptroller and before he resigned as governor, [Eliot] Spitzer was an activist AG for New York.
His basic pattern being to subpoena records from companies, embarrass them, and get a consent decree.
Anyway, one of his targets as NYAG was payola and he found a lot of pretty damning stuff, especially at WKSE, a Top 40 station in Buffalo with an openly venal program director. In our 2008 article, my co-authors and I went through all the subpoenaed documents, coded every song that was mentioned, and then looked at their diffusion curves. It turns out that the songs shown in these documents to have involved payola had diffusion curves that especially follow the Zeno's paradox pattern.
The other piece of evidence is that if you look at songs over a several year period, the Zeno's paradox pattern goes away during the brief periods when people are scared of prosecutors.
That's not to say that it's all payola, in the sense of outright criminal activity, but I think it is fair to say that one of the most important forces in radio is record label promotion, much of which is legitimate.
McArdle: Does it vary between markets? Are some more prone to payola than others?
Rossman: Good question. There's qualitative evidence that some programmers are more reticent to accept influence than others. So in his one-hit wonder memoir, Jacob Slichter says that really big and prestigious stations like KROQ tend to be honest. Likewise the ethnographer Jarl Ahlkvist has done interviews with radio programmers and finds that some of them are much more oriented toward record labels than others (including being willing to accept "gifts"). However, this wasn't the kind of thing that I was able to figure out a way to operationalize in my quantitative work.
McArdle: What are the "gifts?" Cash?
Rossman: Yes, this is fascinating to me and is one of the things I'm interested in looking at in the future. In the Spitzer documents, the gifts are usually some sort of in-kind item to be used in the station's own promotional campaigns. The 10th caller gets concert tickets or a signed T-shirt. That sort of thing.
More than that, if you read the documents closely, there's a strong rhetorical tone of friendship. It's not necessarily "here's the quid pro quo," although you do see that, but often it's "hey, can you do me a favor?" The thing is that the favors are always that the station gets something of pecuniary value and presumably reciprocates with airplay. That's how it works now.
I find this fascinating because it's like something out of the anthropology of gift exchange, but it's being done by very professional business people. In the past it was much more openly venal and involved more cash. From the late 1970s through 1986, radio was controlled by the mob, and you had a lot of people getting hundred dollar bills in record sleeves or cocaine hidden inside cassette tapes.
McArdle: Wait, radio was controlled by the mob? Like, "The Godfather" mob?
Rossman: Yes, the Gambino family. There's a great book on payola in the 1980s by Fredric Dannen called "Hit Men." Basically, there was a cartel of consultants called "The Network" that bribed programmers and was extorting money from record labels. If you didn't pay their billings they'd blacklist your artists.Most famously they had Pink Floyd kicked off the LA airwaves during an extremely successful concert tour as retaliation against their label.This cartel was connected to Piney Armone, a Gambino underboss.
There was also dope, cash and the mob in the mid 70s. The 70s thing came out when the Feds and Mounties caught a gangster smuggling heroin from Montreal …and then searched his papers and found he was also involved with a record label. They flipped his contact at the record label and he told the Feds about payola. The mob hasn't been involved since '86 though.
In the late '50s, there was a lot of cash (and on one notable occasion, hookers) but as far as I know, no gangsters.
McArdle: But now it's just the perverse sense of obligation engendered when someone gives you a freebie. Like people who feel obligated to buy something because they got a free sample of cheese at the supermarket.
Rossman: Yes, that's pretty much it. In my new research I'm just starting I'm interested in how people who would never take a quid-pro-quo bribe don't see anything venal about gift reciprocity.
McArdle: What's the most interesting thing you've discovered so far in this line of research? About the different ways we view cash and gifts.
Rossman: More interesting than heroin? That work is just starting and is largely theoretical at this point. I mostly just find it interesting that whether something is framed as a purchase or as a gift makes a big difference.
McArdle: I suspect that may tell you something about politicians who are now getting embarrassed about spouses getting jobs, and loans at Very Special Rates. For decades, they've been getting these deals, and no one blinked, because that sort of quid pro quo was just the way we do things. Now, folks are starting to ask questions, and they're honestly surprised that we think of it as bribery.
Rossman: Exactly, or you can have people who do sugar daddyism and are shocked when people see it as prostitution. I think there's a lot of gift exchange in politics, much of it effectively a workaround for ethics rules. Politics is one of the areas I want to look at. [Radley] Balko had a good piece on this a few years ago.
McArdle: How effective is payola? Are songs with a Zeno's paradox pattern more or less popular than those that diffuse organically?
Rossman: Payola can be very effective. This was particularly true in the 1980s when it was less a bribe paid by labels than extortion by consultants who controlled the stations. In more recent periods it's hard to say.
There is a sense, though, in which Zeno's paradox has a hard ceiling on how much success it can yield. You typically see Zeno's paradox only in a song's natural fit format. (A format is basically like a genre). When a song crosses over into new formats, stations tend to be skeptical of it. No matter how enthusiastic the promotion guy is, or how many favors you owe him, you've got to be a bit iffy about playing a song that isn't an obvious fit with the mix of genres that your listeners are accustomed to hearing on your station. As such, stations tend to take a wait-and-see approach to crossover singles, to see what their format peers do. If I'm a soft rock station and the label is telling me to play a country song, I don't care that the label says it's good and I don't even really care that country stations thought it was good. I really want to see if other soft rock stations think it's good, and that can only happen organically.
So to have a truly huge mega-hit that crosses over between genres, the only way to achieve that is organically. You also see a similar pattern with viral marketing campaigns. Banner ads and promoted content can get an ad a certain number of page views, but to become ridiculously popular you've got to have people wanting to forward the link to their friends
McArdle: Can you give an example of a crossover song?
Rossman: Sure, on Page 78 of the book I use, "Love Song" by Sara Bareilles. This followed the Zeno's Paradox pattern for AAA and Hot AC and then a few months later, followed S-curves to crossover to AC and Top 40.
McArdle: Sorry, could you break out those formats for my readers?
Rossman: Adult Album Alternative (basically, singer-songwriters), Hot AC (up-tempo soft rock), AC (love ballads), Top 40 (mainstream pop music, much of it influenced by hip hop). All of those formats skew female, especially AC.
McArdle: We do love our love ballads.
One last question, and then I'll let you enjoy your Friday. I want to talk about transparency, because one of the things that really surprised me is that payola isn't technically illegal. It's only undisclosed payola that violates the law.
Is that right?
Rossman: Yes, the statute emphasizes disclosure rather than the act itself. I suspect that's in order to avoid making advertising illegal. And it's interesting that television also has what you might call "payola." Except that product shots aren't really controversial because they disclose them in the closing credits with the "a promotional consideration was provided by ..." language.
So, to me, the really interesting question is not why are record labels willing to pay for airplay. That's fairly obvious, it's a crucial input for their business model.
The really interesting question is why don't the stations disclose?
McArdle: I was just going to ask that!
Rossman: That law was passed in 1960 or 1961. There have been three major payola scandals since then.
I don't think ... that the stations are being subverted by their venal staff. That might explain 1974 and 1986, but by Spitzer most payola was going to the station itself (in the form of promotional support) and not to its staff. Rather I think there are two explanations.
One, there's a certain romanticism about music. The premise of a radio broadcast is that the disc jockey is excited to play these cool records for you. A "we just paid to play that" back-announce would spoil the premise.
Two, it's just plain cumbersome to make a disclosure. TV has fine print, but radio doesn't. The only way to disclose is to interrupt the broadcast by having the disc jockey read an announcement. Stations don't even like to have the DJ back-announce the song and artist, and there's nothing embarrassing about that. They're going to like it even less to have the DJ pause the music for a few seconds to confess the station's venality.
McArdle: So, to tie this interview to a news peg -- we journalists like to do that -- I wonder what this suggests about insider trading. Specifically, the folks who say that we should make insider trading legal, as long as you disclose.
Rossman: Well, yeah, one implication of that is that you'd have to make disclosure as simple as possible. If it requires getting an attorney to spend 15 billable hours, then people might prefer to break the law rather than do that. But if it's just checking off a box, it might get some take-up. I'm not an expert on financial markets or laws so there are probably a lot of other problems with disclosure, but let's just say that making it painless to disclose is a necessary but not sufficient condition to get disclosure to work.