Peer Influence on Music Choices

You are here :: Home » Publications » Press Releases » Peer Influence on Music Choices

Peer Influence on Music Choices

Forecasting which songs will be popular is like hitting the jackpot for music producers. But a new study by Columbia University researchers finds that it is virtually impossible to predict which ones will be successful. This is because of social influence. In a world where individuals make decisions about what they like or don't like without knowing anyone else's opinion, we could easily predict the success of a product. However, in the world as it is, we cannot shut out the influence of others and that influence has a significant impact on how popular songs (or movies or books) become.

The study by Matthew Salganik, Peter Dodds, and Duncan Watts, entitled "Experimental Study of Inequality and Unpredictability in an Artificial Cultural Market," appears in the February 10 issue of Science. The results demonstrate that social influence - that is, knowledge of consumption choices made by others - contributes to both inequality and unpredictability in cultural markets. Popular songs become more popular while unpopular songs become less popular as people accumulate more information about what others have chosen. This is true even with indistinguishable groups of subjects evaluating the same set of songs, so a better understanding of the audience or the product will not eliminate this inequality or unpredictability.

The Experiment

To reach these conclusions, the researchers conducted a web-based experiment with over 14,000 participants, most recruited from a teen-interest web site. The experiment created an artificial music market of 48 unknown songs from unknown bands. Participants were able to listen to any of the songs and, while listening, were asked to rate them on a scale from 1 to 5. They were then given the option to download the song.

Participants were randomly assigned to one of two conditions-an "independent" one where they did not know how others felt about the songs and another where they were exposed to peer opinions. Those in the "social influence" condition were told the number of previous downloads for each song. Additionally, participants in the social influence condition were randomly assigned to one of eight "worlds" which evolved independently of the others.

The Results

The researchers found that, while quality songs (as determined by market share and rank by the independent participants) do better on average, songs of any quality can experience a wide range of outcomes. "The 'best' songs never do very badly, and the 'worst' songs never do extremely well, but almost any other result is possible," they said. A song in the top 5 in terms of quality had only a 50% chance of finishing in the top 5 of success.

An excellent example of this unpredictability is the song "Lockdown" by 52metro. It was ranked 26th in quality (basically right in the middle) by participants who had no knowledge of others' opinions. But in one of the social influence worlds it came in first and in another world the very same song came in 40th out of 48.

While these conclusions could be reached using theoretical models, it would be impossible in a theoretical situation to rule out the possibility that X is popular because X is the best. But this experiment was able to show through direct observation that how popular popular songs are, and even which songs become popular, depends in large part on how much people influence each other's decisions.

Though the experiment cannot reproduce the complex nature of real-world cultural markets, it does yield some important and counterintuitive results that have wide market implications. The authors suggest "that experts fail to predict success not because they are incompetent judges or misinformed about the preferences of others, but because when individual decisions are subject to social influence, markets do not reflect pre-existing individual preferences." Market outcomes are unpredictable, no matter how much skill or information one has.