What Filesharing Studies Really Say: Part 15 – People Who Download Buy More

This is part 15 of the re-publication of my meta-analysis on what filesharing studies really say.

[Originally published on ZeroPaid in May of 2012.]

We are now well into the final stretch of our long-running series of what file-sharing studies really have to say. In our fifth to last study of the series, we cover a study that found that people who download copyrighted material also purchased more music legally.

This particular study goes by the name “Measuring the Effect of File Sharing on Music Purchases”. It was published clear back in 2006 by the Journal of Law and Economics.

Before I go any further here, I just want to point out that because of the age of this particular study (6 years old), there wasn’t as much to work with in terms of deciphering what a lot of the numbers really mean. There was less other published works to work with which can lead to the odd claim here and there that is way out into left field – especially compared to the other studies we have covered previously in this series alone.

So, having said that, we’ll now look at the abstract. This abstract says, in part, the following:

I estimate the effect of music downloads on the probability of purchasing music using a European individual-level cross section of 15,000 people from 2001. A simple comparison of means shows that people who regularly download music online are more likely to buy music. The positive relationship persists when controlling for observed characteristics. However, simultaneity between tastes for music and peer-to-peer usage makes it difficult to isolate the causal effect of music downloads on music purchases. To break that simultaneity, this paper uses measures of Internet sophistication and the speed of the Internet connection as instruments. The results suggest that peer-to-peer usage reduces the probability of buying music by 30 percent. On the basis of my estimates, back-of-the-envelope calculations indicate that without download sales in 2002 would have been around 7.8 percent higher.

This is what we mean by slightly off comments. What this abstract appears to be saying towards the end is that because high-speed internet access exists, the access to file-sharing which causes a 7.8 percent decline in music purchases. That, at least to me, seems like quite a stretch, but we’ll see how this study actually carries out this math. What is also important is what is said towards the beginning of the abstract and that is people who download more also tend to purchase more music. So, that will make how this study found the reduction even though there is a noticeable benefit to having file-sharing around that high-speed internet causes a decline in music sales.

In what is, to me anyway, starting to be a tiring image to see, there’s the graph that shows music sales from 1990 to 2003. It, like a lot of other studies, notes the decline in late 90’s to early 2000’s. The study also covers some of the litigation tactics the industry as well as covers basic architecture of file-sharing. Again, this has been seen throughout numerous studies already.

The study shows how it came about it’s 30% probability:

I employ measures of Internet sophistication and access to broadband as instruments. My estimates indicate that peer-to-peer usage reduces the probability of buying music. Using measures of sophistication as instruments, I find that music downloading reduces the probability of buying music by 30 percent.

What is the math used behind this number? The study didn’t necessarily show it yet. The study then goes on to discuss the background of where some of the numbers of losses come from. One thing that was noticeable was that the studies authors believes that first empirical analysis was used by plaintiffs (who hired Soundscan) when litigating Napster. The study also discusses financial records of big music corporations as well as compression of audio into MP3.

The study then began discussing downloads and sales saying:

Across the overall sample (15,133 observations after dropping individuals with missing values), 39.3 percent bought music during the month prior to the interview, 9.0 percent regularly download MP3 files, and 50.7 percent have Internet access.

In what seems to be evidence towards the concept of people who download music buy more music, the study summarizes the following data:

The percentage of people who bought music is much larger among the group who regularly download MP3 files (55.8 percent) than among those who do not (37.7 percent), which suggests that MP3 downloaders have a strong taste for music. Considering only those people who have an Internet connection at home (5,917 observations), 47.1 percent bought music during the month prior to the interview and 21.0 percent regularly download MP3s. Again, the fraction of people who bought music is higher among those who regularly download music (55.1 percent) than among those who do not regularly download music (45.0 percent).

Next, the study discusses how it came to the conclusion of downloading an MP3 reducing the probability of purchasing music by 30 percent. It seems the data was derived from the survey of 15,000 people.

The study also discusses something that sounds familiar if you’ve heard a talk or two from Cory Doctorow:

Downloading music from an illegitimate site is free of charge. However, there are other costs associated with the process of downloading. An important one is the cost of time involved in this process. The risk is that sophisticated Internet users are people with a low value of time and therefore for a given taste for music more willing to substitute music purchases for freely downloaded music.

What this reminds me personally of is the comment of how younger people are money poor and time rich while older people are money rich and time poor – thus as people get older, the likelihood of them having money in later years tends to insinuate that the younger people are merely future customers. There are similarities to be had here.

When I got to this paragraph, though, I have to admit, I had to laugh a little:

Publishing a Web page is positively correlated with having an MP3 player and a CD writer but is negatively correlated with owning a Walkman and a hi-fi stereo. As suggested before, this may be explained by a negative effect of music downloading on purchases of complements of music in regular formats. Excluding “own a Web page” from the list of instruments results in a similar estimated impact of downloading MP3 on the probability of purchasing music.

When I read this, I envisioned huge billboards posted by the RIAA saying “Beware of the web designing MP3 toting teenagers for they will pillage your music and create a high traffic webpage for your dog!” with a giant picture of a teenager sporting an eye patch and white ear buds with a circle and a line drawn through him. Then in smaller print along the bottom, the billboard saying, “Always trust the teenagers who listens to cassettes” along with a tiny picture of a clean cut teenager giving a thumbs up.

This comment was something I personally found interesting. While discussing some of the mathematics being used as well as describing a large series of tables, the study notes:

If the instruments were selecting individuals with weak taste for entertainment or individuals with low probability of buying entertainment goods, a negative impact of the predicted values of MP3 file downloading on purchases of these goods would be expected. None of the regressions show that downloading music reduces the probability of buying other entertainment goods. This suggests that the negative effect of downloading music on the probability of buying music is not driven by instruments that pick individuals with a weak taste for entertainment goods.

This actually provides an interesting connection to one of the theories I found in earlier studies that an increase in competition might be a negative force on the music industry. One argument could be made (which I have heard before in debates) is that something like a video game is much more interactive and touches on multiple senses (touch, sight, sound) whereas something like listening to music only touches on one sense (the sense of sound) and is a passive experience. I happen to agree with that particular point and would say that this provides quite a challenge for the music industry. We should point out that, yes, that’s not the angle this study intended here even though it touches on this indirectly. The angle that this study seems to suggest is that because the song is much more easily downloaded compared to other forms of entertainment, then people would pirate music and legally consume other forms of entertainment. I don’t think that really holds true anymore, but again, this study is from 2006, so it may have been more truthful then than now.

The study then returns to the 30 percent figure and had an interesting note:

Instrumenting with measures of Internet sophistication, the results in the last section indicate that downloading MP3 files online reduces the probability of buying music during the month prior to the survey by 30 percent. Downloaded music may be shared with people who do not download MP3 files and affect their purchases as well. In this event, the estimates understate the effect of online music downloading on music sales.

To my knowledge, this is considered a slightly more sophisticated form of word of mouth. It’s the concept of “Hey, I heard this song that was great, want to hear it?” Which I would be surprised if artists didn’t like this kind of publicity.

The study then admits:

The database does not contain information on quantities of music purchased or on intensities of music downloads to calculate what music sales would have been in the absence of music downloading. Moreover, the probability of buying music and intensity of purchases may vary in complicated ways.

I don’t think you could be more honest here. There is a lot more going on than just taking a number of purchased music and taking a number of downloading and hit “divide” on the calculator (which was seen in the study that says that losses amounted to less than $2 per album)

What was kind of strange was the combination of these two paragraphs:

Across the overall population, 9 percent regularly downloaded music in 2001. Therefore, if both digital music users and nonusers had the same propensity to buy music, the effect on the music industry would be a reduction in music sales units of 2.7 percent.

But digital music users have a greater propensity to buy music, which indicates that a correction for the heterogeneity in the groups is needed. It was shown in Table 1 that the probability of buying music for people who download music is higher than the probability for nonusers of peer-to-peer systems. Also, using my estimate with a treatment on the treated interpretation, the proportion of music downloaders who would have bought music would have been around 30 percent higher if the possibility of downloading music did not exist. This suggests that the probability of buying music for downloaders would have been around double the probability of buying for nondownloaders

And this right after:

As there are no data on quantities of music purchased, an assumption about the number of units bought is needed. A probably conservative assumption is that users and nonusers of peer-to-peer systems, when reporting that they bought music during the month prior to the survey, bought the same quantity of units. With this assumption, the estimated impact on units sold at the 2001 level of file sharers would be a reduction of 4.9 percent. If users of peer-to-peer systems not only have greater propensity to buy music but also would have bought more units conditional on buying, this would be an underestimate of the impact.

I’m no economist, but I get leery whenever I see a percentage value, then the word “assumption” and then a different percentage value after. I do wonder if we are starting to trail off into the territory of guesswork here. Then, this paragraph came after:

In October 2001 Napster was over and Kazaa was not yet an option. In 2002, the number of users increased from 3 million to 5 million (IFPI 2003). Assuming that this magnitude is representative of the increase in the number of downloaders in the European countries considered, and that these new downloaders have the same probability of buying as people who were downloaders in October of 2001, sales in 2002 would have been around 7.8 percent higher.

So, a network population was added citing IFPI statistics and we’ve arrived at the 7.8 percent seen earlier. Call me dense, but even though I may not know exactly what is going on here, these numbers don’t look entirely trustworthy to me.

With that, we look at the conclusion:

There is controversy about the effect of this strategy on the number of downloads. While suing individuals who offer music as opposed to individuals who download files may reduce the number of files available to download, it is not clear whether this would actually affect the number of downloads. This is important when considering the public-good nature of the files offered online. In addition, while it appears that the number of users has decreased for some popular sites such as Kazaa, the legal strategy appears to have induced individuals to use alternative and less popular sites and forums where the risk of being prosecuted may be lower.

I think the author meant “networks” instead of “sites” here, but minor quibble.

The conclusion, overall, discusses enforcement and how targeting file-sharers may be difficult in various circumstances. While it mentions the 30 percent reduction of probability (probability being a key word here) in purchasing music, the percentage of music that would otherwise be bought that was guessed at didn’t make the conclusion (nor did the positive effects of file-sharing, curiously enough)

Overall, I’d say that some of the rationale of some of the math is rather puzzling. In addition, it’s a little odd that the conclusion pretty much dropped all of the figuring and just talked about how enforcement would be difficult. Still, interesting to know that this study does cover the point that file-sharers can have a positive effect on music sales even though it was only mentioned in passing a couple of times while the focus was trying to find the alleged negative effects of file-sharing.

Drew Wilson on Twitter: @icecube85 and Google+.

Leave a Comment

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Scroll to Top