By Brett Danaher, Michael D. Smith, Rahul Telang

The HADOPI legislation is a significant and controversial measure with passionate supporters and passionate detractors. As co-authors of a study analyzing the impact of the HADOPI law on music sales, we have watched with concern as our finding have been interpreted in ways that are removed from our original intentions. Because of this, we wanted to take a moment to clarify what we believe our study can say about the effectiveness of HADOPI, while also — and maybe more importantly — clarifying what our study cannot say about the broader policy questions raised by this and other similar legislative efforts to reduce piracy.

In terms of what our study can say: simply stated, our research question is “did HADOPI cause an increase in iTunes music sales in France?” In raising this question, we would like to first point out that as academics we have no philosophical or practical predisposition about what the answer is or what it should be. Our sole goal is trying to understand what the data say.

Given this, we also note that, as with most empirical social science research, there is no perfect way to answer this question. Lacking a perfect approach, we used a well established statistical technique that found a group of countries with iTunes sales patterns that were statistically similar to French iTunes sales prior to HADOPI, and then analyzed whether iTunes sales in France diverged from sales in these “control group” countries after HADOPI. Our analysis finds that iTunes sales in France increase by 22.5-25% in France relative to sales in the control group. Further, in the paper we note that if this increase were causally related to HADOPI, one would expect to see a larger increase in sales among heavily pirated genres than for less heavily pirated genres, which is exactly what we find: The post-HADOPI increase in Rap, Hip Hop, Rock, and Pop sales is larger than the sales increase for the Classical, Christian, Folk, and Jazz genres. Finally, we are also able to show that our results don’t appear to be caused by specific price promotions or France-specific advertising by individual labels or bands.

By itself, does this conclusively establish that the observed increase in French sales relative to the control group countries was caused by HADOPI? Not necessarily. Indeed, we note in the paper that any effect that was unrelated to HADOPI and yet led to a disproportionate increase in French music sales in early 2009 relative to the control group countries, and that led to a larger increase in heavily pirated genres than in other genres could have caused the effects we observe.

After we released the initial draft of our findings, Le Monde proposed that iPhone sales could provide just such an explanation if iPhone sales spiked in France relative to the control group countries in early 2009, if iPhone purchases drive iTunes sales, and if iPhone purchasers are disproportionately more likely to purchase heavily pirated music genres than other music genres.

Le Monde’s evidence to support this counter-explanation is available on their website ( Their argument uses Google trend searches for the term “iPhone” in France and in our control group countries and shows a spike in French iTunes sales that seems to correspond with the Christmas holidays and with release of new iPhone models.

However, after looking at this explanation carefully, we don’t believe that it explains our results for two main reasons. First, it’s important to note that our results are driven by relative changes in French sales, not absolute changes; and thus if interest in new iPhones spikes in France at about the same level as in the control group countries, this would not explain our empirical results. Indeed, in the chart below we reproduce Google trends data documenting searches for the term iPhone in France and in our control group countries (normalized to initially fall on the same scale) and this chart seems to show that searches for the term ‘iPhone’ in France do not significantly diverge from the normalized levels of searches in our control group countries.

But even these data only provide a weak proxy for actual iOS device sales. To truly test Le Monde’s hypothesis that iOS device sales led to the observed increase in French iTunes sales, one would need to know the actual penetration levels of iOS devices in France and in the control group countries.

Fortunately, these data are available from IHS Screen Digest. And in an appendix to our paper (available here we analyze these data and find that changes in iOS device penetration do not appear to explain our result. Notably, the change in iOS device penetration in France between 2008 and 2009 is nearly the same as the change in penetration in the control countries, and if anything the change in iOS device penetration is smaller in France than in the control countries when measured from 2008 to 2010.

Of course, this only addresses one potential counter-explanation for our results, and we remain open to exploring other possible counter-explanations that are unrelated to HADOPI and explain the observed results. All we are saying is that based on the best data we have, we continue to believe that HADOPI provides the most reasonable explanation for the increase in French sales we observe and the disproportionate increase in sales of highly pirated genres.

Now, in terms of what our study cannot say: even if one accepts this result, one might still ask “do your findings establish that HADOPI is a ‘good’ law?” As we carefully outlined in the paper, there are several reasons why this question is beyond the scope of our current study. First, even if HADOPI has caused French music sales to go up, we do not have data to answer questions like “what are the direct and indirect costs of HADOPI,” “do the benefits of HADOPI outweigh the costs?”, or “are human rights and due process rights being protected?” Second, we note that we are only able to observe the impact of the enacted HADOPI law, and it’s results to date. We cannot say whether the observed increase in music sales will persist in the future, or whether a different version of the HADOPI law (say one that focused on anti-piracy awareness and education without the threat of penalties) would produce a similar effect. And we certainly cannot use our data to evaluate the potential effectiveness of other forms of anti-piracy regulation (for example, site blocking in the Stop Online Piracy Act proposed in the United States).

Still, in spite of these acknowledged limitations, we continue to believe that our data provide compelling empirical evidence that the HADOPI law made a difference in French music sales. Moreover, we believe that bringing scientific rigor to a challenging policy question provides a useful voice to inform the larger debate. As such we welcome further discussion of our results and their salience for this and other anti-piracy efforts.