Last week I wrote about a report from the US Government Accountability Office (GAO) that examined the reliability of recorded music industry research papers seeking to estimate the loss from “piracy” in the digital field, and found all of them seriously wanting.
As far as I know, no similar analysis has been carried out for European reports. So I thought it might be interesting to look at one particular European report on the subject - not least because I've heard that its findings influenced some of the MPs voting on the Digital Economy Act.
The report is called “Building a Digital Economy”, and it comes from the International Chamber of Commerce (ICC):
A new study entitled “Building a Digital Economy: The Importance of Saving Jobs in the EU's Creative Industries” released today predicts losses due to piracy to reach as much as 1.2 million jobs and €240 billion in retail revenue by 2015 in the creative industries most impacted, based on current trends and assuming no significant policy changes.
The study shows that this sector is already experiencing substantial losses. In 2008 the creative industries most impacted by piracy (film, TV series, recorded music and software) experienced retail revenue losses of €10 billion and losses of more than 185,000 jobs due to piracy.
These are big, frightening figures, and it's easy to understand why politicians might be impressed by them. But of course, anyone can bandy numbers around: the question is, do they have any basis in objective fact?
At 68 pages, the report [.pdf] is quite long, and very professionally produced. It is stuffed full of tables designed to bolster its case. In calculating the claimed loss from piracy, a fairly simple approach is adopted. The number of copyright infringements are multiplied by the substitution rate (the percentage of people who would have bought stuff if they had not downloaded it for free) times the unit retail price. This is calculated for the top 5 economies in Europe, and scaled up proportionally to include all the other EU economies. The claimed job losses caused by that revenue lost is obtained by dividing the latter by an average sales revenue per employee.
Now, you could criticise this for being overly simplistic, but as a first-order estimation, I think it's fair enough. The authors of the report adopt quite reasonable substitution rates, depending on the medium. So, ultimately, the credibility of those big numbers comes down to one factor: the number of copyright infringements that are input into the basic formula. The report is well aware of this, and much of its total length is given over to presenting the background to the figures it uses. But in the main body of the report itself, the figures are simply stated without explanation, and their provenance explained by the footnote “National assumptions for copyright infringement are detailed in Appendix 1.”
Appendix 1 therefore contains the real meat of this research. Naturally, I turned first to the UK market, since it's the one I know best. We find the relevant figures in App. 1.13 – “Assumptions for revenue losses due to pirated music in the UK”. The number of copyright infringements per year, in millions, is given as 1177 (I'll concentrate on digital music downloads for the purpose of this discussion, since it's the one most people focus on in arguments about the scale of piracy.)
Again, the crucial issue is: where does this very exact figure of 1,177 million copyright infringements come from? And this time we finally discover the source: it's the BPI:
the representative voice of the UK recorded music business. We are a trade organisation funded by our members - which include the UK's four major record labels and hundreds of independent music companies.
So, the figure is not from independent research, it's not even from an independent organisation, but from the very industry group that pushed so hard for the Digital Economy Act among other things – and probably did so drawing on this report, whose figures it had supplied. But wait, maybe the BPI obtained those figures from some independent outfit. After a search on the BPI's site, this is all I could find: