In section 7.3, the EC FLOSS Report tells us that firms working on FLOSS are not only faster, they’re also richer. Geez, faster and richer? Are they better looking too? Yet despite the authors’ assertions, I’m not sure we should be green with FLOSS envy just yet.
The authors are trying to erect a foundation upon which to rest the report’s main argument: to grow European economies, the EU should bolster the European ICT sector by betting big on FLOSS. In 7.1 and 7.2 they claimed that FLOSSers work faster (are more productive). As we discussed in our last blog post, FLOSS: The Software Hare that Beats the Proprietary Turtle?, the data didn’t really support the claim. But we were willing to give the authors the benefit of the doubt because FLOSSers should work faster, given the often imitative nature of most FLOSS projects. In this section, however, they are starting to knowingly fill the foundation with hollow claims.
The basic claim of 7.3 is that FLOSS-contributing firms perform better than firms that aren’t FLOSS contributors.
They base these claims on two key findings, based on the firms who contributed code to Debian 3.1, a popular FLOSS distribution. First, they find that FLOSS contributors have more revenue than their industry peers. Second, they find that these firms generate greater revenue per employee than the industry average.
These findings are the result of an analysis of the 986 firms that contributed at least some code to the Debian 3.1 project. Of those firms, the authors found matches for 158 in the Amadeus database of company information (Note: Only a quarter of those matched, or 40 out of 158 firms, have a European presence, but it doesn’t dissuade the authors from suggesting that Europeans should build on their leadership in FLOSS development).
Now comes the great leap of FLOSS faith. The authors count all of the revenue (and employees) for each FLOSS contributing firm when comparing their performance to industry averages. That means all the revenue of giant manufacturers like Nokia, and Siemens are counted as revenue for “FLOSS Contributors”, even if their revenue had no reliance on coding or using FLOSS products. While we don’t know all the firms represented in the 158 (when we requested a list of the firms, the author declined to provide any further information until they complete a more detailed analysis of company contribution to open source), we do know that top contributors include SAP, IBM, Sun, and AT&T.
To give a sense of how ridiculous this analysis is, let’s look at AT&T’s finances, for example. Does anyone want to guess what percentage of telecom giant AT&T’s $47 billion in revenue is attributed to the fact it donated some code to an open source project? What about the percentage of its 160,000 employees that are paid to work on open source projects? Probably close to nil on both counts. Yet, AT&T represents 12.5 percent of the revenue and 28 percent of the jobs the authors attribute to FLOSS contributors.
Nonetheless, the authors include all revenue and employees of FLOSS contributing firms in proclaiming that “firms contributing code to FLOSS projects have in total at least 570 thousand employees and annual revenue of Euro 263 Billion.
But when you start digging into the data underlying these findings, you discover nothing to justify even a hint of cause-and-effect here. It’s as if the authors noticed some puddles on the ground while it was raining, and proclaimed the discovery, “Puddles cause rain! The funny part is that the authors know it and admit it.
In a bit of Orwellian doublethink on pg. 61, the authors write:
So, FLOSS alone is not an explanation for the performance of FLOSS contributing firms – as we noted earlier, we highlight a correlation, rather than identifying any causal relationship. But what is important to recognise is that – most clearly in some sectors such as software producers – firms that contribute to FLOSS perform better, on some measures of revenue per employee, than firms that produce only other kinds of software.
The FLOSS authors are playing an old rhetorical trick. To insulate themselves from criticism, the authors smartly pepper the section with admissions that they have found no causal linkage, but merely a correlation. Yet, in the next breath (and the one after that, and the one after that) they ignore the disclaimers to suggest that this correlation is important enough for Europe to bet its future on FLOSS.
What exactly is the explanation for this admittedly weak correlation? Perhaps big revenue-rich firms can better engage in extracurricular activities such as FLOSS projects. We won’t know for sure until the authors reveal the identity of the 158 firms in their sample.
In the end, this weak correlation barely deserves a footnote, but the authors dedicated an entire section to it. Moreover, it raises questions about the reliability of the FLOSS report. Especially considering the authors’ own mantra on sound methodology:
The design of the research methodology provides for an approach based on sound expertise in economic analysis…backed by high standards of academic rigor….
[p.16]The authors will need to rely on richer analysis and better looking data to convince Europe to pursue pro-FLOSS industrial policies, because the findings in this section do not create the strong edifice for FLOSS that this report seeks to erect.