I love the spring. Not, of course, because of the glorious weather, since we don't have any. But because it's time for the annual BSA report on piracy, which is guaranteed to provide me with hours of innocent fun as I go through finding its methodological errors and dodgy data.
Long-suffering readers of this column may recall my previous discussions of these reports and their egregious flaws. For example, back in 2010, I pointed out that the BSA's claim that reducing PC piracy by 10% would create $142 billion in new economic activity was nonsense - the money saved by piracy does not simply disappear, but is spent elsewhere. In 2011, I noted that the BSA used the misleading phrase "commercial value of software piracy", something repeated in 2012, when the BSA spoke of the "commercial value of this shadow market of pirated software" as if that had any relevance to what was happening on the ground.
So, with a certain excitement I began reading this year's report, intriguingly entitled "Competitive Advantage: The Economic Impact of Properly Licensed Software". Here's the summary:
properly licensed software has a positive impact on national economic activity that is more than three times the impact of pirated software, according to a new study from BSA | The Software Alliance. And the additional economic value associated with lawful software use is especially pronounced in developing markets: Every dollar invested in properly licensed software in low-income countries yields an astounding $437 in additional national production, on average.
Well, yes, that is pretty astounding. Indeed, so astounding it clearly deserves closer scrutiny. Here's what seems to be the core reason that licensed software is preferable to pirated copies:
Fully licensed software, in particular, improves effectiveness and efficiencies in enterprises by reducing exposure to viruses and other security vulnerabilities, meaning fewer system malfunctions, downtime, and IT repair costs. This is because licensed software comes with value-added services that provide access to upgrades, patches, and manufacturer support such as training and problem resolution.
What's amusing is that this is something I noted in a reply to a comment on my 2011 article:
it could be argued that the widespread use of unauthorised copies is beneficial to the West, since companies in emerging economies must spend far more time dealing with viruses, Blue Screens of Death etc., all of which lowers their competitivity.
So, obviously, I'm delighted that the BSA is finally catching up with what I said two years ago. But how does this belated understanding lead to that "astounding $437" figure quoted above? It all comes down to mathematics – which is good news, since I have a couple of degrees in the subject lying around somewhere. Here's the basic idea:
This study seeks to quantify, at the national level, the economic value derived from the productivity gains and operational efficiencies that enterprises achieve with software. To that end, researchers at INSEAD employed a production function model (see figure 3), which relates changes in various economic inputs — such as labor, physical capital, or PC software — to national output. A production function determines the "elasticities" of national production (GDP) associated with each of these variables — that is, the extent to which changes in each input correlate with changes in GDP.
Once those elasticities have been calculated, they are then used as the basis for claims about the beneficial effects on the overall economy of using licensed software. But more revealing, I think, is the alternative approach used in the report:
Another way to assess the value of software to national economies is to calculate the return on investment (ROI) in GDP from spending on software — or the additional GDP created by each extra dollar invested in software. Again, properly licensed software delivers a consistently higher return than pirated software.
While the preceding analysis of overall GDP benefit finds that higher-income markets have the most to gain by increasing spending on legal software, it is emerging economies that see the greatest returns on a dollar-for-dollar basis. A one-dollar investment in fully licensed software is associated with an average return of $437 in additional GDP for low-income countries, compared to $140 for middle-income countries and $117 for high-income countries.
What that says is that if an extra $100 is spent on a legal copy of Microsoft Windows, say, it will produce an additional $43,700 GDP for a low-income country. Who would have thought that a single copy of Windows could produce so much extra wealth? In fact, another table in the report shows that some countries are even more fortunate.
According to the BSA, a one-dollar investment in fully licensed software in Zimbabwe produces the rather extraordinary return of $1,918 – which means that buying a single copy of Windows would add $191,800 to the Zimbabwean economy: I imagine Robert Mugabe is rushing to his local software shop in Harare at this very moment. This is, of course, utter nonsense, and is born of using economic models without really thinking about whether the results make sense. Let's look more closely at the details of how the BSA's figures were derived.
To its credit, the BSA have provided a brief description of the methodology employed:
To conduct the research, INSEAD, one of the world's leading business schools and research institutions, utilized an established macroeconomic approach known as a production function model to analyze the changes in national output (GDP) that result from changes in various economic inputs. A production function is used to determine the elasticities in GDP associated with those inputs — that is, the responsiveness of GDP to a given change in a particular variable.
The input variables included labor, physical capital, information and communications technology (ICT) capital, and PC software, as shown in the figure below. For the purposes of this analysis, PC software was extricated from the broader ICT capital variable and demarcated as licensed and pirated software in order to provide specificity in terms of the discrete GDP impacts associated with properly licensed and pirated software. The commercial values of licensed and pirated software serve as proxies for software use in the model.
The relationship between output and the various inputs, including software, is expressed by the following national production equation:
Qit = Ait.F(Lit, Kit, KITit, Xitl, Xitu)
That simply says that the GDP depends in some (unspecified) way on Labour (Lit), Physical Capital (Kit), non-software ICT Capital (KITit), Licensed Software (Xitl), Pirated Software (Xitu) and a factor Ait (these all use subscripts and superscripts in the original – apologies for the less elegant format adopted here.) The report goes on to give the more specific formula employed:
To compute the elasticities, the production function is converted into an additive linear empirical equation using the logarithmic form of a Cobb-Douglas function:
Log Qit = a + b1Log (Lit) + b2Log(Kit) + b3Log(KITit) + b4Log (Xitl) + b5Log(Xitu) + eit
In this equation, b1, b2, b3, b4 and b5 are parameters to be estimated and represent the elasticities of output Q with respect to each input. The term eit represents an error term, or the amount by which the observed values of Q deviate from the estimated values as a result of characteristics in a country (such as different work ethics) that may affect national output but are not measured in the equation; and a is a constant term.
The Cobb-Douglas function goes back to 1927, and originally only applied to the simpler situation of GDP depending on just two factors: Labour and Capital. Moreover, it makes the huge assumption that overall GDP depends on those variables in an extremely simple way – that it is proportional to powers of them multiplied together. There is no reason why that assumption is correct, and even less reason why it should be correct for a more complex 21st-century market involving licensed and pirated software, which are then pulled out from the more general Capital variable (mathematically, separating them in this way produces quite different equations).
Not only that, the equation itself has two fudge factors, called "a" and " eit" in the BSA study. That means that the accuracy of the results is already compromised even before a mere eight years of figures are plugged into the equation in order to estimate the elasticities. Moreover, the figures for licensed and pirated software are based on BSA estimates, and are therefore not independently verified, unlike the more general economic variables. The BSA report itself admits:
The impact of pirated software use is less clear. Across all countries studied, it is associated with an elasticity of between 0 and 0.03 percent — at most. There is too much variability in the results to confirm the results more precisely than that.
Against that background, to claim as the report does that the US's GDP benefit if the use of licensed software increased by one percent, is precisely $15,094 million, is absurd: giving five figures of accuracy would require at least six figures in the numbers used to calculate it. But the GDP benefit is based on multiplying the GDP with the respective elasticities (licensed and pirate), and then calculating the difference. The elasticities for both licensed and pirated programs depend entirely on BSA's figures, which, as we've seen in previous years, probably aren't reliable to more than one significant figure either, with the pirated figure particularly uncertain.
As usual, then, the BSA's claims about piracy turn out to be based on exaggerated claims arising from simplistic models using limited data, the most important of which comes from the BSA itself. But of course the flaws in this study go much deeper, because it totally overlooks the effects – or even the existence – of free software.
For example, in the Cobb-Douglas function, no account is taken of the effect of free software on GDP. And yet we know that open source is having a massive impact around the world as more and more companies switch to it. To omit it simply renders the model even more inaccurate.
But in a sense, that doesn't matter. The premise of the latest BSA report is subtly different from those published in previous years. Where those tried – unsuccessfully – to argue that pirated software was damaging the overall economy because of money not spent, this year's takes a different tack. Here, the argument is that licensed software is better than unlicensed software because of the hidden costs of using pirated copies.
We learn more about that in a Microsoft-sponsored study that is referenced by the BSA one [.pdf]. Here are some of its key points:
Counterfeit users put themselves at risk by using pirated software, regardless of whether their software is obtained through Peer-to-Peer file sharing, from a street market, or from a hardware retailer that fails to install genuine Microsoft products on the machines they sell.
Price excluded, the attribute that users value the most in their software is protection against viruses and hackers, yet millions of consumers and businesses still choose to install potentially-infected counterfeit software on their computers.
Of course, open source software like GNU/Linux and LibreOffice is almost totally unaffected by viruses (I've certainly never seen one on these systems in 18 years of using free software). Which means that it has all the benefits of licensed Microsoft software – but is free.
Regularly downloading updates for the software that users employ is a vital part of maintaining a secure and efficient computing experience. One in four of the counterfeit versions of Microsoft software tested in our study was unable to download automatic Windows and Office updates, and one in five was also unable to manually install Windows updates.
Again, that's never a problem for open source, which usually offers automatic – and free – updates as a matter of course.
Perhaps the most universal and annoying wait of all is the time spent waiting for a computer to boot. In boot time tests, genuine Windows machines outperformed those with pirated installations 60% of the time by an average of 56%. Recapturing this time on a daily basis at work and at home would be a boon for most users. Less time spent waiting is more time spent being productive, connecting with friends and family, or having fun on your PC.
Certainly true – as is the fact that GNU/Linux installations typically boot in a few seconds compared to the aeons of even the allegedly-swift Windows installations. Again, open source wins hands down over Microsoft.
Whether you are a road warrior who depends on that extra bit of battery life, or you are just trying to control costs, energy efficiency is an area of concern due to its ramifications on battery life, energy costs, and the environmental impact of computing. A majority of the PCs installed with genuine Microsoft Windows and Office software in our study consumed less energy and exhibited superior battery life than their pirated counterparts
That seems unlikely, but I'm prepared to take their word for it. But what it overlooks is that GNU/Linux has always been able to run on machines with far lower hardware specifications, thus saving energy. Moreover, open source typically runs well on older systems, which means you don't have to throw them out as unusable as is the case with Microsoft's demanding software. Less waste means less pollution and less money that you need to spend.
So what this all boils down to is that the fundamental premise of the latest BSA study – that licensed proprietary software is better in many ways than pirated copies – actually applies to open source software even more strongly, with the added virtues that the software is free to try, to use and to modify. That means the potential economic impact of free software is also even greater than that offered by both licensed and unlicensed proprietary software. It's yet another reason for governments around the world to promote the use of open source in their countries by everyone at every level.
In conclusion, I have to say I'm really grateful that the BSA has made the case for free software so cogently in its report, and look forward, as ever, to next year's edition.