Open vs closed source software: The quest for balance

Sebastian v. Engelhardt, Andreas Freytag, Stephen M Maurer, 29 October 2010

a

A

Open source software (OSS) like the operating system Linux is marked by free access to shared source code that is developed in a public, collaborative manner. While most of this activity was originally non-commercial, over the past decade companies have been asking themselves whether similar OSS methods can be made to earn a profit. This has led to an explosion of OSS-based business models and investments throughout the information and communications technologies sector (Ghosh et al. 2002, Dahlander and Magnusson 2005, Lerner et al. 2006).

Governments are similarly intrigued and have begun experimenting with various pro-OSS measures including procurement preferences, tax breaks, and grants (Lerner and Tirole 2005, CSIS 2008). At first, the implicit policy assumption seemed to be that OSS was inherently more efficient than proprietary, or “closed source”, software (CSS)1. This argued for almost any policy that promised to increase the amount of OSS. More recently, however, some politicians have begun to argue that society needs a “balance” of CSS and OSS firms (CSIS 2008). But how can policymakers recognise the right “balance”? Pro-OSS interventions make very little sense if there are too many OSS firms already.

The threshold question

The threshold question, of course, is whether governments can influence OSS at all. Ten years ago, most scholars were pessimistic. This was sensible in an era when OSS was driven by non-commercial incentives like altruism, reputation, and signalling. How do you influence a “movement” dominated by college students? (Schmidt and Schnitzer 2003).

Since then, however, things have changed dramatically. Deshpande and Riehle (2008) report that the OSS sector grew from about 500 projects in 2001 to 4,500 in 2007. Furthermore, this growth was dominated by business models in which companies contribute to a shared code base in hopes of increasing consumer demand for some related product (e.g. hardware, software) or service. This for-profit outlook is clearly responsive to government’s traditional tax-and-spend policy levers.

Western governments, then, should have little difficulty influencing OSS development. Governments in the developing world will, as usual, face bigger challenges. von Engelhardt and Freytag (2010) study differences in OSS activities across 70 countries. They find that the main predictors of OSS activity are generalised cultural factors like interpersonal trust, favourable attitudes toward scientific progress, and a culture of self-determination/individualism. No government can supply these preconditions overnight. More encouragingly, there is some evidence that strong intellectual property rights and/or deregulated markets promote OSS. Developing world governments will presumably find such interventions easier to manage.

Getting the right mix

But having power is only half the analysis. How, if at all, should governments use it? One important theoretical insight starts from the observation that OSS and CSS are both imperfect tools, each of which has distinct areas of advantage and disadvantage (von Engelhardt 2008). This implies that large modern economies will usually require a mix of both methods.

Furthermore, von Engelhardt and Maurer (2010) provide an important clue to choosing this mix. They point out that the existence of CSS code increases OSS output and vice versa. To see why, consider an all-OSS world in which each company offers consumers exactly the same shared code as every other company. By definition no company can then compete by writing more OSS code than its rivals. This lack of competition suppresses code production for the same reason that cartels suppress output. Conversely, a wide range of generic models predict that software production should peak when roughly 15% to 20% of all companies adopt OSS methods.

Without specific evidence to the contrary, this suggests that competition policy should not normally challenge OSS collaborations whose members comprise less than about 20% of the market. (Larger collaborations may also be acceptable and would have to be examined on a case-by-case basis). In the US, authorities have long applied a similar "Five Effort" rule to joint R&D ventures. This, then, is the “balance” that politicians speak of.

That would be the end of the story if we could rely on markets to deliver the right mix of OSS and CSS firms. However, there are at least two reasons to doubt this.

  • First, there are many cases in which would-be OSS companies cannot profitably enter all-CSS industries and vice versa. For this reason, we expect a significant number of pure-state industries to become “locked in” indefinitely.
  • Second, consider markets where OSS and CSS business models co-exist. We have already pointed out that OSS operates as a de facto cartel. This will normally make OSS firms more profitable than an equivalent number of CSS firms. As a result, we expect most markets to host far more OSS firms than policymakers would like.

How should governments intervene?

One obvious solution is to promote OSS (or CSS) development in markets that appear to be “locked in,” i.e. have remained CSS (or OSS) for a long time. These interventions will have to be made on a case-by-case basis. Much more general policies will be needed to adjust the number of OSS firms in mixed markets. Detailed analysis suggests that the best solution would be to tax OSS firms and use the funds to provide tax breaks for CSS firms. Of course, we recognise that OSS’s altruistic image may make such policies politically unrealistic. This, however, could easily change as the public learns to associate OSS with companies like IBM, Sony Ericsson, and LG. In the meantime, our analysis suggests that current proposals advocating pro-OSS procurement preferences are poorly motivated and should be viewed with suspicion.

In theory, government can also intervene by purchasing more OSS output than the private sector is willing to fund. Theory suggests that this may often improve welfare. That said, direct funding of OSS production would yield a host of familiar problems.

  • First, policymakers seldom, if ever, know which projects are likely to deliver the most social benefit per euro invested.

Historically, this has notoriously persuaded governments to invest in projects with little or no value.

  • Second, government investment inevitably breeds lobbying expenditures that offer little or no value to society.

On the other hand, the argument that government investment has been mishandled in the past is not definitive. The economics literature is filled with clever “mechanism design” papers that explain how government investment can be made to track private sector judgements about which projects do (and do not) offer value. Probably the most straightforward scheme would be for government to withhold funds unless and until companies made matching investments (Maurer and Scotchmer 2004).

Concluding remarks

Twenty years on, the OSS revolution is stronger than ever. In the meantime, scholars have learned a great deal about this new production method. Governments need to be aware of these insights.

References

Benkler, Y (2002), “Coase’s penguin, or, Linux and the nature of the firm”, Yale Law Journal, 112(3):369-437.

CSIS (2008), “Government open source policies”, Centre for Strategic and International Studies.

Dahlander, L and MG Magnusson (2005), “Relationships between open source software companies and communities: Observations from nordic firms”, Research Policy 34(4):481-493.

Deshpande, A and D Riehle (2008), “The total growth of open source”, in B Russo, E Damiani, S Hissam, B Lundell, and G Succi (eds.), Open Source Development, Communities and Quality, Vol. 275 of IFIP International Federation for Information Processing, Springer, 197-209.

von Engelhardt, S and A Freytag (2010), “Institutions, culture, and open source”, Jena Economic Research Papers 2010-010, Friedrich-Schiller-University Jena and Max-Planck-Institute of Economics.

von Engelhardt, S, SM Maurer (2010), “The new (commercial) open source: Does it really improve social welfare?”, Berkeley Goldman School of Public Policy Working Paper 10-001.

von Engelhardt, S (2008), “Intellectual property rights and ex-post transaction costs: the case of open and closed source software”, Jena Economic Research Papers 2008-047, Friedrich-Schiller-University Jena and Max-Planck-Institute of Economics.

Ghosh, RA, R Glott, B Krieger, and G Robles (2002), “FLOSS final report, part 3: Basics of open source software markets and business models”, International Institute of Infonomics, University of Maastricht.

Lerner, J and J Tirole (2005), “The economics of technology sharing: Open source and beyond”, Journal of Economic Perspectives, 19(2):99-120.

Lerner, J, PA Pathak, and J Tirole (2006), “The dynamics of open-source contributors”, American Economic Review, 96(2):114-118.

Lessig, L (1999), Code and Other Laws Of Cyberspace, Basic Books.

Lessig, L. (2006), Code: Version 2.0, Basic Books.

Maurer S and S Scotchmer (2004), “Procuring Knowledge”, in G Libecap (ed.), Advances in the Study of Entrepreneurship, Innovation and Growth: Vol. 15, JAI Press.

Schmidt, KM and M Schnitzer (2003), “Public subsidies for open source? Some economic policy issues of the software market”, Harvard Journal of Law & Technology, 16(2):473-505.


1 Such a view is to some extend based on arguments by e.g. Benkler (2002) and Lessig (1999, 2006)

Topics: Competition policy, Productivity and Innovation
Tags: Competition policy, intellectual property rights, Open source software, technology

Comments

Comment on the comments

We are very glad to see these ideas debated, the subject (and particularly its competitions policy implications) has not been discussed nearly enough.
It seems like we should clarify more what the cartel effect is all about. 
First: Firms can cartelize regarding various variables, not only prices and quantities but also regarding quality or sales territories. The main point is that firms manage to collaborate instead of competing, and that this has a negative effect for customers.  
Second: where is the cartel effect we referred to located? As several readers noted, OS businesses sell products (hardware or software) that comes with OS software or offer complementary services. Of course these firms compete with their “bundles”. The cartel effect only relates to the shared software, not the complementary product or service that's sold with it. But that still matters. If every manufacturer knows it will have the same shared software as every other manufacturer, then it will invest less in software. And consumers will get less value when they purchase the bundle. Another way to say this is that software output is restricted, a very classical cartel symptom. 
A related point that several of you mentioned is that the cure for cartels is competition, either from closed source firms or open source collaborations. Absolutely. Competition does indeed help. However, there are good market reasons to think that there won't be enough competition. This is something that competitions authorities should watch closely. But if you find a market where multiple OS collaborations are fighting it out, then the competitions authorities should be happy with that.  
Obviously, it is difficult to explain these details in a short essay. Readers interested in the full argument can consult: 
S. Maurer, "The Penguin and the Cartel: Rethinking Antitrust and Innovation Policy for the Age of Commercial Open Source," GSPP Working Paper (2010).
S. Maurer and S. von Engelhardt, "The New (Commercial) Open Source: Does It Really Improve Social Welfare?" GSPP Working Paper (2010).

A final remark: Of course, OSS has its advantages. Sharing code avoids wasteful duplicated efforts while each CSS firms have to write the whole code own its own. Furthermore, OSS benefits from user-producer innovations and other voluntary contributions. This is common knowledge today. That's why we do not stress this aspect in the article. But this effect is clearly taken into account in the underlying economic research papers: This is the reason why it is good to have OSS in the market – although it has the “cartel effect”. To put it short: From social perspective, the code sharing of the OSS principle has its benefits and costs when it comes to commercial OSS.
 Thanks again to everyone for their comments.  

reality shows a different picture

"OSS operates as a de facto cartel. This will normally make OSS firms more profitable than an equivalent number of CSS firms"

?? While a cartel can dictate prices, OSS firms can't. They have to sell services around the product or a related product (as the paper itself states above). So how can they be more profitable?

This also thrashes the argument, that 100% OSS firms would halt software output. OSS firms differentiate themselves through services and related products, and the innovations are in the related products (for example support for new hardware or a new software feature that differentiates until it is copied, similar to CSS).
Also OSS has the advantage that part of the consumers are also producers and they will always innovate because they want to use these new features. There is bigger involvement of companies in OSS, but that doesn't mean that the "college students" or their ideals vanished.

In my opinion the most important role of governments in the software market should be to insure competition irrespective of OSS and CSS. The strong inclination of software markets to degenerate to monopolies must be counteracted with an unconditional support for open standards.

Are you sure you meant cartel?

>> To see why, consider an all-OSS world in which each company offers consumers exactly the same shared code as every other company. By definition no company can then compete by writing more OSS code than its rivals. This lack of competition suppresses code production for the same reason that cartels suppress output.
 
If software doesn't exist, you expend resources to get it written. Frequently this means you pay someone to accomplish it if you can't or don't want to do it yourself. Since the software didn't exist before, those writing it are providing unique value. By the time others can take that software and offer it to the customer a second time, the customer already has free access to it and already paid for the service of getting it written.
 
Software development in this scenario is a service business not a retail business. There is no cartel. The prices for a person's time and skill is not controlled by any collaborating group.
 
By the way, it is odd to call the opening up of information to the public a cartel? Cartels are involved with limiting supply to increase prices and profit. Revealing blueprints and other information has the exact opposite effect, supply increases and prices drop.

You are kidding, right?

I find this article difficult to be taken seriously. (I work at a Fortune 100 which uses both OSS and CSS).
Quotes from the article:
"Furthermore, this growth was dominated by business models in which companies contribute to a shared code base in hopes of increasing consumer demand for some related product (e.g. hardware, software) or service. This for-profit outlook is clearly responsive to government’s traditional tax-and-spend policy levers."
Ok, so you are saying that govt can penalize a business model by taxing it. True. Taxes are used for social engineering purposes.
"One important theoretical insight starts from the observation that OSS and CSS are both imperfect tools, each of which has distinct areas of advantage and disadvantage (von Engelhardt 2008). This implies that large modern economies will usually require a mix of both methods."
I haven't read von Engelhardt 2008, but let's assume that it is true that each tool as distinct areas of advantage and disadvantage. This does not, by any means, imply that large modern economies will usually require a mix of both methods. One may have a net advantage over the other, even if there is a local optimum in one area. It does indicate that you will usually find a mix of both methods, but I see nothing that indicates it <strong>requires</strong> both methods.
"They point out that the existence of CSS code increases OSS output and vice versa. To see why, consider an all-OSS world in which each company offers consumers exactly the same shared code as every other company. By definition no company can then compete by writing more OSS code than its rivals. This lack of competition suppresses code production for the same reason that cartels suppress output."
So the first sentence proves that the hypothetical example can't exist. Therefore, you cannot have an OSS cartel.
Actually, if you really looked at the number of forks of OSS projects, you would realize that there are constantly competing versions of OSS software. And companies that use one or another variation will contribute something to the variation they use, in order to help sell hardware. This creates additional differentiation between different OSS approaches to the same problem domain. There is no singular "cartel" selling an undifferentiated product.
"Conversely, a wide range of generic models predict that software production should peak when roughly 15% to 20% of all companies adopt OSS methods."
Cite please.
"Without specific evidence to the contrary, this suggests that competition policy should not normally challenge OSS collaborations whose members comprise less than about 20% of the market. (Larger collaborations may also be acceptable and would have to be examined on a case-by-case basis)."
What??? Competition policy is about antitrust, not sufficient research and development into new software. OSS, by definition, mean that anyone can look at the source code and become a competitor. So you aren't talking about "competition", you are claiming that the world would have more software production if there was more closed source software. That claim is, at least, arguable, because closed source software tries to create monopoly profits and open source software does not create monopoly profits. If you assume that programmers are incentivized by the potential for monopoly profit, then those companies/programmers would gravitate toward producing closed source software. If it is actually better and is marketed better, then it will generate more consumer demand, attract more programs and your production problem is solved. On the other hand, if it is not better and marketed better, the closed source software won't generate more consumer demand and the open source software will control that market segment. Notice that there are both quality and marketing points here. The better product will typically lose to the better market product.
"Second, consider markets where OSS and CSS business models co-exist. We have already pointed out that OSS operates as a de facto cartel. This will normally make OSS firms more profitable than an equivalent number of CSS firms. As a result, we expect most markets to host far more OSS firms than policymakers would like."
What?? First, you made a claim that OSS operates as a de facto cartel. You have absolutely no facts to back that up. Second, because OSS is available to anyone, there are no entry costs to newcomers to the producers. Monopoly/Oligopoly profits can only exist if there are barriers to entry. There are no barriers to entry with OSS. To me, this implies that you can't have monopoly profits. Third, if you have "far more OSS firms than policymakers would like", that further implies that the cartels can't hold together for pricing purposes, again ending the excessive profits. Finally, I have no idea what your decision tree is for policymakers preferences. 
"Detailed analysis suggests that the best solution would be to tax OSS firms and use the funds to provide tax breaks for CSS firms."
What?? My company, which makes hardware and sells hardware and services uses both OSS and CSS software. How are you planning on taxing us? What possible justification can you have for telling us what components to use in our production?
 
 
 

Academics without information

There is so very little actually correct content in this self-referential abstract of a thesis that it begs some correct information.  First, is the false assumption that OSS is a monolithic, homogeneous block "cartel".  Withness Red Hat, Suse, Ubuntu, Debian, Mint, et cetera.  All Linux distributions, all different, competing sources.  Then add FreeBSD, OpenBSD, et al., and you have choice in OSS operating systems, nowhere near a cartel.  On a project level, you have competing projects like Gnome and KDE - where competition within OSS drives improvement.  At an application level, there are choices like Firefox & Konqueror, Postgres & MySQL, Apache & lighttpd & boa - you name it, examples abound of competing OSS applications.  They are all based on public standards, allowing freedom and ease to choose between any of them. Any deviation from standards by a package can be corrected by users. Again,there is no cartel.  Organizational adoption of OSS allows an entity to control its destiny because it can tailor the software it uses to its needs - these changes can be incorporated upstream to provide this functionality to benefit other organizations. Everyone wins. Commercial OSS entities generally make their living from supporting OSS-based solutions, be it hardware, software suites or user support.
Let's shift focus to the CSS world.  A world where a single vendor can promulgate its own "standard" reducing interoperability to its own eco-system.  This is a  predatory practice leading to "vendor lock-in."  There is no need to name names, because some obvious ones come to mind.  CSS tends to eschew public, international standards because that would force them to compete on an equal footing.  Every vendor is looking to offer that "nifty feature" or "unique file format" that forces customers to continue to use their software.  They will go so far as to lock customer data in their proprietary formats so that there is no choice other than to start from zero.  "One size fits all" is the emperor's clothes of the proprietary model - the organization must adapt to suit the vendor's model.  Oh, and CSS companies, because they charge their customers large sums of money, are often swimming in cash.  They have enough money to bribe public officials, journalists and, yes, even academics to support them.  As well as fund extravagent astroturfing campaigns.  These are not hypotheticals, the proof exists.  But I will not be doing your reasearch for you.
The very idea of "taxing OSS" to "support CSS" is risable - it is, in effect, robbing the poor to feed the rich.  Governments have the most to gain from an OSS model.  With scarce resources, the community efforts of OSS are greatly beneficial to agencies, with the final benefit to the citizen.  CSS benefits executives, shareholders, investors, and other private parties, at the expense of tax dollars from the public.  Governments are not corporations with a profit motive (nor are they the masters of a fuedal era): they are supposed to represent their constituents to their collective benefit.  Governement involvement in OSS will directly benefit the public, as the (a) the direct costs are far lower and (b) the improvements sponsored by the public sector enhance OSS for the individual end user.  Sure, this model does not result in a handful of individuals becoming wealthy billionaires, but skewing wealth distribution by transferring it from the many to the few seems to be a regressive system that puts history on its head.
 

An error I spotted, and some further details please !!

Interesting article.
Some points:
 
You say: "To see why, consider an all-OSS world in which each company offers consumers exactly the same shared code as every other company. By definition no company can then compete by writing more OSS code than its rivals. This lack of competition suppresses code production for the same reason that cartels suppress output."
 
So, you suggest that in an "all-OSS" world code production would be suppressed, since the OSS community will be analogous to a cartel.
 
You then, later on, say:
"Second, consider markets where OSS and CSS business models co-exist. We have already pointed out that OSS operates as a de facto cartel."
 
With respect, you have only pointed this out in all-OSS world, not a world where OSS and CSS coexist.
 
Your proposal is to tax OSS production and use this to subsidise CSS code generation. Given the nature and wide source of inputs to a complete OSS project, the devil is really in the detail of this proposal. On what basis would you set this proposed taxation ??
 

LWN and some thoughts

You really should have a look at the comments on LWN about your article. They raise many good points.
As for my personal opinion, 'strong IP laws and/or unregulated markets'... this is exactly the confusion the folks who introduced the term IP wanted. Patents are about government-granted monopolies, which are as far from an unregulated market as is imaginable. Software patents are hugely detrimental to the industry, serving only to allow large companies to eliminate their smaller competition, who can't afford to defend against patent lawsuits regardless of merit. And merit there generally isn't any of -- most patents are for blatantly obvious things, often things that have been done for 10 or 20 years prior to the patent application. That doesn't matter if you can't afford to defend in the legal system. Copyright, sure, but using the word IP conflates things that are entirely unrelated.
The barrier to entry for the software industry is essentially zero, and the cost of one unit of software is zero. Nothing but fixed development costs. Prices for mass-market software _should_ be approaching zero, and they are.
There's nothing unnatural or harmful about the way OSS is going. Apple WebKit, for example, derives from KHTML, and Google Chrome uses a modified WebKit as well. You get Apple, Google, and others all funding the improvement of WebKit. WebKit improves. Because all costs in software are fixed, it's best to sell a lot of copies. There's a tendency toward natural monopoly. With open source software, the same development occurs, *but no one company owns it*. As a result there is no growth of power of any one company, and as a result no risk of arbitrary behavior. If an open source project moves away from the desires of any particular group, an opportunity opens up in the market for serving those people, and the cost for a company to enter the market supporting that group is zero. With closed-source software, you're stuck. Look at Microsoft's attitude with respect to Windows XP. If it were open source, there would be no way to force upgrades. One needn't live in fear of Microsoft following their self-interest and destroying them in the process. VB6 being another good example. The development costs are paid in either model, one just promotes concentration of power, lock-in and risk, and the other doesn't.

Sebastian v Engelhardt
PhD Student and Research Assistant, DFG Graduate School
Andreas Freytag
Professor of Economics, Friedrich-Schiller-University Jena; Honorary Professor; University of Stellenbosch
Stephen M Maurer
Adjunct Associate Professor of Public Policy and Director, Information Technology and Homeland Security Project, University of California, Berkeley