Antirival goods, network effects and the sharing economy
First Monday

Antirival goods, network effects and the sharing economy by F. Xavier Olleros

Nothing facilitates large-scale collaboration like the prospect of inclusive, all-win games. Modern humans have gotten much better at large-scale collaboration because they have discovered, or invented, a broad range of collective goods that are easy to share and become more valuable the more they are shared, thus multiplying the opportunities for all-win outcomes. Steven Weber (2004) and Mark Cooper (2006a, 2006b) have drawn our attention to ‘antirival goods’ — subject to increasing returns to shared use — to differentiate them from ‘rival goods’ — subject to decreasing returns to shared use — and ‘nonrival goods’ — subject to constant returns to shared use. Unlike Weber and Cooper, I argue that nonrivalness and antirivalness are orthogonal properties of some collective goods, rather than stages along the same continuum away from rivalness. Collective goods, I also argue, are most inclusive when they are both nonrival and antirival. In an economy rich in both nonrival and antirival goods, the collaborative stance will often be the default collective choice, at large and small scales alike. Digital technologies are ushering in a transformative age as they expand the cornucopia of nonrival and antirival goods available to us. This inclusiveness of many digital goods eliminates the free-riding problem and mobilizes large amounts of volunteer work.


Rival, nonrival and antirival goods
Antirivalness revisited
Some characteristics of antirival goods
Types of antirival goods
Antirivalness beyond the ‘sharing economy’
Why is the notion of antirivalness important and necessary?
Antirival goods in the digital economy




In the last two millennia, and especially in the last two hundred years, humans have learned to better collaborate at very large scales. According to a number of researchers (Seabright, 2004; Ridley, 2010; Henrich, 2015; Muthukrishna and Henrich, 2016), this largely explains the accelerated improvement of average human prosperity, as measured by a steeply rising trend in global per capita GDP since around 1800 (Maddison, 2001).

How have we learned to master large-scale collaboration despite our well-entrenched tendency toward self-interest and opportunism? Robert Wright (2001) eloquently argued that we have done so by discovering, inventing and exploiting a broadening range of opportunities for positive-sum interactions. According to Wright, we come from a world of zero-sum, or even negative-sum, interactions and we are moving towards a world much richer in positive-sum interactions. This in turn would explain why we are slowly but steadily transitioning from a culture of exclusion to one of inclusion (Turchin, 2015).

But where have so many new opportunities for positive-sum interactions come from? In this paper, I argue that the types of economic goods (e.g., competences, tools, recipes, data and institutions) that we share determine to a high degree the inclusiveness of our interactions and our capacity to collaborate effectively at large scale. Along with Steven Weber (2004) and Mark Cooper (2006a, 2006b), I highlight the importance of antirival goods, which they implicitly define as goods subject to increasing returns to shared use. While agreeing with this definition, I see antirivalness differently from Weber and Cooper, and, unlike them, I argue that nonrivalness and antirivalness are orthogonal properties of some collective goods, rather than stages along the same continuum away from rivalness. Collective goods, I claim, are fully inclusive only when they are both nonrival and antirival.

Thus, in my view, we have become more adept at large-scale collaboration with total strangers thanks to the discovery, invention and deployment of economic goods that are both nonrival and antirival, particularly in the digital domain. Going forward, in relative terms, we will depend progressively less on rival goods (e.g., the private car) and more on nonrival and antirival goods (e.g., fleets of autonomous connected vehicles).

In the sections that follow, I will focus on the following questions: What are antirival goods? What are some of their characteristics? What are the main types of antirival goods? How are sharing and trade intertwined in the world of antirivalness? Why is antirivalness conceptually and practically important? How do digital technologies impact the world of nonrival and antirival goods?



Rival, nonrival and antirival goods

Two different notions of ‘rivalness’ coexist in economics. Firstly, rivalness between human or physical resources in the creation and appropriation of value (e.g., competition between similar experts to get a job, competition between similar products to serve a market). In this sense, the opposite of a physical rival is a ‘complement’ and the opposite of a human rival is a ‘partner’. Secondly, rivalness in the usage of a good or service, that is, the degree of shareability of that good or service. In this sense, the opposite of a perfectly rival (i.e., non-shareable) good is a perfectly ‘nonrival’ (i.e., shareable) good, one subject to zero costs of sharing and constant individual benefits, regardless of the scale of sharing.

Rivalness in usage, or weak shareability, comes in different forms. An economic good can be rival because it is inseparable from the human body and essential to its functioning (e.g., our eyes), because it is meant for private use only (e.g., our contact lenses, underwear or smartphones), because it is a consumable (e.g., an apple), or because it is subject to congestion and therefore only shareable at a small scale and for short periods of time (e.g., a private car). In all such cases, sharing is either impossible or subject to increasing costs and decreasing benefits.

An economic good is considered non-rival if the cost of providing it to an additional individual is zero. Thus, a perfectly non-rival good could be consumed simultaneously by an unlimited number of people, without any loss to any of the people involved in the sharing. Perfectly nonrival goods — neither consumable nor subject to congestion — are rare in the analogue economy. Many analogue goods that were once thought to be nonrival (e.g., suburban highways, the air we breathe, the fish stock in ocean waters, etc.) tend to become precarious or congested with more intensive usage. The digital economy, on the other hand, is rich in nonrival goods — from tweets and MP3 files to Angry Birds games and Spotify playlists. Digital goods, with the notable exception of proof-of-work cryptocurrencies, have negligible reproduction and distribution costs. They are infinitely shareable, enjoying ever-constant returns to sharing, regardless of the scale of sharing.

According to the theory of public goods, a good’s rivalness and its excludability are orthogonal properties (Cornes and Sandler, 1996; Laffont, 2008). Figure 1 shows the four types of economic goods that result from combining these two dimensions. As already indicated, the rivalness dimension reflects the degree to which a good is shareable. Being impossible or difficult to share, rival goods are subject to increasingly higher costs and lower benefits from shared use. In other words, consumption externalities amongst the people sharing the good become increasingly negative with the sharing. Contrarily, nonrival goods are subject to low costs and constant benefits from shared use, regardless of the scale of sharing. In other words, consumption externalities amongst the people sharing the good are negligible. As for the excludability dimension of the table, it reflects the degree to which access to an economic good can be denied to some people, typically those unwilling to pay for it. According to the theory of public goods, while a ‘private good’ is rival and excludable, a ‘public good’ is nonrival and non-excludable, which explains why the economics of those two types of goods are so different.


The classic matrix of economic goods
Figure 1: The classic matrix of economic goods.


This classification of economic goods is a foundational piece to a voluminous literature spanning more than 60 years (Laffont, 2008). But as the economic world has become richer in the types of goods that it offers, this conceptual framework has become much too narrow and constraining, particularly as we move into the digital realm. Two authors have sketched a way out of this predicament.

In 2004, Steven Weber, a professor of political science at Berkeley, described the open source software movement in his book The success of open source (Weber, 2004). While most people thought, and still think, of digital tools and creations as nonrival — i.e., infinitely shareable — Weber coined a new term and argued that open source software is not simply nonrival, but ‘antirival’. To him, that difference is the key to understanding the success of open source. As he put it:

“I believe the solution to the [open source software] puzzle lies in pushing the concept of nonrivalness one step further. Software in many circumstances is more than simply nonrival. Operating systems like Linux in particular, and most software in general, actually are subject to positive network externalities. Call it a network good, or an antirival good. In simpler language, it means that the value of a piece of software to any user increases as more people use the software.” [1]

Following up on Weber, in two papers published in 2006, Mark Cooper — then director of research at the Consumer Federation of America — explicitly argued that the classic typology of economic goods (rival, nonrival, excludable and non-excludable) is incomplete. As shown in Figure 2, like Weber, Cooper sees ‘antirival goods’ as more than simply nonrival. He also sees ‘inclusive goods’, his own term, as more than simply nonexcludable. As per Figure 2, Cooper is more explicit than Weber in arguing that antirivalness is a state further away from rivalness on the same dimension and trajectory as nonrivalness. There is therefore a single path away from a world of rival goods and zero-sum interactions: that of nonrivalness and, beyond it, antirivalness.


Mark Cooper's (2006b) revised matrix of economic goods
Figure 2: Mark Cooper’s (2006b) revised matrix of economic goods. The overlays in red and green are mine.




Antirivalness revisited

My view of antirivalness is different from Weber’s and Cooper’s. As I see it, rather than being a state further away from rivalness on the same dimension as nonrivalness, antirivalness belongs on another dimension, orthogonal to the first. Both qualities are opposed to rivalness, but in different ways. Firstly, while rivalness and nonrivalness are qualities of the units being shared, antirivalness typically is a quality of the system enabling the sharing. Secondly, while nonrivalness only implies a fall in the costs of sharing, antirivalness also implies a rise in both the scalability and benefits of sharing [2]. Thirdly, while in the digital realm, the road to nonrivalness is generally quick and straightforward, the road to antirivalness is not. I will illustrate these three points with a well-known example.

In September 1999, Shawn Fanning, an American teenager, developed and launched Napster, the first peer-to-peer platform for large-scale music sharing. Napster is widely thought to have triggered the transition of the music industry from the age of expensive CDs to that of inexpensive streaming (Witt, 2015). But the success and impact of Napster would have been impossible without three previous events: the creation of the MP3 format for compressed digital music by the Fraunhofer Institute (in 1993), the release of an unprotected MP3 codec by an Australian hacker into the public domain (in 1995) and the broad deployment of free CD-to-MP3 transfer software such as Winamp (starting in 1997). Together, these three innovations rendered private collections of music files potentially nonrival.

Within this burgeoning value ecosystem, Napster’s P2P platform quickly became viral and transformative in part because it facilitated the costless sharing of MP3 files at any scale, but also because Shawn Fanning made it antirival by default — i.e., unless you changed the settings, you automatically shared all your downloads with everybody else — and launched it at a time when enough people already had enough Internet bandwidth to be able to afford accepting the default. This is how Napster rapidly scaled up from Fanning’s own music collection to millions of songs for free download. Unfortunately, unable to find a route to becoming a legitimate community or a legitimate business, Napster was shut down by U.S. authorities in July 2001.

The Napster case shows that while nonrivalness is a quality of the units being shared — i.e., the MP3 files — antirivalness typically is a quality of the system enabling large-scale sharing — i.e., the Napster platform. It also shows that, while nonrivalness only implies a fall in the costs of sharing — an MP3 file can be shared without any loss to the music sharer — antirivalness implies a rise in both the scalability and benefits of the sharing. As regards scalability, before Napster, sharing mp3 files was quite laborious and only viable at small scale, even online (Witt, 2015). Despite being nonrival, MP3 files only became shareable at any scale within a sharing platform easily accessible to millions of music fans. And as regards the benefits from sharing, the Napster platform became more valuable to sharers the greater their numbers and the richer and more varied their music collections became. Lastly, the Napster case also shows that while in the digital realm the road to nonrivalness is generally quick and straightforward — it only takes a few seconds to transform a song from a rival CD into a nonrival MP3 file — the road to building and sustaining a viable antirival platform is not.

Napster and many other similar cases show that there are indeed two distinct and complementary paths away from a world of zero-sum interactions: the path of nonrivalness — toward easily shareable modules — and that of antirivalness — toward scalable infrastructures subject to increasing returns to shared use, partly due to further reductions in the costs of sharing and partly due to a rise in the benefits of sharing. Thus, only by combining nonrivalness and antirivalness do goods become shareable at any scale and fully inclusive, subject to increasing returns for all the people involved in the sharing.

Nonrivalness alone cannot ensure inclusiveness. MP3 files alone could not have triggered the transformation of the music industry toward a global ecosystem of inexpensive downloads and streaming. Antirivalness alone cannot ensure inclusiveness either. Consider two classic antirival but costly-to-share goods: the traffic light and the English language. Installing traffic lights in large cities and learning English as a second language are very useful endeavors, and their usefulness rises with the number of cities and people who undertake them. But neither one is fully inclusive, as the network of traffic lights remains too expensive for poor cities and learning English remains too difficult for poor people living in non-English speaking countries. Likewise for the International Space Station, for example. Were it open and accessible to many more experts and tools, it would probably be far more useful to humanity. But space limitations and the considerable costs of traveling there and back limit such possibilities.

I consequently disagree with Mark Cooper’s understanding of inclusive goods. While he sees inclusiveness as a state orthogonal to nonrivalness and antirivalness, I see inclusiveness as the emergent property of goods that are both nonrival and antirival. Thus understood, an inclusive good is naturally viral: its sharing is easy, organic and self-reinforcing, since it is increasingly inexpensive and valuable for all (think of Wikipedia, Google Search, Waze, YouTube, etc.).



Some characteristics of antirival goods

As already mentioned, unlike rivalness and nonrivalness, antirivalness tends to be a systemic, network property. While the individual element can be rival (e.g., the credit card or the dollar bill), the overall system (e.g., the Visa System or the U.S. currency) is antirival. Likewise, while the individual element can be nonrival (e.g., a French word, an academic citation, a videogame or an MP3 file), the overall system (e.g., the French language, App Store or iTunes Store) is antirival. This in turn determines what we can share at scale and what we cannot. Sharing is becoming increasingly infrastructural because infrastructures are becoming increasingly antirival. We rarely share our smartphones, but we do share the Internet, the Web, the smartphone’s platform and its ecosystem of apps and software updates.

Rivalness, nonrivalness and antirivalness often apply to different aspects of the same good. Consider, for example, the economic nature of a traffic light. The physical object is rival: it is alienable and cannot be replicated at zero marginal cost. Its usage is nonrival: although subject to occasional congestion, any given traffic light can serve many drivers and pedestrians, concurrently or in sequence. But the three-color code is strongly antirival, at both the local and the global scale. Since it has a very high critical threshold at the local level — you wouldn’t want to live in a city where only 90 percent of drivers stop at red lights — local authorities need to enforce it. Similar distinctions apply to a US$10 bill and to a printed copy of Hamlet. In the former case, the antirival good is the American currency, and in the latter case, the antirival good is the English language.

The world of antirival goods is very heterogeneous. Some antirival goods are spontaneous, organic and robust (e.g., French cuisine, the jazz repertoire and most natural languages), while others are designed, contested and fragile (e.g., the Euro currency, most cryptocurrencies and the Affordable Care Act in the U.S.). Consequently, the degree of difficulty in deploying and managing antirival ecosystems can vary a lot. Natural languages can use some institutional help in order to thrive (e.g., public schooling and alphabetization programs), but they don’t absolutely require it — they certainly didn’t before the seventeenth century. Traffic lights have been deployed across many countries and cultures at some expense but without great difficulty. Bernie Ecclestone singlehandedly created the Formula 1 value ecosystem, and he managed it quite well for more than 25 years. The Eurozone institutions, on the other hand, could have been better designed and managed (Baldwin and Giavazzi, 2015).

Antirivalness and inclusiveness. Why are some positive-sum interactions so challenging, while others are so easy to create and manage? I think that the natural inclusiveness of some systems explains their easy deployment and resilience. If antirivalness is strong, it’s not just that everybody wins from the sharing; it’s that everybody wins in rough proportion to the sharing. So, while the typical way to manage a rival good is exclusion, and the typical way to manage a nonrival good is mere permissiveness, the optimal way to manage a strongly antirival good is inclusiveness, perhaps to the point of subsidizing early users and penalizing deviants.

The rivalness-nonrivalness-antirivalness interaction often is cyclical. While a fertile new idea or discovery (e.g., graphene or gravitational waves) is antirival, an old one (e.g., The Pythagorean theorem) is merely nonrival: easy to share but no longer subject to increasing returns to sharing. While emerging technologies (e.g., graphene supercapacitors and quantum computing) and markets (e.g., today’s cryptocurrency markets) are antirival — they invite multiple entries and bets, the more diverse the better — mature technologies and business opportunities tend to become increasingly rival. Early in its lifecycle, a complex tool (e.g., the Large Hadron Collider or the International Space Station) is antirival, for at least two reasons: its funding is easier to justify if many experts are able to share it, and its potential for improvement is typically vast. This antirivalness will tend to fade away as the tool becomes more perfected and codified and as its full potential becomes more nearly realized.

Nonrivalness, antirivalness and the free rider problem. Nonrivalness and antirivalness affect a good’s shareability in very different ways. By lowering the cost of sharing and/or reproduction, nonrivalness implies that opening the door to freeloaders could be costless. Antirivalness, on the other hand, may raise the benefits of large-scale sharing to the point that it will altogether eliminate the free-riding problem. The very act of consumption may become so valuable to the shared experience that passive consumers will add value to the good without having to pay for its consumption — e.g., by attracting eager advertisers and by providing the user data needed to improve the performance and profitability of the overall platform. Thus, an all-around win-win dynamic ensues (e.g., Google Search, YouTube, TED conferences, Duolingo, Waze, etc.).

History shows that given the increasing returns to sharing proper to antirival networks, even illegal free-riding can be beneficial to everyone in surprising ways. Thus, for example, the Fraunhofer Institute initially opposed the sharing of pirate MP3 files, while it failed to attract multimedia content providers to adopt its music compression format and was unable to monetize a shareware version of their MP3 codec and player aimed at individual music fans. By the end of 1996, the situation was so dire that the project’s team was ready to give up on the MP3. But then, to everyone’s surprise, the surge in popularity of pirated MP3 files turned the tide, and the Fraunhofer ended up making a considerable fortune by selling MP3 licenses to the same content providers and distributors who had previously shunned the format (Sterne, 2012; Witt, 2015).



Types of antirival goods

Antirival goods differ in the many productive or expressive forms that they can take: they can be discoveries and inventions (e.g., perovskite solar cells and CRISPR/Cas9 gene editing), fertile ideas (e.g., superstring theory), cooking recipes (e.g., tiramisu), communication networks (e.g., Internet, the Web), communication codes (e.g., the English language), complex games (e.g., chess, go), complex collective tools (e.g., the Large Hadron Collider), protocols (e.g., the blockchain protocol), algorithms (e.g., the backpropagation algorithm), technical standards (e.g., Wi-Fi, TCP/IP), social conventions (e.g., driving on the right side of the road), organizations (e.g., the Linux project), art forms (e.g., the jazz repertoire), institutions (e.g., the U.S. dollar, Creative Commons set of licenses), or even urgent collective problems and emergencies.

It might seem paradoxical to consider collective problems as ‘collective goods’. But, as has been well documented, necessity — e.g., a viral epidemic or any other collective crisis — often is the mother of collaborative invention and innovation, particularly in recent times (Mokyr, 2002; Levitt, 2013). Think of the multi-decade search for remedies to cancer. The more widely a particular type of cancer spreads, the easier it will be to fund the requisite R&D efforts to counter it; also, the richer and more varied will be the pertinent available data and the tools and perspectives brought to bear in trying to fight it. Thus, tragic as fatal diseases are, there is a silver lining to many of them. Were cancer not so prevalent, for example, we would not know so much today about the latent power of our own immune system to fight such a disease (Mukherjee, 2010; National Cancer Institute, 2017).

Antirival goods can also differ in the way they accumulate value, in whether or not there is a minimal floor to their value creation process and in whether or not there is a maximal ceiling to that process. Consider the contrast between smartphones and traffic lights. The network of traffic lights in a city accumulates value in a linear and predictable way, one traffic light installation at a time. The network’s value has a very low floor regarding the number of installed lights, but a very high floor regarding the social acceptance of the color code: it plummets if the percentage of drivers respecting red lights falls significantly under 100 percent. And the network has a ceiling: once all the major intersections of the city have traffic lights, there will be few gains from installing additional ones. At that point, the value-creation capacity of the network will have reached its zenith.

The network of smartphones in a country, on the other hand, accumulates value in an unpredictable combinatorial process, as new types of applications converge on every machine. The network’s value has a very low local floor: given its global connectivity, even a single smartphone in a city would already be useful. And the network’s value has no predictable ceiling: the day that every citizen has a smartphone at hand need not mark the zenith of the network’s value, as still newer and more powerful applications may keep converging on every machine.

Wikipedia and YouTube are more similar to the smartphone network than to the network of traffic lights. But they also differ from each other in an important way: while the quality of YouTube’s modules (the individual videos already uploaded) doesn’t improve with a greater number of viewings, the quality of Wikipedia’s articles tends to improve with a greater number of readers, some of whom will take time to update them or correct mistakes.



Antirivalness beyond the ‘sharing economy’

Economic goods can be shared in various ways, and differences in shareability are more than just differences of degree. The concurrent sharing of a car trip or taxi ride (e.g., BlaBlaCar, UberPOOL) is not the same as the serial sharing of a car (e.g., Car2Go), which is not the same as the sharing of a parking lot or a highway. Likewise, we do not share the theory of relativity, the Web or the English language the way we share MP3 files. More generally, there is strong sharing — deliberate and communitarian — and weak sharing — unconscious and automatic, often the unintended byproduct of self-interested actions. Importantly, the latter can be far more scalable and consequential than the former. Let me elaborate on this point.

Beyond strictly communitarian sharing, there is sharing by commercial propagation. As regards the latter, there are at least three complementary ways of generating positive-sum interactions: the improvement, the repurposing and the recombination of shared economic goods.

Improving shared goods

The value of some antirival goods rises the more widely a collective good is shared simply because its intended functionality improves with sharing (e.g., Google Search, the English language, Bitcoin or the jazz repertoire). There may be several aspects to such improvements, some of them commercial and others communitarian. Consider the English language, for example. The more people that master this language, the more useful it becomes as a means of expression and communication, in at least three ways: the number of people who can communicate effectively increases, the speed and fluidity with which new words and expressions are added to the language increase, and the quantity and variety of cultural content (e.g., music, films, blogs, novels, technical reports, academic articles, etc.) generated in this language also increase. The first two types of language network improvements are strictly communitarian, but the third one is not; and it is increasingly powerful in a global and connected economy (Bragg, 2003).

Repurposing shared goods

The value of other antirival goods (e.g., novel tools and inventions) rises the more widely a good propagates because it is eventually repurposed for uses that were not originally intended or even imagined. Think of the history of laser technology, for example. The first lasers were ingenious solutions in search of a problem; no one quite knew what to make of them (Bertolotti, 2005). But as laser technology has propagated, the body of theoretical and practical knowledge about lasers has become increasingly extended, repurposed and reinvented. Today, lasers are more varied, versatile and ubiquitous than ever. We find them enabling much of the Internet’s file transport infrastructure, scanning the surroundings of autonomous vehicles and facilitating surgical interventions. We find them within multimedia recorders and players, at the airport check-in stand and at the supermarket check-out counter, among many other places.

Recombining shared goods

The value of still other antirival goods rises the more widely the good propagates, because it can eventually be recombined into new syntheses with entirely new properties. History shows that innovation is indeed largely combinatorial (Burke, 2007). Think of the semiconductor and the cardiac pacemaker. Alone, a semiconductor cannot save lives; but properly combined, implanted and monitored, the ensemble of sensors, computerized generator, biocompatible metal case and long-lasting electric batteries saves lives. As semiconductor technology has propagated, its body of theoretical and practical knowledge has become increasingly recombined, enabling the invention of a vast array of powerful new tools, such as the cardiac pacemaker.

Late in the lifecycle of an economic good, its improvement potential will tend to decrease, but the opportunities for repurposing and recombination may well multiply. Think, for example, of the electric motor. By now, the most obvious and consequential opportunities to improve the technology have all been explored and exploited. But precisely because electric motors have never been as efficient and versatile as they are today, their opportunities for repurposing and recombination keep multiplying: from Nespresso coffee machines to cars, trucks, drones and robots (Shoults, 1942).

As shown earlier, the sharing economy and trade economy have subtle interconnections that need to be fully recognized. By trading goods and services, formally or informally, we often share their blueprints, or are simply inspired by them, thereby opening the door to further improvements, repurposing and recombinant innovation (Ridley, 2010). Thus, ideas, designs, styles, recipes, protocols, algorithms, etc., often become shareable and antirival by the commercial propagation of the goods and services that embody them. A scriptwriter or movie director, for example, can walk out of a cinema with several ideas for a better movie than the one she just watched. This is not sharing in a strong, communitarian sense, but it is a crucial aspect of our progressive capacity to generate positive-sum interactions.



Why is the notion of antirivalness important and necessary?

Today, 14 years after Steven Weber coined the term ‘antirival good’ and 12 years after Mark Cooper’s elaboration of the concept and its connection with the classic typology of economic goods, there is much talk and thinking about ‘shared goods’ and ‘network effects’, but there is no vibrant conversation about ‘antirivalness’, even though those three concepts are so closely related. In the previous sections, I have already hinted at the fact that ‘antirivalness’ is an economic quality that merits its own name and discussion. In what follows, I will offer other, more explicit arguments about the importance and usefulness of having a conversation about antirival goods.

A fully developed theory of antirivalness would complete the classic typology of economic goods. This has been Cooper’s main argument on the topic and, although I do not agree with his characterization of antirival goods and inclusive goods, I find his call for expanding and updating the mainstream classification of economic goods timely and convincing.

A greater awareness of antirivalness would keep us from misunderstanding and misclassifying antirival goods. Even though we increasingly live in a world of antirival goods, we still have some difficulty in understanding the contours and dynamics of such a world. In 1813, Thomas Jefferson, already retired from the U.S. Presidency, wrote a letter to Isaac McPherson where he argued for a broader sharing of new ideas, since, he said, ideas are like the light that spreads from one lighted candle to another: it is to everyone’s advantage to share a light that doesn’t degrade as it spreads (Jefferson, 1813).

In 2009, Alex Tabarrok, professor of economics at George Mason University, quoted Jefferson approvingly: “Ideas have this amazing property. Thomas Jefferson, I think, expressed this quite well. He said, ‘He who receives an idea from me receives instruction himself, without lessening mine. As he who lights his candle at mine receives light without darkening me.’ Or to put it slightly differently: one apple feeds one man, but an idea can feed the world” (Tabarrok, 2009).

As I see it, Tabarrok’s endorsement of Jefferson’s analogy reflects a misunderstanding of the economic nature of ideas, more surprising in 2009 than in 1813. A new idea is not comparable to a lighted candle. Firstly, because not all new ideas are easy to share at any scale (hint: do you have a good understanding of the intricacies of quantum computing?). And secondly, because even easy-to-share new ideas behave very differently from lighted candles. As it is shared, a candle’s light spreads costlessly; but as they are shared, fertile ideas improve and recombine. In other words, whereas the light of a candle is nonrival, a fertile idea is strongly antirival. Thus, while the spread of candlelight has fairly predictable consequences — if the sharing takes place indoors, with every lighted candle the luminosity in the room will increase in a perfectly linear way — the spread of a fertile idea, in conjunction with the spread of other complementary ideas, has unpredictable consequences. Hence, an easy-to-share and fertile new idea (e.g., Louis Pasteur’s germ theory of disease, or Charles Townes and Arthur Schawlow’s invention of the laser) deserves to be spread far more rapidly and widely than the light of any candle.

A good grasp on antirivalness can help us to better understand the economics of ‘network goods’. Not all network goods are antirival, though some may become so. The Angry Birds game and any music file sold at the iTunes Store are broadly considered to be network goods (cf., Shapiro and Varian, 1998; Liebowitz, 2002), but they don’t get better the more people use them.

Likewise, not all antirival goods are network goods. Some collective goods only become scalable and antirival in the measure that they develop well-designed and well-managed networks (e.g., Facebook, Waze, etc.). But many others are antirival well before they become ‘network goods’, if they ever do. From the moment that it was invented, the game of chess was strongly antirival: the average mastery and intricacy of the game improved with a rising number of expert players committed to it. But it took a very long time for chess to become a vibrant network good (Shenk, 2006). Likewise for the saxophone (Segell, 2005), jazz repertoire (Gioia, 2011) and any natural language (McWhorter, 2003). This is also true of fertile new ideas. A fertile idea always sprouts in one brain; unfortunately, sometimes it stays there, failing to develop a value network beyond the original inventor or discoverer. This, of course, doesn’t make the idea any less fertile and antirival — think of Gregor Mendel’s theory of genetics, dormant and unexplored from 1866 to 1900 (Edelson, 1999).

Many collective problems and widely shared afflictions (e.g., smallpox before 1966) become antirival well before they find a network capable of solving or eradicating them. A widespread form of cancer, for example, is an antirival good even if no effective cancer-fighting network has yet arisen. Complex collective tools too can be strongly antirival without being fully networked. Although singular, expensive to replicate, subject to congestion and scheduled to cease operating in 2024, the International Space Station (ISS) is strongly antirival: its performance and usefulness would surely improve if more experts were able to use it concurrently. But in this respect, the ISS is a long way from becoming a truly networked good. One can imagine future networks of connected space stations, operated by semi-autonomous robots constantly sharing their findings and best practices, with expert humans interacting with them from an Earth-based station (Ackerman, 2017).

So, figuratively speaking, we could say that all antirival goods ‘want to become’ network goods, but only some succeed at it. Evidently, antirival goods that are also nonrival — as tends to be the case with digital platforms — have a much better chance of scaling up rapidly and becoming vibrant network goods.



Antirival goods in the digital economy

Digital technologies enable at least three new kinds of antirival goods: multisided platforms, learning algorithms and the massive databases that the latter exploit.

Multisided platforms are lean intermediaries specialized in facilitating transactions and interactions between two or more complementary constituencies that create value for each other. In the last 20 years, they have become the workhorses of the digital economy — e.g., Google Search, YouTube, eBay, LinkedIn, Uber, Airbnb, AliExpress, etc. Their rise is closely tied to that of machine-learning algorithms capable of rapidly analyzing very large databases in order to optimize the platform’s core mission — e.g., online search, medical diagnosis, teaching a new language, industrial logistics, movie recommendations, etc. Crowdsourcing and datasourcing are essential elements of such digital platforms (McAfee and Brynjolfsson, 2017).

Like any other practical recipe, an algorithm can become antirival by combining it with a complementary algorithm to achieve a result that neither one alone could achieve (Auerswald, 2017). But learning algorithms are antirival in another way. Being data-driven and able to learn from their errors and approximated guesses, their performance will improve in rough proportion to the quantity and variety of the pertinent data they analyze (Domingos, 2014).

On its own, a piece of data is not naturally antirival; it only becomes so if processed by expert analysts in conjunction with complementary pieces of data. Digital technologies are transforming data in two important ways: by rendering their collection reliable and inexpensive at any scale, and by partly delegating their analysis to powerful machines and algorithms. Rendering such databases fully secure still remains a challenge (Mayer-Schönberger and Cukier, 2013).

In the digital realm, the transition to antirivalness generally involves the use of large databases and learning algorithms. Khan Academy only became antirival — i.e., started to enjoy strong network advantages, in addition to scale economies — when it started using learning analytics in following up and coaching individual students. Likewise, Gmail only became antirival when it started improving its anti-spam software using learning algorithms. As for Netflix, it became an antirival platform in two distinct stages. Since 2008, when its recommendation engine started to use learning algorithms. And more so since 2011, when most of its clients switched from DVD rentals to streaming content. Streaming allows the collection of much more precise usage data, and hence a more optimized process. In the words of two insiders:

“Streaming has not only changed the way our clients interact with the service, but also the type of data available for our algorithms. For DVDs our goal is to help people fill their queue with titles to receive in the mail over the coming days and weeks; selection is distant in time from viewing, people select carefully because exchanging a DVD for another takes more than a day, and we get no feedback during viewing. For streaming, clients are looking for something great to watch right now; they can sample a few videos before settling on one, they can consume several in one session, and we can observe whether a video was watched fully or only partially.” (Amatriain and Basilico, 2012)

Being both nonrival and antirival, digital platforms excel at turning the dreaded free-riding problem into a driver of sharing and collaboration. In so doing, many of them enable a shift toward free access of what used to be exclusive club goods.

Consider the TED talks, a series of exclusive and expensive annual conferences — US$8,500 for regular attendees and US$17,000 for donors, in 2015 — on a variety of leading-edge topics. In 2006, Chris Anderson, TED’s owner and curator, made the then-shocking decision to start releasing all their recorded conferences into the public domain, both on TED’s own Web site and on YouTube. Three years later, Anderson launched TEDx, a decentralized franchise for organizing similar conferences at any time and in any city across the world. In 2015 alone there were around 2,500 TEDx conferences organized in dozens of cities. Today, TED is by far the world’s most popular source of conferences, with well over 100 million monthly viewings, if we include those on YouTube. A critical element of this success is the work of more than 9,000 translators who provide, free of charge, subtitles for the online TED talks in 120 languages (Weinberg, 2016).

Figure 3 tries to capture how the TED talks reach and leverage a global audience. Again, the key element in their transition to a new business model was the 2006 decision to place all of the recorded conferences in the public domain. The wealth of carefully curated, top-of-the-line free conferences attracts a large crowd of enthusiastic, unremunerated translators. The rising stock of attractive content with subtitles in multiple languages ensures a global and loyal audience for TED talks. The online audience would in past times be branded as ‘free riders’, since they contribute nothing other than their consumption of free content. But given the nonrivalness of recorded conferences and the antirivalness of the TED and TEDx platforms, their promise of a global audience acts as a powerful bait to attract the very best speakers on any topic. In the old days, such speakers would have settled for a well-remunerated conference to a few hundred people. Today, the TED ecosystem offers them a more enticing proposition: the chance of reaching hundreds of thousands of people across the world — a conference by Ken Robinson holds the record with 45 million views. For top-notch professionals, nothing compares with this opportunity to broadcast their ideas and creations to so many people at no cost to themselves. Not getting paid by TED for their service is immaterial to them. Unsurprisingly, the list of people queuing up to be invited to speak at, or attend, a TED conference is very long. Consequently, the TED curators can be even more selective of their speakers and on-site audiences today than they were when their conferences catered to a small American elite (Weinberg, 2016) [3].


How became an antirival platform
Figure 3: How became an antirival platform.


This double accomplishment by TED of attracting hundreds of unremunerated contributors — speakers and translators — and turning millions of passive consumers into the pillars of its attractiveness has been replicated, to various degrees and in different ways, by many other platforms, from Linux, Wikipedia and YouTube to Stack Overflow, TripAdvisor, Instagram and Duolingo.

Antirivalness and platform architecture. In digital platforms, antirivalness and inclusiveness are certainly enabled by nonrival modules (e.g., the zero marginal costs of reproduction of PDF or MP3 files), but an optimal platform architecture is just as important a prerequisite. Consider again peer-to-peer networks for music sharing. The first-generation P2P network (i.e., Napster) decentralized the file storage but centralized the overall registry in Shawn Fanning’s servers, which rendered the network much easier to prosecute and shut down. The second-generation networks (i.e., Gnutella, Kazaa and their imitators) decentralized both the file storage and the overall registry. But then Bram Cohen figured out a way to do better than that. In July 2001, he launched BitTorrent, an open-source third-generation P2P network that decentralized the file storage and the overall registry, but also enabled concurrent downloads from many repositories, rather than only one, as had always been done before. Witt (2015) explains the impact that such a novel platform architecture had on the shareability of music files:

“The greatest benefit of the torrent approach was the way it solved one of the Internet’s long-standing problems: the traffic bottleneck. Historically, popular files tended to crash servers, as millions of users crowded around a narrow doorway and tried to push their way in. But the matching schematic of torrents opened hundreds of doors at once, taking pressure off a single server and transferring it to many individuals. This inversion of the traditional paradigm of file distribution had a startling result: with torrents, the more people who attempted to simultaneously download a file, the faster the download went.” [4]

Witt’s explanation highlights well the importance of an optimal architectural design for enabling the sharing of digital goods at any scale. Although its share of Internet traffic has declined since its 2008 peak, BitTorrent technology remains a critical element of Internet infrastructure, particularly the Micro Transport Protocol for traffic congestion management (Baer, 2014). Moreover, BitTorrent has inspired several leading-edge initiatives to reinvent and decentralize the Web, such as ZeroNet (Sarkhel, 2017) and IPFS (Kahle, 2015).

The surge of digital sharing and collaboration

The digital economy is proving to be much more collaborative than the purely analog economy. While facilitating the sharing of some rival goods (e.g., private rooms and cars), digital technologies have mostly boosted the creation and sharing of nonrival goods (cultural goods, reliable data points, 3D designs, etc.) and antirival goods (learning algorithms, multisided platforms and big data). Digital platforms also allow us to mobilize a broader range of motivations to collaborate, some of which (e.g., the improvement of collective tools and infrastructures, or the accumulation of human, civic or reputational capital) can minimize or eliminate both the need for remuneration and the likely drawbacks of free-riding. Digital technologies have also amplified the scope of generalized, often unconscious, mutualism through crowdsourcing and datasourcing. Importantly, while crowd-sourcing can be laborious and expensive, data-sourcing is increasingly automatic, especially online (Mayer-Schönberger and Cukier, 2013). Thus, thanks to digital technologies, the optimal scale of coordination, collaboration and sharing often is the entire world.




What we share affects the way we interact. This is true for all kinds of goods and services, from the spontaneous and emergent (e.g., the English language) to the deliberate and designed (e.g., the traffic-light color code), and from the abstract and generic (e.g., The Pythagorean theorem) to the more tangible and concrete (e.g., a taxi ride).

In this paper, I have advocated a reprise of the conversation on antirival goods that Steven Weber and Mark Cooper tried to start in 2004–2006. Like them, I too think that such a conversation would help us to transcend the outdated strictures of the classic typology of economic goods. But I have also argued for a different understanding of antirivalness and its relationship with inclusiveness: not as a further stage along the rival-nonrival continuum, but as an altogether different vector that tends to emerge at a higher and more systemic level than nonrivalness.

It seems perfectly logical to see antirivalness as a further stage in the rival-nonrival continuum. While rival goods are subject to decreasing returns to shared use (I am penalized if I share such a good), nonrival goods are subject to constant returns to shared use (I am not penalized if I share such a good) and antirival goods are subject to increasing returns to shared use (I gain if such a good is shared by more people). But as I have argued here, this view fails to take account of the fact that nonrivalness and antirivalness tend to operate at different levels of reality: often, while modules and components can be rival or nonrival, the overall system is antirival. Hence, neither one — nonrivalness or antirivalness — is a necessary precondition for the other. Furthermore, both must be combined to ensure that economic goods become inclusive and shareable at all scales, to everyone’s advantage.

I have also argued that in order to better understand and explain our increasing ability to invent and exploit positive-sum interactions, we need to develop a nuanced and expansive theory of shareable goods (cf., Benkler, 2004; Sundararajan, 2017). To do so, we need to go beyond the so-called ‘sharing economy’ to encompass weaker, but very fertile, forms of sharing that owe more to commercial enterprise, trade and competition than to communitarian sharing.

In closing, I should note that while the digital transition has in many ways helped us to move in the direction of a more sharing and collaborative society, the current crop of digital antirival platforms has a bias toward higher concentrations of economic power that needs to be acknowledged and corrected (Ezrachi and Stucke, 2016; Stucke and Grunes, 2016; Srnicek, 2017). Thankfully, a new generation of digital platforms, already in the making, might give us the tools to further decentralize our markets, institutions and societies (Tapscott and Tapscott, 2016). Were this to happen, it would render the digital economy a lot more inclusive than it is today. End of article


About the author

F. Xavier Olleros is an associate professor of innovation management at the University of Quebec Business School (ESG-UQAM), in Montreal. He has published articles in Research in marketing, Industrial and corporate change, International journal of innovation management, International journal of technology management, Technovation, Research and technology management, Long range planning, Journal of business research and Journal of product innovation management. He is coeditor of the Research handbook on digital transformations (Edward Elgar, 2016).
E-mail: olleros [dot] xavier [at] uqam [dot] ca



1. Weber, 2004, pp. 153–154.

2. It is important to note that the costs and benefits of sharing considered here are individual, rather than collective. Sharing a nonrival good (e.g., music MP3 files) will almost always increase the collective benefits of the shared good. And even sharing rival goods can, in some cases, increase the collective benefits of the shared good — e.g., by shifting scarce resources to people who are temporarily in dire straits. But only the sharing of antirival goods raises individual benefits from sharing for all the people involved.

3. While carefully curated (only about 40 percent of TEDx projects are approved by the mother organization, to begin with), TEDx conferences are not as selective of speakers and on-site audiences as the main TED conferences.

4. Witt, 2015, p. 167.



Evan Ackerman, 2017. “How NASA’s Astrobee robot is bringing useful autonomy to the ISS,” IEEE Spectrum (9 February), at, accessed 30 December 2017.

Xavier Amatriain and Justin Basilico, 2012. “Netflix recommendations: Beyond the 5 stars,” Netflix Technology Blog (5 April), at, accessed 9 November 2017.

Philip H. Auerswald, 2017. The code economy: A forty-thousand year history. Oxford: Oxford University Press.

Richard Baldwin and Francesco Giavazzi, 2015. “Eurozone crisis: A consensus view of the causes and a few possible solutions,” Vox: CEPR’s Policy Portal (7 September), at, accessed 12 January 2018.

Drake Baer, 2014. “How BitTorrent rewrote the rules of the Internet,” Fast Company (5 March), at, accessed 9 November 2017.

Yochai Benkler, 2004. “Sharing nicely: On shareable goods and the emergence of sharing as a modality of economic production,” Yale Law Journal, volume 114, number 2, pp. 273–358, and at, accessed 9 November 2017.

Mario Bertolotti, 2005. The history of the laser. Bristol: Institute of Physics.

Melvyn Bragg, 2003. The adventure of English: 500AD to 2000: The biography of a language. London: Hodder & Stoughton.

James Burke, 2007. Connections. New York: Simon & Schuster.

Mark Cooper, 2006a. “The economics of collaborative production: A framework for analyzing the emerging mode of digital production,” draft for presentation at the Economics of open content: A commercial–noncommercial forum, MIT, at, accessed 12 January 2018.

Mark Cooper, 2006b. “From WiFi to wikis and open source: The political economy of collaborative production in the digital information age,” Telecommunications and High Technology Law, volume 5, pp. 125–157, and at, accessed 12 January 2018.

Richard Cornes and Todd Sandler, 1996. The theory of externalities, public goods, and club goods. Second edition. Cambridge: Cambridge University Press.

Pedro Domingos, 2014. The master algorithm: how the quest for the ultimate learning machine will remake our world. New York: Basic Books.

Edward Edelson, 1999. Gregor Mendel, and the roots of genetics. Oxford: Oxford University Press.

Ariel Ezrachi and Maurice E. Stucke, 2016. Virtual competition: The promise and perils of the algorithm-driven economy. Cambridge, Mass.: Harvard University Press.

Ted Gioia, 2011. The history of jazz. Second edition. Oxford: Oxford University Press.

Joseph Henrich, 2015. The secret of our success: How culture is driving human evolution, domesticating our species, and making us smarter. Princeton, N.J.: Princeton University Press.

Thomas Jefferson, 1813. “Letter to Isaac McPherson” (13 August), at, accessed 9 November 2017.

Brewster Kahle, 2015. “Locking the Web open: A call for a decentralized Web” (11 August), at, accessed 9 November 2017.

Jean-Jacques Laffont, 2008. Fundamentals of public economics. Revised English language edition. Translated by John P. Bonin and Hélène Bonin. Cambridge, Mass.: MIT Press.

Alexandra M. Levitt, 2013. Deadly outbreaks: How medical detectives save lives threatened by killer pandemics, exotic viruses, and drug-resistant parasites. New York: Skyhorse Publishing.

Stan Liebowitz, 2002. Rethinking the network economy: The true forces that drive the digital marketplace. New York: AMACOM.

Angus Maddison, 2001. The world economy: A millennial perspective. Paris: OECD Publishing.
doi:, accessed 12 January 2018.

Viktor Mayer-Schönberger and Kenneth Cukier, 2013. Big data: A revolution that will transform how we live, work, and think. Boston, Mass.: Houghton Mifflin Harcourt.

Andrew McAfee and Eric Brynjolfsson, 2017. Machine, platform, crowd: Harnessing our digital future. New York: Norton.

John McWhorter, 2003. The power of Babel: A natural history of language. New York: Perennial.

Joel Mokyr, 2002. The gifts of Athena: Historical origins of the knowledge economy. Princeton, N.J.: Princeton University Press.

Siddhartha Mukherjee, 2010. The emperor of all maladies: A biography of cancer. New York: Scribner.

Michael Muthukrishna and Joseph Henrich, 2016. “Innovation in the collective brain,” Philosophical Transactions of the Royal Society B: Biological Sciences, volume 371, number 1690 (19 March), at, accessed 12 January 2018.
doi:, accessed 12 January 2018.

National Cancer Institute, 2017. “Immunotherapy: Using the immune system to treat cancer” (8 September), at, accessed 30 December 2017.

Matt Ridley, 2010. The rational optimist: How prosperity evolves. New York: Harper.

Aritra Sarkhel, 2017. “ZeroNet, an encrypted Internet for the post Snowden world,” ETtech (23 June), at, accessed 9 November 2017.

Paul Seabright, 2004. The company of strangers: A natural history of economic life. Princeton, N.J.: Princeton University Press.

Michael Segell, 2005. The devil’s horn: The story of the saxophone. New York: Farrar, Straus and Giroux.

Carl Shapiro and Hal R. Varian, 1998. Information rules: A strategic guide to the network economy. Boston, Mass.: Harvard Business Review Press.

David Shenk, 2006. The immortal game: A history of chess: A history of chess or how 32 carved pieces on a board illuminated our understanding of war, art, science, and the human brain. New York: Doubleday.

D. R. Shoults and C. J. Rife, 2042. Electric motors in industry. New York: Wiley.

Nick Srnicek, 2017. Platform capitalism. Cambridge: Polity Press.

Jonathan Sterne, 2012. MP3: The meaning of a format. Durham, N.C: Duke University Press.

Maurice E. Stucke and Allen Grunes, 2016. Big data and competition policy. Oxford: Oxford University Press.

Arun Sundararajan, 2017. The sharing economy: The end of unemployment and the rise of crowd-based capitalism. Cambridge, Mass.: MIT Press.

Alex Tabarrok, 2009. “How ideas trump crises,” TED2009, at, accessed 9 November 2017.

Don Tapscott and Alex Tapscott, 2016. Blockchain revolution: How the technology behind Bitcoin is changing money, business, and the world. Toronto: Portfolio/Penguin.

Peter Turchin, 2015. Ultrasociety: How 10,000 years of war made humans the greatest cooperators on Earth. Chaplin, Conn.: Beresta Books.

Steven Weber, 2004. The success of open source. Cambridge, Mass.: MIT Press.

Samantha Weinberg, 2016. “18 minutes to change the world,” Economist 1843, at, accessed 9 November 2017.

Stephen Witt, 2015. How music got free: The end of an industry, the turn of the century, and the patient zero of piracy. New York: Viking Press.

Robert Wright, 2001. NonZero: The logic of human destiny. New York: Vintage Books.


Editorial history

Received 9 November 2017; revised 30 December 2017; accepted 3 January 2018.

Creative Commons License
This paper is licensed under a Creative Commons Attribution 4.0 International License.

Antirival goods, network effects and the sharing economy
by F. Xavier Olleros.
First Monday, Volume 23, Number 2 - 5 February 2018

A Great Cities Initiative of the University of Illinois at Chicago University Library.

© First Monday, 1995-2018. ISSN 1396-0466.