Crowdsourced Digital Goods and Firm Productivity: Evidence from Free and Open Source Software DRAFT

As crowdsourced digital goods become more widely available and more frequently used as key inputs by firms, understanding the impact they have on productivity becomes of critical importance. This study measures the firm-level productivity impact of one such good, non-pecuniary (free) open source software (OSS). The results show a positive and significant return to the usage of non-pecuniary OSS that has gone unmeasured in prior studies of the economics of IT. The study addresses the endogeneity issues inherent in productivity studies by using inverse probability weighting, an instrumental variable approach, and firm fixed effects to add support for a causal interpretation. Across firms, a 1% increase in the amount of non-pecuniary OSS used by a firm leads to a .073% increase in productivity. This translates to a $1.38 million increase in value-added production for the average firm in the sample. This effect is greater for larger firms and for firms in the services industry. These findings show that firms willing to take on the risks associated with non-pecuniary OSS reap benefits from collective intelligence and labor spillovers. Further, the results indicate that existing studies underestimate the amount of IT at the firm, and therefore underestimate the importance of IT to firm productivity. 1 [email protected]. The author is grateful for helpful comments from Shane Greenstein, Carliss Baldwin, Yochai Benkler, Raj Choudhury, Anil Doshi, Marco Iansiti, Ohchan Kwon, Karim Lakhani, Kristina McElheran, Hart Posen, Scott Stern, Neil Thompson, Mike Toffel, Joel West, and Feng Zhu. Additional helpful comments were received from participants at ACAC 2014, AOM 2014, AOM 2014 BPS Dissertation Consortium, CCC 2014, Charles River Conference 2014, DRUID 2014, HBS TOM DBA Seminar 2014, HBS TOM Alumni Conference 2014, NYU Engelberg Center Conference on Knowledge Commons 2014, OUI 2014, SMS 2014, and ZEW ICT Conference 2014. All mistakes remain the author’s own. Nagle -­‐ Crowdsourced Digital Goods and Firm Productivity 2 I. Introduction As the digital age progresses, information goods are easier and easier to reproduce at costs that are rapidly approaching zero. Coupled with decreases in communication costs, this has made it easier for groups of individuals, frequently referred to as the crowd, to produce digital goods that are freely distributed to users who do not pay a monetary price. Wikipedia, the online crowdsourced encyclopedia, is a frequent example of this phenomenon. However, there are many other examples including open source software (OSS), crowdsourced innovation tournaments, and the digitization of consumers’ opinions via online review sites and social media. The same information cost decreases that enable the production of these goods also enable firms to use these crowdsourced goods as inputs into production. Recent research has shown that firms are increasingly relying on these types of goods to drive innovation and production (Baldwin and von Hippel, 2011; Lakhani, Lifshitz-Assaf, and Tushman, 2012; Corrado and Hulten, 2013; Altman, Nagle, and Tushman, 2014). This trend is also widely discussed in the popular press as technology giants like Apple, Google, and Facebook increase their reliance on crowdsourced digital goods to grow their innovative and productive efforts. However, it is not only technology focused companies that are relying on the crowd Ford, Pepsi, and a host of other well-known non-IT brands use free inputs from the crowd to help drive their bottom line. Further, these same crowd based technologies are allowing small startups to have a large impact, even when they are capital constrained, due to a reliance on free crowdsourced digital goods as inputs. Despite the growing importance of crowdsourced digital goods as inputs into production, measuring the value they help create can be difficult. In a classic Schumpeterian creative 2 Apple’s Mac OSX and iOS are based on the OSS operating system BSD (http://www.wired.com/2013/08/jordan-hubbard/), Google’s YouTube and other products rely heavily on user generated content (UGC), and Facebook is increasingly relying on OSS to power it’s infrastructure and giving back code to the open source community (http://readwrite.com/2013/10/17/is-facebook-the-worlds-largest-open-source-company). 3 Ford has relied on the open source community to help develop both hardware and software (http://www.forbes.com/sites/tjmccue/2013/01/10/ford-motor-company-sees-open-source/), while Pepsi has used crowdsourcing to develop ad campaigns (http://www.usatoday.com/story/money/business/2013/01/19/crowdsourcing-super-bowlcommercials-doritos-lincoln-pepsi/1842937/). Nagle -­‐ Crowdsourced Digital Goods and Firm Productivity 3 destruction process (Schumpeter, 1942), these new goods destroy old business models while creating new opportunities for growth. For example, the introduction of Wikipedia destroyed much of the market for pecuniary encyclopedias (both paper and digital). At the same time, Wikipedia has provided great societal value. However, as with all crowdsourced digital goods, this value is difficult to measure for two primary reasons. First, because these goods are frequently free, standard productivity measures, which rely on price to reflect value, do not properly capture these increasingly critical inputs. Second, because such goods are often distributed under licenses that allow for unlimited copying, it is unknown exactly how widespread such goods are. Despite the increasing prominence of crowdsourcing, these measurement challenges have prevented research from analyzing how its impact varies across different firms and market environments. As such, the goal of this paper is to answer the following question: what is the impact of free crowdsourced digital goods on firm productivity? After answering this broad question, the paper seeks to answer the related question: What are the firm-level determinants of the productivity related returns to such goods? As the production, and productive use, of such goods increases, the answer to these questions becomes more interesting and more important. Recent research has shown that the increased use of unpriced goods of both a digital (Greenstein and Nagle, 2014) and non-digital (Bridgman, 2013) nature may be an important factor in understanding recent trends in Gross Domestic Product (GDP). Whether measuring GDP via a country’s production or its expenditures, non-pecuniary digital goods can cause such a measure to greatly underestimate the true productivity of a nation and its firms. Further, these same issues can lead firms and managers to underestimate the importance of including crowdsourced digital goods as key inputs into their productive and innovative processes. While some leading firms, like Google and Facebook, have embraced the crowd and the free labor and content it provides, others have shied away from relying on such inputs due to concerns about reliability, sharing with competitors, and restructuring business models to add the user directly into the production and innovation process. Crowdsourced inputs are also allowing increasingly smaller firms to be valued at prices that seem absurd by pre-Internet standards. For example, WhatsApp, a five-year old firm bought by Nagle -­‐ Crowdsourced Digital Goods and Firm Productivity 4 Facebook for $19 billion, had only 55 employees at the time of acquisition and relied heavily on crowd-developed OSS to power its technology and on the crowd of users to drive its value. In addition to productivity-related implications, the reliance on, and contribution to, crowdsourced goods also has implications for firm competitive strategy. In a world where a firm must rely on actors outside of its boundaries for valuable inputs, and at the same time must consider contributing internally developed code to the world, co-opetition (Brandenburger and Nalebuff, 1996; Afuah, 2000) becomes an increasingly important concept. As firms’ competitors increase their reliance on crowdsourced digital goods, understanding how these goods contribute to productivity and for what types of firms they are the most useful becomes increasingly important to allow managers to make the right decisions regarding the crowd. Further, understanding the productive implications of free digital goods scratches the surface of the broader issue of all digital goods, which essentially have a marginal cost of zero, and are therefore likely priced below their actual value. To understand how usage of such non-pecuniary digital inputs affects firm productivity, the paper first discusses why such goods should have a positive impact on productivity and then considers what firm characteristics determine the degree of this impact. To test the resultant hypotheses, it utilizes a dataset that measures the usage of one particularly important non-pecuniary crowdsourced digital good, open source software (OSS) operating systems. OSS is a frequently used example of a digital good that is produced by a community of users and is a particularly interesting example due to the lack of price that is normally, although not always, associated with this model of production. This data is combined with firm financial data and productivity measures to allow for the application of a classic Cobb-Douglas production function analysis to understand the role of non-pecuniary IT inputs in firm-level productivity. This is a standard methodology for estimating the value of IT (Brynjolfsson and Hitt, 1996; Dewan and Min, 1997; Tambe, Hitt, and Brynjolfsson, 2012; Huang, Ceccagnoli, Forman, and Wu, 2013), although nonpecuniary OSS is normally not accounted for in such frameworks. Due to sample 4 WhatsApp acquisition and employee information obtained from http://blogs.wsj.com/corporateintelligence/2014/02/19/facebooks-whatsapp-price-tag-19-billion/. WhatsApp Open Source usage obtained from http://www.whatsapp.com/opensource/. Nagle -­‐ Crowdsourced Digital Goods and Firm Productivity 5 selection and endogeneity concerns, inverse probability weighting, a method similar to propensity score matching, is used to construct a setting more like that of an experiment. Panel fixed effects and instrumental variables are also utilized to allow for a more causal interpretation of the results. The results show that firms that use non-pecuniary OSS have higher levels of productivity than those that do not. Further, they show that increased usage of nonpecuniary OSS has a positive and significant impact on firm productivity. This makes intuitive sense since firms that use non-pecuniary IT are able to tap into the collective intelligence of the crowd through spillovers from free labor. The primary effect is robust to various endogeneity concerns, allowing for a causal interpretation of the results. The estimates indicate that a 1% increase in the amount of non-pecuniary OSS used by a firm leads to a .073% increase in productivity when comparing firms against other firms. This translates to a $1.38 million increase in production output for the average firm in the sample. This effect is greater for larger firms and for firms that are less technologically intense. Further, this effect is greater for firms in the services sector than for those in the manufacturing sector. The main effect is of a similar order of magnitude as other ITrelated inputs. Finally, because the study measures only non-pecuniary OSS operating systems, it does not capture other firm investments in non-pecuniary OSS. Therefore, the main effect is likely a lower bound for the true effect of all non-pecuniary OSS on productivity. Further, the results indicate that current studies underestimate the amount of IT at the firm, and therefore underestimate the importance of IT to firm productivity. This paper seeks to add insights to two important bodies of literature: the user innovation literature and the returns to IT literature. The user innovation literature (e.g., von Hippel, 1986, Chatterji and Fabrizio, 2014), in particular that which is centered on OSS (e.g., Kogut and Metiu, 2001; Lerner and Tirole, 2002; Lakhani and von Hippel, 2003; West and Lakhani, 2008), focuses primarily on supply side questions, e.g. why do individuals and firms contribute time and resources to the development of OSS, with almost no literature focusing on the demand and usage side of the OSS market. At the same time, the literature on the returns to IT investment (e.g., Brynjolfsson and Hitt, 1996; Tambe Nagle -­‐ Crowdsourced Digital Goods and Firm Productivity 6 and Hitt, 2012; Huang, Ceccagnoli, Forman, and Wu, 2013) focuses almost exclusively on IT investments that are of a pecuniary nature, completely missing investments in nonpecuniary IT, such as OSS. This paper contributes to both of these bodies of work by filling these important gaps in the literature and shedding light on the underestimation of IT used by the firm, and therefore the underestimation of the productivity impact of nonpecuniary IT. Understanding the impact of such goods on firm productivity not only helps to contribute to the broad literature on the determinants of productivity, but also shows that user innovation is no longer a rare phenomenon and is becoming a key input into firm productivity and innovation. Additionally, the paper offers insights for practitioners that can be utilized to increase the profitability of the firm’s operations and gain competitive advantage by using crowdsourced goods as inputs. Finally, for policy makers, the results encourage policies that incentivize production of public digital goods as a method for increasing firm and, in turn, national productivity. This paper is laid out as follows. Section II discusses the existing gap in the user innovation and productivity of IT literatures and then presents a brief history of OSS operating systems. Section III develops a theoretical rational for how overcoming the risks of using non-pecuniary OSS allows for firms to reap the benefits from collective intelligence, hypothesizes how OSS may affect productivity, and considers firm characteristics that are determinants of this effect. Section IV constructs the models used in the estimation and discusses strategies for dealing with the sample selection and endogeneity issues in the study. Section V details the dataset on OSS usage and firm production and the construction of the main variables. Section VI presents the results and discusses their implications, and Section VII concludes. II. Crowdsourced Digital Goods and the Returns to Information Technology One of the oldest and most successful crowdsourced digital goods is open source software and this will be the setting of the empirical analysis. Therefore, this section first reviews prior research on crowdsourced digital goods and user innovation as well as that on the returns to IT investments. In doing so, an important gap is identified at the 5 See Syverson, 2011 for an over view of this literature. Nagle -­‐ Crowdsourced Digital Goods and Firm Productivity 7 intersection of these two literatures, motivating the primary research question. Then, this section gives a brief history of the development of the two most widely used OSS operating system, GNU/Linux and BSD, both of which play an integral part in today’s modern IT ecosystem. II.A Free and Open Source Software as an Input into Productivity As early as the 1980’s, production by users has been a topic of interest in the management field (von Hippel, 1986). While such production is by no means limited to the digital world, it is here that user innovation is frequently studied, primarily in the realm of OSS. However, most of the academic work on OSS has been focused on exploring supply side mechanisms – why do users contribute to OSS (Lerner and Tirole, 2002; West and Lakhani, 2008), how do users join OSS projects (von Krogh, Spaeth, and Lakhani, 2003), how do users help each other contribute to OSS (Lakhani and von Hippel, 2003), and how do OSS communities protect their intellectual property (O’Mahony, 2003). Research on the supply side has also been extended to better understand why firms release some of their proprietary code as OSS (Harhoff, Henkel, and von Hippel, 2003; von Hippel and von Krogh, 2003; Henkel, 2006; Lerner and Schankerman, 2010; Casadesus-Masanell and Llanes, 2011). Despite the abundance of literature on the supply side of OSS, there is almost no literature on the demand side of OSS – who uses it, why do they use it, and are there productivity benefits to using it remain unanswered questions. This is despite the fact that OSS, and more broadly – non-pecuniary, community-based user-production, has been identified as an increasingly important input into the business models of firms in both academic literature (Krishnamurthy, 2005; Baldwin and von Hippel, 2011; Lakhani, Lifshitz-Assaf, and Tushman, 2012; Altman, Nagle, and Tushman, 2014; Greenstein and Nagle, 2014) and popular literature (Howe, 2008; Shirky, 2008). Although the productivity related value of OSS usage has not been directly investigated, there is a significant body of literature examining the impact of IT usage on productivity 6 The one notable exception is Lerner and Schankerman (2010), which explores the cross-country differences in demand for OSS usage. However, their analysis does not examine the returns to OSS usage and does not include the US. Nagle -­‐ Crowdsourced Digital Goods and Firm Productivity 8 at both the firm and country levels. This literature has shown that the rate of return for firm investments in IT is positive and significant (Brynjolfsson and Hitt, 1996) and productivity boosts from investments in IT are frequently mistaken for intangible firmspecific benefits (Brynjolfsson, Hitt, and Yang, 2002; Syverson, 2011; Tambe, Hitt, and Brynjolfsson, 2011). Studies have also shown that IT-producing and using industries contributed a disproportionately large amount to the economic growth experienced in the US, particularly from 1995-2004 (Jorgenson, 2001; Jorgenson, Ho, and Stiroh, 2005). In addition to spending on IT capital, spending on IT labor has also been found to boost firm productivity (Tambe and Hitt, 2012). Further, participation in networks of practice adds IT related knowledge spillovers that increase productivity (Huang, Ceccagnoli, Forman, and Wu, 2013). However, it has been found that not all firms receive the same return on IT investment (Aral and Weill, 2007) and that the returns to IT investment are not as strong as they once were (Byrne, Oliner, Sichel, 2013). An important aspect of all such studies is that they measure IT investment via dollars spent on software, hardware, labor, or a combination of the three. Since most OSS does not have a price directly associated with it, it is not properly factored into such calculations. This mismeasurement of “digital dark matter” has been shown to be on the order of billions of dollars for one piece of OSS in the US alone (Greenstein and Nagle, 2014) and the inclusion of intangibles and non-pecuniary production have been shown to significantly alter GDP calculations (Corrado, Hulten, and Sichel, 2009; Bridgman, 2013). Because of this measurement issue, OSS is not properly included in current productivity calculations, and therefore the productive value of OSS is currently unknown. Despite the vast literatures that exist in these two areas, there is a noticeable dearth of literature that addresses the intersection, leaving an open question this paper attempts to answer: What is the impact of OSS on firm productivity? After establishing a baseline 7 Although some literature exists analyzing the total-cost of ownership (TCO) when comparing open and closed source software (e.g., MacCormack, 2003; Varian and Shapiro, 2003; Russo et al, 2005; Wheeler, 2005; Fitzgerald, 2006), a consensus has not been reached and this literature does not explore the productivity implications of the two types of software, just the costs of employing it. The analysis in this study will control for the costs of employing either type of software by including labor and capital costs in the analysis. This allows for the measurement of the impact of the software itself even though the TCO question is not directly addressed. Nagle -­‐ Crowdsourced Digital Goods and Firm Productivity 9 answer to this question, the paper further considers the firm-level differences in extracting productivity value from OSS, allowing for a better understanding of the productivity implications of non-pecuniary crowdsourced digital goods. II.B Institutional Context: The Free and Open Source Software Movement Although the concept of free and open source software developed as part of the early computer culture, it was not formalized until 1983 when Richard Stallman founded the GNU Project to create a computer operating system that gave users the freedom to share and modify the software, unlike the predominant operating system at the time, UNIX, which was proprietary and closed-source software. Two years later, Stallman founded the Free Software Foundation (FSF), a non-profit organization designed to encourage the creation and dissemination of software with unrestrictive licenses, including the GNU General Public License (GPL), which continues to be the most widely used software license for free software. The FSF emphasizes that it uses the word “free” to mean “liberty, not price”, encapsulated in the pithy slogan “free as in free speech, not as in free beer.” However, the software released under this license is frequently also offered at a price of zero. This ambiguity later led to Eric Raymond’s call for the use of the term “open source” instead of “free” (Raymond, 1998). As the GNU Project progressed, it was successful in creating most of the middle and upper layers (user interface) of the operating system. However, very little work had been finished for the lowest layers, known as the kernel, of the operating system. In 1991, Linus Torvalds released the Linux kernel to take the place of the incomplete GNU kernel. GNU developers rapidly latched on to the Linux kernel and the combination of the Linux kernel and GNU software on top of it became the basis for most free and open source operating systems in use today. The other main free and open source operating system is the Berkeley Software Distribution (BSD) operating system, which was initially proprietary until a variant of version 4.3 was released as open source in 1989 under the terms of the BSD License, which allowed for redistribution provided the BSD License 8 GNU is a recursive acronym for “GNU’s Not UNIX”. 9 http://www.gnu.org/philosophy/free-sw.html, retrieved on February 23, 2014. Nagle -­‐ Crowdsourced Digital Goods and Firm Productivity 10 was included. Both GNU/Linux and BSD rely on a community of mostly unpaid contributors to maintain and upgrade the code base. Since these early operating systems were released, there has been a flood of free and open source software projects that are either a variant of these operating systems or are applications that run on top of them, such as the vast array of projects maintained by the Apache Software Foundation. Although unrestricted non-pecuniary software is at the core of the free and open source software movement, many companies have structured profitable business models on top of this software. Common examples of this include Red Hat, which offers its own Linux distribution and charges for customer support, the IBM HTTP Server, which is built on the open source Apache HTTP Server and is included with the IBM WebSphere Application Server, and Apple’s Mac OS X, which is built on the FreeBSD operating system. III. Theory and Hypothesis Development This section constructs a theoretical argument for why firms that take on the risks of using non-pecuniary OSS obtain a positive impact on productivity. It then constructs a number of hypotheses about what characteristics of the firm help to determine the level of impact usage of non-pecuniary OSS has on firm productivity. Since it was first created, there has been a long debate on the value offered by nonpecuniary OSS. Although there is a diversity of opinions, the consensus in the literature on the total cost of ownership (TCO) of software is that the actual cost for software is negligible when compared to the hardware and labor costs of implementing, using, and maintaining it (e.g., MacCormack, 2003; Varian and Shapiro, 2003; Russo et al, 2005; Wheeler, 2005; Fitzgerald, 2006). Beyond potential cost-savings, non-pecuniary OSS is often seen as riskier than pecuniary software for a number of reasons. First, because a 10 The focus of this research is primarily on non-pecuniary OSS. The availability of pecuniary products, like Red Hat Linux, which build on non-pecuniary OSS is important, but the risks associated with these products is lower due to the contractual relationship a customer has with the vendor, which greatly mitigates these risks. Nagle -­‐ Crowdsourced Digital Goods and Firm Productivity 11 collective of users, rather than a central producer, creates non-pecuniary OSS, there is rarely official technical support for the products. While some users do offer help by creating manuals or answering user questions (Lakhani and von Hippel, 2003), there is no guarantee that a user’s question will ever be answered because they do not have a service agreement with any vendor. Relatedly, although larger OSS foundations, like the Linux Foundation and the Apache Foundation, employ commons-based governance structures (Ostrom, 1990; O’Mahony and Ferraro, 2007), there is no guarantee that the OSS project will be continuously developed and supported. Likewise, even if the project is continuously maintained, there is no guarantee about the features and technical path of future versions. From a security standpoint, the openness of the underlying code in OSS allows anyone to examine it for security vulnerabilities. Although Linus’s Law would predict that the open nature of the code would be a benefit from a security perspective, recent widespread vulnerabilities in OSS integral to the operation of the Internet and Linux have shown that these bugs are not always caught early in the development process. Perhaps the most concerning risk of all is the lack of a contractual relationship between a firm using nonpecuniary OSS and any one entity responsible for the development of such software, which leaves the firm with no one to sue when something goes wrong. There are no service level agreements (SLAs) for non-pecuniary OSS, which means the use of such software is riskier than pecuniary software where such agreements exist. The view of non-pecuniary OSS as a risky decision led to the commonly used phrase “No one ever got fired for buying Microsoft.” This phrase became popular in the technology industry 11 Linus’s Law is attributed to Eric Raymond (1999), but named after the founder of Linux, Linus Torvalds. Linus’s Law states “Given enough eyeballs, all bugs are shallow,” which implies that the more people who look at the code, the more likely bugs are to be found and fixed. 12 The Heartbleed security bug was introduced into the OpenSSL cryptography library in December 2011, and was not noticed and fixed until April 2014. As of May 8, 2014, more than 300,000 public web servers were still vulnerable to the issue (Graham, 2014). The Shellshock security bug was introduced into the Bash Shell in 1992, and was not noticed and fixed until September 2014. The Bash Shell is used in nearly all Unix-style operating systems, including Linux and BSD, the latter of which is the basis of the Mac OS X operating system. 13 This phrase actually started about IBM in the 1970’s, long before OSS. However, it was ported to Microsoft in the 1990’s as OSS started to gain traction in the marketplace. Interestingly, IBM later invested heavily in OSS and built some of its products on top of OSS. However, IBM but Nagle -­‐ Crowdsourced Digital Goods and Firm Productivity 12 as customers were increasingly willing to pay a premium for software from big name firms they could trust. However, another pithy quote from the technology industry helps to illuminate the potential benefits of non-pecuniary OSS – “No matter who you are, most of the smartest people work for someone else.” This quote, known as Joy’s Law, highlights the fact that regardless of how big and powerful a company is, it can never hire all of the best and brightest people. In the software development world, this means that code developed within a closed firm cannot benefit from the intelligence of anyone outside of the firm (von Hippel and von Krogh, 2003). Non-pecuniary OSS projects address this problem by allowing anyone to contribute to the development of the underlying code base. Therefore, the use of OSS allows a firm to harness the labor efforts of a wide collective of individuals. Although collective intelligence and the wisdom of crowds is often associated with completing simple problems, recent research has shown that the crowd can also be successful in solving more complex problems (Woolley et al, 2010; Woolley and Fuchs, 2011; Yi et al, 2012), including software development (von Hippel and von Krogh, 2003). Like many investment opportunities a firm must make, the decision to invest in nonpecuniary OSS allows firms that are willing to take on higher levels of risk to obtain higher levels of reward. For many firms, the risks of relying on non-pecuniary OSS are too high and they therefore rely on pecuniary software. However, the firms that are willing to take on the risks associated with non-pecuniary OSS allows them to obtain the benefits of tapping into the collective intelligence of the crowd, leading to productivity spillovers from the free external labor and knowledge that support the non-pecuniary OSS ecosystem. Although such a benefit should occur with any use of non-pecuniary offered large support contracts and SLAs, removing many of the risks associated with the use of non-pecuniary OSS. 14 This statement is from a speech Bill Joy, the co-founder of Sun Microsystems, gave in 1990, and was first mentioned in print by Gilder (1995). 15 While it is true that some firms who use non-pecuniary OSS also contribute back to the creation of these products, even these firms benefit from the external labor contributed by other firms and individuals, which they do not pay for. A deeper analysis of this relationship is left for future research. Nagle -­‐ Crowdsourced Digital Goods and Firm Productivity 13 OSS, the benefit should increase as more non-pecuniary OSS is used. These related hypotheses are presented formally as follows: H1a: The usage of non-pecuniary OSS at a firm has a positive impact on firm productivity. H1b: An increase in the amount of non-pecuniary OSS used at a firm has a positive impact on firm productivity. Due to differences in capital constraints, it is likely that firm size will play a role in determining the productive impact of non-pecuniary OSS. For very small firms, nonpecuniary OSS can play a critical role in allowing the IT capability of the firm to ramp up quickly, without expensive outlays for pecuniary software. However, as firms grow, it is likely they will not be able to fully support a non-pecuniary OSS infrastructure themselves, and will therefore rely on external consulting firms to take the place of the support that comes with pecuniary software. On the other hand, larger firms have the capacity for greater economies of scale and can therefore obtain greater returns from their IT investments as well as any consulting activities to help implement an OSS infrastructure. Together, this implies a quadratic relationship between firm size and productivity returns to non-pecuniary OSS that is high for very small firms, drops for medium sized firms, and increases for larger firms. Due to data restrictions, it is only possible to test the latter portion of this relationship and the former is therefore left for future research. This leads to the following formal hypothesis: H2: Except for very small firms, the productivity impact of non-pecuniary OSS is greater for larger firms than for smaller firms. Firms differ on how intensely they use IT in their production processes. When firms that are less IT-intensive adopt non-pecuniary OSS, this unpriced software is likely to represent a larger portion of their IT usage when compared to a more IT-intensive firm.