'THE CURSE OF THE FOLLOWER': PATH - CiteSeerX

5 downloads 0 Views 269KB Size Report
15 end of the Seventies and became dominated by the Avalon Hill company, .... of products in a specific genre spurs further innovation, attracts free lancers to.
Document de travail du LEM 2008-03

‘THE CURSE OF THE FOLLOWER’: PATHDEPENDENCY AND ITS CONSEQUENCES IN THE ROLE PLAYING GAME INDUSTRY

Benoît Demil*, Xavier Lecocq**, Filippo Carlo Wezel*** * IAE Lille ** IAE Lille and IÉSEG School of Management *** University of Lugano, Switzerland

‘The curse of the follower’: path-dependency and its consequences in the Role Playing Game industry

ABSTRACT The present paper studies the path dependent process that led to the emergence of a dominant market segment. In particular we focus on (i) nailing down the mechanisms that triggered this outcome, and (ii) on its associated organizational consequences – i.e., in terms of behaviour and performance. A qualitative case study and a set of quantitative analyses are carried out making use of the secondary data collected in the Role Playing Game sector during the period 1974-1999. The findings presented shed light on a refined view of the antecedents and organizational consequences of path dependency.

1

INTRODUCTION Scholars from fields as diverse as sociology, history, economics and management have investigated the study of the path-dependent processes underlying organizational, institutional, technological and industrial evolution – for an overview see, e.g., Garud and Karnoe (2001). This growing body of research has heavily drawn on Arthur’s logic (1994) to underscore – sometimes even naively (Mahoney, 2000) - the role of history-dependent mechanisms in shaping economic agents’ choices. The logic of path dependency has been applied to the emergence and the persistence of dominant technologies (e.g., Arthur, 1994, Chap. 2; David, 1985; Cowan and Gunby, 1996; Puffert, 2000), to scientific theories (Sterman and Wittenberg, 1999), to spatial agglomerations (Arthur, 1994, Chap.4), to the evolution of organizational populations (Carroll and Harrison, 1994) and to provide justification to strategic persistence (Karim and Mitchell, 2000; Lamberg and Tikkanen, 2006).

In the present article, we employ the self-reinforcing mechanism fuelling path-dependency to comprehend the evolution of market segments. Indeed, we study how the aggregation of individual choices led to the emergence of a dominant market segment with a specific focus on: (i) nailing down the mechanisms responsible for such outcome and (ii) on the associated organizational consequences – i.e., behaviour and performance. Our argument revolves around the concept of niche density, ultimately stressing the striking similarities between density related models and path dependence ones (see Carroll and Hannan, 2000) – albeit the former adopts a sociological lens while the latter focuses on economic forces. As density acts as a signal of the potential viability of a niche, we argue, it serves as a guide to drive the positioning of first time entrants, ultimately increasing the potential for spillovers (both on the demand and on the supply side) and stimulating further re-entries. Such a reinforcing mechanism (i.e., positive feedback) lies at the origin of what we label as ‘the curse of the

2

follower’: organizations initially attracted in the most visible segment are progressively faced with over-crowding and various lock-ins.

After a qualitative description of the Role Playing Game sector (henceforth, RPG) over the period 1974-1999, the hypotheses advanced will be tested using quantitative methods on secondary data. During the period covered by the study, the fantasy segment progressively gained momentum, but the organizational consequences of joining such segment turned from positive to negative due to increasing competition. The results of our quantitative inquiry indicate that, faced with negative results, incumbent organizations exhibited substantial inertia, whereas new entrants did engage in competitive differentiation. We believe that the findings presented here provide a refined view of the antecedents and organizational consequences of path dependency and may be seen as complementing other approaches that rely on positive feedback mechanisms — e.g., the pioneer/follower framework.

This article is constructed around five sections. The first delineates the theoretical perspective of path dependence reviewing its main definitions and the state of the art. Then, we introduce the hypotheses derived from our conceptual framework. The second section describes the empirical field of our study and the data collection process. The third section introduces the analyses and the models used in this article. The fourth section presents and discusses the results of such analyses. The last section underlines the implications of the results obtained.

3

THEORETICAL BACKGROUND Path dependence as a process of convergence Path dependency appears as an intriguing characteristic of systems with increasing returns where agents have to make choices by allocating their resources to different potential courses of action. The presence of increasing returns to scale implies, for instance, that an increase in output will be proportionally greater than its input additions, resulting in average costs decreases. The more a firm produces, the more it will do it efficiently. Broadly speaking, the logic of increasing returns lies on the existence of a positive feedback on the state of a system. Such a view contrasts with the traditional approach of decreasing returns which tend to stabilize systems via negative feedback (Arthur, 1994). It is worth noticing that positive returns do take place not only on the side of producers but also on that of consumers too. The more consumers adopt a certain technology, in fact, the more such a technology acquires value for its users. Consider, for instance, the case of network technologies requiring compatibility across products or users triggering network externalities (Farrell and Saloner, 1985; Katz and Shapiro, 1985). In these situations, users do exhibit strong incentives to buy the most diffused technology/product to engage in (information) exchanges with peers.

Due to this intuitive feature, any system characterized by increasing returns is potentially conducive to the emergence of dominant solutions. Indeed, such systems exhibit four specific properties (Arthur 1994, Chap. 2): (i) path dependent processes may be inefficient (i.e., the course of action taken is not necessarily path-efficient); (ii) non-predictable results are possible; (iii) path dependent processes are inflexible (i.e., lock-ins unfold and the manifestation of an event depends on its previous occurrence); (iv) nonergodicity prevails (i.e., the role of small events to create irreversible paths). The basic definition of pathdependence refers to the fact that minor historical events are not “forgotten” by the dynamics

4

of the system (Arthur, 1994). Consequently, under conditions of increasing returns, small accidents occurring in the early history of the system are not averaged away and dramatically impact on its evolution. Due to an initial accident, a path-dependent sequence may appear more largely influenced by chance or contingent elements than by systematic forces (David, 1985). The logic underlying increasing returns reinforces the interest to repeat previous choices, and to the emergence of rather inflexible paths even when faced with suboptimal outcomes. For the sake of clarity, while path dependency often result into a convergence of choices around one -or eventually few- solutions, such a convergence does not necessarily lead to the disappearance of other options.

The relevance of path dependency in organizational theory We believe that the concept of path dependence provided three contributions to the organizational theory literature. First, suboptimal outcomes and unintended effects are made less dependent on the intrinsic qualities of a product or a technology. Arthur (1994) illustrates this point by means of simulation experiments and for the specific case of technological races. Sterman and Wittenberg (1999) focused on a different research question (the development and consolidation of theories) and concluded that “new theories with great explanatory power frequently fail to attract a critical mass of adherents, while weaker ones often triumph” (p. 338). Consequently, path dependency presumes that markets fail to promote the most effective solution. Such a debate extended to considerations regarding the lack of historical efficiency in organizational development. As Carroll and Harrison (1994: 745-746) noticed “(…) the social acceptance of a particular organizational form may embed the population defined by it so strongly in the social fabric of its environment as the “right way” to accomplish a particular task or goal that it renders technically superior alternative unthinkable and thus unviable”.

5

Second, several authors have stressed that path dependency brings ‘historicity into economics’ (David, 2000) and enables to take the time dimension explicitly into account (e.g., Pierson, 2004; Lamberg and Tikkanen, 2006). In particular, path dependence illuminates three different but related aspects associated to evolution. In a path-dependent process, past choices heavily influence contemporaneous ones. That is because, for instance, the utility of a technological solution is proportional to its installed base – i.e., the number of previous adopters. For organizational researchers, the inter-temporal correlation between decisions is particularly intriguing because helps interpreting the absence of change in actors’ behaviours. Moreover, according to the path dependence argument, any initial small and contingent perturbation may be viewed as the main driver of a system’s evolution. For instance, the diversity of gauges chosen by North American railway companies has been shown to be driven by contingent considerations (Puffert, 2000). Last, path dependence gives prominence to the order of appearance of actions (Pierson, 2004, Chap. 2). Changing the initial path of a system may radically modify the evolutionary trajectory observed.

The third attractive characteristic of path dependency concerns its associated lock-ins. The existence of lock-ins is particularly interesting to organizational theorists because it points to courses of action spanning beyond ‘pure rationality’– an argument echoing in diverse concepts adopted in organization theory such as institutions, bounded rationality, routines, structural inertia and escalation of commitment. As path dependency limits actors’ rationality, it may be viewed as inhibiting organizational adaptation. Indeed, organizations may be locked in previous strategic choices that may become maladaptive to the environment and to current competitive scenario (Lamberg and Tikkanen, 2006). Above and beyond the negative sides of path dependency, organizational researchers have evoked also its positive features. In

6

particular, similarly to inertia, path dependency is conducive to the consolidation of organizational procedures and to resource accumulation – e.g., Karim and Mitchell (2000).

Research question and hypotheses Our purpose is to test the framework of path dependency in the context of imprinting and subsequent development of an emerging industry with particular attention to the level of market segments. In particular, this research is guided by two broad questions: Why initial strategic moves are determinants for the evolution of an industry? What are the consequences of a path-dependent trajectory for organizational behaviour and performance? To deal with these questions, we introduce the concept of density of organizations as a key measure of aggregation effects occurring within segments. We claim that such a density-guided aggregation may be seen as an antecedent (as it acts as a signal of success and favours increasing returns) and a consequence of path-dependent processes.

Initial moves Arriving in a brand new industry poses great challenges for newcomers. They are requested to invest on products and technologies, to target customers and to position in segments often without much of the relevant knowledge. In this context, the formation of entrepreneurs’ expectations represents a key component of entrepreneurial choice, involving a wide variety of issues ranging from demand and supply conditions to resources availability and to risk propensity (Lomi et al., 2007). However entrepreneurs often lack criteria to evaluate the performance of other companies (Aldrich and Fiol, 1994). Because of the inherent complexity and uncertainty associated with forecasting distant events, entrepreneurs tend to rely on selected signals and use them as anchors to guide their decision making.

7

Under this scenario, mimicry appears an alternative rationale adopted by boundedly rational decision makers (Levitt and March, 1988; Greve, 1998) and new organizations tend to replicate the choices of incumbent organizations (Aldrich, 1999). Haunschild and Miner (1997) classified imitation along three dimensions. The first refers to benchmarking on successful ventures (e.g., Delacroix and Rao, 1994). Nevertheless, the lack of objective criteria for success in emerging industries (Aldrich and Fiol, 1994) renders the success of others difficult to evaluate. A second mechanism consists in targeting organizations with prominent characteristics such as high status (e.g., Baum et al., 2000; Barreto and BadenFuller, 2006). However, this approach may be of limited utility when most of the incumbent organizations are de novo– i.e., with no progressed identity. Thus, a last alternative may be to follow the most frequent choices and practices adopted in the market – what we may conceive as a ‘density-based’ heuristic. Entrepreneurs may have interest in jumping on the bandwagon of an emerging dominant market segment to improve the viability of their ventures. Consequently, frequency based imitation (i.e., density-informed decision making) will constitute an important clue to make sense of the environment and to generate a signal of what may be an appropriate choice for that specific market. Under this scenario, the first emerging segment in a market is likely to receive disproportionate attention from entrepreneurs – for a comparable logic see, e.g., Sorenson and Audia (2000) on the proliferation of geographic agglomerations. Consequently, we propose that:

Hypothesis 1: Firms entering an industry for the first time will target segments with higher visibility.

8

Reinforcing mechanisms A mechanism for the reproduction of the initial moves needs however to be at work to trigger path dependency and overcoming the exploration of alternative options (Schwartz, 2003). Beyond the inertia of individual organizations (Hannan and Freeman, 1984), the persistence of initial choices in a market can be induced by positive feedback mechanisms of several types such as scale economies, learning-by-doing, and commensalism (e.g., Arthur, 1994; Carroll and Harrison, 1994: 748). This line of reasoning suggests that organizations will exhibit convergent choices, further reinforcing the appeal of them. Thus, the more visible a segment becomes, the more it will spur subsequent convergence and it will attract new organizations, further reinforcing the potential spillovers attached to the segment and triggering re-entries of incumbents. Under the presumption of a minimal forward looking activity, incumbent firms in fact will be able to foresee the opportunity of positive spillovers and re-enter the same market segment to improve their survival chances. In presence of increasing returns from adoption (.e.g., low prices, compatibility with other consumers) this process will enact a virtuous circle also on the users side, , attracting more and more customers in the dominant segment and ultimately increasing its carrying capacity. Building on these considerations, we advance:

Hypothesis 2a: Firms re-enter segments where they entered before – especially when initially choosing highly visible segments. Hypothesis 2b: Firms re-entering the same segment will benefit from improved survival chances – especially when located in highly visible segments.

9

Lock-in consequences for incumbents and new entrants An open debate in the existing literature concerns the existence of manifest strategic actions. Indeed, some authors have suggested that path dependency promotes a mechanistic view of markets and technologies, not leaving much room to strategic thinking (e.g. Cusumano et al., 1992; Schilling, 1998). In particular, Liebowitz and Margolis (1995) have questioned some of the assumptions of Arthur’s models. According to these authors, economic agents do not actively take part in technological races and seemingly lack entrepreneurial skills. Conversely, Liebowitz and Margolis (1995) argue that the existence of suboptimal solutions offer to producers the opportunity to detect new profit opportunities and to develop alternative paths of growth. In the similar vein, Schwartz (2003) has also argued that path dependency focuses on lock-ins and not enough on the degree to which endogenous changes lead to the emergence of alternative scenarios. These critics open the door to the inquiry of lockouts, i.e. escapes from a specific course of action. In response to this critique, other authors have argued that path-dependency does not imply an infinite self-reinforcing sequence but, above all, provides an explanation for the emergence of persistence in social life (Pierson, 2004).

Empirically, Puffert (2000) have provided mixed evidences of these divergent arguments. He proves the validity of a path-dependence process in the early stages of development of the North American railways, but he simultaneously show how incumbent firms exhibited strategic and purposeful behaviours. The interest of companies in compatibility of gauges grew as interregional rail traffic increased and the final 4’8.5’’ gauge standard overcame other options due to its early adoption by important lines imitating the British standard, creating a large installed base for this standard and favouring further adoptions to increase compatibility. While strategic considerations inspire these choices of adoption, the standard in question was not intrinsically technologically or economically optimal – see also Arthur (1984).

10

We address these seemingly conflicting findings by drawing on Pierson’s discussion of institutional enhancement (2004: 122-129). In particular, two mechanisms may be at work against the persistence of inefficient path-dependent trajectories. The first refers to competitive pressures. It implies that, eventually, an inefficient solution is out-competed by better ones. For instance, Sterman and Wittenberg’s simulation (1999) incorporates competition among diverse theories to argue that the dominance of a given perspective conforms to Kuhn’s argument on paradigm shift. In our study, competition entails that organizations choosing an inefficient solution will suffer from their choice (i.e., low performance) more than those diversifying. In the present study, we read competitive pressures as coming from the overcrowding in the dominant segment. The second mechanism refers to learning processes. If negative consequences of a given choice are observed, purposively rational actors learn from them and opt for more rewarding solutions. If so, we should observe incumbents and new entrants learning about the negative consequences of a path-dependent competition (i.e., overcrowding here) and progressively targeting other segments. Consequently, we hypothesize:

Hypothesis 3a: Competitive pressure is greater in the highly visible segment than in any other one, leading to higher mortality rates. Hypothesis 3b: Competitive pressure in the highly visible segment will lead incumbent firms to avoid it and to move to other market segments. Hypothesis 3c: Competitive pressure in the highly visible segment will lead new entrants to avoid it and to enter into other market segments.

11

DATA COLLECTION AND RESEARCH FIELD In this section of the paper, we introduce our research setting, the role-playing game (RPG) industry, and the data collected. We will also describe the emergence of this industry and its later developments. Our aim is to make use of qualitative data to discuss how a given market segment (i.e. the fantasy genre) has evolved and how its dominance got reinforced through diverse increasing returns mechanisms.

Data collection The population under investigation is that of tabletop RPG publishers using English language and releasing printed products (and not PDF ones, appeared towards the end of the Nineties). While some publishers are English or Canadian, the vast majority of them is based in the United States of America. Our empirical study concerns the analysis of 25 first years (19741999) of this industry1.

In the first place, we collected qualitative data on the industry from RPG publications (Comics and Games Retailer, D20 Magazine, Dragon Magazine), books (Schick, 1991) and Internet websites (D20 Reviews, Game Manufacturers Association, Game Publishers Association, Game Spy, Gaming Report, RPGA Network, RPGNow, RPG Planet, Wizard’s Attic). A great deal of factual data and interviews were collected during this initial phase. Academic literature is also available , with specific emphasis on its role as a social practice (e.g. Kociatkiewicz, 2000; Fine, 2002; Booker, 2004). The purpose of this qualitative section of the paper is twofold. Firstly, it provides an introduction to the RPG industry, its market segments, its products and consumers. Secondly, it enables us to specify the underlining 1

We are concerned here only with paper-and-pencil RPGs (also called “tabletop RPGs”). This category should be neither confused with computer role-playing games, which make use of video, nor with live-action role-playing games where costumed players act out the actions associated with their characters.

12

mechanisms responsible for increasing returns – and ultimately behind path dependence evolution.

In a second stage of our research design, we built a dataset encompassing 444 RPG publishing companies and 6,112 products including core rules and complementary products. To collect the data, we crosschecked several archival sources: a book on collectibles (the “Official price guide to Role Playing Games” by Brown and Lee (1998)), two databases found on the Internet and elaborated by German fans (RPG Index and Drosi), an English online encyclopaedia (Darkshire) and a French online encyclopaedia (FFJDR) dedicated to the industry. We found these sources congruent and allowed us imputing the missing data of the “Official price guide”. A brief description of each product allowed us to classify each of them across the 12 genres identified. Double coding has been led and litigious cases were discussed among the authors to improve the coding.

Finally, our dataset contains variables not only concerning products but also editors (date of entry, genre of each product, year of release of each product, price, kind of product (core or complementary), number of pages, price, authors of each product). Thus, this database encompasses every publishing company and its products for the period 1974-1999.

What is a roleplaying game? In the RPG industry, products are different types of books containing rules of the game, and setting descriptions/scenarios. This hobby is usually defined as “an activity in which a group of people (called the players) creates and role-plays characters in a world devised by another participant, called the Game Master, who describes the results of their actions as well as the actions themselves of everything and everybody else in this created world” (Kociatkiewicz,

13

2000: 71). Players and the Game Master (or the storyteller) follow rules of the game to create characters and to decide players’ actions. The game consists in a storytelling adventure generally involving 3 to 6 people who take on the roles of selected alter egos and interact orally with other players around a table for several hours. The objective is, above all, to share the experience and not necessarily to win. Players need just paper and pencil to keep records of their adventure under the guidance of the Game Master and throw of a dice determines the specific events, ranging from testing their skills to fighting.

We selected the RPG industry for two reasons. First, it provided an opportunity to study the emergence and development of an industry where qualitative data is plethoric. It is worth noticing, however, that because of the small size of most of the publishers, financial data are lacking. Second, in the RPG industry we were able to identify all the relevant actors, and in focused on firms for great part undiversified into other activities. Thus, the consequences of the strategic choices of the firms involved are relatively transparent. The choice to study a 25 years period, from the inception of the industry in 1974 to 1999 is motivated by the fact from 2000 onwards, various technological and market shakeouts have drastically modified the industry and its structure. For instance, the year 2000 witnessed the emergence of electronic RPGs sold in PDF format through Internet platforms and the diffusion of a new property right regime (an open source) which attracted several new entrants (see Lecocq and Demil, 2006).

From Napoleon to the Dragon: The origins of a new business The emergence of RPG is closely linked with the wargame field. Indeed, Gygax and Arneson -the inventors of Dungeons and Dragons (D&D)- were themselves war-game players and initially belonged to this community. Wargames were born in 1953 and aimed at reconstructing historical battles with tokens or miniatures. These kind of game boomed at the

14

end of the Seventies and became dominated by the Avalon Hill company, at least up until the point in which computer wargames took over – i.e., in the Eighties (Dunningan, 1992).

Under the umbrella of the International Federation of Wargamers (IFW) - composed by several hundred members at the beginning of the Seventies--, Gygax created the “Castle and Crusade Society” subgroup dedicated to games set during the medieval period. Gygax met Arneson (who belonged to the IFW too) during the annual convention of enthusiasts, the GenCon. They began to collaborate on various wargames rules and, as from 1971, they began adapting the rules to individual players and not only to armies, opening the path for RPG. In Gygax’s recollection of the events “in 1971 I published a fantasy supplement for that company called Guidon Games: Chainmail, which had all the rules that we’d been using was what became Dungeons & Dragons, because in order to accommodate under a gable we couldn’t be everybody on tabletop, so we just started “well, we could do this with paper and pencil”, and instead of having lots of figures everybody just has one, and instead of ten people each with sixty figures, we could have thirty people which had one” (Kliehm, 1993). Moreover, instead of being focused on a specific historical period and to keep up with the interests of their reference group, the two creators also introduced fantasy elements into their games. Arneson’s recollection stresses how accidental was the choice of the fantasy setting: “Some other people in my group set up rules for modern games, or even back in the age of Napoleon. We would get in these arguments, though, about historical accuracy, the latest translation of the latest book, and what was “real” (…) I didn't want people always coming up with some new book saying we just had to use because it was right and the old one was wrong. This was a fantasy world, so who could come up with anything to prove that he was lying or that a monster wasn't accurately represented? [Laughs]” (Rausch, 2004b).

15

The huge success of Tolkien’s novel at the end of the Sixties in both the United States and in England became a good justification for the choice of a medieval setting —flowered with magical items and wizards. This was the recipe for Dungeons & Dragons (D&D) the first ever published role played game (Booker, 2004). The above elements show how contingent was the choice of the fantasy genre for the industry. This point is crucial as any path-dependent explanation has to envisage « what might have been » (Cowan and Gunby, 1996: 521).

The take-off of the Dragon In the October of 1973 Gygax and his friends founded the Tactical Studies Rules (TSR) publishing company and started selling their games. The same year, Gygax and Arneson created D&D. At this time, Gygax thought that “all we would get was the hardcore gaming hobbyist and maybe science-fiction fans” (Rausch, 2004a) despite encouraging signs: “At that time people were calling me day and night, and they said “hey, we are six guys playing your game and have a problem here, can you tell us how this works”. So I had to do two things: First change my phone number and have it not listed anymore, second write down the rules. So in 1972 I did a 50 page set of rules called “Dungeons & Dragons”. And it by 1973 became 150 pages. And that was basically what got published in 1974” (Kliehm, 1993). Interestingly, the authors tried first to sell the game to a game publisher, Avalon Hill, which declined under the conviction that there was no market for games without winners. Indeed, in the RPG players have to cooperate to deal with specific quests.

However, following the 1974 release by TSR, the D&D’s fame spread rapidly. As Arneson recalls, “we sold the five [first] hundred copies in just a few months, which was amazing at that period. Usually, you had a game that sold a thousand copies and was regarded as being successful, and we did five hundred in three months. We reprinted again, I think we did a

16

thousand on the second run, and that sold out in a couple months. Then we did two thousand and that sold out in a couple months until it became a hit” (Rausch, 2004b). TSR sold 1000 copies of D&D in the first ten months of 1974. 2,000 others were sold by April 1975 and 3,000 units more by the July of that year2. Thanks to word of mouth D&D became a cult not only in the United States but also across the globe, particularly among school and university students, with the result that fans started demanding for more materials. D&D’s success was phenomenal: translated into more than dozen languages and sold in 50 countries, it inspired television shows and movies, and spawned a large number of imitators and competitors. Meanwhile, TSR moved from a tiny two-people firm to a 300 employees in 1982 and illegal photocopies of D&D started floating around.

The rise of new market segments The success of D&D attracted new entrants into the field and led to the emergence of a new industry of RPGs, with its publishers, fan clubs, hobby magazines and the development of its own sub-culture (Fine, 2002; Booker, 2004). The RPG players became organized in small communities and conventions or tournaments became the meeting point for fans and enthusiasts. Today an estimated five million people worldwide play RPG at least once a month —half of them in the U.S.— (Dancey, 2000).

After the release of D&D, several other RPG rule systems were created by rival companies, drawing inspiration from a variety of source materials. While jumping on the fantasy bandwagon, they explored other settings for their rule-sets, giving birth to alternative genres. Consequently, the unfolding aggregation cannot be attributed to a voluntary collective choice

2

A successful wargame achieves several thousand copies with a few blockbusters hitting up to 300,000 copies after several years of exposition to the market (Dunningan, 1992). Conversely, thirty years of success led D&D to sell several million copies.

17

but rather to the accumulation of agents’ decisions. Altogether, 11 different genres (which are the market segments of the RPG industry) have emerged since 1974: “science-fiction”, “western”, “modern setting”, “horror”, “humour”, “super-heroes”, “historic setting”, “postapocalyptic”, “pulp”, “all-settings” (universal products that may be used with any core rules), and “other” (including specific themes such as time travel) (see Appendix 1). Altogether these 12 market segments are nowadays recognized by the publishing companies, players and experts of the industry. Moreover, the different encyclopaedias we reviewed exhibit wide consensus on the above classification (e.g. Schick, 1991; Brown and Lee, 1998).

Fantasy, therefore, was the initial segment explored. Congruent with a path dependency framework, the choice of this genre is explained by the context within which the decision was taken as well as by the personal preferences of the creators. The dominance of this genre — measured by the number of products released compared to the other genres - remained unchallenged for more than twenty years. Indeed, the proportion of products yearly released in the fantasy varies from 27% in 1997 to 88% in 1978, with a mean proportion of 44% over all the period under study (1974-1999). These figures invite three remarks. First, pathdependent explanations do not predict the monopoly of a technology or a choice, but rather its dominance. The diversity of consumer tastes may contribute to preserve product level differentiation. Second, 1997 marks the year of collapse of TSR, one of the main providers of fantasy products, after a financial crisis which led to a takeover. This year TSR released only 9 products way below their annual mean of 40. Finally, as the proportion of games belonging to the fantasy segment decreased over time, an homogenous distribution of games across different genres should be observed whereas a 30% to 40% of supply is still concentrated in the fantasy segment at the end of our window of observation.

18

The business model in RPG sector The industry structure is characterized by low entry barriers, due to minimal capital requirements to enter. Like writing a novel, creating a RPG primarily requires time and a good imagination. Low entry barriers encourage new entrants, increasing the proliferation of RPGs, often incompatibles due to proprietary core rules.

The RPG sector comprises numerous small publishing companies, in many cases set up by hobbyists. These firms release products in the form of books containing the rules of play, the descriptions of the setting and the diverse adventure scenarios. These modules constitute a RPG. The RPG can be considered as a technological system, with different modules in need of being compatible to be jointly used. Traditionally, each firm develops its own set of core rules and complementary products to create a system partially incompatible with those released by its competitors. As game rules are not standardized, buyers have buy several books (the game’s rules of play as well as complementary products such as adventure scenarios) to play each RPG. Often like, however, users of the same genre exchange products from different publishers. At least, during the period under study, these books were sold in specialized stores for around $25 (core rules) and for $10 in case of complementary product.

The arguments for increasing returns in the RPG sector As suggested by Schwartz (2003) and Pierson (2004), any path dependence explanation relies on the existence of small, contingent events creating a path (see above) and on the identification of the mechanisms sustaining its reproduction. Several positive feedbacks in the RPG sector rendered the fantasy segment more attractive than other ones for consumers and for publishers. RPG products are informational goods, a typology of products often associated

19

with increasing returns. Indeed, consumers buy rules of the game helping a game master to narrate a story and stimulating interaction with other players.

Arthur (1994: 112) has proposed that four features trigger increasing returns leading to a wider number of adopters – which in this industry could be publishers and consumers. Indeed, increasing returns and convergence towards fantasy concerned these two types of agents — although we will be able to test our hypotheses only in the case of publishers due to the lack of data concerning consumers. The first feature – large set-up or fixed costs - implies that an option will gain momentum by lowering unit costs. In this case, producers gain an initial advantage increasingly difficult to be filled by followers. However, RPG can hardly be considered a sector where economies of scale prevail but rather as a sector where differentiation is more relevant than cost domination. Therefore, consumer behaviour is less influenced by prices than by features such as, for instance, the number of alternative products available. The high number of entries during history of the sector supports a narrative of low barriers to entry and no cost advantage for incumbents.

However three other features of Arthur’s (1994) discussion appear visible in our setting. The learning argument refers to a process where, as adoptions cumulate, the feedback obtained is incorporated into new variants3. Such learning effects are at work in the RPG sector . The proliferation of products in a specific genre spurs further innovation, attracts free lancers to write scenarios, to draw illustrations and to develop new settings. Each new product released in a genre potentially brings other complementary products to be used in combination with others, ultimately increasing the choice available to consumers. This learning accumulation occurs at the genre level and reinforces agglomeration effects. Indeed, learning at the genre 3

Theoretically, this mechanism may lead to cost advantages but such benefits can hardly be considered determinant for the survival of a RPG publisher.

20

level can hardly be transferred in other genres. For instance, Schick (1991) suggests that during the Seventies the initial failure of several products launched in genres diverse from fantasy may be attributable to the lack of comparable conditions. While a fantasy product may build on previously developed knowledge of that genre, the exploration of other segments exposes firms and consumers to the need to accumulate genre-specific knowledge.

The third feature, coordination effects, consists in increasing the utility of a solution as increasingly others adopt it. Indeed, RPG is a collective activity requiring physical proximity between players4. As fantasy games soon benefited from a critical mass of consumers (especially due to the success of D&D, but also thanks to the numerous me-too products), a player had a higher probability of finding other players in that segment and to be subject to information contagion (Arthur and Lane, 1993). Needless to say, the more a technology or a product is diffused, the more it becomes easy to learn and to gain access to information about it. Moreover, RPG could be conceived as experiential products: their qualities are ex ante difficult to assess and players rely on the experience their peers to choose among competing products. Under the minor assumption that consumers are risk averse, they will tend to favour products and segments for which they possess the relevant information.

The last, but related feature favoring increasing returns refers to self-reinforcing expectations, meaning that adopters are incited to “pick the right horse” (Pierson, 2004: 24). In the case of RPG, two elements influence this anticipative behavior made by players. First, mastering a rule of the game requires a huge time investment. A market survey in 1999 led by Wizards of the Coast –the leader after the decline of TSR- and involving 65,000 respondents found that several years are needed to learn and master the core rules of a specific RPG (increasing 4

One of the main reason to abandon the RPG hobby is that players do not find enough peers to play with; another one is to become too much attached to the game – i.e., a time-consuming hobby (Dancey, 2000).

21

switching costs). For this reason Dancey (2000), for instance argues that high concentration of players on few products should unfold5. Secondly, the business model of RPG relies on buying a core rule and several complementary products (scenarios, campaigns, new characters …). In the long run, the financial investment of a player may be non trivial. The Wizards of the Coast’s market study, for instance, shows that the total spending in RPG by age cohorts was $297 for players ranging between 12 and 17, $850 between 18-24 and $2,213 for those above 25 years (Dancey, 2000). These investments (time and money) could be lost if an adopter does not find other players accepting to play the same game or at least the same genre, ultimately leading to a natural reinforcement of the prevalent solutions.

Other characteristics played an important role in the emergence and domination of fantasy too. First, the founders of several editing company were previous players (and they generally don’t stop to play after creating their firms). So, if players were subject to some information contagion, new companies too. Secondly, during the early stage, information on games as well as on the industry was very scant, except for D&D (in 1980 reviewed even in the New York Times). Before the advent of Internet there were few occasions to gain such information. Even game conventions drew few hundreds of attendees in the early years of the industry. Regarding specialized publications, Dragon Magazine (released in 1976 by TSR) essentially covered only TSR’s products (and above all D&D) and focused primarily on the fantasy genre. We advocate that these effects reinforced the path dependency trajectory of development of the industry and, in particular, the reiterated dominance of the fantasy genre.

5

When asked what games players play in 1999, the answers were (multiple choices allowed): D&D: 66%; Vampire: The Masquerade: 25%; Star Wars: 21%; Palladium: 16%; Werewolf: The Apocalypse: 15%; Shadowrun : 15%; Star Trek: 12%; Call of Cthulu: 8%; Legend of the five rings : 8%; Deadlands: 5%; Alternity: 4%; Gurps: 3% (Dancey, 2000). This distribution appears rather skewed especially when considering that several hundred RPGs have been released since the inception of the sector. We interpret this evidence as providing support to our idea of increasing returns taking place at the product/genre level.

22

DATA ANALYSIS

Variables Besides the qualitative discussion concerning the evolution of the RPG sector, the secondary data collected allow us to proceed to an empirical test of the hypotheses advanced. The models estimated in the empirical part of the paper include as independent variables, (i) segment-level density, (ii) environmental variables measuring institutional and competitive effects (e.g., periods, lagged entries), and (iii) a set of controls accounting for systematic differences across firms and for significant changes in the institutional setting. The dependent variables in our three sets of models are the number of first time entries into the segment, the likelihood that the firm will offer a further core product in the same segment (i.e., re-entry) and the length of time (i.e., number of years) that the firm survived in the segment. Given the limited information on the exact date of the exit from the segment (in most of the cases correlated with the exit from the industry as a whole), in a limited set of cases we coded this event as happening two years after the launch of the last product ever offered – and as censored those offering a product from 1998 onwards.

The independent variables used to test the theoretical hypotheses proposed are three. The density of games offered in the segment – Segment Density - is introduced to estimate the level of legitimation and of unobserved spillovers taking place in the focal genre. This variable is included in the model to test Hypotheses 1 and 2. The squared term of segment density – Segment Density2 – controls for the level of competition in the segment and is added in the analysis of mortality rates. Hypothesis 2a argues that firms exploit the existence of spillovers on the demand side by re-entering a segment where they are already located (Hypothesis 2) and will benefit from doing so. To test such hypothesis we coded a variable –

23

Re-entry - as 1 each time the focal firm launches a new core product in the focal segment, 0 otherwise. This is the same variable used as dependent in the analyses testing Hypothesis 2. Last, a variable called Number of exits from the Fantasy segment, counting the yearly number of firms exiting from the most popular segment in the industry, is introduced to test Hypotheses 3b & 3c.

Five types of controls were included in the analyses to account for organization-specific and environmental effects. A set of dummies controlling for the most prolific firms in the industry were added into (i) the re-entry analyses (i.e., Hypothesis 2 & 3b) and (ii) when estimating the effect of re-entry on the probability of exit from the segment of the focal firm – and the existence of overcrowding (i.e., Hypotheses 2a and 3). We identified these firms (i.e., TSR, Fasa, Game Designer’s Workshop, Iron Crown and Chaosium) as those offering more than 10 core products or re-editions in the same niche. The age of the focal firm was also added as a control – Age. The variable labeled as lagged number of entries in the segment controls for the degree of autocorrelation observed when carrying out the entry and in the re-entry analyses. Periods effects were also added to include variations in the evolution of the segments analyzed. We sliced the history of the industry into three periods: the early years from 1975 until 1979, marked by the unchallenged success of TSR; a mid-period marked by the growth of the opening up of diverse new segments (1980-1983) – in the empirical analyses used as a reference group; and a last period including the years following 1983 until 1999 that controls for the increasingly high rates of exits and stronger competition. Last, a set of dummies at the genre level rules out any time-invariant unobserved segment- level effects. The covariates were lagged one year to avoid problems of simultaneity.

24

Models and methods of analysis We design the validation of our theoretical model along three sets of analyses which differ from the traditional methodologies used to study the emergence and development of path dependence: case studies and computer simulations (for a notable exception see, e.g., Karim and Mitchell, 2000). We carried out the analyses at two different levels of analysis: the segment and the firm. As for the segment level we first aim at measuring the impact of density and lagged foundings on number of entries in the segment for the first time – see Hypotheses 1 & 3c. Second, the logic we advanced imposes a more refined analysis at the firm-segment level concerning the forces stimulating re-entries in the same segment (Hypothesis 2 – but see also Hypotheses 3b&c) – i.e., conditional upon being positioned in that specific segment. Third, and last, as we argued that the firms included in our sample are forward looking and re-enter a focal segment to take advantage of the potential market spillovers, we should observe a reduction of the exit rates upon re-entering the fantasy segment – i.e., see Hypothesis 2a-, ultimately triggering overcrowding (Hypothesis 3a).

As far as the first entry into the segment is concerned, we defined it as a realization of an arrival process (Barron and Hannan, 1991). The Poisson regression model represents the most appropriate solution for studying dependent variables that take similar integer values. Nevertheless, under the assumption that the process of segment entry follows a Poisson distribution, overdispersion may be a problem, i.e., the tendency of the variance of the entry rate to increase faster than its mean. Although this problem would not affect the coefficient estimates, standard errors might be underestimated, and chi-square values overestimated. To avoid this, a negative binomial model specification is thus preferred. By using a log-linear function to link the covariates to the rate, the formulation of the model becomes:

λi (t) = exp(β ′N it + ϕ ' zit + ε it ],

25

[1] where Nt is the measure of density for the segment i, zit is a vector of control variables, and exp(εit) ~ Γ [1, α]. In this formulation of the negative binomial model, the parameter alpha, estimated directly from the data, captures the overdispersion. The final data set (with all the genres pooled together) includes 300 observations (25 years*12 segments). To deal with autocorrelation, we used the Generalized Estimating Equations (GEE) method. GEE requires a specification of a working correlation matrix. After a preliminary analysis of correlograms, an AR(3) correlation among the observations was chosen, providing the best fit to our data.

Second, re-entries could be multiple because firms may decide to offer an increasing number of products into the niche – a choice triggered – we argue -- by spillovers effects both on the side of producers and on that of consumers. We coded the number of the firm-segment reentries according to an ordinal response and carried out the re-entry analysis via an ordered logit model specification. The ordered logit (or cumulative logit) model builds on the idea of cumulative probability. The cumulative probability Cijk is the probability that the kth firm located in the ith market segment offers the jth or higher number of core games. It is possible to turn this cumulative probability into the cumulative logit:

 C ijk C ijk = log  1− C ijk 

   

[2] The models presented here estimate the cumulative logit as a linear function of several independent variables – with particular attention to segment level density: Cijk = α j − β ik

[3]

26

Note that each α jk indicates the logit of the odds of being equal to or less than category j for the baseline group (when all independent variables are zero). The β jk tells us how a one-unit increase in the independent variable increases the log-odds of being higher than category j (due to the negative sign). Because, this β jk is not indexed by j we are assuming that the one unit increase affects the log-odds the same regardless of which cut-point we are considering – an assumption of as parallel regression that seem to hold in our dataset. The final data set (with all the genres pooled together) includes 1005 firm-segment observations.

Last, regarding the analyses of mortality rates, the final data set (with all the genres pooled together) includes 429 firms and 215 exits, implying 1882 firm-year observations. We obtained such a dataset after divided the life of each producer in organization-years through the spell-splitting technique (Allison, 1984; Tuma and Hannan, 1984). Different time functions and different covariates can be used to model the hazard rate of segment ith exit for the kth organization. We model the rate at which failure events occur at a particular time, t, conditional upon the values of the observed covariates and upon the event not having occurred prior to time t. This rate, ri[t|X(t)] , is known as the hazard rate, defined as:

ri (t | X[t])=

Pr (t + ∆t > T ≥ t | T ≥ t ) . ∆t → 0 ∆t lim

[4] A log-linear function is used to relate observed covariates, X(t), to the hazard rate. To allow the baseline hazard rates of the different producers to vary in an unrestricted fashion, we modeled this rate according to a piecewise exponential formulation. More precisely, the tenure of an organization in the segment is divided into intervals, and the hazard is constant within each

27

interval but can vary across intervals. We define a set of J intervals, dividing the tenure in the segment at precise points (a1, a2, a3, a4,…, aj), where a0 = 0 and aj = ∞. The interval J is given by [aj-1, aj), and the hazard of the firm i is defined by: r(t) = µ j exp [ β ′x], per a j-1 ≤ a < a j , [5] where αj = log µj. This formulation allows the intercept of the log-hazard function to vary at logr1 (a)=αj + β' xi

different cut-off points. After examining life tables, we divided this variable in the following three segments: Age 0-2 (0.5-2 years), Age 3-10 (3-10 years), and Age 10-∞ (10-onwards).

To estimate the hazard rate (r) for the organization k, in the segment i, we modeled it as a function of a set of tenure-specific effects (a), firm’s characteristics (X) and a vector of controls (V):

rik( u) = µj(a) exp(Xika'α+Vit'β). [6] Because the observations cannot be assumed to be independent, we report robust standard errors, using the sandwich estimators developed by Huber (1967) and White (1980). All the estimates presented below were obtained using the software package STATA 9.

RESULTS Table 1 reports the GEE estimates of the negative binomial models concerning the number of segment level first time entries in our sample during the period 1975-1999. The first column presents the estimates of the model including all the segments; the second column reports those related to the fantasy genre; the third column shows the estimates concerning the other genres excluding fantasy – i.e., Hypothesis 1. Table 2 follows the same logic in reporting the estimates of the cumulative model concerning re-entries into segments – i.e., Hypotheses 2,

28

3b & 3c. Last, Table 3 reports the estimates concerning the piecewise exponential models of

organizational mortality rates in the focal segment – once again by pointing out the differences across genres – i.e., Hypotheses 2a and 3a.

---Insert Table 1 about here--Table 1 shows the estimates regarding the segment level drivers of first time entrants. Column 1 pools together the different genres and suggests that entries were low in the early phase (until 1979), but high during the years following 1983. Density and lagged entries did not seem to have stimulated entries into the focal segment. The estimates reported however appear to be heterogeneous across niches. In particular, while the above commentary agrees with those obtained for all the other genres but Fantasy (i.e., third Column of Table 1), it does not describe properly the dynamics of first time entry into the fantasy segment (i.e., second Column of Table 1). Within this niche some indications concerning the existence of a positive feedback between lagged entries, density and contemporaneous entries becomes manifest. In particular, the autocorrelation between lagged entries and current ones is such that one further entry the year before increased by 8% the chances of new entries. In a similar vein, when the density of incumbent products is for instance set at 20, the multiplier of the entry rate increases by about 10% and when it is set at 100 it improves by 50% compared to the zerodensity scenario. Interestingly, and in agreement with a learning perspective, newly entering firms into this industry progressively interpreted the competition in the fantasy segment as a deterrent to enter that portion of the market and as an incentive to explore alternative options.

---Insert Table 2 about here--Table 2 reports the estimates regarding the drivers of re-entries into the same segment. As before, Column 1 pools together the different genres, Column 2 shows the estimates concerning the fantasy genre and Column 3 those related to the other genres. Again, the heterogeneity of the findings across niches is remarkable. This time, the variety of results

29

does not concern the estimates of previous entries – which on average do exhibit a positive impact on the re-entry in all the niches – but the impact of density. In particular, the number of products existing in the niche seems to negatively (but not significantly) influence reentries into the segments, but the opposite result holds true in the fantasy niche. This set of results seems aligned with the hypothesis that suggests the existence of a positive relation between density and re-entries in the fantasy segment via positive spillovers – which, however, remain unobservable here. This set of findings is robust to the addition of the second order effect of density (i.e., statistically not significant). The imprinting of previous choices becomes clear when exploring the effect of competition in the fantasy segment on the incumbents’ conduct: contrary to Hypothesis 3b, incumbents reduce the intensity of their innovative efforts but did not use engage in the exploration of alternative options – here notice the sharp contrast with the same findings on first time entrants. We interpret this result as aligned with an idea of cognitive lock in – a classical element of path dependence.

---Insert Table 3 about here--Table 3 reports the estimates regarding the firm-level hazard rate of exit from the focal segment. Also here, the results reported in Column 1 appear to be misleading and the heterogeneity of the findings obtained across the niches is worth consideration. The fantasy niche is clearly exposed to the negative consequences of overcrowding (i.e., density dependent) on mortality rates: the hazard rate of exit initially declines with density, but then augments with further increases of this variable. Almost the opposite result is obtained when fitting this model to the data that exclude the above segment. As Column 3 shows, the positive estimate of the first order effect of density signals the difficulty of building up the market in absence of sufficient spillovers (see also Pólos et al., 2002). Only at relatively high values of density this competitive effect disappears and the segment gains that legitimacy that support mutualism among producers. The effect of re-entries on survival is equally intriguing:

30

re-entries do improve survival rates only in the Fantasy segment. This finding seems to point to the existence of a direct and indirect effect of re-entries in this segment: while re-entering it improves survival, re-entries contribute to increase density which ultimately induces crowding. The aggregate trend of results provides wide support to our reasoning – with the sole exception of Hypothesis 3b.

DISCUSSION Path-dependency in the market structure of a new sector In the present paper we have been concerned with a series of questions concerning the impact of new entries (but also re-entries) on the evolution of market segments and the associated organizational-level consequences in terms of behaviour (lock-ins) and performance (mortality). Arguing that significant increasing returns occur in the RPG field, we have tested several hypotheses leading to reconstruct the sequence of events that led to what we have labelled “the curse of the follower”. In RPG industry, the launching of a new game in the fantasy setting has attracted an increasing number of firms sustaining its dominance. This segment grew thanks to the signals regarding profitability and the beneficial spillovers effects. On the consumer side, the enlargement of the community of players contributed to enrich the supply of products. On the producer side, the dominance of fantasy favoured the constitution of an increasing pool of writers, illustrators or project managers able to create new products. The positive feedback above-mentioned helps interpreting the high rate of entries in the fantasy segment and the choice made by incumbents to re-enter such segment. Over time however, incumbent firms in the fantasy segment became confronted with increasing competitive pressures due to overcrowding. Interestingly enough, this pressure did not entail a massive diversification into other segments by incumbent firms but rather by new entrants.

31

Contributions to the path-dependency literature We advocate that our study makes several contributions. At a methodological level, we have not only provided evidence on the emergence of a dominant segment during the early years of a sector, but also shed light on the mechanisms behind its reiterated dominance. In so doing we have stressed the similarities between a density dependent approach and path dependence research and recurred to the concept of density which provides a parsimonious (and rather unused) measure of path dependency. So far path dependence theory has been essentially analyzed by means of qualitative case studies (e.g. David, 1985; Cusumano et al., 1992; Cowan and Gunby, 1996; Puffert, 2000) or through simulations (e.g. Arthur, 1994; Sterman and Wittenberg, 1999). In this paper, we made a substantive effort to move beyond qualitative analysis and to provide a quantitative test of the hypotheses proposed. In particular, we tracked the unfolding of path dependency by means of longitudinal data shedding light on its emergence and pointing also to the forces behind its potential decline.

On a theoretical standpoint, the evolutionary sequence identified provides also a partial reconciliation of the lock-in with the lock-out mechanisms. Density generates positive spillovers in terms of installed base of consumers but contributes also to increase the competitive pressure faced by the firms involved in that segment. Interestingly, such a process is not much different from that advanced by a series of studies regarding the emergence of technological standardization (e.g., Katz and Shapiro, 1985; Besen and Farrell, 1994): a convergent solution (the adoption of a shared standard) produces positive externalities but promotes intra-standard competition. This balancing effect stabilized the fantasy segment providing a ceiling to its evolution. It has also encouraged more and more firms to explore other market segments eventually leading to the flourishing of twelve different niches. Our results differ from those of Puffert (2001) because we are able to show that some

32

organizations learned to differentiate their products and to explore new market segments. So, where Puffert identifies a sequence leading to a convergent choice of railway gauges, we observe an increasing diversity in the RPG sector to escape the ‘curse of the follower’. Interestingly, differentiation appeared to be easier for new entrants than for incumbents. While the latter are stuck in their initial choices, the former have more leeway to take advantage of differentiation. Two explanations can be advanced to make sense of this finding. Firstly, incumbent firms may be subject to organisational inertia (Hannan and Freeman, 1984) – potentially due to their installed base of consumers. Secondly, the choices of incumbents may also be bind by resources constraints. Indeed, the organizations included in our sample are very small in size and any initial mistake may be difficult to overcome.

Beyond these direct (increased legitimacy and increased carrying capacity) and indirect (increased competitive pressure) effects of density identified here, several cognitive mechanisms may also be evoked as antecedents of path dependence, – see also the argument of Arthur and Lane (1993) on information contagion. Indeed, in the RPG industry, the release of D&D me-too products (such as “Chivalry and Sorcery” or “Tunnels and Trolls” evoking the same denomination of “Dungeon and Dragon”) a few months after the entry of D&D have contributed to legitimate this segment in the eyes of current players and newcomers. The growth of the fantasy segment may have lead actors to consider RPG as associated to fantasy settings. This phenomenon echoes the literature on the first-mover advantages, particularly when a pioneer defines the categorization of products (e.g., Lieberman and Montgomery, 1988) and receives disproportionate attention from consumers (e.g., Carpenter and Nakamoto, 1988, 1990). During the Seventies and Eighties the information on the industry was limited. The most important game conventions (the Gencon at Lake Geneva, Wisconsin) were organized by TSR (the company publishing D&D). Moreover, the first published magazine

33

dedicated to RPG, Dragon Magazine is released at the end of the 1970’s. It is sponsored by TSR and was described as “the official source of information regarding all aspects of the

Dungeons & Dragons hobby”. Observers describe this magazine as “one of the defining icons of gaming culture.” Sponsored magazine and conventions contributed to reinforce the cognitive lock-in of players and publishers in the fantasy segment. Thus, while Arthur (1994) insists on increasing returns to explain the path dependence evolution of a technology, we advance that a specific solution (e.g., technology, standard, market segment…) may also diffuse and reach taken-for-grantedness via a series of cognitive and sociological mechanisms. Such arguments favouring the assimilation of RPG with fantasy RPG seems to be corroborated by the development of two other geographical markets. Albeit we have not displayed the data, similar phenomena can be evoked for the markets of German and French language, the most important in Europe. In Germany, the market emerged in 1981 and started with the translation of D&D in the local language. This event oriented the market towards fantasy in a comparable way to the American market discussed here. Conversely, the French market was established in 1982 with the release of a game situated in a modern setting, albeit in the same year an unauthorized French translation of D&D appeared too (but rapidly removed from the market). Interestingly, a far greater diversity in genres emerged and the fantasy segment became soon one of the many.

The strategies to enter a new sector Our results shed also light on the strategies of new entrants in an emerging sector. A vast literature has debated on the most effective strategy to penetrate a new market. Some authors have argued that pioneers gain a decisive advantage over followers while others have contended that followers display a better performance as they do not bear the costs of discovering consumer preferences (Lieberman and Montgomery, 1988). Our results advance a

34

more fine-grained hypothesis. The key question to us concerns when the benefits of joining the bandwagon are outweighed by those of differentiation. Mimicking TSR early on in the industry provided legitimacy to new entrants and access to a large installed base of players. Compared to fantasy, other segments had difficulties to take off and to acquire an attractive position. Over time, however, the scenario changed. Some segments acquired sufficient visibility offering new opportunities to differentiate; conversely, the fantasy segment began to exhibit more and more exits due to the competitive pressure generated by overcrowding. Presuming a decent external validity of the present study, we may conclude that in emerging industries, entering the dominant segment may increase mortality rates in the long run when competitive pressure will unfold. ‘Exploring the unknown’ and being able to differentiate conversely may increase mortality rates in the short term up until legitimacy is established – and the carrying capacity of the segment augmented.

Limitations and further research To conclude we would like to stress a few limitations and to highlight potential avenues of further research. First, we have focused here only on de novo organizations marked by limited resources. The nature of the sample may then be responsible for the difficulty encountered by incumbent firms to alter their initial choice. Different results could emerge when studying organizations diversifying from other fields or simply having access to more resources. Second, in the present paper we focused on producers as triggers of market segment emergence and not on consumers due to a lack of data. More refined data are needed to further explore this issue. Finally, our study relied on a clear segmentation of the sector relatively stable over time. This feature may be hard to find in other empirical settings.

35

Regarding further research, two avenues may be pursued. Firstly, we have limited our data to the entry choices of core rules producers to compare entry and re-entry movements across an homogeneous sample and to concentrate on ‘substantial’ firm level innovations. However, several firms choose to enter the RPG sector with complementary products only. While we believe that this exclusion does not represent a serious threat to our results and to the evolutionary dynamics illustrated here, it provides an opportunity to extend our empirical test collecting new data. Last but not least, this article focuses only on English language products. The collection of data in other empirical settings may be essential to validate our findings and provide a comparison across geographical markets.

36

REFERENCES Aldrich, Howard E. 1999 Organizations evolving, Sage Publications Ltd. Aldrich, Howard E., and Marlene C. Fiol 1994 “Fools rush in? The institutional context of industry creation”, The Academy of Management Review, 19 (4): 645-670. Allison, Paul D. 1984 Event History Analysis: Regression for Longitudinal Event Data, Beverly Hills CA: Sage. Arthur, Brian W. 1994 Increasing returns and path dependence in the economy, Ann Harbor, The University of Michigan Press. Arthur, Brian W., and David A. Lane 1993 "Information Contagion", Structural Changes and Economic Dynamics, 4: 81-104. Barreto, Ilidio and Charles Baden-Fuller 2006 “To conform or to perform? Mimetic behaviour, legitimacy-based groups and performance consequences”, Journal of Management Studies, 43 (7): 1559-1581. Barron, David N., and Michael T. Hannan 1991 “Autocorrelation and Density Dependence in Organizational Founding Rates”, Sociological Methods and Research, 20: 218–241. Baum, Joel A.C., Stan X. Li, and John Usher 2000 “Making the Next Move: How Experiential and Vicarious Learning Shape the Locations of Chains’ Acquisitions”, Administrative Science Quarterly, 45(4): 766-801. Besen Stanley M., and Joseph Farrell 1994 “Choosing how to compete: strategies and tactics in standardization”, Journal of Economics Perspectives, Vol.8 (2), p. 117-131. Booker, Courtney M. 2004 “Byte-Sized Middle Ages: Tolkien, Film, and the Digital Imagination”, Comitatus: A Journal of Medieval and Renaissance Studies, 35: 145-174. Brown, Timothy , and Tomy Lee 1998 Official price guide to Role Playing Games, The Ballantine Publishing Group, first edition. Carpenter, Gregory S., and Kent Nakamoto 1988 “Market pioneering, learning and preference”, Advances in Consumer Research, 15: 275-279. Carpenter, Gregory S. and Kent Nakamoto 1990 “Competitive strategies for late entry into a market with a dominant brand”, Management Science, 36 (10): 1268-1278. Carroll, Glenn R. and Michael T. Hannan 2000 The Demography of Corporations and Industries. Princeton University Press, Princeton, N.J. Carroll, Glenn R., and Richard J. Harrison 1994 “On the historical efficiency of competition between organizational populations”. The American Journal of Sociology, 100 (3): 720-749.

37

Cowan, Robin, and Philip Gunby 1996 “Sprayed to Death: Path Dependence, Lock-in and Pest Control Strategies”, The Economic Journal, 106 (436): 521-542. Cusumano Michael A., Yorgos Mylonadis, and Richard S. Rosenbloom 1992 “Strategic Maneuvering and Mass-Market Dynamics: The Triumph of VHS over Beta”, The Business History Review, 66 (1): 51-94. Dancey, Ryan S. 2000 “Adventure game industry market research summary”, February 7th, v1.0. David, Paul .A. 1985 “Clio and the economics of QWERTY”, American Economic Review, 75 (2): 332-337. David, Paul A. 2000. “Path dependence, its critics and the quest for ‘historical economics’”, in P. Garrouste and S. Ioannides (edited by), Evolution and Path Dependence in Economic Ideas: Past and Present, Edward. Elgar Publishing, Cheltenham, England. Delacroix, Jacques, and Hayagreeva Rao 1994 “Externalities and ecological theory: unbundling density dependence”, in J.A.C. Baum and V. Singh (eds), Evolutionary dynamic of organizations, 255-268, New York: Oxford University Press. Dunnigan, James F. 1992 The Complete Wargames Handbook, 2nd edition, online version. Farrell, Joseph, and Garth Saloner 1985 “Standardization, Compatibility and innovation”, The RAND Journal of Economics, 16 (1): 7083. Fine, Gary A. 2002 Shared Fantasy. Role-Playing Games as Social Worlds, The University of Chicago Press. Garud, Raghu and Peter Karnoe 2001 Path Dependence and Creation, Lawrence Erlbaum Associates, London. Greve, Heinrich A. 1998 “Managerial Cognition and the Mimetic Adoption of Market Positions: What You See is What You Do”, Strategic Management Journal, 19 (10): 967-988. Hannan, Michael T., and John Freeman 1977 “The Population Ecology of Organizations”. American Journal of Sociology, 83: 929-984. Hannan, Michael T., and John Freeman 1984 “Structural inertia and organizational change”, American Sociological Review, 49: 149-164. Haunschild, Pamela R., and Ann S. Miner 1997 “Modes of interorganizational imitation: the effects of outcome salience and uncertainty”, Administrative Science Quarterly, 42: 472-500. Huber, Peter J.

38

1967 The behavior of maximum likelihood estimates under nonstandard conditions. In Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability. Berkeley, CA: University of California Press, vol. 1: 221–223. Karim, Samina, and Will Mitchell 2000 « Path-dependent and path-breaking change: reconfiguring business resources following acquisitions in the U.S. medical sector, 1978-1995 », Strategic Management Journal, 21: 1061-1081. Katz, Michael and Carl Shapiro 1985 « Network Externalities, Competition and Compatibility », American Economic Review, 75(3): 424–40. Kliehm, Martin 1993 “Interview with a legend”, Salon des Jeux in Paris, April, 17th. Kociatkiewicz, Jerzy 2000 “Dreams of Time, Times of Dreams: Stories of Creation from Roleplaying Game Sessions”, Studies in Cultures, Organizations & Societies 6(1): 71-86. Lamberg, Juha-Antti, and Henrikki Tikkanen 2006 « Changing sources of competitive advantage: cognition and path dependence in the Finnish retail industry 1945-1995 », Industrial and Corporate Change, 15 (5): 811-846. Lecocq, Xavier, and Benoît Demil 2006 "Strategizing industry structure: The case of open systems in a low-tech industry", Strategic Management Journal, 27 (9): 891-899. Levitt, Barbara, and James G. March 1988 “Organizational learning”, Annual Review of Sociology, 14: 319-340. Lieberman, Marvin B., and David B. Montgomery 1988 “First Mover Advantages“, Strategic Management Journal, 9: 41-58. Liebowtiz, Stan J., and Stephen E. Margolis 1995 “Path dependence, lock-in, and history”, Journal of Law, Economics, & Organizations, 11 (1): 205-226. Lomi, Alessandro, Erik Larsen and Filippo Carlo Wezel 2007 ‘Getting there’: exploring the role of expectations and pre-production delay in processes of organizational founding. Working paper, University of Lugano. Mahonney, James 2000 “Path dependence in historical sociology”, Theory and Society, 29 (4): 507-548. Pierson, Paul 2004 Politics in time, Princeton, NJ: Princeton University Press. Pólos, Laszlo, Michael T. Hannan, and Glenn R Carroll 2002 “Foundations of a theory of social forms”, Industrial and Corporate Change, 11: 85-115. Puffert, Douglas J. 2000 “The standardization of track gauge on north American railways, 1830-1890”, The Journal of Economic History, 60 (4): 933-960.

39

Rausch, Allen 2004a “Gary Gygax Interview”, GameSpy, August 15th /articles/538/538817p1.html. Rausch, Allen 2004b “Dave Arneson Interview”, /articles/540/540395p1.html.

GameSpy,

and 16th, http://uk.pc.gamespy.com

August

19th,

http://uk.pc.gamespy.com

Schick, Lawrence 1991 Heroic worlds: A history and guide to Role Playing games, NY: Prometheus Books. Schilling, Melissa A. 1998 “Technological lockout: an integrative model of the economic and strategic factors driving technology success and failure”, Academy of Management Review, 23: 267-284. Schwartz, Herman 2003 “Down the Wrong Path: What’s Wrong with Theories of Path Dependence”, American Political Science Association annual meeting, August. Sorenson, Olav, and Pino G. Audia 2000 “The social structure of entrepreneurial activity: geographic concentration of footwear production in the U.S., 1940-1989”, American Journal of Sociology, 106 (2): 324-362. Sterman, John D., and Jason Wittenberg 1999 “Path dependence, competition, and succession in the dynamics of scientific revolution”, Organization Science, 10 (3): 322-341. Tuma, Nancy B., and Michael T. Hannan 1984 Social Dynamics: Models and Methods, New York: Academic Press. White, Halbert 1980 “A heteroskedasticity-consistent covariance matrix heteroskedasticity”, Econometrica, 48: 817–830.

40

estimator

and a

direct

test for

Figure 1. The persistence of fantasy in the RPG sector over the period 1974-1999 (number of products released and proportion of them included in the fantasy segment)

400

100,0 90,0

350

80,0 300 70,0 250

60,0

200

50,0 40,0

150

30,0 100 20,0 50

10,0

41

98

19

96

94

19

92

19

19

90

88

19

86

19

19

84

82

19

80

19

19

78

19

19

19

76

0,0

74

0

fantasy total proportion

Table 1. GEE estimates of negative binomial models for number of first time entries in the segment during the period 1975-1999.6

All genres

Fantasy

Other genres

Constant

2.02** (.51)

1.08** (.33)

-2.54** (.65)

1974-1979

-.82** (.34) .48** (.27) .006 (.047)

-.21 (.28) -.75** (.38) .12** (.045)

-.92** (.39) .51** (.30) .021 (.071)

-.005* (.003) .14** (.04)

.005** (.002) -.074 (.067)

-0.007 (.005) .16** (.047)

Number of observations

Included 300

Not included 25

Included 275

Wald Chi-squared

154.43

47.01

95.6

Variables

1983-1999 Lagged number of entries in the segment Segment density (H1) Number of exits from the Fantasy segment (H3c) Genre level dummies

6

** p