Dominance in the prototyping phase—The case of hydrogen passenger cars

Research Policy 41 (2012) 871–883 Contents lists available at SciVerse ScienceDirect Research Policy journal homepage: www.elsevier.com/locate/respol Dominance in the prototyping phase—The case of hydrogen passenger cars Sjoerd Bakker a,∗ , Harro van Lente b , Marius T.H. Meeus c a Delft University of Technology, OTB Research Institute, PO Box 5030, 2600 GA, Delft, The Netherlands Utrecht University, Department of Innovation and Environmental Sciences, PO Box 80115, 3508 TC, Utrecht, The Netherlands c Tilburg University, Department of Organization Studies, Center for Innovation Research, PO Box 90153, 5000 LE, Tilburg, The Netherlands b a r t i c l e i n f o Article history: Received 12 July 2010 Received in revised form 19 September 2011 Accepted 18 January 2012 Available online 11 February 2012 Key words: Dominant design Innovation Expectations Prototyping Hydrogen Automotive industry a b s t r a c t The notion of dominant designs refers to dominance in the market, hence the literature on dominant designs ignores the selection process that already takes place in pre-market R&D stages of technological innovation. In this paper we address the question to what extent pre-market selection takes place within an industry and how this may lead to dominance of one design over others before the market comes into play. Furthermore we study what selection criteria apply in the absence of actual market criteria. We do so through a historical analysis of design paths for hydrogen passenger cars. We argue that prototypes are used by firms in their internal search process towards new designs and at the same time as means of communicating technological expectations to competitors and outsiders. In both senses, prototypes can be taken as indicators of design paths in the ongoing search process of an industry. We analyzed the designs of prototypes of hydrogen passenger cars from the 1970s till 2008. A database is compiled of 224 prototypes of hydrogen passenger cars, listing the car’s manufacturer, year of construction, hydrogen conversion technology, fuel cell type, and capacity of its hydrogen storage system. The analysis shows to what extent one design gained dominance and which strategies were adopted by the firms in their search processes. We conclude that indeed a dominant prototyping design has emerged: the fuel cell combined with high pressure gaseous storage. Actual and expected performance acted as selection criterion, but so did regulation and strategic behaviour of the firms. Especially imitation dynamics, with industry leaders and followers, is a major explanatory factor. Our main theoretical claim is that the selection of a dominant prototyping design is based on an interaction of sets of expectations about future performance of technological components and regulatory pressure that results in herding behaviour of the firms. © 2012 Elsevier B.V. All rights reserved. 1. Introduction This paper pursues the identification of dominance in prototype designs of hydrogen vehicles and more specifically of hydrogen passenger cars (HPCs). Theoretically our analysis starts from dominant design theory. Abernathy and Utterback (1978) originally defined a dominant design as the most commonly used configuration for serving a purpose by using technology. The concept has proved to be a fruitful source of innovation research and many scholars have used the dominant design concept (Anderson and Tushman, 1990; Henderson and Clark, 1990; Suarez and Utterback, 1995; Murmann and Frenken, 2006). The most salient aspect of the distinct definitions of dominant designs is that it has a market share larger than 50%. ∗ Corresponding author. Tel.: +31 15 2784542. E-mail address: s.bakker-1@tudelft.nl (S. Bakker). 0048-7333/$ – see front matter © 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.respol.2012.01.007 Our paper pursues a further development of the implicit assumptions regarding the dominant design perspective. The general dominant design perspective assumes that selection between designs only happens after market introduction and that performance improvement is an important criterion. Our main conjectures are: (a) already in the pre-market introduction prototyping phase variation and selection processes take place, which do lead to a dominant prototype design and (b) this dominance can only very partially be explained in terms of alignment with user preferences or updates of price and performance characteristics. Innovation trajectories have histories preceding market introduction of assembled products and many designs are tried and tested in the form of prototypes. An evolution of designs can be expected as well as an evolution of the selection criteria that apply to those designs (Garud and Rappa, 1994). Arguably, selection also takes place in that phase, since not all prototyped designs make it to the market. So our general research question reads as follows: how do the dynamics of variation, selection, and retention unfold in the prototyping phase, and can we extend dominant design thinking 872 S. Bakker et al. / Research Policy 41 (2012) 871–883 into the pre-market stage? More specifically we ask three questions: (1) to what extent do HPC prototype designs evolve into dominant prototype designs already during the pre-market phase, (2) what kind of selection criteria and mechanisms apply in the absence of market forces that explain dominance of a HPC prototype, and (3) how do strategic manoeuvres of HPC prototype producers impact the dominance process? Our main theoretical claim is that the pre-selection of designs in the prototyping phase of the innovation process is based on an interaction of sets of expectations about future performance of technological components and regulatory pressure that results in herding behaviour of the firms. The emphasis on expectations and imitation amongst car manufacturers is derived from some specific features of our case: prototypes of hydrogen passenger cars (HPC). We study the pre-market selection through a historical analysis of hydrogen vehicle prototyping in the automotive industry. The analysis contains both an overview of the designs used in the industry as a whole as well as an analysis of individual firms and their prototypes. This paper is structured as follows: first we will further discuss the assumptions and main focus of dominant design theory, as well as the selection mechanisms that are deemed relevant in explaining dominance. Next, we contrast this framework with our case of hydrogen car prototypes and develop an alternative framework for understanding the emergence of a dominant design in the prototyping phase. In Section 4 we analyse forty years of hydrogen prototypes to derive the selection mechanisms that have taken place at the industry level as well as those that applied in the individual firm strategies. 2. Theoretical framework 2.1. Dominant design theory Dominant design literature has challenged and developed the variation assumptions of much innovation research by asking: why do designs become dominant? Table 1 provides an overview of selection criteria and mechanisms before and after market entry, that are presented in the dominant design literature. For selection in the market a distinction can be made between competition amongst successive dominant designs and the competition between multiple contenders in a new product class, for which no old design needs to be battled. For the pre-market phase, some selection mechanisms have been identified, but thorough insight is still lacking. The most straightforward explanation of dominance of a design is that the best design outperforms any other design in terms of performance and price (Abernathy and Utterback, 1978; Anderson and Tushman, 1990; Christensen and Rosenbloom, 1995). Anderson and Tushman (1990) have argued that the emergence of a new dominant design is preceded by a technological discontinuity that brings an order of magnitude improvement in price versus performance over the existing dominant design (Tushman and Anderson, 1986). From the moment that the new design, or regime of designs, is introduced on the market, competition takes off between the old and the new regime, but also within the new regime. This so-called era of ferment results in the emergence of the new design that will dominate until the next technological discontinuity (Suarez, 2004). The notion of technological discontinuities assumes on the one hand that performance can indeed be defined and measured for a given design and its competitors. On the other hand it assumes that performance measures are robust over time. For cases described in the dominant design literature, such as storage capacity for digital hard drives (Christensen et al., 1998), this might be correct. However, in the case of many emerging technologies, performance on a single measure cannot be the explanation for the selections made and often some sort of negotiations take place to determine the sets of relevant criteria (Garud and Rappa, 1994; Das and Van de Ven, 2000). Besides performance, other selection criteria have been proposed in the dominant design literature. These include economies of scale in niche markets in the early stages of diffusion (Frenken et al., 1999); the advantage of a design gained through network externalities in its early stage of market adoption (Hagedoorn et al., 2001; Rosenkopf and Nerkar, 1999; Hounshell, 1985; Klepper, 2002). These mechanisms can only take effect once a (niche) market has been established and occupied. The same holds in that respect for licensing of standards and strategic manoeuvring as in the case of VHS video recorders (Cusumano et al., 1992). Also, regulatory changes can create relative advantages for a specific design over others, and foster the emergence of a new dominant design (Islas, 1999; Miller et al., 1995). Islas, for instance, concludes from a case on gas turbine development that newly imposed regulation hampered the existing designs and favoured the new technology. The aforementioned studies deal mainly with designs that are lined-up for market entry and focus more on conditions for successful entry (such as standardization efforts and other forms of cooperation between firms) than on the battle for optimal designs as such (Grindley, 1995; Wade, 1995). These studies deal with the period right before and after market introduction. This is the phase that Islas,1 citing Willinger and Zuscovitc (1993) and inspired by Kuhn (1992), describes as the pre-paradigmatic phase; the one that precedes the establishment of a new paradigm in a certain market. During this phase, firms seek support for their design and seek alliances with other firms to make sure that their common design makes a strong competitor. There is only one paper, by Suarez (2004), on competition between multiple designs in the pre-market introduction phase in which the role of prototypes is taken into account. Suarez (2004, pp. 279–283) describes a five stage process of becoming dominant. In the first phase pioneering firms or research groups start doing applied R&D, pursuing the production of new commercial products. The second phase is marked by the appearance of a first working prototype of the new product. The third phase in the dominance process is the launching of the first commercial product, and finally there are phases four and five in which a clear early frontrunner appears and one of the alternative designs becomes dominant. In phase II, technological feasibility sets the stage for competing actors to show their technological superiority and introduce the best performing prototype design. As in most of the dominant design literature, Suarez explicitly claims that technical superior performance is decisive in the selection of a dominant design (Suarez, 2004, p. 282). We continue our theoretical discussion in the following section by posing the question how superior performance cannot be the only criterion for selection in the case of hydrogen prototype cars and why other criteria should be added to explain the emergence of a dominant design in the prototyping phase. 2.2. Challenging dominant design theory Our focus on selection in the pre-market stage derives from the specific nature of our case: prototypes of hydrogen passenger cars (HPCs) and the behaviour of car manufacturers in their development. The emergence of hydrogen technologies in the car 1 In the case of gas turbines, market entry is not a sharply defined moment in time, the diffusion process is slow and development and adjustment of designs continues. In such a case pre-market selection of viable designs coincides with market selection and the selection of dominant designs is difficult to ascribe to either a pre-market or a market phase. S. Bakker et al. / Research Policy 41 (2012) 871–883 873 Table 1 Selection of dominant designs in different phases. Phase Successive dominant designs in the market Dominant design in new product class in the market Pre-market emergence of dominant design Selection criteria and mechanisms - Superior performance and price (Abernathy and Utterback, 1978; Anderson and Tushman, 1990; Christensen and Rosenbloom, 1995) - Compatibility with existing DD (Hagedoorn et al., 2001) - Regulation (Islas, 1999) - Superior performance (Frenken et al., 1999; Rosenkopf and Nerkar, 1999) - First to market or initial (niche) market leader (Hounshell, 1985; Klepper, 2002) - Network externalities (Frenken et al., 1999; Rosenkopf and Nerkar, 1999) - Strategic manoeuvring (Cusumano et al., 1992; Garud and Rappa, 1994; Liebowitz and Margolis, 1995) - Standardization/regulation (Hounshell, 1985; Klepper, 2002; Suarez and Utterback, 1995) - Promising breakthroughs in performance (Grindley, 1995) - Regulation (Islas, 1999; Miller et al., 1995) - Organizational support (Rosenkopf and Nerkar, 1999; Wade, 1995) - Strategic manoeuvring (Grindley, 1995; Das and Van de Ven, 2000; Garud et al., 2002; Funk, 2003) - Demonstration of technical feasibility in a prototype (Suarez, 2004) industry reveals two issues that have not been addressed so far by explanations on the emergence of dominant designs. First, hydrogen has emerged as a potential technological discontinuity that might challenge the old dominant design in the future. However all HPCs presented at car shows and demonstration projects are underperforming by at least an order of magnitude in comparison to the current design of passenger cars in terms of price and performance (Romm, 2004). Hence, this potential discontinuity derives its attractiveness from either high expectations of technological progress or from expected changes in market selection criteria that would benefit this option to a great extent. Second, there is no consumer market for hydrogen passenger cars but nonetheless the different designs are in competition amongst each other. Although hydrogen as energy carrier is not a technological discontinuity at present, it could very well become one. First of all, the technology itself may improve to a great extent, both in terms of performance and production costs. Second, with increasing pressure on the car industry to develop more sustainable cars, the hydrogen option might be a necessity in the future. Being a radical move away from the gasoline powered internal combustion engine paradigm, the hydrogen prototypes demonstrate a set of new opportunities and heuristics that gain momentum. As a consequence HPCs have become technological platforms that define technological opportunities and barriers for further technological and market trials (Dosi, 1988). Car manufacturers have experimented with the technology and have shown their results to the wider public during the big motor shows and through press releases. Even though no commercial hydrogen passenger cars are available today and no one can tell for sure whether there will ever be one, car manufacturers do claim that they are serious about HPCs and that they consider hydrogen to be a very serious contender in the race for the fuel of the future (Van den Hoed, 2005). This is exemplified by the fact that in September 2009, Daimler, Ford, GM, Honda, Hyundai, Renault/Nissan and Toyota released a Letter of Understanding on the development and market introduction of fuel cell vehicles in 2015. Given this early stage of HPC development, hydrogen is a prospective rather than an actual technological option and includes many different configurations of the production of hydrogen (natural gas, coal, electrolysis, etc.) and use of hydrogen in cars. There are various configurations of storage methods and conversion technologies, and envisioned hydrogen energy systems differ greatly (McDowall and Eames, 2006). So, while hydrogen is promoted as the fuel of the future, the different technological configurations that can constitute the hydrogen car of the future compete amongst each other. According to Suarez’ phase II of the dominance process, prototypes serve as a test bed and demonstrator of technological progress. This phase sets the stage for competing actors to show their technological superiority, and to select and introduce the best performing prototype design (Suarez, 2004, p. 282). R&D management literature is slightly more realistic about the ‘superiority’ claims of prototypes that are so prominent in the dominant design literature. In the R&D management literature prototypes are viewed upon as steps in problem solving cycles in the research and development activities of firms (Clark et al., 1987; Thomke et al., 1998; Thomke, 1998). Prototypes are part of an iterative learning process, with trial and error. Outcomes of research or engineering activities are applied in a prototype and used to test or prove a scientific or engineering concept: does it work? Or, when a number of technological novelties are fitted together in a prototype to test or prove their compatibility: does this configuration work? Both the R&D literature and the dominant design literature concentrate on the in-firm use of the prototype. However, many prototypes leave the laboratory and industry gates and serve a purpose beyond the firm or research department (Rip and Schot, 2002). Intended or not, they give a high cost signal about the promising technologies that a firm considers valuable for its strategy and this holds especially for the hydrogen prototypes. Even when a prototype does not show superior performance with respect to the existing dominant design, it can still demonstrate the high potential of the technology as promising technologies are not necessarily superior performing technologies. A prototype can display a proof of principle of a promising technology and serve as a platform for wider expectations of the technologies that are applied. Each prototype, we argue, embodies a set of technological and functional expectations. Prototypes thus convey messages to a large external audience about the useful properties of an innovative technology or about the contours of a future market and can be used to convince other actors within or outside the firm. This typically applies in the case of hydrogen passenger cars, which are taken to large fairs and exhibited as a serious bid in the hydrogen future race. The science and technology studies (STS) literature emphasizes that prototypes configure future alignments between the materiality of the technology and the beliefs and behaviour of its users, the non-material. They are, in other words, considered to be working artefacts that help to shape future worlds (Suchman et al., 2002; Danholt, 2005; Wilkie, 2008) by engaging new technologies with its future context of use. Prototypes can even be used to influence 874 S. Bakker et al. / Research Policy 41 (2012) 871–883 policy making by, for instance, convincing regulating bodies that firms already work on zero-emission vehicles (ZEV’s) and that stringent market policies, such as the well known Californian ZEV regulation (CARB, 2008; Collantes and Sperling, 2008), are not necessary. To sum up, there is no such thing as a fully naïve, non-strategic prototype that only serves the company’s internal research trajectory. Also there will be no solely strategic prototype that has no relation with the actual research trajectories of the company. Prototypes are too expensive to function only as a marketing tool and presented with too much enthusiasm to be mere R&D tools. Therefore we have to ask: what drives selection in this pre-market, and R&D driven stage, given the speculative nature of technological opportunities? 2.3. Towards an alternative framework: technological expectations and strategic manoeuvring Given the multiple roles of prototypes in the innovation process, how can we extend dominant design theory in order to integrate prototyping in its framework? As said, the HPC prototype case cannot easily be fitted with Suarez’s dominance process. First, at present there is not a single measure that uniquely describes the technological performance of a hydrogen car. HPC prototype cars must meet several types of performance requirements such as: speed, range, various emission levels, high(-er) energy efficiency, and safety. These performance requirements constrain each other until today. And second, the measures that are available may change due to changes in consumer preferences and governmental regulations. Hence benchmarking the design options on current performance and future targets remains rather complicated and optimization of the diverse characteristics almost impossible. Instead, one can expect a co-evolution of the designs in the making and the targets that these designs should be able to meet. In the prototyping phase, increasing returns to adoption such as scale effects and network externalities will only play a marginal role and will not directly guide the industry towards a single design. Firms will however do so indirectly as they anticipate the selection that will ultimately take place in the market. Hence, in the prototyping phase they will adhere to those designs that are selected by a significant number of other firms as well in order to increase the chances of eventual market success of their hydrogen cars. In For instance, the refuelling infrastructure will be geared towards the dominant storage option and other options would then require a separate infrastructure. Therefore, firms must monitor each other’s activities and design choices and will most probably follow the leading firms with the most promising designs. For us the question is thus whether such a convergence in hydrogen car designs has occurred and, if so, how the prototyping strategies of individual car manufacturers have evolved to let one prototype design become dominant. In our framework we focus on two interrelated factors: first, the role of technological expectations, and second how these expectations are moulded, and translated in the strategic manoeuvring of the firms. Technological expectations, dealing with expected future levels of performance of different design options and components, guide firms in their selection of technologies to develop and incorporate in prototype cars. Strategic manoeuvring deals mostly with the amount of variations, or technological paths, that any firm may choose to pursue and the extent to which they either lead or follow the developments in the industry. These factors are interrelated because technological expectations gain traction and attractiveness when they are shared by a majority of significant actors (Borup et al., 2006; Konrad, 2006). Thus, if a group of firms chooses to develop a specific design, others might derive high expectations of that design and follow their lead. The other way around, a design that is endorsed by a single firm alone is less likely to generate a following and therefore also less likely to gain success after eventual market introduction. Two basic components in any hydrogen car define the design space for HPC prototypes; the conversion of hydrogen to power and the storage of hydrogen on-board the vehicle. Within the design space, eight different configurations or technological paths are possible as there are two conversion options and four storage options or each. Car manufacturers do have distinct choices in moving through the design space, and can combine distinct paths in distinct ways. We discern the following dimensions in the strategies of prototype manufacturers: 1. The number of prototype models produced: higher numbers represent a sincere investment in the HPC trajectory and are indicative of stronger expectations as to the HPC trajectory. 2. The technological paths represented by prototypes: where do car manufacturers place their bets, in which technologies do they firmly believe? – The sequence in which car manufacturers enter or exit these paths, reveal the stability of their beliefs and expectations towards a certain existing or ‘new’ pathway. – Whether they engage in multiple paths in parallel or sequentially: car manufacturers can stick to one or move over different paths and this is another dimension of the stability of technological expectations. Being in more paths simultaneously implies that technological expectations are still fermenting. – The timing of these choices in terms of being first movers, early movers or late followers. Being a first mover implies that a company has identified a meaningful new technological opportunity, which extends the set of technological expectations. Late followers are merely imitators that confirm the feasibility of a path. A large number of imitators is indicative of the transformation of a technological expectation into a confirmed feasible option. These dimensions enable us to distinguish between distinct types of strategic manoeuvring in the HPC prototype design space. We discern two main strategies that can be refined on several dimensions. A deep, and specialized strategy means that companies release prototypes that focus on one path only and that there is a large number of prototypes produced, which also means that prototypes are updated at a relative high frequency. A broad and diversified strategy means that a company has prototypes in distinct paths, and regularly updates its prototype portfolio. A broad and diversified strategy implies that car companies exhibit and release two or more prototypes in parallel. Because not all firms will develop the same designs and underlying technologies from the start, it is likely that the early years of prototyping activities show high variation. A number of first movers will open a certain technological path with their vehicles and other firms may choose to enter that path later on or open up another path altogether. For a dominant HPC prototype design to arise, the majority of car manufacturers should converge from a rather broad spectrum of paths to a single path. As a consequence, there will be a shakeout of other designs that are no longer regarded as promising options. 3. Methodology To establish our unit of analysis, we follow Murmann and Frenken (2006) and distinguish between two levels at which the competition can take place. At the system level, hydrogen challenges the gasoline powered car and all its surrounding elements S. Bakker et al. / Research Policy 41 (2012) 871–883 875 Table 2 Storage and energy conversion technologies. Liquid hydrogen (LH) Metal hydrides (MH) Internal combustion engine (ICE) Bivalent ICE (Biv ICE) such as fuel infrastructures, maintenance stations, etc. However, until now there is no full or limited competition for dominance, as defined by Anderson and Tushman (1990), at the system level. At the subsystem level, shaped by the components that make up the hydrogen car, we do see competition even while there is no product on the market. Our study focuses on this level of aggregation: the core components of the emerging technological system. First, we analyse the extent to which the population of manufacturers converges or diverges on the technologies and configurations they explore and whether or not a dominant design arises. And second, we study the selection criteria that apply by tracing technological performance, strategic manoeuvring and governmental regulations that apply, or were expected to apply, to the to the car industry. The diversity in proposed hydrogen technologies is reflected by the range of hydrogen prototypes that have appeared over the last forty years. In the prototypes, different technologies have been used in different configurations and designs. We compiled a database of 224 prototypes of hydrogen cars. These data were gathered through an online search, using mainly websites dedicated to hydrogen vehicles,2 car manufacturers websites and general car news sites. Additionally, this search was supported by already existing overviews3 of hydrogen models, with visits to car shows (Amsterdam, Geneva) and validated by industry researchers. Buses, trucks, and utility vehicles were excluded from the database. Several technological specifications were included in the database; brand, year, storage method, amount of hydrogen stored, conversion technology used, manufacturer of the fuel cell or engine, range, and maximum speed. In this paper we consider the hydrogen storage method and the energy conversion technology as variables (see Table 2). The specifications of the prototypes were then plotted on a time line, showing either convergence (towards a majority of car manufacturers enter into a design path) or divergence over time (many different designs, all with few adherents). The limitations of our data are the following. Our search method does not necessarily generate all prototypes ever built, and for some prototypes not all specifications are available. This could be the result of secretiveness on the side of manufacturers, although manufacturers use their prototypes as communication tools as well and share most of the information. Especially for the more recent models this data is freely available. For the older prototypes some of the data is missing. Still, we hold that the database is adequate for our purposes and is accurate for the last ten years, when the majority of models were produced. For the industry-wide analysis of the selection of technological components, we used all prototypes in the database. For the analysis of configurations we used only those prototypes of which both the storage and conversion technology are known and we limited our study to the prototypes from incumbent car manufacturers. The prototypes from different brands are, if applicable, taken together under the header of the mother firm instead of the separate brands (e.g. BMW/Mini, GM/Chevrolet). Some of the prototypes were produced in small series, most were produced individually. In our quantitative 2 www.netinform.net/h2/, www.hydrogencarsnow.com, consulted from May 2008–March 2009. 3 A near complete overview was made by Walter Thomas (Thomas, 2007). Methanol (Meth) 250 no. of vehicles produced, cumulative Storage method Gaseous hydrogen (GH) Energy conversion technology Fuel cell (fuel cell) 200 150 100 50 0 1967 1977 1982 1989 1994 1999 2003 2007 Fig. 1. Cumulative number of hydrogen prototypes. analysis we do not take this into account and each prototype model counts as one data point. The production of small series is taken into account in the analysis of individual firm behaviour. 4. Results: components and configurations While the first hydrogen car, consisting of a single piston combustion engine, was already developed in 1807, serious development of hydrogen vehicles started only in the mid 1970s. In those days most hydrogen vehicles were existing models, adapted or retrofitted to run on hydrogen. Only since the 1990s have manufacturers begun to develop dedicated hydrogen vehicles of which the whole design is based on its hydrogen drivetrain. The number of prototypes developed shows a steady growth up to the mid 1990s and since the end of the 1990s the number of prototypes developed each years increases sharply. All major car companies are involved from there onwards. The cumulative prototype production increased from 32 in 1997 up to 224 in 2008 (Fig. 1). The fast majority of the prototypes is developed by the incumbent firms in the automotive industry whereas only 18% of the prototypes is built by universities or as part of research networks. Apparently, car manufacturers share a sense of urgency to develop alternatives for personal mobility and hydrogen is one of their options. The bulk of the prototypes was developed after 1999. As from 1967 until 1999 48 prototypes were build, which means that there is an average production of 1.5 prototypes per annum. Between 1999 and 2008 the bulk of the prototypes has been developed; 192 out of 224. In a relatively short period of time the growth of the number of prototypes has exploded, and went from 1.5 per annum to 20.2 prototypes on average per year. The sudden surge of hydrogen prototypes is most probably caused by announced or expected governmental regulations, of which the Californian emissions directive in the 1990s4 was the most prominent. Hydrogen was seen as the most promising answer to these requirements because of its versatility in terms of both production and use. Furthermore, the development of fuel cells raised 4 Personal communication with former Daimler fuel cell engineer. 876 S. Bakker et al. / Research Policy 41 (2012) 871–883 120 160 140 Internal Combustion Engine (ICE) 120 Fuel Cell (FC) 100 Gaseous Hydrogen Liquid Hydrogen 80 100 Metal Hydride Methanol Reformer 80 60 60 40 40 20 20 0 1967 1978 1986 1993 1999 2004 Fig. 2. Cumulative numbers of different types of conversion technology used. technological expectations as to their efficiency and the possibility of lowering the cost of production. 4.1. Energy conversion technology From our database we conclude that there are two main options for the conversion of hydrogen into power. These are: (1) the hydrogen fuel cell (fuel cell) as main energy convertor and (2) the internal combustion engine (ICE). While the fuel cell is potentially very energy-efficient, it is also rather complex and costly. The internal combustion engine is inefficient, but is also less costly and better understood. Next to that, internal combustion engines can run on hydrogen and hydrocarbon fuels and therefore offer a bi-fuel of flexfuel solution that is less dependent of the build-up of a hydrogen refuelling infrastructure. In the second half of the 1990s a sharp increase in the number of fuel cell prototypes sets in, and in 1998 the cumulative number of fuel cell designs overtakes the combustion engines. From then on, the fuel cell is dominant in the prototypes (Fig. 2). While a number of fuel cell types are available, the preferred option throughout the automotive industry is the proton exchange membrane (PEM) fuel cell. The reason is its high efficiency at low temperatures of operation (<100 ◦ C), whereas other types operate only at temperatures of 400 degrees and higher. It clearly is the dominant choice and it could be interpreted as an example of a dominant design in the pre-market phase. Dedicated fuel cell producers like Ballard and UTC Power together provide 38% of the fuel cells in our database’s prototypes. Car manufacturers themselves have also developed and produced fuel cells, to be used in their own prototypes or in those of others. Most notable here is GM: 12% of all prototypes use a GM fuel cell. Also Honda and Toyota develop fuel cells themselves that are used in various other car brands’ prototypes. 4.2. On-board storage While the fuel cell is often seen as a true technological enabler (creating an opportunity) of the hydrogen vision, on-board hydrogen storage is seen as a constraint. Because of the low energy density (per volume) of hydrogen as a gas under ambient conditions, it is a challenge to take enough hydrogen on board to provide the vehicle with an acceptable drive range without refuelling. The two obvious ways of doing so are pressurising or liquefying the gas. Both require enormous amounts of energy, resulting in energy losses of up to 20% for compression and about 30% for liquefying (Department of Energy, 2002). On top of that, gaseous hydrogen under high pressure is considered as a safety hazard. Liquefied hydrogen suffers losses due to so-called boil-off: it is impossible to prevent any hydrogen to evaporate and the resulting gas 0 1967 1978 1986 1993 1999 2004 Fig. 3. Cumulative numbers of storage methods used. has to be released. A number of more innovative and complex solutions have been proposed as well. Most attention is given to storage in so-called metal hydrides. When hydrogen gas is fed to a tank containing a metal powder, the hydrogen molecules split into hydrogen atoms that are absorbed in the metal’s atomic lattice to form a metal hydride. By forming metal hydrides, hydrogen can be stored with a higher volumetric density than that of liquid hydrogen. The main backdrop, however, is the weight of the total storage system, due to the weight of the metal used. Also the rate of the ab- and desorption (increasing refuelling time), and operating temperatures are still problematic. To circumvent the storage issue, the firms have also experimented with reformers that produce hydrogen from methanol (Meth) on-board the vehicles. This method allows the vehicles to be fuelled with a liquid hydrocarbon and takes away the need for a dedicated hydrogen infrastructure. Even though long ranges could be achieved with the reforming options, all firms have dropped this option from their portfolios. Other competition for gaseous and liquid storage comes from storage in chemical hydrides (bonding the hydrogen to a liquid chemical substance such as ammonia or hydrazine), solid storage in nanomaterials or rather exotic methods such as clathrates (icelike structures capturing the hydrogen). These solutions however are far from practically usable and seldomly used in prototypes. In the meantime, while research is conducted on the alternatives, the automotive industry has predominantly used liquid and gaseous storage systems in its prototypes. Metal hydrides (MH) have been used, but it seems that the industry has discarded them. Nonetheless, as can be seen from the research activities in the US and the EU, expectations of metal hydrides are still very much alive (van Lente and Bakker, 2010). A closer look at the storage methods, as displayed in Fig. 3, reveals an initial dominance of liquid hydrogen storage (LH). Since the late 1990s this dominance was taken over by compressed gas. Between 1999 and 2008, 69% of all prototypes hold a high-pressure tank. This coincides with the increase of the use of fuel cells during that period. 4.3. Configurations The conversion and hydrogen storage configurations that are used in the prototypes vary over time. Fig. 4 shows that since the turn of the millennium one design has emerged as the dominant prototype design. This is the fuel cell vehicle with gaseous hydrogen storage. More specifically, a PEM fuel cell is used in combination with hydrogen storage at either 350 or 700 bar. In the following section we will clarify the selection criteria that were applied to hydrogen prototypes. S. Bakker et al. / Research Policy 41 (2012) 871–883 90 ICE/LH 80 70 ICE/GH ICE/MH ICE Biv 60 50 40 FC/GH FC/MH FC/Meth FC/LH 30 20 10 0 1990 1995 2000 2005 Fig. 4. Configurations used in hydrogen prototypes. 5. Selection criteria and mechanisms In this section we analyze the criteria and mechanisms of the selection of designs that have led to the pre-market emergence of a dominant design of the hydrogen passenger car. The prototypes of the 14 most active firms are taken into account in this analysis. The first mechanism that we investigate is the firms’ strategies of moving between the different design paths. Second, we investigate the role of technological performance and expectations of further progress. 5.1. Technological performance as selection criterion In the case of hydrogen vehicles one performance measure stands out as the most discussed and, next to production costs, most challenging barrier to commercialization; the maximum range of the vehicle. Automakers shared the notion in the 1990s that any car should be able to go at least 480 kilometers without refuelling (Eisler, 2009). Success on this measure may not be critical for all purposes of the prototype (a test of the prototype as ‘configuration that works’ does not necessarily rely on the range), but it is a success in terms of engineering and it is used to communicate the firms’ achievements to a wider audience. The challenge to increase the range of the vehicle is twofold. First, the car manufacturers have tried to improve the overall efficiency of the vehicles. Second, the manufacturers have explored a multitude of storage options in order to store enough hydrogen on-board the vehicle to give it as much range as possible. Another important performance criterion is set by governmental regulations. It was the Californian Air Resources Board (CARB) that initially set the stage for the car industry’s attempts to develop zero-emission vehicles. Battery-electric and hydrogen vehicles were developed in response to this regulation. As the CARB called for zero-emission vehicles, not all hydrogen vehicles were able to pass the requirements. For instance, the methanol reformer option was ruled out and also the ICE vehicles were not seen as true zero-emission as they still emit NOx. More recent are Japanese and EU regulations, and also in some states in the US that favour lowemission vehicles, which in theory widens the scope for hydrogen vehicles. Nonetheless, car manufacturers have opted for the development of zero-emission hydrogen vehicles (on a CO2 basis at least) in the last years. They have done so to increase the impact of hydrogen vehicles on the average emission levels of their full fleets of cars they sell. Table 3 summarizes the developments that have taken place at the fourteen leading manufacturers. It is clear that those firms that have entered the dominant FC/GH trajectory have made most 877 progress in terms of range and speed. This was done through increased hydrogen pressures, but also through increased efficiencies of the entire drivetrain. Even though top speeds are still rather low (compared to conventional cars and BMW’s ICE models) these firms have made progress in that respect as well. The very fact that they have invested in fuel cells development signals the positive expectations these firms have held of this initially and to some extent still underperforming option. The (potential) highefficiency of fuel cells has kept them on that trajectory, knowing that future regulation and fuel price developments would shape different selection environments than is the case for current cars. Also, the firms that have embraced the FC/GH design seem most determined to stay on their trajectory. This is signalled through public statements, manufacturer’s coalitions such as the Californian Fuel Cell Partnership, but also materialized in the form of small series of vehicles, lease-programs and most notably perhaps by Honda’s dedicated hydrogen platform. Progress in range and top speeds could have been achieved through liquid storage and onboard reforming, but these were dropped by most firms due to poor efficiency and strict emission standards. Apart from the prototyping leaders, the rest of the industry has merely followed and these firms have not produced fleets of hydrogen vehicles or developed dedicated hydrogen platforms. While the drive range of fuel cell vehicles has increased, a number of barriers to commercialization remain. First and foremost is the cost of the fuel cell and storage systems. Current prototypes cost about one million US dollars to produce.5 Lower costs can be achieved through higher production volumes and much is done to decrease the amount of platinum catalyst in the fuel cells. Other issues remain as well; these include the lifetime of fuel cells (as measured in operating hours), size and weight of the cells and their cold start capability. 5.2. Strategic behaviour of firms Different firms have taken different routes and applied different strategies. Some have been leaders, others have merely followed. In the following we combine both a historical analysis of the firms’ search routes and a qualitative analysis of their statements, to reconstruct what strategies they have applied to their development trajectories and to what extent technological performance played a role in these. The fact that a dominant prototype design has emerged cannot be explained only as a consequence of the adoption of similar strategies by all firms. The 14 car manufacturers have all chosen their own unique strategies and have unique patterns of followed paths. Fig. 5 displays the different configurations that car manufacturers applied in their HPC prototypes over a period of thirty years 1967–2007. The co-evolution of different paths of HPC trajectories can be divided in three phases. Phase I covers the period from 1967 up to and including 1993. The first phase is marked by the initiation of three paths: FC/LH, ICE/MH, and ICE/Biv. These paths are started by Daimler, GM, and BMW. In phase II, from 1994 to 2000, there is a sharp increase in the number of paths that were explored. The broadening of the number of paths is considerable and grows from three in phase I to six in phase II. Daimler exits the ICE/MH path, and acts as a first mover in three different paths: ICE/GH, FC/GH, and FC/Meth. Between 1994 and 2001, 22 prototype HPCs were developed in total. This relatively higher number in a shorter time period is indicative of quite some new entrants such as Toyota, Honda, Mitsubishi, Hyundai, VW, and Ford. These firms all enter with prototypes in the fuel cell paths. 5 Toyota targets $50,000 price for first hydrogen car (Update2) Businessweek, May 06, 2010. 878 S. Bakker et al. / Research Policy 41 (2012) 871–883 ICE/LH ICE/GH ICE/MH ICE Biv FC/GH FC/MH FC/Meth 1967 1974 FC /LH GM 1975 Daimler 1976 1977 Daimler GM 1978 1979 BMW BMW 1980 1981 1982 1983 1984 Daimler BMW 1985 1986 Daimler 1987 1988 BMW 1989 1990 1991 Daimler Mazda 1992 1993 Mazda 1994 Daimler 1995 1996 BMW Daimler 1997 1998 BMW 1999 BMW 2000 Daimler 2001 Ford 2002 Daimler 2003 Ford 2004 BMW 2005 BMW 2006 BMW 2007 BMW BMW Ford Mazda Ford GM Mazda Daimler BMW Ford Fiat Toyota Mazda Daimler Toyota Daimler Ford Honda Mazda Toyota Daimler Honda Hyundai Nissan Daimler Fiat Ford GM Honda Hyundai Peugeot Toyota Daimler Ford GM Honda Nissan Toyota VW Daimler Fiat Honda Mitsubishi Nissan Suzuki Toyota GM Honda Hyundai Audi Daimler Ford GM Honda Nissan Suzuki Toyota GM Peugeot Toyota Toyota Honda Mitsubishi Nissan Daimler Ford Mazda VW Hyundai Daimler Mitsubishi GM GM Ford Honda Hyundai Toyota VW Fig. 5. Technological trajectories of individual firms. Daimler GM VW GM S. Bakker et al. / Research Policy 41 (2012) 871–883 879 Table 3 Summary of hydrogen development and progress in 14 firms, the number of models and the progress figures refer to the preferred design of the manufacturers as listed the table. Both Ford and Fiat have no clear preferred design, but FC/GH is prominent in their portfolios. Preferred design BMW Daimler GM/Opel Mazda Toyota Ford Honda VW Audi Daihatsu Fiat Hyundai Peugeot Suzuki Biv ICE/LH FC/GH FC/GH Biv ICE/GH FC/GH FC/GH FC/GH FC/GH FC/GH FC/GH FC/GH FC/GH FC/GH FC/GH No. of models 13 8 6 2 5 5 7 3 2 3 2 4 2 2 Progress on range Progress on speed Platform Fleet Use 300 150 → 400 130 → 480 200 175 → 500 160 → 560 180 → 430 150 → 140 220 → 250 120 → 153 140 → 220 160 → 576 250 → 350 130 → 200 215 → 230 120 → 170 128 → 160 60 → n/a 100 → 150 128 → 140 140 → 160 115 → 150 175 → 160# 105-? 100 → 130 124 → 160 95 → 130 110 Topclass sedans Midsize SUV/midsize Midsize Midsize SUV Midsize, dedicated SUV Midsize SUV MPV Small (Panda) SUV Midsize Midsize 100 300 110 n/a 60 30 200 no no no no no no no Loan Lease Demo Lease Lease Demo Lease test test test test test test test Phase III, 2001–2007, is characterized by a remarkable deepening of paths in general and the addition of the ICE/LH by BMW. All in all 65 new prototypes were developed and a large majority (41) of these is found in the dominant FC/GH path. Only 16 are powered by combustion engines. The shear existence of a dominant prototype design of HPCs is a little miracle looking at the enormous diversity of release strategies of HPC prototypes over time. The most prominent car manufacturers in terms of the first mover frequency and the number of HPC prototypes are BMW and Daimler. These two companies have the most distinct strategies and, remarkably, BMW has never presented prototypes in the paths entered by Daimler. If strategies of all passenger car manufacturers would have evolved in this way, a dominant prototype design would not have emerged. In the following we elaborate on the manoeuvring of BMW and Daimler first and, second, we describe the manoeuvring of the other firms to explain why and how the dominant design did emerge as a result of their strategies, technological progress and expectations of further progress. A summary of the strategies is provided in Table 4. 5.2.1. BMW From the start BMW has opted for the use of combustion engines and it has stuck to the ICE paths. It opened the ICE/Biv path and as of phase III BMW differentiated with the ICE/LH path. BMW really behaves conform a path-dependency model, and sticks to a specific passenger car template that fulfils standards range and speed requirements of the customers. BMW’s strategy is partially parallel, but within ICE paths only. BMW kept on updating HPC prototypes in the ICE/Biv path at a high rate; 10 prototypes in 28 years, and in the ICE/LH path it presented 3 prototypes in three years. BMW’s update frequency seems to be affected little by rivalry. It had only 5 competing prototypes with the ICE/Biv design, from Ford, Mazda and Fiat. In the ICE/LH path, there were no early or late followers at all. Compared to the paths that were opened up by Daimler as a first mover the numbers of early and late followers of BMW HPC prototypes did not come close to those generated by the prototype designs of Daimler. To provide an explanation of BMW’s focus on the ICE paths, its tagline for hydrogen related matters is a good starting point: “We stop emissions. Not emotion6 ”. All of BMW’s hydrogen vehicles were based on their high-end luxury vehicles and were powered by an ICE because of its power output and because of the relative low cost of these engines. The power provided by the engine is reflected in their relatively high top speeds (200–230 km/h) as compared to fuel cell vehicles and this is the emotional aspect of cars that BMW has 6 www.bmw.com. tried to preserve. In doing so, it avoids the sobering, or more generally reshaping of customer preferences. The downside of ICE’s is their low energy efficiency, which is reflected in the short ranges of the BMW models. To overcome this barrier, the company has developed a number of bivalent vehicles, which are able to run on both hydrogen as well as gasoline. With these fuels combined, top ranges of 500 km were achieved and drivers are able to refuel in the absence of hydrogen filling stations, making them an interesting option for a transition phase in which filling stations are low in numbers. In contrast, BMW’s pure (liquid) hydrogen cars have not been able to go further than 300 km without refuelling. The inefficiency of ICE’s and the remaining NOx emissions have pushed BMW to the point that the firm is reluctant to stay on this path7 and it might switch to fuel cells like the rest of the industry. 5.2.2. Daimler Amongst the big HPC prototype producers Daimler turns out to have been first mover in four of the eight paths.8 The spreading of bets starts in 1994 when Daimler produces its first fuel cell prototype. Partly this was the result of Daimler’s acquisition of the Dornier company that was involved in fuel cell research as part of it aerospace activities (Steinemann, 1999). Eventually Daimler differentiates into three distinct paths; FC/GH, FC/Meth, and FC/LH. During phase II Daimler explored four paths simultaneously. As per 2002, in phase III, Daimler significantly narrows down its prototype production to two main paths and produced updates of FC/GH prototypes on an annual basis. In parallel, Daimler moves back to its initial ICE path but it does so in vans rather than compact cars. Compared to BMW’s imitators, Daimler first moves generated much more followers that indeed also updated their prototypes at high frequencies. To illustrate, the FC/GH path accounts for 48 out of the 105 HPC prototypes. There is a notable difference between the follower dynamics in the paths that Daimler first moved into. The most successful path, FC/GH, takes off relatively slow with only 7 prototypes in phase II (1994–2000). There were only three companies that followed suit: Ford (with its first HPC prototype ever in 1999), Honda and Hyundai. The FC/Meth path develops much faster in terms of the number of released prototypes, which amounted to 11 between 1996 and 2000. Yet, the last prototypes in this path were released in 2001. Another fuel cell path, FC/MH that was opened up by Toyota, produced 6 prototypes between 1996 and 2000, which is a rate 7 Hesitation towards the ICE designs speaks from a BMW press statement that was released through the Clean Energy Partnership on 10-12-2009: “BWM setzt weiter auf Wasserstoff”. 8 We have included here the Chrysler prototypes that were developed under the heading of Daimler Chrysler between 1998 and 2007. 880 S. Bakker et al. / Research Policy 41 (2012) 871–883 Table 4 Summary of the strategies applied by the 14 firms. BMW Daimler GM/Opel Mazda Toyota Ford Honda VW Audi Daihatsu Fiat Hyundai Peugeot Suzuki No. of models No. of paths First mover 13 20 12 7 10 11 9 5 2 4 3 5 2 2 2 5 5 4 3 4 3 4 2 2 2 2 1 1 +2 +4 +1 +1 +2 +1 Early follower +1 +3 +1 +1 +2 +1 +2 +1 +1 comparable to the FC/GH path. Yet, the higher initial production rates in both the FC/MH and FC/Meth did not continue after the swift take-off. This is quite a contrast compared to the FC/GH path. In fact, the large majority of car manufacturers that had been active in FC/MH, FC/Meth, and FC/LH moved over to the FC/GH path from 2001 onwards, with only one exception and that was Mazda. Being the main pioneers of hydrogen technology, Daimler already presented a prototype model in 1975; a small van equipped with an internal combustion engine and metal hydride hydrogen storage system. Daimler continued experimenting with vans, as they provide a spacious platform for hydrogen storage systems. But in 1984 and 1986 it also used two luxury cars as platform. The range of these vehicles was rather limited, 70 km with the 1986 model whereas and 150 km with the 1984 model. Clearly, performance was not the major issue here and these vehicles served as experimentation platforms for the use of hydrogen and the hydride storage systems. Daimler’s last experiment with hydrides was presented in 1991 in a futuristic experimental model of which no performance numbers were released. Afterwards, hydrogen combustion engines were only used in the Sprinter type van, all equipped with high pressure storage. From 1994 Daimler began to use fuel cells as well, at first in vans and from 1997 onwards in passenger cars; the A and later on the B-class models. Daimler has tried all storage options in combination with fuel cells, except for hydrides which could not provide the range that was needed. The divergence in storage methods is indicative of the pressure that the company felt from expected emissions regulations as well as of the open-minded research driven innovation strategy of the firm (van der Duin, 2006). In terms of performance, the different models have typical top speeds of 145 km/h and their driving range has increased from 150 to 480 km. The biggest range is achieved by the liquid hydrogen model. However, here it shows that speed and range performance is not the only relevant criterion as the liquid hydrogen option was used only once. The same goes in that respect for on-board reformers. Both options provided longer range than highpressure storage, but problems remained with energy-efficiency and the remaining (CO2 and NOx) emissions which would not comply with expected regulations. The use of fuel cells provided higher energy-efficiencies as compared to ICE’s, but the inefficiency of the storage systems minimized the overall efficiency gains. Gaseous hydrogen was used in vans at first and it was only in 2000 that an A-class was used as platform. Earlier gas cylinders stored hydrogen at 200 bar, but now 350 bar was used and the tanks were small enough to be fitted in a passenger car. The range was still limited at 200 km but it proved to be a start of Daimler’s main trajectory with 5 models following. Including the 2009 B-class model of which 200 will be produced for large test programs in the US and Europe. The pressure is increased to 700 bar, giving the vehicles a 400 km range and enabling a higher top speed of 170 km/h. Late follower +1 +3 +1 +2 +3 +2 +1 +1 +1 +1 Strategy Deep Broad, Deep Broad Broad Deep Broad Deep Broad Deep Deep Broad, Deep Deep Deep Deep 5.2.3. GM/Opel The prototype development strategy of GM and Opel can be characterized as overall broad, and as both sequential and parallel. GM/Opel has been a first mover once, in the FC/LH path and this is the earliest hydrogen vehicle in our data set. Three times it was an early follower and once it was a late follower. GM/Opel took part with 12 prototypes in 5 distinct paths during the whole time span from 1967 up to and including 2007. It switched paths sequentially two times in its earliest designs. In 2001 and 2004 it developed multiple designs simultaneously. In 2001 GM/Opel developed one in FC/LH and one FC/GH. And, in 2004 it released one prototype with the ICE/GH design and one with FC/GH. In most of the paths GM/Opel had one prototype (ICE/GH, ICE/MH, FC/Meth), in FC/LH it had three prototypes. GM/Opel left this path after 2001 and switched to FC/GH and produced 6 out of the 48 prototypes in that path. GM’s first hydrogen fuel cell vehicle, the Electrovan was presented in 1967 already. It could go 200 km with its supply of liquid hydrogen. For a publicity stunt, involving Jack Nicholson as driver, GM presented a Chevrolet hydrogen car in 1978 with an ICE and presumably a metal hydrides storage system. It was only in 1998 that GM took up hydrogen for real, with a modified EV1 (its notorious electric vehicle) with a methanol reformer and a fuel cell. In the same year, Opel presented the same configuration in its Sintra and Zafira models. In 2000 and 2001 two liquid hydrogen fuel cell cars were presented, both with 400 km ranges. In 2001 however, a gaseous hydrogen vehicle was also released, starting GM’s dominant trajectory from there on. This vehicle only had a range of 130 km, but in later models this gradually increased to 480 km with increased hydrogen pressures. Amongst these fuel cell vehicles was the 2005 Sequel with a dedicated ‘skateboard’ chassis in which the full drivetrain was incorporated (McConnell, 2007). The US developed prototypes are based on large SUV platforms, whereas the Opel prototypes are somewhat smaller multi-purpose vehicles. The only divergent model was a one-off hydrogen Hummer that was developed for Californian governor Schwarzenegger. The Hummer could only run 80 km on its supply of gaseous hydrogen. Currently, GM has 110 Equinox fuel cell/high pressure cars in a test and demonstration project. 5.2.4. Mazda An example of a company that has a broad, diversified, and completely sequential strategy is Mazda. Mazda produced seven prototypes and started as a late follower of Daimler in the ICE/MH path. In this path Mazda produced a prototype in 1991, and in 1993. Mazda then switched to FC/MH and also here it produced 2 prototypes in 1997, and 1998. Next it moves in 2000 to FC/Meth, but this proved to be its last fuel cell vehicle and Mazda switched back to S. Bakker et al. / Research Policy 41 (2012) 871–883 the ICE/Biv design. Mazda together with BMW is the only company that never produced a FC/GH prototype. The main reason for Mazda’s commitment to hydrogen combustion engines ever since, next to lower cost of combustion engines as compared to fuel cells, is its proprietary rotary engine. This engine design allows for more controlled combustion of hydrogen than in conventional engines. To compensate for the lower fuel efficiency of the combustion engine, Mazda has chosen bivalent engines (like BMW did as well). Nonetheless, its prototypes have never gone further than 230 km without refuelling. Mazda has started a lease program, but the number of cars was never revealed. 5.2.5. Toyota Being rather active and seemingly dedicated firms, Toyota, Ford, and Honda have all entered the prototyping race in phase II. Toyota produced 10 prototypes and has exclusively focused on fuel cells. This clear focus derives from the fact that fuel cells have been developed in-house from 1992 onwards. Toyota opened up the FC/MH path, and in parallel it produced 2 FC/Meth prototypes. Toyota first presented a hydrogen prototype in 1996 and its fuel cell was powered from a metal hydrides storage system, the first in the industry. The following year, the same platform was modified and could then run on reformed methanol, improving the vehicle’s range from 175 km to 500 km. Together with Daimler it was also the pioneer for this design. In the year 2001 it strikingly presents three different prototypes, next to the metal hydrides models, it also showed an on-board reformer (which would be its last) and its first prototype with a 250 bar gaseous system. The latter could achieve a 250 km range. While reformers fell out of favour due to emission standards, metal hydrides proved incapable of increasing range. Toyota did announce a doubling of storage capacity from 1.5 to 3 percent hydrogen-to-system weight ratio, but further increases proved impossible without breakthrough materials. From there on Toyota remained on the high-pressure path and ranges were increased to 500 km with increased hydrogen pressures. Top speeds remained roughly the same at about 150 km/h. Toyota’s platforms for its hydrogen activities have consistently been their SUV/crossover models, from the RAV4 to the Highlander. The size of the vehicles has allowed for the size of the storage systems. Recently, Toyota announced a test program with 100 vehicles in the US. 5.2.6. Ford A first glance at the trajectories followed by Ford reveals two points: first it started relatively late with HPC prototypes (only in 1999) and second, it explored a wide range of designs. Until 2007 Ford released 11 prototypes. Its strategy evolved into a broad diversified portfolio spread across four different paths including parallel development of ICE and FC prototypes. In that sense, Ford’s strategy has some resemblance to Daimler’s approach, be it at lower speed and intensity. Both combustion engines and fuel cells are exhibited in respectively 2006 and 2007. The platforms have also varied to a large extent for both conversion options, from the electric vehicle startup Th!nk, with which Ford was shortly partnering, to SUV type models such as the Explorer and Edge. Apart from two vehicles in 2000, which had an on-board reformer, Ford has shown a strong preference for high-pressure storage. Whereas in early models 250 bar was used, Ford moved to 350 and later 700 bar storage systems. The latter provided their 2006 SUV fuel cell type a 560 km range, but in 2007 Ford returned to 350 bar which set back the range of both fuel cell SUV’s of that year to 360 and 500 km. Clearly, range was not the only criterion used and like other manufacturers Ford focused on fuel cell performance and overall efficiency of drivetrain as well. In its ICE prototypes, ranges vary as well. Top speeds are only scarcely communicated and never exceed 140 km/h. Ford 881 did however set a speed record for hydrogen fuelled vehicles at 333 km/h, but this was merely a publicity item. Ford was early with hydrogen test fleets such as a 30 unit program in 2001 and another fleet of 30 HPCs in 2006. 5.2.7. Honda Honda is the third company that entered the HPC arena in phase II. Honda developed a highly focused strategy and clearly has a preference for the FC/GH path in which it produced 7 prototypes. In 1999 Honda introduced one prototype in FC/MH and FC/Meth each. This makes Honda’s strategy the deepest in the industry in the sense that all its prototypes are in FC, and the largest share (7 out of 9) of prototypes is produced within the FC/GH path. The 5.2.8. Other firms The group of companies with 5 or less HPC prototypes consists of Suzuki, Fiat, Nissan, Peugeot, Hyundai, Mitsubishi, and VW/Audi. They entered relatively late in hydrogen vehicles either in the last part of phase II, but mostly in phase III. The thickening of the FC/GH path is explained mainly by the fact that the majority of Japanese and Asian HPC producers Toyota, Honda, Mitsubishi, Nissan and Hyundai exited the FC/MH path and FC/Meth path and entered the FC/GH path. On top of that also the car companies that produced smaller numbers of HPC prototypes and were mostly not early followers, but merely late followers, like VW/Audi, Hyundai, Peugeot, Nissan, Fiat, Suzuki, always released FC/GH prototypes and some of them the larger share (VW/Audi: 3:4; Hyundai 4:5; Peugeot 2:2; Nissan 3:5; Fiat 2:3; Suzuki 2:2). 6. Conclusions We conclude that competition amongst prototype designs is not confined to the market and that competition can commence beforehand. Moreover, pre-market competition amongst different prototype designs can also result in the dominance of a single design in that phase. Eventually, this dominant prototype design is likely to be the challenger of the old dominant design during an era of ferment. In our study we traced the competition in the pre-market phase in the case of hydrogen passenger cars. One design has emerged as the dominant prototype design throughout the industry: the fuel cell vehicle with high-pressure gaseous hydrogen storage. Before this design became dominant, a wide variety of options had been developed and tested. One of the major rationales for the industry to work on hydrogen vehicles, and to explore such a wide variety of designs, has been regulatory pressure from governments such as the Californian standards for zero-emission vehicles and the announced EU regulations on CO2 for 2012. Another factor is the expected rise in fuel costs, making that fuel efficiency will be of greater concern to future customers than it is today. Cost of ownership and especially cost of use will become much more important in the future than it is today, according to manufacturers. Our findings reveals how in the absence of market forces, the selection of the emerging dominant design is guided by a combination of current and expected technological performance characteristics, anticipated regulations and strategic manoeuvring of the firms that guides. These pressures generated the challenge of developing a vehicle that performs in accordance with customer preferences and complies with the expected regulations. A complicating factor was, and still is, the fact that there is no single performance measure and that the different performance characteristics constrain each other; e.g. a greater drive range would compromise the top speed of the car for instance and those vehicles that had the greatest driving range – those with on-board methanol reformers – did not comply with CO2 emission standards. From the start of hydrogen vehicle 882 S. Bakker et al. / Research Policy 41 (2012) 871–883 development, it has been uncertain what design would fit best with the complex set of performance measures, and future regulations and customer preferences. We have identified eight main paths that were explored throughout the industry and for each of those paths the individual firms must have had positive expectations of further progress. Over time, some of the paths were abandoned because of changing regulation or because it became clear that further progress was unlikely with the given design. We have shown in our analysis that mimicking behaviour by a large number of firms must have played a role in the process of variation and selection. Only seven firms can be said to be first movers, the rest has acted as followers. From those seven firms, BMW, Daimler and GM have been the most prominent leaders. The paths that they have opened were promising to themselves and in most cases to some other firms as well and the followers have not only mimicked the design but they have adopted the positive expectations as well. BMW is an exception to this rule. Although BMW clearly had high and stable expectations of the hydrogen combustion engine, these expectations were only shared by Mazda that explored this option in the majority of its prototypes as well. It was the FC/GH path that was opened by Daimler that did attract a lot of followers. This design was able to gain dominance, we argue, because it showed most progress and therefore remained promising. That is, the FC/GH design showed progress terms of driving range and energy efficiency and the expectations were strengthened by a number of demonstration projects with tens or hundreds of vehicles. These projects have signalled the dedication of those firms with respect to that design and this has helped to transform this promising design into a truly feasible option. These patterns of reinforcement of technological expectations by R&D progression and the entry of a number of late followers is a clear herding pattern of followers mimicking first movers. These technological and strategic factors have co-evolved and reinforced each other for two reasons. First of all, there is an incentive for agreement on the best design before market entry. The (technologically) best performing design can only succeed when the majority of firms adopt it. This incentive for agreement on the design before market entry is presented by network externalities and economies of scale. For instance, there is the need to build up an infrastructure for the refuelling of the vehicles. Strong commitment on the part of BMW to a liquid hydrogen vehicle will only result in market success if other firms adopt this option as well and a liquid hydrogen infrastructure is build up as a result. Second, those firms that are most active in the development of hydrogen passenger cars give the strongest signals about their expectations of their designs and they produce the greatest numbers of prototypes and engage in demonstration projects with their fleets. For other firms, the followers, this is reason to adopt those promising designs as well. These followers do not explore the full variety of paths but in order not to miss out, they develop a small number of models in the path that is surrounded by collectively held positive expectations. Their conformity amplifies the expectations held on a certain configuration. Presumably, the dominancy of a design in the prototyping phase is less stable as compared to a dominant design in the market. It lacks the positive feedbacks from economies of scale and from network externalities such as extensive supportive infrastructures. Yet, a dominant prototype design is not completely volatile, either. Dominancy in the prototyping phase builds mostly on converging technological expectations and newly acquired knowledge and capabilities within the firms. Regulation may also be aligned with the dominant prototype design and thereby effectively rule out other designs. Moreover, the dominancy may be reinforced by the early deployment of a supportive infrastructure for the demonstration projects in which the prototypes are used. This infrastructure is specifically designed for the dominant prototype design (in the hydrogen case: to deliver gaseous hydrogen at appropriate pressures). Diverging designs with different storage technologies will not profit from this infrastructure and would therefore face an extra barrier. In this paper we focused on the car industry, which provides an excellent environment to analyse the pre-market competition as its prototypes are clearly visible and its future direction is so much part of public debate. We assume, however, that such competition is not unique. While in other industries such battles are often not as visible and prototypes are perhaps primarily built for internal use, similar dynamics are to be expected as well. Suarez (2004, p. 280), for instance, mentions the signalling function of a prototype in the case of HDTV developments. For industries that are guided heavily by standardized or regulated products, mimicking behaviour and other ‘follow the leader’ strategies are very likely since those R&D trajectories represent the most promising options for future markets. Further research is needed however to uncover these pre-market battles in other industries. Another question for further research concerns the relation between the existing capabilities of incumbent firms and the dynamics of expectations in the case of emerging radical innovations. While radical innovations in general, and by definition, build on capabilities outside the incumbent industry, some of the firms’ capabilities may be of relevance nonetheless. We have witnessed two instances in which an incumbent firm entered a technological trajectory on the basis of its existing capabilities. Daimler mobilized its expertise of fuel cells (through its acquisition of Dornier) and Mazda was able to use its proprietary rotary engine that appeared useful in combination with hydrogen. These two examples are the exception rather than the rule in our case as most capabilities needed for the prototype development are either newly acquired by the firm itself or through alliances with (new) suppliers. Yet, in other cases, in other industries, such existing capabilities may have a bigger role. To conclude, we have shown that the prototyping phase may result in a dominant prototype design before market forces are at play. Tracing the configuration of prototypes appears to be helpful to untangle the dynamics of variation and selection and of technological trajectories before they enter the market. Our main theoretical claim is that the pre-selection of designs in the prototyping phase of R&D trajectories is based on the way in which sets of expectations about future performance of technological components and regulatory pressures reinforce each other and results in herding behaviour of the firms involved. Our study of prototypes therefore provides an alternative to the ex-post conclusions that have been drawn in literature so far. Acknowledgements We would like to thank Floris Prager for his contributions to the database and Jochen Markard, Roald Suurs, and Carolina Castaldi for their comments on earlier versions of this paper. Funding for this research project is provided by the Netherlands Organisation for Scientific Research (NWO), within the framework of the Advanced Chemical Technologies for Sustainability program. References Abernathy, W.J., Utterback, J.M., 1978. Patterns of industrial innovation. Technology Review 50, 41–47. Anderson, P., Tushman, M.L., 1990. Technological discontinuities and dominant designs: a cyclical model of technological change. Administrative Science Quarterly 35 (4), 604–633. Borup, M., Brown, N., Konrad, K., Van Lente, H., 2006. The sociology of expectations in science and technology. Technology Analysis & Strategic Management 18 (3–4), 285–298. CARB, 2008. ZEV Program Fact Sheet, California Air Resources Board. S. Bakker et al. / Research Policy 41 (2012) 871–883 Christensen, C.M., Rosenbloom, R.S., 1995. Explaining the attackers advantage – technological paradigms, organizational dynamics, and the value network. Research Policy 24 (2), 233–257. Christensen, C.M., Suarez, F.F., Utterback, J.M., 1998. Strategies for survival in fastchanging industries. Management Science 44 (12), 207–220. Clark, K.B., Chew, W.B., Fujimoto, T., Meyer, J., Scherer, F.M., 1987. Product development in the world auto industry. Brookings Papers on Economic Activity 1987 (3), 729–781. Collantes, G., Sperling, D., 2008. The origin of California’s zero emission vehicle mandate. Transportation Research Part A 42 (10), 1302–1313. Cusumano, M.A., Mylonadis, Y., Rosenbloom, R.S., 1992. Strategic maneuvering and mass-market dynamics: the triumph of VHS over Beta. The Business History Review 66 (1), 51–94. Danholt, P., 2005. Prototypes as Performative. ACM, New York, NY, USA, pp. 1–8. Das, S.S., Van de Ven, A.H., 2000. Competing with new product technologies: a process model of strategy. Management Science 46 (10), 1300–1316. Department of Energy, 2002. National Hydrogen Energy Roadmap. Office of Hydrogen, Fuel Cells, and Infrastructure Technologies, Washington, DC. Dosi, G., 1988. The nature of innovative process. In: Dosi, G., Freeman, C., Nelson, R., Silverberg, G., Soete, L. (Eds.), Technical Change and Economic Theory. Frances Pinter, London, pp. 221–238. Eisler, M.N., 2009. Getting power to the people: technological dramaturgy and the quest for the electrochemical engine. History and Technology 25 (1), 49–68. Frenken, K., Saviotti, P.P., Trommetter, M., 1999. Variety and niche creation in aircraft, helicopters, motorcycles and microcomputers. Research Policy 28 (5), 469–488. Funk, J.L., 2003. Standards, dominant designs and preferential acquisition of complementary assets through slight information advantages. Research Policy 32 (8), 1325–1341. Garud, R., Jain, S., Kumaraswamy, A., 2002. Institutional entrepreneurship in the sponsorship of common technological standards: the case of Sun Microsystems and Java. The Academy of Management Journal 45 (1), 196–214. Garud, R., Rappa, M.A., 1994. A socio-cognitive model of technology evolution: the case of cochlear implants. Organization Science 5 (3), 344–362. Grindley, P., 1995. Standards Strategy and Policy: Cases and Stories. Oxford University Press, USA. Hagedoorn, J., Carayannis, E., Alexander, J., 2001. Strange bedfellows in the personal computer industry: technology alliances between IBM and Apple. Research Policy 30 (5), 837–849. Henderson, R.M., Clark, K.B., 1990. Architectural innovation – the reconfiguration of existing product technologies and the failure of established firms. Administrative Science Quarterly 35 (1), 9–30. Hounshell, D.A., 1985. From the American System to Mass Production, 1800–1932: The Development of Manufacturing Technology in the United States. Johns Hopkins University Press. Islas, J., 1999. The gas turbine: a new technological paradigm in electricity generation. Technological Forecasting and Social Change 60 (2), 129–148. Klepper, S., 2002. The capabilities of new firms and the evolution of the US automobile industry. Industrial and Corporate Change 11 (4), 645–666. Konrad, K., 2006. The social dynamics of expectations: the interaction of collective and actor-specific expectations on electronic commerce and interactive television. Technology Analysis & Strategic Management 18 (3–4), 429–444. 883 Kuhn, T.S., 1992. The Structure of Scientific Revolutions. University of Chicago Press, Chicago. Liebowitz, S.J., Margolis, S.E., 1995. Path dependence, lock-in, and history. Journal of Law, Economics, and Organization 11, 205–226. McConnell, V.P., 2007. Downsized footprint and material changes for GM’s fourthgeneration fuel cell technology. Fuel Cells Bulletin 2007 (1), 12–15. McDowall, W., Eames, M., 2006. Forecasts, scenarios, visions, backcasts and roadmaps to the hydrogen economy: a review of the hydrogen futures literature. Energy Policy 34 (11), 1236–1250. Miller, R., Hobday, M., Leroux-Demers, T., Olleros, X., 1995. Innovation in complex systems industries: the case of flight simulation. Industrial and Corporate Change 4 (2), 363–400. Murmann, J.P., Frenken, K., 2006. Toward a systematic framework for research on dominant designs, technological innovations, and industrial change. Research Policy 35 (7), 925–952. Rip, A., Schot, J.W., 2002. Identifying loci for influencing the dynamics of technological development. In: Sørensen, K.H., Williams, R. (Eds.), Shaping Technology, Guiding Policy: Concepts, Spaces and Tools. Edward Elgar, Cheltenham. Romm, J.J., 2004. The hype about hydrogen. Issues in Science and Technology 20 (3), 74–81. Rosenkopf, L., Nerkar, A., 1999. On the complexity of technological evolution: exploring coevolution within and across hierarchical levels in optical disc technology. In: Variations in Organization Science: In Honor of D. T. Campbell. Sage Publications. Steinemann, P.P., 1999. R&D strategies for new automotive technologies: insights from fuel cells. In: International Motor Vehicle Program (IMVP), Massachusetts Institute of Technology (MIT). Suarez, F.F., 2004. Battles for technological dominance: an integrative framework. Research Policy 33 (2), 271–286. Suarez, F.F., Utterback, J.M., 1995. Dominant designs and the survival of firms. Strategic Management Journal 16, 415–430. Suchman, L., Trigg, R., Blomberg, J., 2002. Working artefacts: ethnomethods of the prototype. The British Journal of Sociology 53 (2), 163–179. Thomke, S., von Hippel, E., Franke, R., 1998. Modes of experimentation: an innovation process – and competitive – variable. Research Policy 27 (3), 315–332. Thomas, W., 2007. Fuel Cell and Hydrogen ICE Vehicles, Data Base 04-20-2007. Thomke, S.H., 1998. Simulation, learning and R&D performance: evidence from automotive development. Research Policy 27 (1), 55–74. Tushman, M.L., Anderson, P., 1986. Technological discontinuities and organizational environments. Administrative Science Quarterly, 439–465. Van den Hoed, R., 2005. Commitment to fuel cell technology? How to interpret carmakers’ efforts in this radical technology. Journal of Power Sources 141 (2), 265–271. van der Duin, P., 2006. Qualitative Futures Research for Innovation. Eburon Uitgeverij BV. van Lente, H., Bakker, S., 2010. Competing expectations: the case of hydrogen storage technologies. Technology Analysis & Strategic Management 22 (6), 693–709. Wade, J., 1995. Dynamics of organizational communities and technological bandwagons – an empirical investigation of community evolution in the microprocessor market. Strategic Management Journal 16, 111–133. Wilkie, A., 2008. Prototyping health care users. In: 4S/EASST 2008 Joint Annual Meeting Rotterdam. Willinger, M., Zuscovitc, E., 1993. Efficience, irréversibilités et constitution des technologies. Revue d’économie industrielle 65 (1), 7–22.