LAST GASP OR CROSSING THE CHASM? THE CASE OF THE CARBURETOR TECHNOLOGICAL DISCONTINUITY

Nathan R. Furr*Marriott School of Management

Brigham Young UniversityProvo, UT

Tel: (801) 422-1814e-mail: nfurr@byu.edu

Daniel C. SnowMarriott School of Management

Brigham Young UniversityProvo, UT

Tel: (801) 422-2409e-mail: dsnow@byu.edu

Aug 15, 2012

Under review at Strategic Management Journal

Keywords: technology strategy; strategic change; strategic renewal; industry evolution; technology evolution

*Authors listed in alphabetical order

1

Although technological discontinuities are often described in terms of a rapid transition from one technological generation to the next, we reexamine this conclusion in light of the tendency of technology to experience a last gasp—a sudden leap in performance—when threatened. Although the last gasp pattern has been suggested in prior research, it has not been sufficiently explored empirically. In this paper, we attempt to validate the presence of a last gasp and also suggest the sources of such a last gasp beyond prior explanations that incumbents simply try harder. Specifically, we explore how incumbent technology choices influence the emergence of a last gasp as well as the impact of the last gasp on the technological discontinuity and incumbent performance. We test these assumptions among the full population of carburetor manufacturers during a technological discontinuity as the industry transitioned to electronic fuel injection. We find evidence of a last gasp but, contrary to prior assumptions that the last gasp comes from incumbents ‘trying harder,’ we find a much more nuanced story of why extant technologies demonstrate a sudden performance leap. More importantly, we observe a unique pattern of incumbents exploring hybrid technologies that both contribute to the last gasp but also form the bridge for incumbents to cross to the next technical generation. These findings contribute to the technology strategy, innovation, and organization change literatures.

2

INTRODUCTION

Firms face many strategic threats, but technological discontinuities

are some of the most profoundly challenging. In the canonical description of

technology discontinuities, existing firms respond rigidly to such technology

changes and are swept aside, sometimes even blindsided, by new

generations of technology (Christensen and Bower, 1996; Henderson and

Clark, 1990). However, firms face significant uncertainty in the technology

contests that constitute the transition from one technical generation to the

next, and their outcomes are not always certain (Adner and Kapoor, 2012).

For example, although some industry observers predicted that hybrid

electric vehicles would replace the more inefficient combustion engine,

when threatened by the new technology, automobile manufacturers

managed to squeeze so much extra fuel efficiency from the seemingly

technically exhausted combustion engine that advantage of hybrid vehicles

over ‘old’ technologies may be in question (Naughton, 2012). Similarly,

although silicon semiconductors have been threatened for decades by

gallium arsenide, silicon semiconductor firms have somehow improved

performance of the ‘old’ technology enough to defer the threat, even though

silicon seems to have exhausted the very limits of its capability. In each of

these cases, in the face of a threat, incumbent actions improved the

technical trajectory of an older technology and allowed incumbents to

collectively stop, or at least significantly delay, a threatening technology

transition.

3

The observation that incumbents often fight back raises significant

questions for the study of technology discontinuities. For instance, existing

literature highlights the tendency of firms to focus on existing technology

when threatened by a technical transition (Benner and Tushman, 2002;

Gilbert, 2005; Tripsas and Gavetti, 2000). According to this view,

incumbents focus on their existing technology because they are constrained

by capability, resource, and cognitive inertia that limits their willingness

and ability to respond to a threatening technology by leaping to the next

generation (Rosenbloom, 2000; Sull, Tedlow, and Rosenbloom, 1997;

Tripsas, 1997b). While certainly an accurate description in many respects,

such ex post observations have led many to conclude that incumbent efforts

in extant technology are the death throes of rigid incumbents. Yet there

may be rational reasons, ex ante, that incumbents focus on their existing

technology beyond the traditional capability, cognition, and market power

explanations. Namely, in the light of uncertainty before a technology

transition occurs, incumbents fight back because, as in all previous threats

they have faced and beaten back, they expect that their actions will affect

the technology trajectory and, thus, their survival chances.

Indeed, early observers of technology transitions noted not only the

tendency of incumbents to fight back, but also the surprising effects such

actions have on seemingly exhausted technologies. For example, in

observing technology transitions, such as the shift from ice harvesting to

mechanical refrigeration or from sailing ships to steam ships, researchers

4

have observed that not only do firms fight back (Harley, 1988; Utterback,

1996), but that rather than inducing irrelevant inertial efforts, such threats

‘appear to induce vigorous and imaginative responses’ (Rosenberg, 1976:

205). Researchers have suggested that incumbent actions lead to an

unexpected improvement in the rate of performance improvement—a ‘last

gasp’—that reshapes the technology trajectory (Foster, 1986; Henderson,

1995; Tripsas, 1997b). Such a last gasp can have significant implications for

firm and industry evolution and, more importantly, an incumbent’s

interpretations of their actions in the face of threat. In cases such as the

silicon semiconductor or even the incandescent lightbulb, a last gasp may

forestall competitors for decades or even indefinitely. The last gasp is not a

panacea, however, as it may provide incumbents false hope that by fighting

they can improve their technology sufficiently to defeat a threatening

technology, leading to the observations of inertia and failure previously

described in the literature.

Therefore, the last gasp has important strategic implications for

incumbents, shaping whether and when incumbents should respond to a

technology threat—whether to fight a new threat, retreat from a threat, or

leap to the next generation (Adner and Snow, 2010c). In this paper, we

attempt to answer questions related to the last gasp that affect incumbent

strategic responses to a technology threat. First, although the last gasp may

has been suggested by several technology and strategy scholars (Adner et

al., 2012; Henderson, 1995; Tripsas, 1997b), the phenomenon of a last gasp

5

has received little empirical validation. Second, we know relatively little

about the sources of the last gasp, other than suggestions that, when

threatened, incumbents simply try harder (Utterback, 1996). Finally, there

has been limited attention to the performance implications of the last gasp:

both at the level of the technology trajectory and the level of the firm. We

attempt to address these gaps by examining a technology transition at the

time of threat and, using a data rich industry setting, provide some

quantitative validation of a last gasp. Third, beyond the existing explanation

that last gasps are the result of simply trying harder, we suggest two

additional sources of a potential last gasp. Specifically, we hypothesize that

firms take many actions—efforts to innovate, reconfigure, and recombine—

that improve the existing technological trajectory and lead to a last gasp of

technology performance. Finally, we conduct an exploratory analysis to

examine the performance implications of a last gasp, both at the industry

level, by estimating the effect of a last gasp on a technology transition, and

at the firm level, by examining the effect of technology choices on

incumbent survival.

We examine these questions quantitatively in a unique data set

capturing the population of passenger automobile carburetor manufacturers

over two decades during which incumbents were threatened by a new

substitute technology—electronic fuel injection (EFI). Carburetors represent

an ideal setting in which to test the tendency and effects of incumbent

efforts to fight a threat because carburetor manufacturers had been

6

threatened by potential replacement technologies in the past, but

incumbent efforts had forestalled a technology transition (carburetors had

prevailed against wick carburetors, the rotating brush carburetors, catalytic

carburetors, vaporizers, and mechanical fuel injectors, among others). EFI

technologies posed an uncertain threat because of their high cost and

delicate nature, leading many carburetor manufacturers to fight back,

which allows us to observe the relationship between incumbent actions and

any potential last gasp. In addition, because we have detailed data on a

critical carburetor performance indicator, we can observe the effect of

incumbent actions on the technology trajectory.

In the empirical analysis, we find that incumbents were surprisingly

vigorous in their response to the threat of EFI, leading to a significant last

gasp—a sudden leap in performance that does not match the canonical S-

curve of technology evolution (Foster, 1986). While prior literature has

observed this pattern qualitatively, we provide one of the first, robust

empirical validations of this phenomenon. Furthermore, whereas prior

literature has largely ignored what incumbent actions contribute to a last

gasp (suggesting that the last gasp results from simply ‘trying harder’), we

observe that trying harder contributes to the last gasp in only a subset of

cases. Instead, the combined efforts of innovation, reconfiguration, and

recombination contribute to the last gasp, but in ways that depend on the

technology choices of individual firms. Finally, we provide some of the first

empirical validation of the performance implications of a last gasp. At the

7

industry level, we estimate the rate of substitution of EFI versus

carburetors and find results that suggest the carburetor’s last gasp deferred

the EFI’s eventual victory by approximately two years. At the firm level, we

find that the technology choices made by individual incumbents reshape the

type of ‘last gasp’ they experience and also their survival chances.

Specifically, while no incumbents made the leap directly to EFI, a subset of

incumbents engaged in a unique type of action: the creation of a hybrid

between the old and new technical generation. Although prior literature has

suggested such hybrids may be a manifestation of organizational

dysfunction (Christensen et al., 1996; Tripsas, 1997b), we find that hybrids

both make a significant contribution to the last gasp and, at least in the case

of EFI, may act as a stepping stone for surviving the technological

discontinuity. In fact, only firms that focused primarily on hybrids actually

survived to the next technical generation.

These results have significant implications for our theories of strategy

and organization in technology settings. First, we provide evidence of a last

gasp, which contributes to the enrichment of our understanding of industry

and technology evolution (Adner et al., 2012; Anderson and Tushman, 1990;

Henderson, 1995). Second, by showing how a broader range of incumbent

responses contribute to the last gasp, we enrich the growing body of

research on incumbent responses to technology threat, showing a

multiplicity of important incumbent actions in the face of threat that affect

competition and survival (Agarwal and Gort, 2002; Hill and Rothaermel,

8

2003; Lavie, 2006; Tripsas, 2009). Third, in providing some of the first

detailed evidence of how the last gasp affects firm and industry

performance, we enrich our understanding of technology strategy. By

showing how a last gasp can defer a technology transition, we move beyond

the observation that technology transitions can take time, to highlight

specific innovation dynamics that affect the nature and timing of incumbent

responses (Adner et al., 2012; Adner et al., 2010c; Cohen and Tripsas, 2012;

Eggers, 2012; Henderson, 1995). Furthermore, by showing the crucial role

hybrids played in leaping to the next generation, we help resolve a debate

in the literature about hybrids while suggesting rich avenues for future

research. Indeed, hybrids represent an understudied but potentially

important artifact during eras of ferment, in industries that can

accommodate hybrids. Finally, this research helps recharacterize the

stereotype of inertial incumbents, demonstrating vigorous incumbent

responses that illustrate another reason that incumbents fight back in

addition to existing capability and cognitive explanations—namely, their

actions have an effect.

THEORY

Although the organization and strategy literatures address issues of

change and renewal (Agarwal and Helfat, 2009; Brown and Eisenhardt,

1997; Eisenhardt and Martin, 2000), responding to technological

discontinuities represents one of the most challenging events in the life of a

firm (Gilbert, 2006; Lavie, 2006). When such threats emerge, prior research

9

suggests that incumbents focus on existing technology because of cognitive

(Gilbert, 2005; Kaplan, Murray, and Henderson, 2003; Tripsas et al., 2000)

and capability inertia (Benner, 2009; Christensen et al., 1996; Thomke and

Kuemmerle, 2002). While such observations are often accurate, the rigidity

explanation has largely led to the dismissal of incumbent actions as doomed

and irrelevant.

However, incumbent efforts focusing on extant technology may not

always be irrelevant. Indeed, rigidity interpretations are often made ex post

a technology transition and, therefore, overlook the fact that ex ante

technology transitions are often characterized by significant uncertainty.

The emergence of a threatening technology, promising though it may be, is

not guaranteed to spell the end of the old technology for many reasons

(Adner and Kapoor, 2010a; Rosenkopf and Tushman, 1998). First, it is often

unclear whether the technology can overcome the associated technology

risks to become a viable substitute (Adner et al., 2010a). For example,

consider home-sized nuclear power plants from the 1950s or turbine-

powered cars of the 1960s, which promised to outperform existing

technologies but failed to do so. Second, even when technologies do

materialize, they often do not cross the price/performance threshold to

displace extant technology (Anderson et al., 1990). For example, consider

composite airframes pioneered (at vast expense) in the 1970s that are just

today becoming economically viable substitutes for aluminum ones. Finally,

even when technology transitions do begin to occur, there are often

10

questions of which market niches will be replaced and which will be left

untouched. For example, consider the parallel operation of hybrid and

combustion engines for well over a decade or the parallel existence of

online media portals and newspapers as customer niches retain their

preference for one technology over another (Adner and Snow, 2010d).

In light of such uncertainty before a transition occurs, when

threatened, incumbents may also choose to focus on extant technology

because they believe they can defer or even defeat the threat. Indeed, prior

observers have noted that when threatened, the trajectory of an extant

technology can exhibit a surprising and unexpected leap in performance

uncharacteristic of the traditional S-curve pattern described in prior

literature (Utterback, 1996). In many ways, it would seem surprising that

incumbents investing in extant technology late in the life of the technology

could make many improvements in the technology itself. According to the

basic principles of technology evolution, a technology’s performance

evolves according to an S-curve pattern as efforts to improve a given

technology are eventually exhausted, leading to a decline in the rate of

technical improvement (Dosi, 1982; Foster, 1986). The S-curve pattern of

evolution is driven as much by the exhaustion of the technology as by the

nature of competition itself, as firms locked in intense competition along a

technology trajectory near its zenith should have competed away any

innovation and performance advantages (Barney, 1986; Porter, 1980;

Rumelt, 1984).

11

Nonetheless, economic historians and innovation scholars have

observed such surprising leaps in sailing ships (Gilfillan, 1935; Harley,

1988), typesetters (Tripsas, 2008), semiconductor equipment (Henderson,

1995), alkali (Rothwell and Zegveld, 1985), and ice harvesting (Utterback,

1996), labeling such unexpected leaps ‘last gasps.’ For example, both

Tripsas (2008) and Henderson (1995) qualitatively observed that

incumbents were able to extend the life of extant technologies in surprising

ways before a technological discontinuity and even afterward. Modern

examples of technologies exhibiting a performance improvement that

extends their life include CISC processor architecture, incandescent

lightbulbs, steel bicycle frame materials, coronary artery bypass graft

(CABG) surgery, silicon semiconductors, and many other technologies that

appear to exhibit a sudden burst in performance when threatened. Although

last gasps have been qualitatively observed across many industries, rarely

have they received significant empirical validation. Therefore, we

hypothesize that when threatened by a new technology, the technical

trajectory of an industry may exhibit a last gasp, or sudden leap in

performance beyond the expected technology trajectory improvement rate.

Hypothesis 1 (H1): When threatened by a new technology generation,

the technology trajectory of an existing technology may exhibit a last

gasp (a sudden increase in product performance in excess of existing

technology trajectory).

12

There may be many reasons for a last gasp, although the predominant

explanation to date has been that when threatened, incumbents simply ‘try

harder,’ leading to an unexpected leap in the technology trajectory. While

such a performance leap may seem counterintuitive from the competitive

and technology evolution standpoints, there are both economic and

organizational reasons for incumbents to try harder. From an economic

perspective, incumbents locked in fierce competition at the margin may

have little incentive to invest in improving existing technologies; but when

threatened by substitution, incumbents may engage in innovation projects

that may have been economically infeasible when competing for marginal

cost advantage. However, such projects suddenly become feasible in the

face of the devaluation of existing assets (for a review of the economic logic

of investment in innovation, see Henderson, 1993, or Tripsas, 1997). In

other words, innovation efforts previously not justified by a marginal

innovation return may seem reasonable if they could preempt substitution,

leading incumbents to invest in innovations they may have once ignored

(Henderson, 1993; Martin and Mitchell, 1998; Mitchell, 1991). In more

practical terms, carburetor producers battling competitors on cost may not

be able to justify the expense of certain innovation improvements but when

threatened by total replacement, the investments become reasonable in the

face of losing the entire asset base.

From an organization perspective, incumbents locked in fierce

competition among similar competitors may have cognitive and

13

organizational impediments that lead them to overlook innovations that

could potentially increase technology performance (Gilbert, 2005; Porac et

al., 1995; Tripsas et al., 2000). However, both organization and innovation

theorists have suggested that above all else, organizations are motivated to

survive (March and Simon, 1958; Thompson, 1967). Therefore, when

threatened, organizations may suddenly recognize innovations that were

previously unseen. For example, Utterback (1996) described how when

threatened by mechanical refrigeration, ice harvesters switched from

removing ice with horses to tractors—an innovation within reach but unseen

until threatened with substitution. Similarly, when threatened,

organizations, may consider alternatives that were previously ignored. For

example, in a study of change in the medical imaging industry, Martin and

Mitchell (1998) observed that incumbents tended to introduce new product

designs only when threatened by declining market share or substitute

product design. In support of this view, Tripsas (2009) noted that the digital

photography company Linco failed to recognize the opportunities in the

parallel flash drive market until their business began to be threatened by

their competitor, Sysco, who introduced a flash drive product. Even when

thinking of organizations more broadly, threat can lead to the recognition of

innovation that may have been previously overlooked. For example, in 1870,

when the City of Paris was suddenly threatened with invasion for the first

time in centuries, despite France itself having been engaged in a war for

almost a millennium, citizens suddenly came up with so many innovations

14

that the city had to establish a scientific committee to process the burst of

new ideas, some as familiar today as tanks, germ warfare, chemical

warfare, petroleum napalm, and handheld grenades (Horne, 2002).

Although the City of Paris ultimately capitulated (as did sailing ships and ice

harvesting), there are times when the incumbents do defend against the

threat by improving technology (e.g., silicon semiconductors, WAP

protocols, many nanotechnology applications). Therefore, although the

carburetor industry appeared to have reached the apex of its innovation

capability after decades of competition, we hypothesize that incumbents

facing a threat from a new technology will expend extra efforts to innovate,

leading to a sudden leap in the technology trajectory.

Hypothesis 2 (H2): When threatened by a new technology generation,

incumbents’ efforts to extract greater performance from existing

technology will contribute to a last gasp in the technology trajectory.

Although the innovation efforts of an incumbent has been the primary

explanation for a last gasp, there may be other previously unexamined

reasons contributing to a last gasp. A second potential source of a last gasp

may simply be incumbent reconfiguration. Often over the course of firm

growth, a firm may diversify into market niches they can profitably serve

but where their product or processes are comparatively poorly suited. For

example, although carburetors tend to operate more efficiently in smaller

vehicles, manufacturers also produced carburetors for heavy machinery,

such as dump trucks, where the carburetor operates much less efficiently.

15

Prior work suggest that when facing a potential threat, such as a

technological discontinuity, incumbents may be forced out of market

segments where their products are the least competitive (Christensen,

1997; Christensen et al., 1996). In addition to forced retreat, other work

suggests that, when threatened, incumbents may choose to reconfigure

their resources to respond to the threat, often by recalibrating around their

sources of advantage (Adner and Helfat, 2003; Adner and Snow, 2010b;

Lavie, 2006). As an illustration in the carburetor market, when EFI first

appeared in the auto industry, it emerged in high-end, large, expensive car

models that could absorb the significant increase in cost and where, due to

the large vehicle weight, carburetors were particularly inefficient solutions.

Some carburetor firms actively retreated from these segments in order to

focus on small and moderate weight vehicles where they had an advantage

relative to the threatening technology. Although ex post such a strategy

may appear to result in a death spiral for the firm (Leonard-Barton, 1992),

ex ante such a strategy appears rational and may have worked in some

technological transitions where next-generation technologies coexist

alongside incumbents that have entrenched in their areas of comparative

advantage (e.g. e-books versus traditional books, laptops versus notebooks,

and so forth). More importantly, retreat and reconfiguration may be an

overlooked source of the apparent last gasp. Specifically, as incumbents pull

out of market segments where they are less competitive, the ‘performance’

of a technology trajectory may appear to make a sudden leap simply

16

because incumbents selling the technology have retreated from the poorest

performing applications to the highest performing applications.

Hypothesis 3 (H3): When threatened by a new technology generation,

incumbents reconfiguring to market segments where they have

comparative advantage relative to the threatening technology will

contribute to a last gasp in the technology trajectory.

Finally, another reason for a last gasp in the technology trajectory

may be that when faced with a potential but uncertain technological

discontinuity, incumbents can defend against the potential disruption by

recombining components from the new technology with older technology to

improve the performance of the older technology. We define such a

combination of technology from different technical areas as a hybrid and, in

this paper, focus on the specific case of hybrids that combine technology

from the old and new generation into one product (Baldwin and Clark,

2000). Such intergenerational hybrids require some level of component

modularity in each technical generation and are often based on the product

architecture of the old technology. The emergence of hybrids has been

observed in the study of several technology transitions, including the

business machines (Rosenbloom, 2000), typesetting (Tripsas, 1997b),

semiconductor equipment (Henderson et al., 1990), and newspaper

industries (Gilbert, 2005). Modern examples—such as the hybrid electric

vehicle, the hybrid SSD / hard drive storage device, the digital SLR camera,

17

or the Microsoft Surface (hybrid between tablet and notebook)—continue to

emerge.

Despite the prevalence of hybrids during and sometimes after

technology transitions, their impact remains a point of contention. One view

seems to suggest that hybrids are the physical manifestation of

organizational dysfunction: rigid incumbents unable to break the frame of

extant technology create clumsy products that are neither here nor there.

For example, Henderson and Clark (1990) observed that semiconductor

equipment producers often struggled to recognize the architectural changes

of the next generation, so they created clumsy hybrids that contained

components from the new generation that were based on the older

architecture, such as Kasper’s attempt to introduce contact aligners with

proximity aligner components. As another example, Tripsas (1997a)

observed that in the typesetting industry an incumbent attempting to

address the shift from hot metal typesetting to phototypesetting created a

clumsy hybrid based on the old hot metal architecture. Therefore, hybrids

may be technologies that are neither here nor there and are simply the

byproduct of organizational rigidity.

However, another view suggests that hybrids may play an important

role in the transition from one technical paradigm to another. Although new

technologies often threaten to displace an older technology, often such

threats do not emerge; even when a discontinuity does occur, it can take a

surprisingly long time to shift from one generation to the next. Although an

18

armchair strategist might suggest that as soon as a new generation

emerges incumbents should leap to the next generation, in practice, firms

leaping too early are often censured for moving too quickly and wasting

precious resources (Benner, 2007, 2010). For example, Microsoft was

criticized for leaping too early into interactive media control with WebTV

and Apple was critiqued for leaping too early into PDAs with the Newton

and into digital cameras with the QuickTake. While those early products

failed, those same industries later emerged as profitable technical

transformations (Kaplan and Segan, 2008). Therefore, during the period

when a transition may be uncertain or in process, incumbents have the

struggle of how to address a potential new generation while an older

generation remains profitable. Under such circumstances, a hybrid may

actually be a sophisticated hedging and learning strategy. Specifically,

borrowing from a future technical paradigm provides a learning option for

firms: if the threatening technology materializes, incumbents have

developed some knowledge about how it operates and are better positioned

to adapt. However, if the threatening technology proves less viable, by

borrowing components from the threatening technology, they can both

preserve existing resources and possibly capture any residual spillovers

from the new generation. For example, gas-electric hybrid vehicles borrow

components from electric vehicles, grafting them into traditional

combustion vehicles. If a future emerges dominated by electric vehicles, the

makers of hybrids will have decades of experience with the design,

19

sourcing, and production of components such as batteries, electric motors,

and electric drivetrains, which could provide a significant advantage.

However, if the future remains dominated by combustion vehicles, the

makers of hybrids may be able to gain advantages over those combustion

vehicles by borrowing the best components from the threatening technical

generation. In the carburetor industry, such hybrids emerged as standard

carburetors equipped with Fuel Feedback System FFS controls—a

component from the EFI technical generation grafted onto the carburetor

architecture. Therefore, we hypothesize that under the appropriate

conditions (technical discontinuity outcome uncertainty, technical

generations with modular components that allow for the development of

hybrids), incumbents are likely to borrow components from a threatening

technology as an option on the future technology, which also improves the

performance of the older technical generation.

Hypothesis 4 (H4): When threatened by a new technology generation,

incumbents recombining components from the threatening technology

with extant technology will contribute to a last gasp in the technology

trajectory.

RESEARCH SETTING AND DESIGN

The research setting is the automobile carburetor industry during a

period of transition to electronic fuel injection technologies from 1978 to

1992. Automobile carburetors served the purpose of mixing gasoline and air

in a ratio that can be burned efficiently by the car’s engine, and a car’s fuel

20

economy performance depends heavily on the carburetor. Carburetors

employed the Bernoulli effect—essentially simple suction—to draw gasoline

into the stream of air entering the engine.

Carburetors were the standard technology for preparing gasoline for

combustion from the invention of the automobile in the late 1800s through

the early 1980s. In the 1960s and 1970s, increasing oil prices and a growing

awareness of air pollution moved U.S. policymakers to regulate automobile

fuel economy and airborne pollutant emissions. In response to those

regulations and market demand for fuel-efficient cars, automakers and

suppliers worked to improve carburetor performance in order to extract

higher miles per a gallon—a critical performance criteria for both suppliers

and automakers enforced as a federal regulation on the overall miles per

gallon (MPG) for the manufacturers’ fleets. In particular, carburetor

manufacturers strove to improve emissions and fuel economy performance

by increasing the precision of control for the ratio of air and gasoline

entering the engine. The closer these proportions matched the ideal so-

called stoichiometric ratio, the better the performance of the engine. But by

the end of the 1970s, carburetor performance seemed to have reached its

limit in terms of consistent delivery of a stoichiometric mixture of air and

fuel.

By 1980, as an alternative to carburetors, EFI was offered for the first

time on mass produced automobiles.1 EFI used sensors and an onboard

1 The history of automobile fuel injection systems starts long before 1980, however. During World War II, the German firm

Bosch invented mechanical fuel injection (MFI) for use in military aircraft. Development continued after the war in high-end and

21

computer to monitor the engine's performance and change operating

parameters in real time to adjust to changing conditions. This allowed

precise control of an automobile’s air/fuel mixture. The increased precision

allowed automakers to use more advanced emissions control devices (such

as the catalytic converter), and it gave cars better fuel economy

performance. Importantly, the sensors and electronics that EFI required

had not been developed before the advent of EFI—existing microelectronics

were too delicate for the harsh environments automobiles presented.

Therefore, at the emergence of EFI, it was not clear to carburetor

manufacturers if EFI could survive in passenger cars or be produced cost

effectively. However, if successful, EFI would completely substitute the

carburetor manufacturers and, therefore, it represented a very real threat

that could destroy their core industry. The plot in Figure 1 shows counts of

automobile models equipped with carburetors and EFI during the 10-year

transition period from the former to the latter.

< Insert Figure 1 about here>

DataThe data come from three sources. The first data set, provided by the

U.S. Environmental Protection Agency, lists fuel economy and tailpipe

emissions performance for each type of car model sold in the United States

racing automobiles. MFI used a complex mechanical pump to deliver pulses of gasoline through nozzles into each cylinder in the

engine. In the late 1960s, fuel injection systems that contained some electrical controls were offered in a small number of car

models. The first modern electronic fuel injection systems, which we refer to as EFI, appeared in 1980. The feature that identifies

them as EFI in this study is the presence of closed-loop controls that sense the engine's performance and that change operating

parameters in real time. The U.S. EPA refers to these as feedback fuel systems. Because MFI and early electrical fuel injection

application was limited to a few automobiles, and it is (literally in this case) an historical footnote, we do not include it in this

analysis.

22

for the automobile model years 1978 through 1992. In these data, a car

model is any available combination of model name, model year, body type,

engine size, transmission type, power output, and carburetor or EFI. This

data set is merged with a data set from the U.S. National Highway Traffic

Safety Administration (NHTSA) and the Department of Transportation that

contains observations of each car model’s physical characteristics—weight,

car class, number of doors, type of engine, size of engine, type of

transmission, type of fuel delivery system, and presence of engine

management computer. Finally, for all car models for which we were able to

find a repair manual with carburetor part numbers, we used repair manuals

to identify the manufacturer of the car model’s carburetor.

The resulting data set contains 10,505 observations, an average of

700 car models per year from 1978 through 1992. It clearly shows the

pattern of substitution that occurred as EFI technology grew to dominate

the car market. In the first year of the sample, carburetors are found in 100

percent of the models. The first EFI systems appeared in 1980 and by 1992,

they had completely displaced carburetors from the market (see Figure 1).

Table 1 provides a description of the variables we used, their construction,

and their summary statistics. Table 2 reports the correlation matrix of these

variables. Table 3 reports descriptive statistics for these variables.

<Insert Table 2 about here>

Dependent Variable

23

Fuel Economy Growth Rate (MPG). We calculate annual fuel economy

growth rate by interacting MPG (miles per gallon) with a model year

counter (years from start of the sample).. As discussed earlier, the (MPG)

performance of a vehicle is strongly influenced by the carburetor and by the

technology it incorporates, and it is the most representative performance

variable for manufacturers driven to increase performance by federal

regulation and competitive pressure.2 The U.S. EPA reports MPG

performance on both city and highway test cycles and on a combined

city/highway test cycle for each car model. We use this combined

city/highway cycle MPG result as the dependent variable because it most

accurately represents real, in-use MPG performance. We estimate fuel

economy performance of an individual carburetor as the MPG of the

automobile after controlling for all observable other-than-carburetor

physical attributes of the car that impact its fuel economy.

Independent Variables

2 Carburetor performance can be measured on several dimensions, but we have chosen to focus on fuel consumption in this study

because it is the clearest and most economically important measure of carburetor performance. Other potential performance

dimensions include fuel consumption, emissions performance, and drivability. Carburetor emissions performance depends

heavily on other components (such as catalytic converters) that we cannot observe in these data, so performance of a carburetor is

impossible to identify independent from the performance of these other components. To add to the difficulty of measuring a

carburetor’s emissions performance, during the period of the transition from carburetors to EFI, there was little useful variation in

emissions performance—cars passed or they did not. A regulatory regime in which manufacturers were not given credit for

overachieving on emissions performance contributed to the lack of variation. Drivability, while important, is difficult to measure

objectively. Fuel consumption is objectively measured. It is possible to estimate a carburetor’s contribution to a car model’s fuel

consumption. Finally, carmakers received credit in the form of avoiding punishment (fines) by overachieving relative to

regulatory standards. Overachievement in one car model was a fungible benefit that could be applied to underachievement in

another of the firm’s models. For these reasons, we use (car-attribute-controlled) fuel consumption to measure carburetor

performance.

24

Car model attributes. The data set contains measures of physical

attributes that influence a car model’s MPG performance. A vector of these

car model-specific attributes is included in all but one of the specifications,

the reason for which will be explained later. The variables measure a car

model’s weight in pounds (WEIGHT), its engine’s horsepower output

(POWER), the presence of an automatic (rather than a manual) transmission

(AUTO), and the engine displacement (a measure of the engine’s size) in

cubic centimeters (ENGINECC).

Carburetor (CARB) or electronic fuel injection (EFI). These dummy

variables indicate the type of fuel delivery system present in a car model. A

fuel delivery system in this data set is either a carburetor (CARB = 1 or EFI

= 0) or an electronic fuel injection system (CARB = 0 and EFI = 1) as

reported by the EPA. The CARB and EFI variables are perfectly inversely

related. Both are used in the specifications for ease of coefficient

interpretation.

Electronic feedback fuel system controls (FFS or NOTFFS). This

dummy variable indicates the presence (or absence) of FFS controls in a

car’s fuel delivery system. Electronic FFS controls enable electronic fuel

injection (EFI) systems to measure engine performance and to adjust

operating parameters in real time, an ability necessary to the proper

functioning of an EFI system. Although such functionality was not a

necessary component in a carburetor fuel delivery system, most

manufacturers adapted FFS for use in carburetor systems. The variables

25

FFS and NOTFFS are perfectly inversely related.

Car model year through 1983 (PRE) and car model year after 1983

(POST). The PRE and POST dummy variables indicate whether a car model

was built before or after the arrival of EFI. Starting in 1980, car models

equipped with EFI start to appear in the sample. The model year 1992 was

the last year for carburetors in the U.S. automobile market (see Figure 1).

The point at which EFI arrived as a threat to the carburetor (and thus the

point at which we would expect the last gasp to occur) presumably falls

somewhere between these two dates. As a practical matter, identifying the

precise date on which EFI began to represent a viable threat to the

carburetor is not possible, so we run our specifications on a range of dates

on which this reasonably may have occurred.3 The methodology used here is

as follows: the pattern in Figure 1 indicates rapid growth in EFI adoption

from 1984 through 1988. Assuming a lag of two to five years from the point

at which an automaker specifies a car model’s fuel system and the point at

which the car model is introduced, carburetor firms may have recognized as

early as 1981 and as late as 1986 that EFI represented a serious threat to

3 A related issue is whether there may have been variation in the time at which individual carburetor firms sensed the threat posed

by EFI. First, the event that seems to have caused the carburetor’s last gasp—the arrival of EFI—happened to all of the players in

the market at the same time. Although carburetor firms may have responded at slightly different speeds, the impetus to respond

impacted them at essentially the same time because product plans are widely known among competitors in the auto industry. This

is, in part, because union negotiations are transparent, there is movement of engineers among firms, and there is a shared supply

base for many components. As an empirical matter, the robustness of the results to moving specific year provides some evidence

that there is not systematic bias introduced by firms having potentially different dates at which they adopted EFI or at which the

threat of EFI materialized for them.

26

carburetors.4 The empirical tests in this section were conducted with a

range of EFI ‘arrival’ dates, starting with an assumed arrival date in the

middle of the 1981-1986 range. Results are substantively similar with

category splits ranging from 1982-83 through 1985-86, so a midpoint split

(1983-84) is used for the presented results. The PRE and POST variables

are perfectly inversely related.

Mixed EFI/carburetor firm (MIXED) or carburetors-only firm (PURE).

These dummy variables are generated by combining information about a car

model (from the EPA data), the manufacturer of the carburetor found in the

car model (from carburetor repair manuals), and the product portfolio of the

carburetor manufacturer as described in industry journals. This

combination generates a sample of car models about which the carburetor’s

manufacturer’s product portfolio is known. This sample contains 595 car

models distributed over 11 years. Table 5 reports descriptive statistics for

this group. Car models containing a carburetor from a firm that produced

primarily carburetors are identified as PURE = 1. Car models containing a

carburetor from a firm that produced both carburetors and EFI systems are

identified as MIXED = 1.

Specifications

To test the hypotheses, we look for relationships between the arrival

of the new technology (EFI) and performance changes in the existing

4 There is qualitative evidence that EFI’s ultimate victory was seen as a fait accompli. As late as the early 1980s, knowledgeable

industry observers were uncertain about whether EFI would cause the eventual death of the carburetor. Carburetors were

improving and, at that point, the future trajectory of EFI’s progress was not clear (Norbye 1981).

27

technology (carburetors). In general terms, the empirical strategy in this

section is to start with a specification that estimates the rate of carburetor

performance improvement before and after the emergence of EFI as a

potentially dominant new technology. This simple OLS model is described

in Equation 1

MPGi=β1 (YEAR∗CARB∗PRE )i+β2 (YEAR∗CARB∗POST )i+β3 (YEAR∗EFI )i+δ ( X )i+ε i

in which a carburetor’s fuel efficiency in miles per gallon (MPG) is

estimated by regressing a car model’s MPG on time trend variables before

and after the arrival of EFI. In this model, the fuel economy (MPG) of car

model i is regressed on interactions between the car model’s YEAR and

three mutually exclusive and collectively exhaustive categorical variables.

This specification permits us to report coefficient estimates for all

categories in the regression tables. This type of specification, commonly

used in labor economics, removes the constant and does not make use of

the typical practice of omitting a base category whose coefficient estimate

is zero (Jacob and Lefgren, 2003; Jacob, Lefgren, and Moretti, 2007). The

attractiveness of this approach is that it makes it possible to interpret the

coefficient estimates for multiple-category interactions like those reported

in Table 5. These three categories indicate the car’s type of fuel delivery

system (CARB or EFI) and, in the case of carburetors, whether the car

model appeared PRE or POST the arrival of EFI. The vector X contains

carburetor fixed effects for the pre-EFI period and EFI and carburetor fixed

effects (the base interaction terms) for the post-EFI period. For the initial

28

results, the counterfactual carburetor fuel economy improvement in this

analysis is the controlled rate of carburetor improvement before the arrival

of EFI. Then, in order to understand technology and firm-type effects, we

compare rates of carburetor improvement across these categories in the

post-EFI arrival era. The focus on rate of change rather than level of fuel

economy provides a clear picture of technological development trends.

To control for alternate explanations beyond the hypothesized effects,

specifically patterns that might cause the improvement observed in this

base specification to be amplified by a change in the population to which

carburetors were used, we specify a second OLS model including a vector of

variables describing attributes of car model i. Although this simple

specification controls for changes in the population of cars in which

carburetors remained, it does not account for nonrandom assignment of

carburetors and EFI to individual car models. As a result, the endogenous

nature of the selection process could cause biased coefficient estimates. To

prevent such biases and address the possibility of this type of selection, we

use an instrumental variables (IV) approach in a two-stage least squares

(2SLS) regression. We instrument for the presence of a carburetor in a

given car model by using the carburetor penetration rate in car models built

by different manufacturers, but that were built in the same year, have

engines with the same number of cylinders, and are about the same size

(within the same 200 cubic centimeter displacement group). This measure

captures variation, ostensibly driven by considerations known to car

29

companies but not to the researchers, in the penetration rate of carburetors

across different types of car engines. In notation, the instrument for the

presence of a carburetor in car model i built by manufacturer j in model

year k with engine type l is

CARBINST ijkl=∑p ≠i

∑q≠ j

MODELpqkl , carburetor

∑p ≠i

∑q ≠ j

MODELpqkl , carburetor∨EFI

where the numerator is the sum of all cars built in model year k with engine

type l that are carbureted, but that are not model i (and, therefore, are not

built by manufacturer j). The denominator is the sum of all cars built in

model year k with engine type l that are carbureted or have EFI, but that

are not model i and are not built by manufacturer j.

The advantage of this instrument is that its validity can be determined

deductively. The instrument is correlated with the propensity of an

individual car model to be equipped with a carburetor. However, the

instrument contains no information about the focal car model, so it is not

correlated with unobserved characteristics specific to that model. This

means there is not a causal relationship between carburetor penetration

among similar engines and the propensity for the focal car model to be

equipped with a carburetor.

One potential problem with this instrument is that carburetor

penetration may be correlated with the physical attributes common to cars

equipped with similar engines. This would be problematic if we did not have

30

detailed controls for the physical attributes of individual car models

(because of remaining engine type group-level endogeneity). However,

because the data contain variables measuring individual car model physical

attributes, we can control for considerations that might change the

likelihood of an individual car model being equipped with a carburetor. As a

result, this instrument is valid under the maintained assumption that there

are no unobserved engine type effects that are correlated with MPG once

we have controlled for individual car model attributes.

Based on the logic of the three hypothesized firm actions that may

contribute to a last gasp, we form three alternative 2SLS specifications that

follow the construction in the base OLS specification (with instruments for

the presence of a carburetor) and that add variables consistent with those

explanations. The specifications in this section implicitly allow the proposed

explanations to operate in concert rather than in a mutually exclusive way.

Therefore, the results are not sensitive to the order in which the controls

are added—the sequential construction of the specifications is for

expositional clarity only. Again, for ease of coefficient interpretation where

there are many interacted variables, the regressions are specified with

mutually exclusive and collectively exhaustive categories with no omitted

variable. As a result, each coefficient is interpretable as a stand-alone

estimate for that category—it is not interpreted as a difference from an

omitted category and it is not necessary to add it to any other category. The

limitation to this strategy is that a post-estimation Wald test must be used

31

to determine whether coefficients are significantly different from one

another. These are reported where relevant. All regressions are reported

with robust standard errors.5 As additional robustness tests we also

constructed models estimating controlled annual MPG growth within each

category independently (i.e. without interaction terms). Comparison across

these models replicates those reported in the fully interacted models

reported here.

RESULTS

Last Gasp

In Hypothesis 1, we proposed that when threatened by a potential

technological discontinuity, the technology trajectory of an older technology

would experience a last gasp, or unexpected increase in the technology

performance trajectory. From a data description standpoint, a plot of fuel

efficiency over time shows that after the introduction of EFI, cars equipped

with carburetors exhibited dramatically increased fuel economy. In Figure

2, the mean annual fuel economy of car models equipped with carburetors

is plotted against the mean annual fuel economy of car models equipped

with EFI. The plots, de-meaned to remove changes in the overall fleet of

cars and to highlight the relative performance of the two technologies in

5 The data are not arranged as a panel of car models because a car’s model name (e.g., Mustang) does not convey useful fuel

economy information about a car year over year or even within year. For example, a Ford Mustang shares many components with

other cars from Ford, Lincoln, and Mercury. But a 1992 Mustang shares very little with a 1993 Mustang. Furthermore, a 1992

Mustang with a four-cylinder engine may have more in common (from the perspective of engine and transmission) with another

Ford model than it does with a 1992 Mustang with an eight-cylinder engine. For these reasons, we cluster standard errors on

brand (Ford, Chevrolet, etc.) rather than on model name (Mustang, Celebrity, etc.) and do not include model fixed effects.

Standard errors have been clustered both ways with no significant effect on results.

32

actual use, appear to show a dramatic increase in fuel economy of car

models equipped with carburetors—the last gasp of the carburetor.

<Insert Figure 2 about here>

Statistically, Model 1 in Table 4 reports the results of a simple OLS model

exploring the MPG growth rate before and after the arrival of the EFI

threat. The results show that after the arrival of EFI, the performance of the

carburetor increased from a growth rate of 0.346 MPG to a rate of 0.858

MPG, a difference statistically significant with a t-test at the 1 percent level.

In other words, these results suggest support for H1: that the arrival of the

EFI led to a last gasp in carburetor technology. Multiple robustness checks

of the arrival date confirm these results, as described in the methods

section.

Reconfiguration

However, although we observe a last gasp using a simple OLS

specification, there may be many forces contributing to the last gasp,

including the selection effects described in Hypothesis 3. Specifically, in

Hypothesis 3, we argued that when threatened, incumbents would

reconfigure, or retreat, to areas of comparative advantage, creating an

endogeneous selection effect. For example, as automakers faced pressures

to increase fuel economy, they may have reconfigured their car designs

creating smaller, lighter cars to increase fuel economy—a change which

would make it appear that carburetor performance improved when, in fact,

the characteristics of the car population actually changed. Alternatively, EFI

33

may have forced carburetor manufacturers to retreat to areas of

comparative advantage, displacing carburetors from the market starting in

heavier and more powerful car models (see Table 3) where the cost of EFI

could be more easily absorbed, leaving carburetors in lighter and less

powerful vehicles. In the data, we can observe this shift over time and

control for it by including variables describing the car model’s physical

characteristics relevant to MPG. If changes in carbureted car model

characteristics (measured here by the variables WEIGHT, POWER, AUTO,

and ENGINECC) are responsible for carburetor performance improvement,

then we would expect that the inclusion of these controls should account for

the increase in MPG growth. In support of Hypothesis 3 (retreat and

recombination), we find that after adding these controls to the simple OLS

model in Model 2 of Table 4, estimated annual carburetor MPG growth

(YEAR*CARB*PRE) decreases from 0.80 MPG per year before the entrance

of EFI to 0.44 MPG per year after (YEAR*CARB*POST), a change significant

at the 1 percent level.6 This oversimplified model suggests that a changing

population of carbureted cars accounts for the entire last gasp, providing

support for H3, but calling H1 (last gasp) into question.

However, this simple OLS analysis does not account for a second,

more subtle selection effect shaping a potential last gasp: auto

manufacturers may have chosen to equip car models with EFI according to

6 The R2 levels in these regressions are very high by social sciences standards because the data are generated by mechanical

components in the automobile interacting in a system governed by physical laws.

34

the expected impact it would have on MPG. This selection effect represents

the active effort by manufacturers to reconfigure their automobile portfolio

by putting carburetors in models where they had the best fit. To control for

this type of selection, we employ a 2SLS regression, reported in Model 3 of

Table 5, with the inclusion of the CARB instrument described earlier.

Interestingly, the revised estimates of MPG growth rates in this regression

are substantially different from the OLS estimates: annual carburetor MPG

growth increases by 0.26 MPG more per year after the emergence of the

EFI threat than before the threat, a difference significant at the 5 percent

level (p < 0.05). In other words, reconfiguration appears to account for

some, but not all, of the apparent last gasp, providing support for H3. Most

importantly, even when accounting for the contribution of reconfiguration

to the last gasp, a significant positive change in MPG growth rates remains,

providing support for H1, or a last gasp in the technology trajectory. This

last gasp effect, after controlling for the different types of reconfiguration,

proved robust under all the conditions described in the methods section and

for multiple specifications employing alternate controls.

Recombination

In Hypothesis 4, we argued that when threatened by a new

technology, incumbents recombine new technologies with older

technologies to improve performance. In the case of carburetors and EFI, a

prime candidate for an intergenerational technological spillover is

electronic feedback fuel system (FFS) controls because such controls were

35

not available before the emergence of EFI but they could be integrated into

the architecture of carburetors. Therefore, we investigate incumbents

adopting FFS controls and ask whether FFS-equipped carburetors

experienced greater MPG growth than non-FFS-equipped carburetors and

to what extent this difference can explain carburetor MPG growth. The EPA

data indicate whether a particular carbureted car includes FFS controls,

and this allows carbureted cars to be divided into two categories―those

equipped with FFS and those without FFS. To test whether the addition of

FFS components had an impact on carburetor efficiency growth, we

reestimate the specification in Table 5, but divide carbureted cars in the

post-EFI era into those with and without FFS.

As the resulting estimates in Model 4 of Table 5 show, after the

emergence of the EFI threat, annual MPG growth was, in fact, higher for

carburetors equipped with FFS than it was for those without FFS.

Carburetors with FFS are estimated to have grown more efficient by 1.90

MPG per year during the post-EFI period, and non-FFS carburetor MPG are

estimated to have grown significantly less quickly during this same period,

at 0.73 MPG per year (both estimates significantly different at the 1 percent

level). These results provide support for Hypothesis 4, that recombination

contributed to the last gasp and, at first blush, seem to suggest that the

performance increase for carburetors actually came from hybrid

carburetors, not from simply ‘trying harder’ to innovate in the old

technology.

36

Innovation

In Hypothesis 2, we argued that the primary explanation for the last

gasp offered by prior literature—trying harder to innovate—would

contribute to a last gasp in performance. Under this explanation,

incumbents threatened by the entry of a new technology expend extra effort

to squeeze even more performance from a threatened technology. The

initial examination of Model 4 in Table 5 described earlier suggests that in

fact ‘trying harder’ may not contribute to a last gasp in performance as the

performance of non-hybrid carburetors appeared to drop from 1.231 MPG

to 0.729 MPG rate, a difference significant at the 1 percent level. This

evidence would appear to lack support for the primary hypothesis provided

by prior literature that, when threatened, incumbents try harder to innovate

and thereby extract more performance from the older technology.

<Insert tables 3 - 5 about here>

Exploratory Analysis: Incumbent Technology Choice and the Last Gasp

The surprising lack of support for the primary explanation for last

gasps offered by prior literature led us to the question of individual firm

technology choice. For the most part, prior literature has explored the last

gasp phenomenon at the level of the industry, or technology trajectory,

ignoring the effects of individual firm technology choices on the last gasp.

To better understand the effect of firm technology choice, we categorized

firms based on one of three primary technology responses to the threat of

EFI: (1) retrench in carburetors, (2) invest in hybrids, or (3) invest in EFI.

37

These categories represent primary technology choices. All firms produced

both carburetors and hybrids, but the differences in primary technology

choice were clearly manifest in number of models, stated priorities, and

even in microbehaviors (for example, incumbents focusing on standard

carburetors purchased the hybrid FFS components whereas incumbents

focusing on hybrids chose to manufacture them—a choice reflecting greater

commitment). Because R&D expenditures were not available for many firms

due to their private ownership structure, we categorized firms based on the

investment priorities, reflected in number of new product models offered

(types of carburetors) and in stated priorities noted in industry magazines.

As a robustness check, we examined a continuous measure based on the

number of carburetor model types produced by each firm, which replicated

the results. Therefore, for interpretive simplicity, we chose to use a binary

representation of primary firm technology choice. These large firms and

their technology choices (which make up more than 99 percent of the

market) are reported in Table 6. Descriptive statistics suggest that 50

percent of incumbents retrenched into carburetors and 50 percent chose to

shift investment to hybrid carburetors. Interestingly, no incumbents elected

to leap straight into EFI—a telling technology choice we will return to later.

After grouping incumbents based on their primary technology choice,

we reexamined Hypothesis 1 (innovation) employing the same 2SLS analysis

but grouping the analysis based on firm technology choices. Specifically, in

Model 5 of Table 5, we split incumbents into PURE (firms choosing to

38

retrench in carburetors) and MIXED (firms choosing to invest in hybrids).

We first tested for potential differences in initial capabilities before the

arrival of the threat and we found that although the MPG growth rate was

slightly higher for MIXED firms in absolute terms (1.233 growth rate for

MIXED versus 1.023 growth rate for PURE) before the EFI threat, this

difference was not statistically significant.

Next, Hypothesis 2 argued that when threatened, incumbents will

attempt to squeeze extra innovation out of the existing technology, which

will lead to a last gasp. Arguably, for those PURE firms choosing to focus

primarily on old carburetors, the threat of an EFI technological

discontinuity may have a greater impact than for those MIXED firms

experimenting with hybrids carburetors which could give provide them an

option on a future EFI world. If these threatened PURE firms were trying

harder to innovate, we should expect to find that carburetors from PURE

carburetor firms improved more rapidly after the arrival of EFI than did

carburetors from MIXED firms.

In support of this hypothesis, the estimates presented in Model 5 of

Table 5 provide tantalizing clues about the nature of threat on incumbent

efforts to improve technology. For incumbents choosing to retrench into

standard carburetors, fuel economy growth in their hybrid carburetors was

significantly worse than the growth these firms obtained from their old-

fashioned carburetors before the arrival of EFI, falling from 1.02 MPG per

year to 0.55 MPG per year, a difference significant at the 1 percent level.

39

However, fuel economy growth in their old-fashioned carburetors actually

significantly improved after the arrival of EFI, increasing from 1.02 MPG

per year to 1.48 MPG per year, a difference significant at the 5 percent

level. This suggests that threatened PURE firms may have directed extra

innovation effort according to their primary technology choice—the old-

fashioned carburetors not equipped with the FFS components from EFI. In

summary, we find support for Hypothesis 2, that innovation efforts led to a

last gasp but qualified by incumbent technology choice: the last gasp in the

standard carburetor occurred only for those incumbents who chose to

retrench into carburetors.

By contrast, for MIXED incumbents that chose to make significant

investments into hybrid carburetors, we see a different pattern of last gasp.

The changes observed in MIXED firm MPG growth rates are almost the

opposite from those observed among PURE firms: MIXED firms’ hybrid

carburetors (FFS) improved more rapidly after the threat. By contrast, fuel

economy growth in their non-hybrid carburetors worsened after the threat.

These results also provide support for Hypothesis 2, qualified by incumbent

technology choices: firms investing in hybrid carburetors experienced a last

gasp in hybrid carburetors but not standard carburetors. In summary, the

nature of incumbent technology choice shaped how those actions impacted

the evolution of a last gasp: incumbents choosing to focus on the original

carburetor produced a last gasp in original carburetors, whereas

incumbents focusing efforts on hybrid carburetors produced a last gasp in

40

hybrid carburetors. Overall this supports Hypothesis 2 but with much

greater nuance than established in prior qualitative or empirical work.

Exploratory Analysis: Performance Implications

Finally, the performance consequences of the last gasp are important,

but largely unaddressed, considerations. Does the last gasp affect the pace

of the technology transition overall and how do incumbent technology

choices affect the performance of individual firms? In terms of technology

trajectory, although the primary purpose of the paper was to provide

empirical validation of the last gasp and the sources contributing to the last

gasp, we can provide some initial evidence to the performance

consequences of the last gasp. Arguably it is possible that a last gasp could

defer or delay a technology transition, in some cases for years, as has

occurred in the semiconductor, lightbulb, and electric vehicle industries. To

determine the effect of the last gasp on the carburetor industry, we

estimate a substitution function that follows a Fisher-Pry logistical curve

(Dattee, 2007). We do this by first estimating the logistic growth trajectory

(‘S’-curve) of EFI as a function of time. We then estimate a model in which

yearly splines are included in the logit around the time of the carburetor’s

last gasp. This allows for a delay in growth of EFI adoption. The revised

model increases the pseudo-R2 of the logit by 1.2%. Although this is a

relatively modest improvement, it is in the expected direction, which is

evidence that supports a delay that is visually quite stark in Figures 3 and 4.

This preliminary analysis suggests that while the last gasp in carburetors

41

did not stop the transition to EFI, it did delay the technological discontinuity

approximately 2 years depending on the market segment.

< Insert Figure 3 and 4 about here>

In terms of the effect of firm technology choice on performance

outcomes, we can provide some qualitative evidence. First, we observed

that none of the incumbents choosing to retrench into carburetors

successfully transitioned to EFI (see Table 6). Every incumbent choosing to

retrench effectively failed. Second, we observed that no incumbents chose

to leap straight to EFI technologies. Third, we observed that all firms

choosing to invest heavily in hybrid technologies successfully survived the

transition to EFI technologies. This represents an important observation

with potential implications for the study of technological discontinuities and

firm strategy; we will return to this point later. In conclusion, firm

technology choice appears to have significant effects on the survival of

incumbents during a technical discontinuity as does the last gasp on the

timing of the technical discontinuity.

DISCUSSION AND CONCLUSION

Although in the canonical description of industry evolution,

technology discontinuities follow a rapid transition from old to new

technology during which time rigid incumbents are swept aside, sometimes

such transitions take longer than expected or never occur at all. Prior

literature has suggested that one reason for the delay may be the tendency

of an older technology to exhibit a last gasp—a sudden, unexpected increase

42

in the technology performance trajectory—when threatened. While prior

literature has suggested this pattern, there has been little empirical

verification of the last gasp and almost no attention to the sources of a last

gasp beyond suggesting that incumbents try harder to innovate (Gilfillan,

1935; Harley, 1988).

In this paper, we provide some initial empirical verification of the last

gasp, in the setting of the carburetor industry when threatened by a

potential technical discontinuity with the emergence of electronic fuel

injection. In addition, we suggest two additional potential sources of the last

gasp—reconfiguration and recombination—in addition to the more common

‘trying harder’ explanation offered by the literature. We find support that all

three source—innovation, reconfiguration, and recombination—contribute

to a last gasp, but in some unexpected ways. First, we find evidence that

incumbents retreat and reconfigure, creating the appearance of a last gasp

—although product performance may not improve per se—rather the

apparent performance improvement is the result of the product retreating

from less efficient to more efficient applications. Second, even when

accounting for such selection pressures, we find that recombination, or the

creation of hybrids between the old and new technology generations, makes

a significant contribution to the last gasp. Third, while we initially failed to

find evidence that the primary innovation explanation offered by prior

literature contributed to a last gasp, once we accounted for incumbent

technology choices, we found that incumbents focusing their efforts on the

43

original carburetor contributed to a last gasp in standard carburetors; those

incumbents focusing on hybrid carburetors contributed to a last gasp in

hybrid carburetors. This finding provides support for the effect of

incumbent efforts to try harder, qualified by their technology choices.

Lastly, we estimated that the last gasp did defer the technology

discontinuity for a time. Furthermore, although no incumbents leapt

immediately to EFI, only those incumbents first investing in hybrid

carburetors survived the transition to EFI technology.

These results contribute to our understanding of technology evolution

and the last gasp. Although Tushman and Anderson (1990; 1986) helped

define the broad characteristics of technology transition, recent work has

begun to flesh out the dynamics of these transitions more fully. For

example, Adner and Kapoor (2012) argue that the emergence of a new

technology and the shape of its trajectory depend on the ecosystem

surrounding a particular innovation. Although the last gasp has been

suggested by several prior observers, we add to this emerging discourse by

demonstrating the existence and sources of a last gasp. Furthermore, we

suggest that the last gasp has important performance implications that may

be more or less significant depending on industry. Although in the case we

studied, EFI technologies eventually replaced the carburetors, carburetor

firms had previously successfully defended themselves against the threat of

wick carburetors, rotating brush carburetors, catalytic carburetors,

vaporizers, and mechanical fuel injectors. If the carburetor’s last gasp had

44

more effectively defended its place in the automobile, students of the

industry would likely be praising the innovativeness of incumbents rather

than criticizing their rigidity—much as we do when describing

semiconductor incumbents today. The point is that the last gasp and its

effect on deferring or even defeating technology transitions gets little

attention from scholars, but considering these counterfactuals remains

important to developing a robust theory. For example, over the years, there

have been many technologies that seemed likely, imminent replacements

for the incandescent lightbulb. Neon, halogen, fluorescent, compact

fluorescent, and LED are among the principal substitutes that have been

heralded as the death knell of the incandescent bulb. However, as scholars,

we have few, if any, core narratives that describe doomed executives at

incandescent lightbulb companies, grimly holding on to their losing

technology as the waves of creative destruction drown them. Rather, efforts

to extract additional performance from the lowly incandescent lightbulb

seem to have held off the challengers for now, even though the challengers

have had the force of government regulation on their sides. It is only after a

successful technology transition has occurred that we can tell the rigidity

narrative. And for now, we don’t have that narrative for incandescent

lightbulbs or silicon semiconductors. This work highlights the need to

develop a more robust representation of incumbent actions that goes

beyond the interpretation that incumbents are either rigid or engage in full-

scale transformation. As a first step toward this discussion, we describe and

45

validate three specific responses—innovation, reconfiguration, and

recombination—that contribute to the emerging discussion of how firms

actively respond to discontinuities (Eggers and Kaplan, 2009; Kaplan, 2008;

Tushman and Rosenkopf, 1996).

In addition, the surprising results around the creation of hybrids

spanning technological discontinuities represents an important new topic in

the study of firm strategy and technological evolution. Although hybrids

have been observed in prior studies of technological discontinuities, they

are often discounted as representations of organizational dysfunction.

However, we proposed that hybrids could be a sophisticated learning option

strategy whereby incumbents can learn about a potential future technology

during the gray area between transitions. If the transition fails to occur, the

incumbents engaging in hybrids have not risked everything and have

potentially gained an advantage over competitors in the process. However,

if the transition does occur, the incumbents developing hybrids have a

potential advantage in making the leap to the next generation. We did

observe that the development of hybrids contributed to the last gasp. More

importantly, however, we observed that only those incumbents seriously

developing hybrids survived the technological discontinuity. Qualitatively it

appears that developing hybrids allowed incumbents to develop the needed

capabilities to transition to EFI, during and after the technological

discontinuity occurred. Hybrids, therefore, offer an important opportunity

for study. Clearly the potential for hybrids to occur are limited by certain

46

characteristics: for example, modular components that can be grafted onto

compatible architectures in the context of a period of uncertainty (as

opposed to a short, rapid technical discontinuity). Nonetheless, the role of

hybrids deserves significant further study, which the authors hope to

continue in future work.

Finally, these findings contribute to the literature on change in

dynamic environments (Agarwal et al., 2009; Eisenhardt, Furr, and

Bingham, 2010; Rindova and Kotha, 2001), which emphasizes the need to

quickly adapt to environmental changes such as technological

discontinuities. Whereas one may be tempted to assume that leaping to a

new market or technology is always preferable, our results provide insight

into the dynamic uncertainty that firms face in making such decisions

(Davis, Eisenhardt, and Bingham, 2009). Not only do firms face outcome

uncertainty, but their actions to fight the threat can shift the very frontier of

performance, which further complicates the timing of the decision of

whether to leap to a new market. In this paper, we show that waiting to leap

may be a prudent decision when the technology frontier moves and that a

recombination strategy can act as a hedge against technological

uncertainty. We believe these observations contribute to the conversation

regarding the dynamic forces inside and outside the firm that affect the

timing and outcome of firm strategy, particularly in innovation contests

(Agarwal, Sarkar, and Echambadi, 2002; Benner et al., 2002; Helfat and

Peteraf, 2003).

47

In closing, although our study is limited by the focus on the

carburetor industry, there are many contemporary technology industries

that appear to have experienced last gasps, including CISC processor

architecture, incandescent lightbulbs, photovoltaics, steel bicycle frame

materials, coronary artery bypass graft (CABG) surgery, and silicon

semiconductors just to name a few. In some cases, the last gasp may have

given the incumbents false hope that they might survive a threat, but in

other cases, the last gasp seems to have allowed an extant technology to

defeat a threat. In a world increasingly influenced by technology and

shaped by change (D'Aveni, 1994; Wiggins and Ruefli, 2005), how firms

manage both threat and transition may become as relevant as how firms

maintain advantage once they capture an opportunity (Anderson, 1999;

Eisenhardt and Bhatia, 2002). We hope these findings contribute to the

development of robust theory about how organizations operate in and

manage such environments.

48

Table 1. Descriptive statistics of variables

Variable Construction Obs.Mean

Std. Dev Min

Max

MPG City-highway miles per gallon of gasoline, U.S. EPA

10,505

24.5

6.9 9 69

WEIGHT Weight of the vehicle (pounds)

10,505

3,497

792 1,750

6,000

POWER Engine horsepower 10,505

124 43 41 478

AUTO Dummy var: (1) auto. trans. or (0) man. trans.

10,505

0.52

- 0 1

ENGINECC

Engine displacement in cubic centimeters

10,505

3,110

1,416

802 7,538

CARB Dummy var: (1) carburetor or (0) electronic fuel injection

10,505

0.46

- 0 1

EFI Dummy var: (1) electronic fuel injection or (0) carburetor

10,505

0.54

- 0 1

FFS Dummy var: (1) feedback fuel system or (0) none

10,505

0.71

- 0 1

PRE Dummy var: (1) obs 1978-83 or (0) obs 1984-92

10,505

0.27

- 0 1

POST Dummy var: (1) obs 1984-92 or (0) obs 1978-83

10,505

0.73

- 0 1

49

Table 2. Correlation matrix of variables

MPGWEIGHT

POWER

AUTO

ENG CC CARB EFI FFS PRE POST

MPG 1.00

WEIGHT -0.85 1.00

POWER -0.61 0.65 1.00

AUTO -0.31 0.27 0.19 1.00

ENGINECC -0.80 0.87 0.64 0.27 1.00

CARB -0.13 0.02 -0.29 -0.05 0.19 1.00

EFI 0.13 -0.02 0.29 0.05 -0.19 -1.00 1.00

FFS 0.23 -0.09 0.16 0.03 -0.21 -0.70 0.70 1.00

PRE -0.21 0.02 -0.16 0.01 0.18 0.56 -0.56 -0.65 1.00

POST 0.21 -0.02 0.16 -0.01 -0.18 -0.56 0.56 0.65 -1.00 1.00

50

Table 3. Car model mean values by model year

51

52

YearObs

N

CARB = 1

N

MPG

mean

WEIGHT

mean

POWER

mean

1978 420 420 22.

0 3,370 111

1979 436 436 22.

2 3,234 107

1980 621 564 20.

2 3,737 121

1981 618 542 22.

0 3,664 117

1982 754 633 23.

9 3,491 109

1983 797 637 24.

7 3,478 107

1984 558 378 25.

7 3,307 107

1985 778 448 25.

7 3,376 114

1986 786 322 25.

5 3,436 121

1987 804 208 25.

6 3,441 124

1988 777 96 25.

5 3,499 132

1989 738 59 25.

4 3,522 133

1990 802 24 25.

4 3,561 139

1991 817 15 25.

4 3,549 145

1992 799 11 25.

0 3,647 152

Total or

Wtd.

10505 4793 24.

5 3,497 124

53

Table 4. OLS tests of explanations for last gasps dependent variable: MPG

(1) (2)

YEAR*CARB*PRE

YEAR*CARB*POST

YEAR*EFI

CARB*PRE

CARB*POST

EFI

WEIGHT

POWER

AUTO

ENGINECC

Model degrees of freedomResidual degrees of freedomObservationsAdjusted R2

0.346 (0.093)** 0.858 (0.090)**-0.093 (0.028)**--664.082 (183.870)** -1,678.162 (178.155)** 210.245(54.682)**

62710,5050.93

0.801 (0.045)** 0.442 (0.043)** 0.382 (0.015)** -1,538.486 (89.347)** -826.419 (85.464)** -707.710(28.786)** -0.006 (0.000)** -0.037 (0.001)** -1.254(0.057)** 0.000 (0.000)**102710,5050.97

Notes: * p < 0.05, ** p < 0.01. Robust standard errors in parentheses. Standard errors adjusted for 27 automobile brand clusters. Coefficient

54

estimates for interacted variables are interpretable without being added to other coefficients. The interacted variables create mutually exclusive and exhaustive categories for the observations in the data.

55

Table 5. Second stages of 2SLS regressions for last gasps dependent variable: MPG

(3) (4) (5)YEAR*CARB*PRE

YEAR*CARB*PRE*PURE

YEAR*CARB*PRE*MIXED

YEAR*CARB*POST

YEAR*CARB*POST*FFS

YEAR*CARB*POST*NOTFFS

YEAR*CARB*POST*FFS*PURE

YEAR*CARB*POST*FFS*MIXED

YEAR*CARB*POST*NOTFFS*PURE

YEAR*CARB*POST*NOTFFS*MIXED

YEAR*EFI

WEIGHT

POWER

AUTO

ENGINECC

CARB*PRE

CARB*PRE*PURE

CARB*PRE*MIXED

CARB*POST

CARB*POST*FFS

CARB*POST*NOTFFS

CARB*POST*FFS*PURE

CARB*POST*FFS*MIXED

CARB*POST*NOTFFS*PURE

CARB*POST*NOTFFS*MIXED

EFI

Model degrees of freedom

1.217(0.071)**

1.483(0.125)**

0.530(0.023)**-0.011(0.000)**-0.047 (0.002)**-1.401(0.079)**0.002(0.000)**54.336(0.871)**

50.911(0.739)**

1.231(0.072)**

1.903(0.187)**0.729(0.144)**

0.530(0.023)**-0.011(0.000)**-0.047(0.002)**-1.401(0.079)**0.002(0.000)**54.336(0.871)**

47.875(0.991)**56.854(1.322)**

1.023(0.168)**1.233(0.074)**

0.550(0.197)**1.951(0.198)**1.479(0.251)**0.688(0.147)**0.536(0.024)**-0.011(0.000)**-0.048(0.003)**-1.385(0.081)**0.002(0.000)**

53.782(1.069)**54.660(0.940)**

56.709

56

Residual degrees of freedomObservations

57.649(0.923)**10268,955

58.096(0.923)**12268,955

(1.648)**47.419(1.076)**50.500(1.753)**57.057(1.360)**57.963(0.993)**14268,955

Notes: * p < 0.05, ** p < 0.01. Robust standard errors in parentheses. Standard errors adjusted for 27 automobile brand clusters. Coefficient estimates for interacted variables are interpretable without being added to other coefficients. The interacted variables create mutually exclusive and exhaustive categories for the observations in the data.

57

Table 6. Firm technology choice and outcomes for major carburetor producers

Carburetor company

Primary tech choice during

transitionOutcome* Eventual EFI

company Details of outcome

Aisan Hybrid carburetors

Successful transition (partial)

Aisan/DensoMerge with another firm (electronics) in Toyota family; supply an EFI components

Carter Standard carburetors Failure --

Exit through discounted acquisition (Colt Industries) to become niche parts supplier (making helicopter parts)

Ford/Motorcraft

Hybrid carburetors

Successful transition Ford/Motorcraft Survive as captive parts division

of Ford

Hitachi Hybrid carburetors

Successful transition Hitachi/JECS Joint venture with Bosch and

emerged as major supplier

Holley Standard carburetors Failure -- Exit to niche (aftermarket

supplier)

Mikuni Solex Standard carburetors Failure -- Exit to niche (motorcycles)

Keihin Standard carburetors Failure --

Exit to niche (motorcycles); continue to supply other parts as part of Honda family

Nikki Standard carburetors Failure -- Exit through discounted

acquisition (Hitachi)

Weber Hybrid carburetors

Successful transition Weber Survive to become EFI producer,

survived as supplier to FIAT

GM/Rochester Hybrid carburetors

Successful transition GM/Rochester

Survive to become EFI producer, captive parts division of GM, survived

*Performance outcomes are difficult to categorize because of the variety of firm outcomes. We classify an exit from carburetors as a failure. We also classify the transition to a niche supplier as a failure since this

58

represents a shift from being a multi-billion dollar primary supplier to a niche supplier a fraction the previous firm size (often with revenues less than 5 percent of prior revenues).

59

Figure 1. Count of annual model EFI and carburetor usage

Figure 2. Mean annual carburetor and EFI MPG versus mean annual total MPG for models sold in U.S. (plots of the annual mean MPG of each population of automobile type, measured against the mean annual MPG for the total population of car models)

60

Figure 3. Plot of delayed diffusion of EFI

Figure 4. Plot of delayed diffusion of EFI with delay splines

61

62

REFERENCES

Adner R, Helfat CE. 2003. Corporate effects and dynamic managerial capabilities. Strategic Management Journal 24(10): 1011-1025.

Adner R, Kapoor R. 2010a. Value creation in innovation ecosystems: How the structure of technological interdependence affects firm performance in new technology generations, Strategic Management Journal, Vol. 31: 306-333

Adner R, Kapoor R. 2012. Innovation ecosystems and the pace of substitution: Re-examining technology s-curves. Working Paper

Adner R, Snow D. 2010b. Bold retreat. Harvard Business Review 88(2)Adner R, Snow D. 2010c. Old technology responses to new technology threats: Demand heterogeneity

and technology retreats. Industrial and Corporate Change 19(5): 1655-1675.Adner R, Snow DC. 2010d. Old technology responses to new technology threats: Demand heterogeneity

and technology retreats. Industrial and Corporate Change 19(5): 20.Agarwal R, Gort M. 2002. Firm and product life cycles and firm survival. American Economic Review

92(2): 184-190.Agarwal R, Helfat CE. 2009. Strategic renewal of organizations. Organization Science 20(2): 281-293.Agarwal R, Sarkar M, Echambadi R. 2002. The conditiong effect of time on firm survival: An industry life

cycle approach. Academy of Management Journal 45(5): 971-994.Anderson P. 1999. Complexity theory and organization science. Organization Science 10(3): 216-232.Anderson P, Tushman M. 1990. Technological discontinuities and dominant designs: A cyclical model of

technological change. Administrative Science Quarterly 35(4): 604-633.Baldwin C, Clark KB. 2000. Design rules. MIT Press: Cambridge, MA.Barney J. 1986. Strategic factor markets: Expectations, luck, and business strategy. Management Science

32(10): 1231-1241.Benner MJ. 2007. The incumbent discount: Stock market categories and response to radical

technological change. Academy of Management Review 32(3): 703-720.Benner MJ. 2009. Dynamic or static capabilities? Process management practices and response to

technological change. Journal of Product Innovation Management 26(5): 473-486.Benner MJ. 2010. Securities analysts and incumbent response to radical technological change: Evidence

from digital photography and internet telephony. Organization Science 21(1): 42-62.Benner MJ, Tushman M. 2002. Process management and technological innovation: A longitudinal study

of the photography and paint industries. Administrative Science Quarterly 47(4): 676.Brown SL, Eisenhardt K. 1997. The art of continuous change: Linking complexity theory and time-paced

evolution in relentlessly shifting organizations. Administrative Science Quarterly 42: 1-34.Christensen C. 1997. The innovator's dilemma. Harvard Business School Press: Boston.Christensen C, Bower JL. 1996. Customer power, strategic investment, and the failure of leading firms.

Strategic Management Journal 17(3): 197-218.Cohen S, Tripsas M. 2012. Managing technical transitions: The importance of disengaging from the old.

Working PaperD'Aveni RA. 1994. Hypercompetition: Managing the dynamics of strategic maneuvering. The Free Press:

New York.Dattee B. 2007. Challenging the s-curve: Patterns of technological substitution. Working PaperDavis J, Eisenhardt K, Bingham C. 2009. Optimal structure, market dynamism, and the strategy of simple

rules. Administrative Science QuarterlyDosi G. 1982. Technological paradigms and technological trajectories: A suggested interpretation of the

determinants and directions of technical change. Research Policy 11: 147-162.

63

Eggers JP. 2012. Reversing course: Ibm’s strategic recovery in the flat panel display industry. Working Paper

Eggers JP, Kaplan S. 2009. Cognition and renewal: Comparing ceo and organizational effects on incumbent adaptation to technical change. Organization Science 20(2): 461-477.

Eisenhardt K, Bhatia M. 2002. Organizational complexity and computation. In J Baum (Ed.), The blackwell companion to organizaitons: 442-466. Blackwell Publishing: Malden, MA.

Eisenhardt K, Furr N, Bingham C. 2010. Micro-foundations of performance: Balancing efficiency and flexibility in dynamic environments. Organization Science 21(6)

Eisenhardt K, Martin J. 2000. Dynamic capabilities: What are they? Strategic Management Journal 21: 1105-1121.

Foster R. 1986. The s-curve: A new forecasting tool, Innovation: The attacker's advantage: 88-111. Simon and Schuster: New York.

Gilbert C. 2005. Unbundling the structure of inertia: Resource versus routine rigidity. Academy of Management Journal 48(5): 741-763.

Gilbert C. 2006. Change in the presence of residual fit: Can competing frames coexist? Organization Science 17(1): 150-167.

Gilfillan SC. 1935. Inventing the ship: A study of the inventions made in her history between floating log and rotorship. Follett: Chicago.

Harley CK. 1988. Ocean freight rates and productivity, 1740-1913: The primacy of mechanical invention reaffirmed. Journal of Economic History 48(4): 851-876.

Helfat CE, Peteraf MA. 2003. The dynamic resource-based view: Capability lifecycles. Strategic Management Journal 24(10): 997-1010.

Henderson R. 1995. Of life cycles real and imaginary: The unexpectedly long old age of optical lithography. Research Policy 24(4): 631-643.

Henderson RM. 1993. Underinvestment and incompetence as responses to radical innovation: Evidence from the photolithographic alignment equipment industry. RAND Journal of Economics 24(2): 248-270.

Henderson RM, Clark KB. 1990. Architectural innovation: The reconfiguration of existing product technologies and the failure of established firms. Administrative Science Quarterly 35: 9-30.

Hill CWL, Rothaermel FT. 2003. The performance of incumbent firms in the face of radical technological innovation. Academy of Management Review 28(2): 257-274.

Horne A. 2002. Seven ages of paris. Vintage Books: New York, NY.Jacob B, Lefgren L. 2003. Are idle hands the devil’s workshop? Incapacitation, concentration, and

juvenile crime. American Economic Review 93(5): 17.Jacob B, Lefgren L, Moretti E. 2007. The dynamics of criminal behavior: Evidence from weather shocks.

Journal of Human Resources 42(3): 38.Kaplan J, Segan S. 2008. 21 great technologies that failed, PC Magazine: Kaplan S. 2008. Cognition, capabilities, and incentives: Assessing firm response to the fiber-optic

revolution. Academy of Management Journal 51(4): 672-695.Kaplan S, Murray F, Henderson R. 2003. Discontinuities and senior management: Assessing the role of

recognition in pharmaceutical firm response to biotechnology. Industrial & Corporate Change 12(2): 203-233.

Lavie D. 2006. Capability reconfiguration: An analysis of incumbent responses to technological change. Academy of Management Review 31(1): 153-174.

Leonard-Barton D. 1992. Core capabilities and core rigidities: A paradox in managing new product development. Strategic Management Journal 13: 111-125.

March J, Simon H. 1958. Organizations. Wiley: New York.

64

Martin X, Mitchell W. 1998. The influence of local search and performance heuristics. Research Policy 26(7/8): 753.

Mitchell W. 1991. Dual clocks: Entry order influences on incumbent and newcomer market share and survival when specialized assets retain their value. Strategic Management Journal 12(2): 85-100.

Naughton K. 2012. Hybrid's unlikely rival: Plain old cars, Bloomberg Businessweek: Porac JF, Thomas H, Wilson F, Paton D, Kanfer A. 1995. Rivalry and the industry model of scottish

knitwear producers. Administrative Science Quarterly 40(2): 203-227.Porter ME. 1980. Competitive strategy. Free Press: New York.Rindova V, Kotha S. 2001. Continuous morphing: Competing through dynamic capabilities, form, and

function. Academy of Management Journal 44(6): 1263-1280.Rosenberg N. 1976. Perspectives on technology. Cambridge University Press: New York.Rosenbloom RS. 2000. Leadership, capabilities, and technological change: The transformation of ncr in

the electronic era. Strategic Management Journal 21(10/11): 1083.Rosenkopf L, Tushman M. 1998. The coevolution of community networks and technology: Lessons from

the flight simulation industry. Industrial & Corporate Change 7(2): 311-346.Rothwell R, Zegveld W. 1985. Reindustrialization and technology. M.E. Sharpe: Armonk, NY.Rumelt RP. 1984. Towards a strategic theory of the firm. In R Lamb (Ed.), Competitive strategic

management: 556-570. Prentice Hall: Englewood Cliffs, NJ.Sull DN, Tedlow RS, Rosenbloom RS. 1997. Managerial commitments and technological change in the us

tire industry. Industrial & Corporate Change 6(2): 461-501.Thomke S, Kuemmerle W. 2002. Asset accumulation, interdependence and technological change:

Evidence from pharmaceutical drug discovery. Strategic Management Journal 23(7): 619.Thompson JD. 1967. Organizations in action. McGraw Hill: New York.Tripsas M. 1997a. Surviving radical technological change through dynamic capability: Evidence from the

typesetter industry. Industrial and Corporate ChangeTripsas M. 1997b. Unravelling the process of creative destruction: Complementary assets and incumbent

survival in the typesetter industry. Strategic Management Journal 18(Special Issue): 119-142.Tripsas M. 2008. Customer preference discontinuities: A trigger for radical technological change.

Managerial and Decision Economics 29(March-April): 79-97.Tripsas M. 2009. Technology, identity, and inertia through the lens of "the digital photography

company". Organization Science 20(2): 441-460.Tripsas M, Gavetti G. 2000. Capabilities, cognition and inertia: Evidence from digital imaging. Strategic

Management Journal 21: 1147-1161.Tushman ML, Anderson P. 1986. Technological discontinuities and organizational environments.

Administrative Science Quarterly 31: 439-465.Tushman ML, Rosenkopf L. 1996. Executive succession, strategic reorientation and performance growth:

A longitudinal study in the u.S. Cement industry. Management Science 42(7): 939-953.Utterback JM. 1996. Mastering the dynamics of innovation: How companies can seize opportunities in

the face of technological change (paperback). Harvard Business School Press: Boston.Wiggins RR, Ruefli TW. 2005. Schumpeter's ghost: Is hypercompetition making the best of times shorter?

Strategic Management Journal 26: 1249-1259.

65