Technological Discontinuities and Complementary Assets: A Longitudinal Study of Industry and Firm Performance
Technological Discontinuities and Complementary Assets: A Longitudinal Study of Industry and Firm Performance
Technological Discontinuities and Complementary Assets: A Longitudinal Study of Industry and Firm Performance
informs
doi 10.1287/orsc.1040.0100
2005 INFORMS
Charles W. L. Hill
Department of Management and Organization, Business School, University of Washington, Seattle, Washington 98195-3200,
chill@u.washington.edu
e suggest that the type of complementary assets (generic versus specialized) needed to commercialize a new technology is critical in determining the industry- and rm-level performance implications of a competence-destroying
technological discontinuity. At the industry level, we hypothesize that incumbent industry performance declines if the
new technology can be commercialized through generic complementary assets, whereas incumbent industry performance
improves if the new technology can be commercialized through specialized complementary assets. At the rm level, we
posit that an incumbent rms nancial strength has a stronger positive impact on rm performance in the postdiscontinuity
time period if the new technology can be commercialized through generic complementary assets. We hypothesize, however,
that an incumbent rms R&D capability has a stronger positive impact on rm performance in the postdiscontinuity time
period if the new technology can be commercialized through specialized complementary assets. Drawing on multi-industry,
time series, and panel data over a 26-year period to analyze pre- and postdiscontinuity industry and rm performance, we
nd broad support for our theoretical model.
Key words: technological discontinuities; complementary assets; incumbent industry and rm performance heterogeneity;
time series and panel data analyses
53
54
Industry-Level Hypotheses
Competence-destroying technological discontinuities
are generally initiated by new entrants; they tend
to be favored over incumbents, frequently triggering a Schumperterian process of creative destruction
(Tushman and Anderson 1986). The demise of the
incumbent industry appears to be particularly salient
if the complementary assets needed to commercialize
the new technology are generic. In this scenario, new
entrants are able to either build the necessary complementary assets internally or contract for them in the
market on competitive terms. Incumbents, often hamstrung by inadequate innovation processes and systems
(Dougherty and Hardy 1996), have little of value to offer
to new entrants. New entrants, however, have a strong
incentive to operate alone in order to capture monopoly
rents accruing from the successful commercialization of
the new technology (Hill 1992).
Examples of this process are numerous and have
been well documented (Foster 1986). For example, the
transitions from vacuum tubes to transistors, and then
from transistors to semiconductors, were accompanied
in each wave by a complete turnover in industry leadership. More recently, the emergence of electronic calculators destroyed the entire set of competencies held
by incumbents within the electromechanical paradigm.
In particular, electronic calculators devalued both the
upstream technological competencies and the downstream complementary competencies of manufacturers
of electromechanical calculators (Majumdar 1982). The
more reliable electronic calculators did not need to
be serviced and could be distributed through general
ofce equipment retailers. Thus, if the commercialization of a competence-destroying technological discontinuity is possible through generic complementary assets,
we expect the performance of the incumbent rms as a
group to decline.
Hypothesis 1. Following a competence-destroying
technological discontinuity, the performance of the incumbent industry declines if the complementary assets
needed to commercialize the new technology are generic.
Not all technological discontinuities necessarily lead
to the decline of incumbents accompanied by the rise
of new entrants to dominance. By drawing on examples from the automobile industry, Abernathy and Clark
(1985) argued that low transilience innovationsi.e.,
innovations that only affect either upstream or downstream competencies of incumbents, but not bothwill
not lead to the displacement of incumbents by new
entrants. As emphasized above, Teece (1986) highlighted
the importance of complementary assets as a critical
factor in determining who benets from innovation. In
illustration, Mitchell (1989) found that the possession of
specialized complementary assets enhanced the probability that incumbents in the medical diagnostic imaging industry would enter a newly emerging subeld;
in fact, the greater the incumbents specialized complementary assets, the faster the entry would be. More
recently, Tripsas (1997), in her study of the typesetter
industry, showed that incumbents may be buffered from
the negative effects of technological discontinuities if
they possess specialized complementary assets.
A technological discontinuity that destroys upstream
competencies of incumbents may simultaneously enhance the value of their downstream complementary
assets if the complementary assets necessary to commercialize the new technology are specialized. This effect
is particularly pronounced in a regime of weak intellectual property protection because specialized complementary assets may allow incumbents to appropriate innovation rents from new entrants (Teece 1986). Frequently, a
specialization-based division of labor occurs, where new
entrants focus on the upstream, technological competencies and incumbent rms focus on the downstream, complementary assets. Such a scenario nds its expression
in extensive interrm cooperation between new entrants
and incumbents. Some observers have suggested that a
competence-destroying technological discontinuity that
simultaneously enhances the specialized complementary
assets of incumbents may lead to a symbiotic coexistence that may benet both new entrants and incumbents
(Pisano 1991, Rothaermel 2000). Although incumbents
and new entrants cooperate to create value, they also
simultaneously compete to divide up the value created
(Brandenburger and Nalebuff 1996). Here, incumbents
who possess specialized complementary assets necessary
to commercialize the new technology are frequently in
a stronger bargaining position to appropriate the joint
value created (Lerner and Merges 1998, Teece 1992).
Thus, we argue that the apparently adverse effect of a
competence-destroying technological discontinuity may
actually have a positive impact on incumbent industry
performance if the complementary assets necessary to
commercialize the new technology are specialized. If
the technological discontinuity merely devalues R&D
and production activities, the challenge faced by incumbent rms is less severe. Moreover, if the marketing
and sales activities of incumbents are largely unaffected
by the change, the value of these activities may even
be enhanced if they are specialized assets, making the
incumbents more attractive as alliance partners to new
entrants. Put differently, incumbents have something to
offer to new entrantsthey have specialized complementary assets that can be joined with the assets of new
entrants to increase the probability of success for both
(Teece 1992). Prior empirical work has demonstrated
that specialized complementary assets held by incumbent pharmaceutical rms enabled them to establish
alliances with biotechnology rms; this not only aided in
adapting to the new technology, but also aided in extracting innovation rents (Rothaermel 2001a). We suggest
that in the face of a competence-destroying technological
55
56
Research Setting
Technological discontinuities are rare events in the evolution of an industry. Tushman and Anderson (1986)
studied three industries, from their inceptions over a
cumulative 165 years of industry history, and identied
eight technological discontinuities. To test our hypotheses, we identied four industries that each experienced
one technological discontinuity in a 26-year time frame,
for a cumulative 104 years of industry history. Each of
the industries was characterized by its well-documented
discontinuity and by the availability of longitudinal and
homogenous data necessary to conduct industry- and
rm-level analyses across multiple industries. Accordingly, we chose the following industries (with the respective discontinuity shown in parentheses): the computer
industry (PC), the steel industry (electric arc furnace),
the pharmaceutical industry (biotechnology), and the
telecommunications industry (wireless telephony).
Tushman and Anderson used drastic advancements in
the respective price-performance frontier as the criteria to
identify technological discontinuities: discontinuities
offer sharp price-performance improvements over existing technology (1986, p. 441). All four of the technologies considered in this study have advanced the
price-performance ratio signicantly in their respective
industries. The PC and wireless telephony have caused
exponential performance increases while simultaneously
drastically lowering prices. The electric arc furnace has
brought exponential performance improvements and signicant price cuts because it allows for smaller batches
of customized steel. Biotechnology has advanced the
performance trajectory tremendously: Scientists are now
able to discover and develop new drugs that were previously impossible to create. Furthermore, many drugs
that could previously be procured only in small quantities
(e.g., insulin) can now be harvested cost effectively in
large quantities. Applying the Tushman-Anderson (1986)
model of technological change, we submit that all four
technologies in this study are considered discontinuities.
More precisely, we argue that when applying the
Tushman-Anderson framework to the discontinuities
chosen in this study, all four technologies would be
classied as competence-destroying for incumbent
rms. Tushman and Anderson described competencedestroying technological discontinuities as requiring
new skills, abilities, and knowledge in both development and production of the product (1986, p. 442).
This denition concerns the upstream R&D activities
of incumbent rms, and implies that they are devalued
or destroyed by such technologies. Prior literature
57
Table 1
Technological Discontinuities, Complementary Assets, and Incumbent Industry and Firm Performance
Impact on
incumbent
upstream
technological
competencies
Type of
complementary
assets needed to
commercialize
new technology
Impact on
incumbent
downstream
complementary
assets
Effect on
incumbent
industry
performance
Stronger positive
effect on
incumbent rm
performance
Industry
examples
Technological
discontinuity
Destroying
Generic
Destroying
Decline
(H1)
Financial strength
(H3)
Computer, steel
Destroying
Specialized
Enhancing
Improvement
(H2)
R&D capability
(H4)
Pharmaceutical,
telecommunications
Biotechnology,
wireless telephony
58
59
First conceived by Bell Labs in 1947, cellular telephony makes it possible to drastically increase a systems subscriber capacity by using many low-powered
transmitters that cover a geographical area, which is
in turn divided into smaller cells. Each system has
a transmission-switching ofce that receives calls and
in turn sends them through the cells to another cellular phone, or, more frequently, to the local telephone
exchange. From there the cellular call is fed into the
traditional wireline telephone communication network.
Although the technology has been available since the
late 1940s, the rst cellular mobile telephone system was
not introduced until after the breakup of the Bell System
monopoly in the early 1980s. Ameritech introduced the
rst cellular network to Chicago in 1984, followed by
a second system in the Baltimore-Washington area. The
number of subscribers in the United States grew exponentially, from a base of 91,000 customers in 1984 to
roughly 33 million in 1995.
Although cellular services provide users with mobility
that wire-based lines cannot offer, the more traditional
network has not been left entirely behind by the popularity of the new technology. To route their calls, radiobased cellular systems still rely heavily on the switching
networks held by incumbent communications rms. As
a result, the growth of the cellular sector has created a
symbiotic relationship between the incumbent rms and
the new entrants: The incumbent rms need access to
the new radio-based technology to develop their own
cellular systems, and the newer rms are even more tied
(in the short run) to the traditional switching networks
dominated by the incumbents (Ehrnberg and Sjberg
1995). In other words, the technological discontinuity
has not made the complementary assets of incumbent
rms (i.e., their switching networks) obsolete. Moreover,
with respect to the commercialization of wireless telephony, the switching networks are specialized complementary assets because they can be used to transmit cellular phone calls without making any additional investment in these downstream assets.
In theory, the cellular providers can build out their own
switching networks and cut the incumbents out of the
market altogether. However, this would be both capitaland time-intensive. Building out cell sites is an expensive
endeavor due to the capital required for equipment, the
cost of acquiring licenses to use the radio spectrum in
metropolitan areas, and the cost of acquiring customers
(which typically includes giving each customer a free
cell phone). The capital commitments required to commercialize the new technology are clearly substantial. For
example, one of the early pioneers of the technology,
McCaw Cellular, had to resort to junk bond nancing
to raise the requisite capital. Already saddled by high
debt, cellular providers have chosen to use the switching networks of incumbent enterprises rather than build
their own. Due to the specialized complementary assets
60
Research Design
Sample and Data
We focused on incumbent rms in the four industries
discussed above. To test our hypotheses, we drew on
the Standard & Poors Compustat and DRI databases, as
well as on the Standard & Poors industry reports from
1972 through 1997. This 26-year time frame reects the
starting and ending date of the industry- and rm-level
analyses, which are determined by the availability of
homogenous data across all four industries. The Compustat database includes nancial, statistical, and market
information on publicly held companies. The Standard &
Poors DRI database of economic indicators is a standard
database for economic research and serves as the source
for relevant macroeconomic data. At the industry level,
we used quarterly incumbent industry performance data
to ensure sufcient observations in constructing the time
series. At the rm level, we obtained annual observations to set up pooled time-series, cross-sectional panels
of data. The majority of empirical work in strategic management relies on a cross-section of data rather than on
longitudinal panel data, and does not allow, therefore,
61
(1995) and Rothaermel (2001a). We applied the following regression model to each of the industry-level time
series, assuming a deterministic trending process:
yt = + t + DT t + HHI t +
gdpY t
+ quarterit + t
(1)
62
Table 2
1.
2.
3.
4.
5.
6.
7.
8.
9.
Return on equity
Return on assets
Financial strength prediscontinuity
Financial strength postdiscontinuity
R&D capability prediscontinuity
R&D capability postdiscontinuity
Firm capital structure
Industry concentration HHI
GDP growth rate
Mean
Std. dev.
1.
2.
3.
4.
5.
6.
7.
8.
1836
761
38054
77710
14073
60970
048
437933
289
1790
503
165660
135673
62984
159326
063
194525
222
069
019
062
019
070
010
037
001
008
009
002
002
029
052
006
013
018
008
003
021
004
010
053
006
004
009
007
029
052
006
004
014
006
017
007
003
Note. N = 566
Y1t
i1t 0 0
1t
=
Ynt
int
X1t
X2t
nt
Xnt
C1t
C2t
0
0 2t
1t
nt
1t
1t
0
2t 2t
+
Cnt
nt
nt
(2)
Results
ROA
ROE
ROA
ROE
ROA
Steel
Pharmaceutical
Pharmaceutical
Telecommunications
Telecommunications
078
209
115
073
311
358
018
102
113
122
247
042
328
076
912
864
Intercept
006
001
003
001
004
003
002
001
002
001
001
001
908E-4
001
014
020
Time trend t
089
009
038
004
154
037
023
010
038
005
011
002
026
004
034
024
Indicator
variable
break date
1988
1985
1985
1986
1981
1981
1983
1983
Break date
p < 001;
ROE
Steel
ROA
Computer
ROE
Computer
DV
Industry
001
509E-4
715E-4
200E-4
347E-4
001
179E-4
360E-4
001
001
393E-4
347E-4
205E-4
213E-4
266E-6
319E-4
Control
variable
HHI t
005
016
007
006
021
031
006
010
001
008
001
003
003
006
001
013
Control
variable
gdp Y t
225
052
084
018
015
090
011
026
006
022
063
008
039
020
annual data
Control
variable
Quarter 1
154
046
079
016
159
091
030
027
025
023
048
008
061
018
annual data
Control
variable
Quarter 2
250
054
092
019
002
092
002
027
044
022
047
008
047
021
annual data
Control
variable
Quarter 3
3643
2155
073
068
010
158
2619
527
039
067
1376
3440
2140
F -stat.
066
074
060
Adj. R2
63
64
Industry-Level Results. Table 3 depicts the industrylevel results. In all four industries, we observe statistically signicant breaks in the incumbent industry
performance time series after the introduction of the
respective technological discontinuity. Hypothesis 1
states that, following a competence-destroying technological discontinuity, the performance of incumbent
rms declines if the complementary assets needed to
commercialize the new technology are generic. The
qualitative data presented in the research setting indicate
that the PC in the computer industry and the electric
arc furnace in the steel industry were technological
discontinuities that devalued the upstream competencies of incumbent rms, and whose commercialization
was possible through generic assets. As predicted in
Hypothesis 1, we would expect a decline in the incumbent industry performance in the postdiscontinuity time
period following the emergence of the PC or the EAF.
The empirical results indicate that 1983 marks the structural break in the performance time series for the computer industry for both the ROE p < 0001 and the
ROA p < 0001 models. As expected, the sign of the
break date variable is negative in both models, indicating a decline in overall computer industry performance.
Similarly, 1981 marks the structural break in the performance time series for the steel industry for both the ROE
p < 0001 and the ROA time series p < 005, and
the sign of the break date variable is negative in both
cases, indicating a decline in overall steel industry performance. As anticipated, both the computer and steel
industries experienced a signicant decline in incumbent
industry performance after the emergence of the respective technological discontinuities. This result provides
support for Hypothesis 1.
Hypothesis 2 suggests that, following a competencedestroying technological discontinuity, the performance
of incumbent rms improves if the complementary
assets needed to commercialize the new technology are
specialized. Above, we identied biotechnology in the
pharmaceutical industry and wireless telephony in the
telecommunications industry as technological discontinuities that devalued the upstream competencies of
incumbent rms. They also simultaneously enhanced
their downstream competencies because the commercialization of these new technologies depended on specialized assets. As predicted by Hypothesis 2, we would
expect an improvement in incumbent industry performance in the postdiscontinuity time period.
Our empirical results show that the structural break
in the pharmaceutical industry occurs in 1986 in the
ROE time series p < 0001, and in 1985 in the ROA
time series p < 0001. As expected, the sign of the
break date variable is positive in both regression models, indicating an improvement in overall pharmaceutical industry performance. In the telecommunications
industry, the break date occurs in 1985 based on the
65
Table 4
Firm-Level Results
Dependent variable ROE
Model 1
base
Controls
Firm-xed effects
Financial strength
(prediscontinuity)
R&D capability
(prediscontinuity)
Firm capital
structure
Industry
concentration HHI
GDP growth rate
Model 2
Signicant
Included
Signicant
Included
Included
Included
Included
Included
471
027
546E-3
367E-4
008
005
497
051
685E-4
610E-4
006
004
6930
1803
45065
032
p < 001;
Test 2:
Average
impact
002
296E-3
018
004
1068542
637528
056
068
004
146E-3
859E-5
009
001
Test 3:
Difference in
average
impact
003
142E-3
127
007
048
011
319E-4
267E-4
006
002
Test 1:
Joint
impact
1148
1077
72422
055
Test 2:
Average
impact
955E-4
607E-4
001
487E-3
371477
201649
Test 3:
Difference in
average
impact
106E-3
290E-4
011
563E-3
067
performance if the complementary assets needed to commercialize the new technology are specialized.
We conducted several robustness checks. First, we
assessed a potential specication bias arising from unobserved heterogeneity through inclusion of the by one
year lagged dependent variables on the right-hand side
of the regression models. Second, we tested for potential
serial correlation. The results were consistent, and thus
reveal neither a specication bias nor serial correlation.
Discussion
Model 4
Signicant
Included
Test 1:
Joint
impact
Model 3
base
Signicant
Included
Independent
variables
Financial strength (H3)
(postdiscontinuity)
R&D capability (H4)
(postdiscontinuity)
Chi-Square
Improvement
over base
Adjusted R2
time period if the new technology can be commercialized through generic complementary assets. By contrast,
we hypothesized that an incumbent rms R&D capability has a stronger positive impact on incumbent rm
performance in the postdiscontinuity time period if the
new technology can be commercialized through specialized complementary assets. Based on an analysis of
longitudinal data across four industries, we found broad
support for the rm-level hypotheses.
These ndings tie in with Abernathy and Clarks
(1985) concept of low transilience innovation, which
suggests that incumbent rms can benet from technological discontinuities if their market-related competencies remain unchanged. Our results also reinforce the
necessity of comprehensively analyzing an innovations
impact on incumbent enterprises, including all links
between different rm competencies. Here, we focus
on the link between technological and nontechnological
competencies. In their study of the semiconductor photographic alignment equipment industry, Henderson and
Clark (1990) showed that seemingly minor technological advances can have severe consequences for incumbents if the new technology changes the architecture in
which the components are integrated. Although we focus
on major rather than minor advances in technology, our
results resonate with Henderson and Clarks (1990) nding that the consequences of technological change can
66
67
68
Organizational Implications
An important nding of this study is that rm-level
competencies are important because rms differ in their
ability to adapt to a competence-destroying technological discontinuity. Thus, managers need to be cognizant
about the technological and nontechnological competencies inherent in their rms. Echoing Teece (1986),
managers need to recognize the different types of
complementary assets and their differential importance
in commercializing a new technology. Moreover, a rms
nancial strength and R&D capability are not entirely
independent of one another, so it is possible for managers to make adjustments in their ratio. If a discontinuity is commercialized via generic assets, managers
should ensure that they have sufcient nancial strength
to acquire new entrants, and thereupon the new technology. If a discontinuity is commercialized via specialized
assets, however, managers can redirect free cash ow to
nance additional R&D activities.
Whereas R&D capabilities are built over time, the
time horizon for transforming a scientic invention into
a commercialized innovation (gestation period) generally takes multiple years (Hill and Rothaermel 2003); it
tends to be longer the more complex the technologys
underlying science. Thus, free cash ow to strengthen
R&D capability should be applied as soon as the impact
of the new technology on the complementary assets can
be understood. This allows managers to take full advantage of the gestation period for the new technology.
Depending on the magnitude of the discontinuity, it can
then take a few more years after the successful development of a new technology before its performance impact
becomes apparent. Thus, incumbents may be able to a
obtain a rst-mover advantage when adapting to a new
technology if they are able to understand the impact the
new technology is likely to have on their technological and nontechnological competencies prior to its manifested impact on performance. In sum, managers can
substitute free cash ow for R&D expenditures and vice
versa, depending on the type of complementary assets
necessary to commercialize the new technology.
Acknowledgments
Endnotes
1
The alliance data are drawn from the MERIT CATI database
(Maastricht University Economic Research Institute
Cooperative Agreements and Technology Indicators database).
This is likely to be one of the most comprehensive databases
covering strategic alliances worldwide over the last several
decades (for a more detailed description of MERIT CATI see
Hagedoorn 2002).
2
Because a time series of quarterly ROA was not available in
the telecommunications industry, we resorted to annual ROA
data.
3
There was no need to explicitly control for rm size because
our dependent variables are ratios and thus are already adjusted
for rm size.
4
The possibility that any of the investigated incumbent industry performance time series could exhibit more than one break
date was ruled out in all eight cases because the t-statistic for
the indicator variable break date increased with time before
peaking and subsequently declining. Thus, all eight time series
exhibited exactly one statistically signicant structural break.
5
The marginally signicant result for a structural break in the
ROA time series and the three year difference in the break dates
between the ROE and ROA time series are partly explainable
by the fact that we had to resort to annual instead of quarterly data for the ROA time series, and thus signicantly fewer
observations constituted the time series.
References
69
Majumdar, B. A. 1982. Innovations, Product Developments and Technology Transfers: An Empirical Study of Dynamic Competitive
Advantage. University Press of America, Washington, D.C.
Ehrnberg, E., N. Sjberg. 1995. Technological discontinuities, competition and rm performance. Tech. Anal. Strategic Management
7 93107.
Mowery, D. C. 1983. The relationship between intrarm and contractual forms of industrial research in American manufacturing,
19001940. Explorations Econom. History 20 351374.
Freiberger, P., M. Swaine. 2000. Fire in the Valley, 2nd ed. McGrawHill, New York.
Giovannetti, G. T., S. W. Morrison. 2000. Convergence. The Biotechnology Industry Report. Ernst & Young, Palo Alto, CA.
Powell, W. W., K. W. Koput, L. Smith-Doerr. 1996. Interorganizational collaboration and the locus of innovation: Networks of
learning in biotechnology. Admin. Sci. Quart. 41 116145.
Quandt, R. E. 1960. Tests of hypothesis that a linear regression system obeys two separate regimes. J. Amer. Statist. Assoc. 55
324330.
Rothaermel, F. T. 2000. Technological discontinuities and the nature
of competition. Tech. Anal. Strategic Management 12 149160.
Hill, C. W. L. 1992. Strategies for exploiting technological innovations: When and when not to license. Organ. Sci. 3 428441.
Hill, C. W. L., F. T. Rothaermel. 2003. The performance of incumbent rms in the face of radical technological innovation. Acad.
Management Rev. 28 257274.
Scherer, F. M., D. Ross. 1990. Industrial Market Structure and Economic Performance. Houghton Mifin, Boston, MA.
70