Upload
vuongkiet
View
220
Download
6
Embed Size (px)
Citation preview
1
WHAT IS OPEN INNOVATION, REALLY?
BRUNO CASSIMAN
IESE Business School, KU Leuven, and CEPR
GIOVANNI VALENTINI1
Bocconi University
ABSTRACT
The open innovation paradigm involves the internal and external organization of the innovation process of firms. Proponents and opponents have different views on the actual contribution of this paradigm to the research community. We argue that given the prior literature on the subject, the main conceptual contribution of the open innovation paradigm is the claim that firms with a successful innovation process are simultaneously active on both sides of the market for technology by buying and selling technology. We derive a testable implication that buying and selling technology should be complementary in the production of innovations but find no evidence for such complementarity in a data set from the Community Innovation Survey for Belgian firms. The productivity of R&D for firms active in selling technology, buying technology, or doing both is not significantly different. Notably, firms moving toward both buying and selling increase their sales of new or improved products, but their R&D costs increase more than proportionally. The results suggest that more research is needed into the costs of organizing for open innovation to establish its actual contribution.
Keywords: Open Innovation, Research & Development, Licensing, Technology acquisition, Complementarity.
1 Corresponding author. Bocconi University, Department of Management and Technology. Via Roentgen, 1 –
20136 Milan (Italy). Ph. (+39) 0258362526. E-mail: [email protected].
The authors are grateful for comments received on prior drafts of this article from Juan Alcacer, Andrea Fosfuri,
Bronwyn Hall, and seminar participants at Copenhagen Business School, SKEMA Business School, the
Strategic Management Society Conference, and the Academy of Management Meeting.
2
INTRODUCTION
Innovation is crucial for firms to create and sustain competitive advantage. Yet academic
research and business practice alike indicate that it is a difficult task. Over the past decade,
Open Innovation has been increasingly indicated as a panacea to this problem. Defined as
“the use of purposive inflows and outflows of knowledge to accelerate internal innovation,
and expand the markets for external use of innovation, respectively” (Chesbrough, 2006a: 1)
and presented as a new ‘paradigm’, Open Innovation has attracted considerable attention from
both academics and practitioners since Chesbrough’s (2003) seminal contribution.
This study identifies, operationalizes, and tests the key theoretical insight of the Open
Innovation framework. We argue that many of the traditional claims grouped under the
umbrella of Open Innovation are not really new, but we also contend that this framework
might contain a novel and relevant implication not made explicit before. Our econometric
analyses, however, do not support this implication but rather reject it, raising doubts about the
actual contribution of the Open Innovation framework.
So, what is innovative—and distinctive—in the Open Innovation framework, really?
From a theoretical perspective, Trott and Hatmann (2009) argue that the principles at the core
of the Open Innovation framework (see Chesbrough, 2003) are actually not new. For
example, the finding that any organization should realize that there is relevant and useful
knowledge beyond its boundaries and try to tap into it was clear at the very least from the
early studies on gatekeepers and boundary spanners (e.g., Allen and Cohen, 1969; Tushman,
1977). Similarly, the concept of firms profiting from a novel idea that they did not necessarily
discover and develop was already supported by early literature on alliances and networks
(e.g., Rothwell et al., 1974). At a more general level, the point that technologies could be
disentangled from physical goods and traded had been made by literature on the markets for
3
technology (e.g., Arora et al., 2001). Therefore, Trott and Hatmann (2009) conclude that the
Open Innovation paradigm can be more accurately characterized as “old wine in new bottles.”
From an empirical standpoint, as Open Innovation suggests, there is some evidence that
firms are increasingly opening their internal research and development (R&D) by sourcing
knowledge from and transferring knowledge outside the firm’s boundaries. Inter alia, an
Organisation for Economic Co-operation and Development (OECD) survey covering 105
medium and large firms shows that almost 60% of the firms interviewed reported increased
inward and outward licensing during the 1990s (Sheehan et al., 2004). Similarly, Zuniga and
Guellec (2008), analyzing a representative sample of patent-filing firms in 2007, highlight the
widespread diffusion of licensing among patenting firms. Still, Mowery (2009) contends that
while the structure of industrial R&D has undergone considerable change since 1985, rather
than creating an entirely novel system, this restructuring has revived important elements of
the industrial research system of the United States in the late nineteenth and early twentieth
centuries. Many of the elements of the Open Innovation approach to R&D management were
already visible and relevant in this earlier period.
Thus, it seems there is nothing really innovative in the Open Innovation framework,
possibly downgrading it from a theoretical framework to a managerial fad. However, we
believe there is something new that characterizes this framework that has not been explored
previously. Specifically, we contend that the most fundamental element of novelty of the
Open Innovation framework is the (implicitly) posited complementarity between inflows and
outflows of knowledge—that is, the idea that the marginal return from engaging in one type
of flow increases as the intensity of the other increases.
While several empirical studies have equated Open Innovation with the mere
acquisition of external knowledge (e.g., Asakawa et al., 2010; Laursen and Salter, 2006), the
Open Innovation framework actually aims at integrating inward and outward knowledge
4
transfers, as is also clear from the original definition previously reported. More recently,
Chesbrough has confirmed that “Open Innovation is really about both sides of the R&D
market” (Chesbrough and Euchner, 2011: 14; emphasis added).
Still, extant literature has overlooked the possible interdependencies between inward
and outward flows of knowledge and focused on either inbound or outbound transfers,
ignoring the other direction of flow. The lack of integration of these two perspectives is
evident in several statements. In a recent review, Arora and Gambardella (2010: 776) suggest
that the aspect that has received most attention from technology scholars is technology supply
and that “by contrast, the determinants of the demand for external technology have not
received the same attention.” In another review, Lichtenthaler (2011: 78) contends that “much
of the work on technology transactions focused on inbound open innovation” and that
“outbound open innovation has been relatively neglected.” Therefore, it seems clear that
literature has typically analyzed each Open Innovation activity (i.e., inbound and outbound
flows of knowledge) independently.
In this study, we posit that the distinctive element of novelty in the Open Innovation
framework is precisely the idea that the effect of inbound and outbound open innovation
activities are not independent and that firms that combine inbound and outbound flows of
knowledge (are expected to) outperform those firms that rely on only one type of knowledge
transfer. After having derived this prediction, we test it empirically. Using data from the
Belgian Community Innovation Survey (CIS), we find that firms that engage in both inward
and outward Open Innovation activities do not display an increase in the productivity of R&D
greater than that of those firms that engage in only one of the two activities. We also find
evidence, when disentangling costs and benefits from open innovation, that firms moving
toward both buying and selling increase their sales from new or improved products, but their
R&D investments increase more than proportionally. These results indicate that researchers
5
should more carefully examine the real costs of organizing for open innovation to establish
that the framework provides any real contribution.
OPEN INNOVATION AND THE COMPLEMENTARITY OF ‘BUY’ AND ‘SELL’
Open Innovation literature contends that due to increased competition and shorter product life
cycles, over the past decades, firms have witnessed a general decrease in their top line growth.
At the same time, the cost of R&D has increased significantly (Chesbrough, 2003). The result
has been a decrease in the productivity of the innovation process. To overcome this problem,
companies are advised to become more permeable to the external environment. They should
rely on externally developed knowledge and technologies (henceforth, BUY), as well as allow
their technology to be used and brought to the market by other firms (hereinafter, SELL). In
doing so, they can acquire new revenues from licensing and decrease the cost and time of new
ideas’ development by leveraging external knowledge and technology. Easier access to the
resources needed to innovate, which are increasingly distributed across firms, also allows
firms to develop more and better new products (e.g., Faems et al., 2010). Being open
therefore increases the productivity of the innovation process.
If it is generally agreed that engaging in either BUY or SELL can enhance firms’
innovation productivity, we contend that, according to some arguments (at times implicit) of
the Open Innovation framework, it does so even more for firms that engage in both BUY and
SELL. That is, the returns from BUY are greater for firms that concurrently engage in SELL,
and vice versa. In the following, we synthetically explain why.
Most companies do not exploit all the inventions they produce. Giuri et al. (2007), using
data from a survey on 9,017 European patents, show that more than one-third of them are not
used. While approximately half these unused patents are so-called blocking patents, the other
half are just “sleeping.” Anecdotal evidence on U.S. firms confirms this trend. A few years
ago, when Procter & Gamble surveyed all the patents it owned, it determined that only
6
approximately 10% of them were in active use in at least one P&G business. Dow Chemical
went through an extensive analysis of its patent portfolio starting in 1993. In that year,
approximately 19% of Dow’s patents were in use in one of its businesses, and a further 33%
had some potential defensive use or future business use. The remaining patents were either
licensed to others or simply not being used in any discernible way (Chesbrough, 2006b).
Although analyzing the reasons why firms produce inventions that they later do not use
is beyond the scope of this study, it nonetheless appears evident that these unutilized ideas
and technologies constitute a waste of corporate resources. If these ideas are offered for sale
to the external market (i.e., if firms engage in SELL), not only does a company increase its
revenues, but according to Chesbrough (2006b), it can also attain two additional outcomes.
First, it contributes to “decongesting” the internal innovation process. Companies might
actually decelerate their efforts in the innovation process if several internal inventions are still
unutilized, because it diminishes the perceived value of additional investments in research.
Similarly, firms can wait to invest in new and profitable technological trajectories until all
inventions they already own are commercialized. Selling unused ideas can help avoid these
problems, freeing and decongesting the innovation funnel. In turn, firms more effectively
embrace promising novel technological areas: They can attain the needed internal and
external sources of knowledge more quickly and use them productively. Even when firms
acquire external knowledge, they do not necessarily have the incentive to exploit it (Puranam
and Srikanth, 2007). Tapping into external knowledge might not provide the expected benefits
if this knowledge is blocked by a congested innovation funnel. Thus, engaging in SELL may
increase the benefits of engaging in BUY.
Second, we expect that engaging in SELL will renew internal inventors’ motivation, in
that when they know their ideas are used, and not kept sleeping within the firm’s boundaries,
they will work harder on new innovations. Motivated inventors will not only produce more
7
and better inventions but also screen more widely for external sources of knowledge.
Although the “not-invented-here” syndrome is still a threat, inventors whose ideas are
exploited commercially should take more advantage of external sources of knowledge,
because they will be less worried about being substituted for by these external sources.
Moreover, the market for knowledge is imperfect and characterized by asymmetric
information, and these imperfections may prevent firms from accessing the market for
technologies. While firms that BUY can gain access to valuable external knowledge and more
quickly achieve useful external capabilities (Chesbrough, 2003; Laursen and Salter, 2006),
search costs and fear of opportunism in negotiations can prevent firms from actively
searching for technologies (Arora and Gambardella, 2010). This market is far from efficient.
For example, Razgaitis (2004) uses a sample of 229 U.S. and Canadian companies and shows
that for every 100 licensable technologies, only 25 result in a potential license, only six to
seven negotiations are begun, and only three to four deals are eventually concluded. These
problems can be mitigated. Firms that SELL regularly have a significant knowledge of the
technology market, gain a reputation of being open firms, and are in general more effective at
valuing external technologies. The same holds for firms that are already active on the
technology market as buyers. Given their knowledge, capabilities, and reputation in the
technology markets, we also expect that firms that BUY will be more effective at SELL and
can thus obtain additional gains from the latter activity. Research has shown that firms that
are open to transferring knowledge based on their prior experience can identify a large
number of knowledge transfer opportunities (Rothaermel and Deeds, 2006), and their
transaction costs in knowledge markets may decrease because of learning effects. As a
consequence, being active in one type of Open Innovation activity (i.e., BUY or SELL)
decreases the costs of engaging in the other activity as well.
8
In summary, we argue that BUY and SELL should increase the productivity of a firm’s
innovation process as independent activities, and that they are complementary activities. More
precisely, using the operational definition of complementarity of Milgrom and Roberts
(1990):
Let us suppose there are two activities, A1 and A2. Each activity can be performed
by the firm (Ai = 1) or not (Ai = 0) and i {1, 2}. Then the function Π(A1, A2) is
supermodular and A1 and A2 are complements if and only if Π(1, 1) - Π(0, 1) ≥
Π(1, 0) - Π(0, 0).
This definition implies that adding an activity while the other activity is already being
performed has a higher incremental effect on performance (Π) than adding the same activity
in isolation. For the purposes of this study, the performance measure we focus on is R&D
productivity. If BUY and SELL are complementary, the contribution of BUY to R&D
productivity is greater for firms that at the same time engage in SELL, and vice versa. This,
we contend, is the fundamentally new argument in the open innovation framework.
DATA AND MEASURES
We drew the sample for this research from innovation data in Belgian manufacturing
industries, which were collected as part of the CIS conducted by Eurostat in European Union
member states.2 Firms involved in the survey represent a stratified sample of the economy
based on industry and firm size. The method and types of questions used in innovation
surveys are described in the OECD’s Oslo Manual. These data have been used in several
recent academic articles on innovation (e.g., Cassiman and Veugelers, 2006; Laursen and
Salter, 2005).
We focus on data from the fourth wave of the Belgian CIS survey, conducted in 2005,
which is the most recent wave that contains information on not only the inward flow of open
2 We obtained the CIS data from ECOOM, The Centre for Research & Development Monitoring of the Flemish
government, that executes the CIS for the Flemish region in Belgium.
9
innovation (i.e., BUY) but also outward flows of knowledge (i.e., SELL). The latter questions
were added to the Belgian questionnaire and did not appear in the standard EUROSTAT
questionnaire for that year or any later wave. Moreover, following Cassiman and Veugelers
(2006), we restrict our attention to the “innovation active” firms included in the sample. We
identified innovation active firms by their answer to the question regarding whether they had
been actively engaged in introducing new or improved products or processes in the previous
two years (if they had done so, they were considered innovative). We obtained a sample of
810 observations, but because of some missing values in some variables of interest, we
performed most of the econometric analyses using a sample of 681 firms.
We characterize a firm’s Open Innovation activities with two dummies, BUY and
SELL, as follows: A firm is characterized as engaging in BUY if it invests in external R&D
contracting, has acquired external knowledge through licensing or R&D advice, or both. A
firm is characterized as engaging in SELL if it sells to the external environment any kind of
R&D activity, including R&D contracting and advice or receives revenues from licensing.
Table 1 reports the frequency of activity adoption for the firms of our sample.
Insert Table 1 and Table 2 about here
We argue that the Open Innovation framework implies that BUY and SELL are
complementary activities for increasing firms’ innovation process productivity. Verifying a
complementarity relationship as defined by Milgrom and Roberts (1990) is a difficult
econometric task (see Athey and Stern, 1998). To provide some initial econometric evidence
supporting this relationship—or rather, rejecting it—we implement the productivity approach
that Cassiman and Veugelers (2006) suggest. With this approach, we test the existence of
complementarity between two activities by regressing the chosen measure of performance on
a mutually exclusive and collectively exhaustive combination of the two activities. For the
purposes of this study, following the arguments of the Open Innovation framework, we
10
measured performance using the productivity of R&D. To measure a firm’s productivity of
R&D, we calculated the ratio between firms’ sales from new or substantially improved
products and total R&D costs, which includes expenses for both internal and external R&D
activities. As for the key covariates, the two Open Innovation activities we identified, BUY
and SELL, allow us to construct four exclusive (open) innovation strategies firms can adopt,
depending on whether they engage in BUY and/or SELL: (1) No BUY no SELL, (2) Only
BUY, (3) Only SELL, (4) BUY & SELL. Thus, BUY and SELL are complementary activities if
the following inequality holds:
(i1) Productivity(BUY & SELL) – Productivity(Only BUY) ≥
Productivity(Only SELL) – Productivity(No BUY no SELL).
Table 2 summarizes the variables used in this study, along with their descriptive statistics.
EMPIRICAL ANALYSES AND RESULTS
Table 3 reports the average value of the productivity of R&D for each of the four Open
Innovation strategies. The figures in column 3.1 indicate that while engaging in either BUY or
SELL increases the productivity of R&D significantly—increasing the yield on each €1
invested in R&D from approximately €3.7 to approximately €10 and €9, respectively—
engaging in both activities does not seem to provide additional benefits. Although the ratio
may be artificially inflated for firms that do not perform any internal R&D, the results are
qualitatively similar if we consider only firms that carry out internal R&D activities
(hereinafter, MAKE) (see column 3.2 in Table 3). Similar results hold when we split the
sample into high-tech industries and other industries, following the OECD classification.
We then verified whether this relationship holds when we regress R&D productivity3 on
the four exclusive innovation strategies together with industry dummies, firm size, and a
3 In the regression, we use the logarithm of the productivity of R&D as the dependent variable to decrease the
skewness of this variable and thus diminish heteroskedasticity concerns.
11
dummy that captures whether the firm is part of a group. More formally, we estimate the
equation:
Yi=β00 No BUY no SELLi + β10 Only BUYi + β01 Only SELLi + β11 BUY & SELLi + γ Xi + εi
where X is a vector of control variables; the test for complementarity of BUY and SELL,
following inequality (i1) then is:
β11 – β10 ≥ β01 – β00.
Table 4 reports the results of this regression. The parameter estimates of model 4.2
show that engaging in any Open Innovation strategy increases R&D productivity; though the
coefficients of BUY & SELL, Only BUY, and Only SELL are not statistically different at
conventional levels of significance, not only is inequality (i1) not supported, but it is
ultimately significant in the opposite direction with p < .01. Thus, our data do not support the
complementarity hypothesis.
Model 4.3 includes only sample firms that MAKE, and previous results are confirmed.
Furthermore, we ran the same regressions of model 4.2 with robust errors calculated, as
Davidson and MacKinnon (1993) suggest to better control for possible heteroskedasticity, as
well as excluding the 5% of observations with the lowest value of total R&D costs, for which
productivity might be artificially inflated. In unreported results (available on request) we
confirm the main findings.
Insert Table 3 and Table 4 about here
While these initial results seem robust to alternative specifications, two important
concerns arise with regard to the way we measure the dependent variable and the possible
endogeneity of the key covariates. First, we measure the productivity of R&D as firms’ sales
from new products over total R&D costs. A problem with this measure is that it only captures
product innovation and not process innovation, whereas R&D investments might be
directed—and usually are—toward both types of innovations. This is problematic in particular
12
for those firms intensely involved in process innovation if they were concurrently more
involved in open innovation activities, because our measure would underestimate their
productivity. Although we cannot fully control for the presence and weight of process
innovation in sample firms, to address this issue, we ran the productivity regression on
subsamples of firms that indicated they had achieved some process innovations in the period
of analysis and those that did not. As the parameter estimates presented in model 5.1 and 5.2
of Table 5 show, our results are robust to these specifications.
Second, to calculate our dependent variable, we divide current sales from new products
over current investments in R&D. (The survey asks for the value of both variables in the year
previous to the one in which the survey was administered.) However, it is easy to notice that
today’s investments in R&D generate sales only in the future, and therefore, we might
underestimate—or more generally misestimate—the outcome of current investments.
Although there is generally a high longitudinal correlation in R&D investments and
innovation activities, we acknowledge that with our data structure, we cannot fully solve this
issue. However, we tried two mitigate this concern in several ways.
We began by calculating a new measure of R&D productivity, matching data on sales of
our sample with data on R&D costs from the previous CIS wave. Still, (1) there is substantial
(random) attrition between one sample and the other, leaving only 169 observations in both
samples; (2) data on innovation strategies are still related to the second wave; and (3) because
CIS in the previous wave was administered by a different organization, we cannot be fully
sure about data coherence across different survey waves. In any case, the hypothesis of
complementarity is rejected in this case. Another way of addressing the same problem is to
consider a different measure of R&D productivity. In particular, we investigated the
productivity of R&D employees (as opposed to total R&D costs), on the premise that this
measure of R&D input is less volatile than total R&D costs and thus might suffer less from
13
the contemporaneousness of R&D input and output. We then calculated the ratio between
sales from new products and R&D employees (for those firms for that indicated that they had
R&D employees and for which this information was available). Although the sample reduces
to 553 observations, the empirical results (not presented here but available on request) still
suggest the nonexistence of a complementarity relationship between BUY and SELL.
Insert Table 5 and Table 6 about here
A second possible concern involves the endogeneity of the relationships we estimated.
If, in the productivity regression, unobserved factors influencing R&D productivity—which
are captured by the error—are also correlated with the adoption of innovation strategies, the
parameter estimates of these strategies may be biased. Specifically, theory warns of the
eventuality of reverse causality: Chesbrough (2006b) contends that large, mature firms are
more likely to adopt an externally oriented innovation strategy after the internally oriented
strategy is widely considered to have failed. Thus, it is possible that the lack of a significant
relationship of complementarity between BUY and SELL is driven by the reality that firms
that adopted an Open Innovation strategy were unproductive before doing so and have
adopted such a strategy to improve R&D productivity. Firms that BUY & SELL could have
thus been poor performers ex ante, which would not be evident in a cross-section (and past
performance, which is included in the regression error, is correlated both with current
performance and the strategy adopted.) Although a proper solution would require longitudinal
data and exogenous shocks, we tried to mitigate this concern in two related ways. First, we
controlled for past performance in the productivity regression by matching our sample with
the firms’ past financial performance as assessed by ROA.4 Controlling for past performance
4 We obtained these data from ECOOM, which were matched to preserve the anonymity of the firms in the data.
Financial performance is the correct control because if innovation strategies have a direct effect on R&D
productivity, which affects firm overall performance, managers usually decide their strategy on the basis of
firms’ financial performance. We calculated ROA for 2001, the year of the previous CIS wave.
14
in the productivity regression should purify the error from possible correlations with
innovation strategies. The results are virtually unchanged (see Table 5, model 5.3) and, if
anything, reject the hypothesis of complementarity even more strongly.
In addition, we determined whether past performance is actually driving innovation
strategies adoption. To do so, we ran a bivariate probit regression with BUY and SELL as the
two dependent variables. In the bivariate probit regression, we control for firm size, whether
firms are part of a group, and if firms MAKE. To facilitate identification, we inserted as
covariates the following variables:
Firms’ reliance on universities as a knowledge source, because Cassiman and Veugelers
(2006) have shown that reliance on university knowledge is an important contextual factor
that drives the effectiveness of BUY in R&D and might thus explain its adoption;
A dummy that takes the value of 1 if the firm has applied for patents, a proxy for firms’
propensity to adopt formal IP protection, which fosters SELL (Arora et al., 2001); and
The lack of qualified personnel and of financing as substantial hampering obstacles to
innovate, which might explain firms’ decision to BUY or SELL.
Table 6 presents the results of the estimation of the bivariate probit, indicating that past
performance is not significantly driving firms to BUY or SELL, thus reducing the concerns of
reverse causality.
Having established the lack of a complementarity relationship between BUY and SELL,
we proceeded by trying to understand the possible underlying causes of this result. To this
end, we split the productivity of R&D in its two constituent parts: sales from new products
and total R&D costs. Table 7 reports the average share of sales deriving from new or
substantially improved products over total sales, as well as of total R&D costs over sales.
These figures suggest that the key issue is that firms that engage in BUY & SELL are actually
characterized by a higher share of sales from new products, but also by an even higher
15
incidence of R&D costs. In other words, BUY& SELL increases firms’ innovativeness, but
this process is disproportionately costly. These results are confirmed by the regressions in
Table 8.
Insert Table 7and Table 8 about here
DISCUSSION AND CONCLUSION
In this study, we claim that the key novel theoretical insight of the Open Innovation
framework is that BUY and SELL are complementary (innovation) activities, and that
engaging in one activity increases the return from the other in terms of R&D productivity. We
empirically tested this claim and found that the complementarity relationship does not hold
for a sample of Belgian manufacturing firms. The results presented are robust to several
additional analyses.
Therefore, our empirical results indicate that firms should specialize in one of the Open
Innovation activities. Companies such as Cisco have built dominant technological and market
positions through their acquisitions (Mayer and Kenney, 2004). Other companies have
innovated in their business model, specializing in the development and subsequent
commercialization of general technologies (Gambardella and McGahan, 2010).
Furthermore, our results are due to a disproportionate increase in total R&D costs,
which is not mirrored in a more than proportional increase in sales deriving from innovative
products. This result is germane to Faems et al.’s (2010) study, in which they empirically
examine the relationship among technology alliance portfolio diversity, product innovation
performance, and financial performance. While they find that technology alliance portfolio
diversity has an indirect positive impact on financial performance through increased product
innovation performance, they also show that technology alliance portfolio diversity exerts a
cost-increasing effect. What is more, their analyses indicate that the direct cost-increasing
effect of technology alliance portfolio diversity exceeds the indirect value-generating effect.
16
The increase in costs, the authors argue, is because as the complexity of a firm’s alliance
portfolio increases, the firm must commit additional resources to its management, such as a
dedicated alliance function at the corporate level of the organization, which leads to a
consequent increase in costs. Moreover, MAKE and BUY, which are generally considered
substitutes, might not necessarily be so in the case of R&D, in which external activities must
be added to internal ones (Brusoni et al., 2001; Veugelers, 1997).
Notably, as Faems et al. (2010: 793) argue, the potential upside advantages of Open
Innovation have been closely scrutinized, and yet “the open innovation model remains
relatively silent on the cost implications of collaborative strategies.” Our study shows that the
costs of Open Innovation activities should not be underestimated. Therefore, opportunities
exist to investigate the drivers of this increase in costs, which, as prior literature seems to
suggest, is strictly linked to the cost of organizing Open Innovation activities through formal,
additional structures. At the same time, we believe further innovation research would be well
served by emphasizing performance measures that focus on not only the degree of firms’
innovativeness but also the cost/input side of their innovation strategies.
The cost-increasing effect of Open Innovation, however, might only be transitory.
Faems et al. (2010) argue that the negative influence of the complexity of firms’ technology
alliance portfolio on costs can reverse in the long run. Over time, learning might cause
management costs to decrease. This trend may also indicate a shortcoming in our empirical
setting: Our data refer to 2005, right after the success of the Open Innovation framework,
when managers’ learning process might have been just at the initial stages. Therefore, we
acknowledge that our results may pertain to not only a specific country but also a specific
time frame.
Notwithstanding the limitations of this study, which should elicit caution in interpreting
our results, we believe this study contributes to strategy and innovation literature that
17
highlights the lack of clear operationalization and empirical testing of the innovative
dimension of the Open Innovation framework. We proposed a possible contribution of the
Open Innovation framework but found no empirical confirmation. We provide directions that
further studies could usefully explore by investigating the contingencies that make the
contemporaneous use of BUY and SELL advantageous. This study simply shows that their
unconditional joint adoption does not provide the positive results much of the literature on
Open Innovation seems to imply.
REFERENCES
Allen, T.J., Cohen, W.M. 1969. Information flow in research and development laboratories. Administrative Science Quarterly, 14: 12–19.
Arora, A., Fosfuri, A., Gambardella, A. 2001. Markets for technology and their implications for corporate strategy, Industrial and Corporate Change, 10: 419-451.
———. 2010. Ideas for rent: An overview of markets for technology. Industrial and Corporate Change, 19: 775–803.
Asakawa, K., Nakamura, H., Sawada, N. 2010. Firms’ open innovation policies, laboratories' external collaborations, and laboratories’ R&D performance. R&D Management, 40: 109-123.
Athey, S., Stern, S. 1998. An empirical framework for testing theories about complementarity in organizational design. Working paper 6600, National Bureau of Economic Research, Boston, MA.
Brusoni, S., Prencipe, A., Pavitt, K. 2001. Knowledge specialization, organizational coupling, and the boundaries of the firm: Why do firms know more than they make? Administrative Science Quarterly, 46: 597-621.
Cassiman, B., Veugelers, R. 2006. In search of complementarity in innovation strategy: Internal R&D, cooperation in R&D and external technology acquisition. Management Science, 52: 68–82.
Chesbrough, H. 2003. Open innovation. Boston: Harvard Business School Press.
Chesbrough, H. 2006a. Open innovation: A new paradigm for understanding industrial innovation. In: Chesbrough, H., Vanhaverbeke, W., West, J. (Eds). Open innovation: Researching a new paradigm, 1-12. Oxford: Oxford University Press.
Chesbrough, H. 2006b. Open business models. Boston: Harvard Business School Press.
Chesbrough, H., Euchner, J. 2011. The evolution of open innovation: An interview with Henry Chesbrough. Research Technology Management, September-October.
Davidson, R., MacKinnon, J. G. 1993. Estimation and inference in econometrics. New York: Oxford University Press.
18
Faems, D., de Visser, M., Andries, P., Van Looy, B. 2010. Technology alliance portfolios and financial performance: Value-enhancing and cost-increasing effects of open innovation. Journal of Product Innovation Management, 27: 785–796.
Gambardella, A., McGahan, A. 2010. Business-model innovation: General purpose technologies and their implications for industry structure. Long Range Planning, 43: 262-271.
Laursen, K., Salter, A. 2006. Open for innovation: The role of openness in explaining innovation performance among U.K. manufacturing firms. Strategic Management Journal, 27: 131–150.
Lichtenthaler, U. 2011. Open innovation: Past research, current debates, and future directions. Academy of Management Perspectives, 25: 75-93.
Mayer, D., Kenney, M. 2004. Economic action does not take place in a vacuum: Understanding Cisco’s acquisition and development strategy. Industry and Innovation, 11: 299-325.
Milgrom, P., Roberts, J. 1990. The economics of modern manufacturing: Technology, strategy, and organization. American Economic Review, 80: 511–528.
Mowery, D. C. 2009. Plus ça change: Industrial R&D in the third industrial revolution. Industrial and Corporate Change, 18: 1–50.
Puranam, P., Srikanth, K. 2007. What they know vs. what they do: How acquirers leverage technology acquisitions. Strategic Management Journal, 28: 805–825.
Razgaitis, R. 2004. US/Canadian licensing in 2003: Survey results. Journal of the Licensing Executive Society, 34: 139-151.
Rothaermel, F.T., Deeds, D. L. 2006. Alliance type, alliance experience and alliance management capability in high-technology ventures. Journal of Business Venturing, 21: 429–460.
Rothwell, R., Freeman, C., Horsley, A., Jervis, V.T.P., Robertson, A.B., Townsend, J. 1974. Sappho updated: Project Sappho: Phase II. Research Policy, 3: 258–291.
Sheehan, J., Martinez, C. Guellec, D. 2004. Understanding business patenting and licensing: Results of a survey. Paris: OECD.
Trott, P., Hartmann, D. 2009. Why “open innovation” is old wine in new bottles. International Journal of Innovation Management, 13: 715–736.
Tushman, M.L. 1977. Technical communication in R&D laboratories: The impact of project work characteristics. Academy of Management Journal, 20: 624–645.
Veugelers, R. 1997. Internal R&D expenditures and external technology sourcing. Research Policy, 26: 303-315.
Zuniga, M.P., Guellec, D. 2008. Survey on patent licensing: Initial results from Europe and Japan. Paris: OECD.
19
Table 1. Number and Percentage of Sample Firms Engaged in Each Open Innovation Activity
SELL
BUY 0 1 Total
0 296 (43.5%) 70 (10.3%) 366
1 205 (30.1%) 110 (16.1%) 315
Total 501 180 681
Table 2. Main Variables’ Description and Descriptive Statistics
Variable Description
Mean Std. Dev.
BUY Dummy that takes the value of 1 if a firm invests in external R&D contracting, has acquired external knowledge through licensing or R&D advice, or both
0.463 0.499
SELL Dummy that takes the value of 1if a firm sells to the external environment any kind of R&D activity, including R&D contracting and advice or receives revenues from licensing
0.264 0.441
Part of a group Dummy for whether the firm is part of a group
0.608 0.489
Size Logarithm of the number of employees 4.153 1.588 Hampering factor: Lack of qualified personnel
The (perceived) extent to which the lack of qualified personnel is an obstacle to innovation (0–3 scale)
1.301
0.979
Hampering factor: Lack of financing
The (perceived) extent to which the lack of financing is an obstacle to innovation (0–3 scale)
1.362
1.096
University as a source of information
Importance for the innovation process of information from research institutes and universities (0–3 scale)
0.991
0.966
MAKE Dummy indicating whether firms have internal R&D
0.750 0.433
Applied for patents Dummy indicating whether firms have formally applied for patents
0.232 0.422
ROA Return on assets 5.274 15.299 Productivity of R&D Sales from new or significantly improved
products over total costs of R&D 6.253 40.321
20
Table 3. Average Productivity of R&D for Each Open Innovation Strategy
(3.1)
All firms
(3.2)
Only firms with internal R&D expenditures
(3.3)
Firms in high-tech industries
(3.4)
Firms in medium- and
low-tech industries
No BUY no SELL 3.683 4.383 2.208 4.098
Only BUY 10.150 5.658 5.462 11.662
Only SELL 8.940 8.774 13.477 7.805
BUY & SELL 4.199 4.036 7.930 2.45
Total 6.253 5.222 5.383 6.529
N of observations
681
511
164
517
Table 4. Results of Productivity Regressions
(4.1) (4.2) (4.3)
All firms All firms Only firms with internal R&D exp.
No BUY no SELL 0.586*** (0.163) 0.816*** (0.184)
Only BUY 0.849*** (0.181) 0.963*** (0.206)
Only SELL 1.217*** (0.212) 1.312*** (0.236)
BUY & SELL 0.955*** (0.205) 1.097*** (0.226)
Part of a Group -0.010 (0.095) -0.056 (0.094) -0.062 (0.109)
Size 0.071* (0.031) 0.047 (0.031) 0.028 (0.034)
Industry dummies Included Included Included
Constant 0.643*** (0.164)
N of observations 681 681 511
Model’s F 3.17*** 20.34*** 17.79***
Complementarity test
Rejected
Rejected
BUY&SELL-OnlyBUY ≥ OnlySELL-NoBUYNoSELL
F=7.61*** F=3.05*
*, **, *** are significantly different from zero at the 10%, 5% or 1% level respectively
Note: Dependent variable: productivity of R&D.
21
Table 5. Results of Additional Productivity Regressions
(5.1) (5.2) (5.3)
Only firms that do process innovation
Only firms that do not do process
innovation
All firms for which ROA is available
No BUY no SELL 0.685*** (0.254) 0.586*** (0.209) 0.580*** (0.184)
Only BUY 1.164*** (0.295) 0.662*** (0.223) 0.865*** (0.206)
Only SELL 1.407*** (0.331) 1.084*** (0.272) 1.194*** (0.239)
BUY & SELL 1.306*** (0.327) 0.684*** (0.257) 0.882*** (0.233)
Part of a Group -0.027 (0.148) -0.049 (0.120) -0.098 (0.104)
Size 0.016 (0.050) 0.058 (0.040) 0.063* (0.036)
ROA (previous CIS wave) -0.002 (0.003)
Industry dummies Included Included Included
N of observations 352 329 598
Model’s F 10.76*** 10.66*** 17.66***
Complementarity test
Rejected
Rejected
Rejected
BUY&SELL-OnlyBUY ≥ OnlySELL-NoBUYNoSELL
F=3.95** F=3.50* F=7.83***
*, **, *** are significantly different from zero at the 10%, 5% or 1% level respectively.
Note: Dependent variable: productivity of R&D
22
Table 6. Results of Bivariate Probit
BUY SELL
MAKE 0.105 (0.138) 0.394** (0.169)
Size 0.173*** (0.046) 0.115** (0.048)
ROA (previous CIS wave) -0.000 (0.004) -0.002 (0.004)
Part of a group 0.166 (0.132) 0.113 (0.146)
Applied for patents 0.344** (0.151) 0.652*** (0.152)
Hampering factor: Lack of qualified personnel
0.064 (0.060) 0.036 (0.064)
Hampering factor: Lack of financing
0.113** (0.055) 0.160*** (0.059)
Reliance on university as a knowledge source
0.369*** (0.067) 0.179** (0.070)
Constant -1.770*** (0.281) -2.402*** (0.321)
Industry dummies Included Included
N of observations 597
Model’s Chi2 210.01***
Correlation (rho) 0.075 (0.080)
*, **, *** are significantly different from zero at the 10%, 5% or 1% level respectively.
23
Table 7. Average Values for Different Open Innovation Strategies
Percentage of sales deriving from new
products
(All firms)
Total R&D costs over sales
(All firms)
Total R&D costs over sales
(Without outliers with R&D intensity>1)
No BUY no SELL 6.8% 16.9% 7.2%
Only BUY 8.0% 15.2% 6.3%
Only SELL 12.0% 11.8% 10.2%
BUY & SELL 16.0% 20.8% 11.8%
Total 9.2% 16.5% 8.0%
N of observations 681 681 665
Table 8. Innovation Strategies and the Constituents of R&D Productivity
(8.1) (8.2)
Dependent variable: Share of sales deriving from new or substantially improved
products (log)
Dependent variable: R&D costs over sales (log)
No BUY no SELL 0.126*** (0.019) 0.169*** (0.034)
Only BUY 0.146*** (0.022) 0.189*** (0.038)
Only SELL 0.167*** (0.025) 0.159*** (0.045)
BUY & SELL 0.196*** (0.025) 0.200*** (0.043)
Part of a Group -0.012 (0.011) -0.010 (0.020)
Size -0.003 (0.004) -0.200*** (0.007)
Industry dummies Included Included
N of observations 681 681
Model’s F 14.86*** 11.37***
*, **, *** are significantly different from zero at the 10%, 5% or 1% level respectively.