The Strategic Importance of Digital Capabilities in External
Knowledge Search
MSc BA SIM
Faculty of Economics and Business University of Groningen
Arkadius Mueller - S2179989 a.j.muller.1@student.rug.nl
Supervisor: dr. Q. (John) Dong Co-assessor: dr. P.J.O. (Pasi) Kuusela
Abstract
It is not until recently that IT investments have been linked to facilitating a firm’s external search strategy. Yet, it is not clear how distinct technologies may influence external search breadth and depth specifically. Therefore, drawing on the attention-based view of the firm, in this paper the enabling effect of two different digital capabilities, namely digital analytics capability and digital communication capability was investigated. To test the hypotheses that both capabilities are individually and jointly positively associated with external knowledge search, a large-scale panel survey from Germany was used. Applying zero-inflated negative binominal regression (ZINB), the results suggest that both digital capabilities had either an individual or joint contribution on external knowledge search, depending on the employed search strategy. This paper contributes to the literature by highlighting the differential effect of specific technologies on the mode of engaging in external knowledge search. Practitioners who are inclined to source knowledge outside their own firm boundaries should invest in appropriate information technology that matches their desired search strategy. Limitations and directions for future research are discussed.
Introduction
Traditionally, internal research and development practices were considered the main source of a firm’s innovation activities (Kleis, Chwelos, Ramirez, & Cockburn, 2012). However, a variety of factors including rising R&D costs, global market competition, the increasing mobility of knowledge workers, and technological developments, gradually motivated firms to shift towards an open innovation strategy (Ferreras-Méndez, Fernández-Mesa, & Alegre, 2016; Kleis et al., 2012). Open innovation refers to the deliberate opening of a firm’s innovation process to facilitate the inflow and outflow of information across firm boundaries. (Chesbrough, 2003; Huizingh, 2011). The inflowing process of accessing
external information from a firm’s external environment is referred to as external knowledge search (Laursen & Salter, 2006). In open innovation research, external knowledge search has commonly been described along two dimensions - a broadly oriented search strategy (i.e. external search breadth) focusing on the number of diverse search channels, such as
One factor that mitigates the diminishing returns of external knowledge search on innovation performance is the utilization of information technology (IT) (Gomez, Salazar & Vargas, 2017). While IT has frequently been described as a facilitator of external knowledge search, for example through the enablement of a firm’s absorptive capacity (Roberts,
Galluch, Dinger & Grover, 2012; Joshi, Chi, Datta, & Han, 2010), its anteceding function has been largely ignored (Dong & Netten, 2017). In other words, the degree to which IT
instigates external knowledge search and contributes to the number of external search channels warrants additional investigation (Dong & Netten, 2017). Because the benefits that are derived from different technologies may vary (Bharadwaj, Bharadwaj, & Konsynski, 1999), deviating from an aggregated IT investment construct, to more distinct technological classes, may provide additional insights (Aral & Weill, 2007). This is especially relevant considering that information technologies likely yield better results when aligned with a firm’s strategic objectives and the potentially synergetic effects between different
technologies (Aral & Weill, 2007). Hence, the effect of specific technologies on fostering a firm’s external search strategy remains a significant research lacuna that demands further scrutiny. This investigation does not only contribute novel insights to the external search literature but may likewise guide practitioners in their IT investment decisions, assuming they are inclined towards an open innovation strategy.
communication capabilities are defined as the implementation and actual use of big data analytics and software-based communication resource respectively.
Drawing on the attention-based view of the firm (Ocasio, 1997, 2011; Simon, 1947), both capabilities are expected to positively influence the number of external search channels in terms of breadth and depth. This is because of two reasons that are related to each
technology’s unique strength. Firstly, utilizing DAC may improve information processing and prioritization of information. Secondly, using DCC contributes to better information awareness and information sharing. Furthermore, when both digital capabilities are jointly utilized, the firm may benefit from the complementary strength of these capabilities, which will also positively influence external knowledge search. To test the present theory, large-scale panel data including 2349 German firms from all industries was used. The results suggest a differential effect of digital capabilities on external search breadth and depth. With regard to external search breadth, both capabilities were individually contributing to external search breadth whereas the interaction effect was not found to be significant. For external search depth, only the joint effect was significant.
This paper contributes to recent search literature by pointing towards a more nuanced view regarding the strategic impact of different technologies on external knowledge search. More specifically, this paper shows that different digital capabilities may not only affect external search depth and breadth in idiosyncratic ways, but that technologies may only jointly affect firms’ search capacity when external search is more intense. This has also augmented the current understanding of digital analytics capabilities and digital
Levinthal, 1990), the present paper focuses on how IT capabilities may influence external search strategy, given their limited attentional capacity. Thus, the attention-based view of the firm will form the theoretical backbone of this study.
The remaining paper is structured as follows. The theoretical foundation and
consecutive hypothesis development are described first. This is followed by the description of the employed methodology. Next, the results are presented. In the last section, the results are discussed with regard to their theoretical and managerial implication and are followed by the limitations and directions for future research.
Theoretical Background Open Innovation and External Knowledge Search
Open innovation is a familiar innovation management concept that has increasingly gained interest in various fields of research including economics, psychology and
anthropology (Huizingh, 2011). According to Chesbrough, Vanhaverbeke, and West (2006), in open innovation, firms deviate from a vertically integrated model to one in which the role of external knowledge sources becomes increasingly important. In fact, external knowledge may help to overcome organizational inertia by challenging the focal firm’s understanding of cause and effect, uncovering potential competence deficiencies, and providing new insights (Zhou & Li, 2012). Tomalinson (2010) shows that dyadic relations between the focal firm and external knowledge sources positively contribute to innovation performance.
competitors, universities, public research institutes, private R&D institutes and consultants among the various external search channels (Ferreras-Méndez et al., 2016).
Digital Analytics Capability
Increasing technological developments, such as the emergence of smart devices, social networking platforms, and cloud storage, have contributed to an exponential
proliferation of potentially valuable data that might be utilized in external knowledge search (Del Vecchio, Di Minin, Petruzzelli, Panniello, & Pirri, 2018). These large volumes of available information that would require advanced analytical, management and processing techniques to extract valuable knowledge have contributed to the term big data (Gupta & George, 2016). While, originally big data was referring to large (1) volumes of data, it quickly expanded beyond the one-dimensional definition to incorporate factors such as the data’s (2) variety (i.e. types of structured and unstructured data formats), (3) velocity (i.e. the speed associated with data creation), (4) veracity (i.e. the degree to which the data is
uncertain) and (5) value (i.e. the economic benefits that firm’s may derive from analyzing the data) (Gupta & George, 2016). In fact, according to Markus (2015), the number of
components characterizing big data has steadily increased and may now include as much as 10 different factors. Utilizing big data allows managers to make better data-driven decisions which in turn may among others impact firm performance (McAfee & Brynjolfsson, 2012). However, to make these data-driven decisions, analytical software solutions and advanced statistical algorithms need to transform big data sets into relevant insights (Gandomi, 2015; Del Vecchio et al., 2018). Some of the big data analytics methods include cluster analysis, natural language processing, predictive modelling, machine learning, sentiment analysis and data visualization (De Mauro, Greco, & Grimaldi, 2016). While big data analytics resources are the subject of analysis, it is their deployment and use that constitutes a firm’s
capabilities can be a source of competitive advantage. Thus, deviating from the resource-based terminology of big data analytics, the term digital analytics capability (DAC) will be used hereafter. Following Rai et al.’s (2012) notion of IT capabilities, a firm’s digital analytics capability is defined as the implementation and actual use of big data analytics resource. DAC have been referred to as a new frontier in innovation research and are characterized as a distinguishing factor between high and low performing firms (as cited in Côrte-Reala, Ruivoa, Oliveiraa, & Popoviča, 2019). Yet, despite their current momentum and their endorsement by practitioners, theoretical foundations on how DAC contribute to
business value, remain scarce (Mikalef, Bourab, Lekakosb, & Krogstie, 2019).
Digital Communication Capability
communication capabilities are defined as the implementation and actual use of software-based communication tools (Rai et al., 2012).
Attention-Based View
The attention-based view of the firm proposes that a firm’s behavior is the result of how it structurally channels decision-makers attention (Ocasio, 1997, 2011; Simon, 1947). Attention is defined as multidimensional construct that focuses on the way in which environmental/organizational triggers come to a decision-makers awareness, how these triggers are encoded, interpreted, emphasized, and acted upon (Ocasio, 1997). According to Shiffrin and Schneider (1977) attention follows an automatic or controlled information processing mechanism that is associated with different attentional capacity requirements. While automatic processing is the result of long-term learning that requires little cognitive effort, controlled information processing is deliberate and highly demanding on the individual’s attentional capacity. As individuals are exposed to novel information, this
information is typically processed in a controlled manner, which will drain a decision-makers attentional capacity. Given that attentional capacity is limited, individuals are only able to attend to different information selectively. In other words, because there is only a limited number of information that individuals can cognitively process, attentional trade-offs are frequently required (Ocasio, 1997; March, 1978). These trade-offs are based on value, relevance, and legitimacy assessments, which are shaped by contextual factors (i.e.
attentional structures) such as social, economic and cultural elements (Ferreira, 2017; Ocasio, 1997). Additionally, technological, physical, human, financial, and other tangible and
constraints bare a cost to the number of external search channels and the information a firm is able to adequately process (Laursen & Salter, 2006). In other words, firms may select too many or too few search channels given their limited information-processing capability. However, according to Li et al. (2013), it is likely that strategies can be implemented that empower decision-makers to overcome the limitations of their information processing ability. Moreover, because of IT’s leveraged capability pertaining to computational and
communicational tasks (Bakos & Treacy, 1986) its effect on attentional information
processing may yield promising results. Consequently, the attention-based view of the firm will be utilized as a theoretical foundation to integrate and explain the effect of DAC and DCC on a firm’s external knowledge search strategy.
Hypotheses Development The effect of DAC on external knowledge search
Good ideas are widely dispersed, and firms require sophisticated digital analytics capabilities to be able to fully exploit them (Del Vecchio et al. 2018). DAC may be instrumental for organizing external collaborations, absorbing information from the environment, and making strategic decisions (Del Vecchio et al. 2018). However,
more attentional capacity is freed up, resulting in a path dependent external search process (Lewin, Massini, & Peeters, 2011).
Moreover, due to the improved insights derived from using digital analytics capabilities, firms can shift their attention to strategically more relevant behavior enabling firms to make more efficient and effective decisions (Shamim, Zeng, Shariq, & Khan, 2018; Janssen, van der Voort, & Wahyudi, 2017). Thus, in the context of external search breadth, digital analytics capabilities can be compared to a filter that directs attention to relevant information and reduces the cognitive strains associated with traditional search. Because the insights that are derived from using digital analytics capabilities can be re-deployed across the entire network of partners, as well as in more in-depth collaborations, using DAC will significantly amplify the number of breadth and depth respectively. Therefore, the hypotheses pertaining to the application of DAC are as follows:
H1a: Digital analytics capabilities are positively associated with external search breadth.
H1b: Digital analytics capabilities are positively associated with external search depth.
The effect of DCC on external knowledge search
DCC are faster aware of potential issues pertaining to their external search channel and would be able to respond to those issues more dynamically. Because firms are better able to direct their attention to where it is needed, they will be more efficient in operating an
increasing number of breadth and depth channels. Consequently, the hypotheses pertaining to the application of DCC are as follows:
H2a: Digital communication capabilities are positively associated with external search breadth.
H2b: Digital communication capabilities are positively associated with external search depth.
The interaction effect of DAC and DCC on external knowledge search
Information is in a constant flux and need to be continuously reanalyzed to remain valuable (Bumblauskas, Bumblauskas, & Igou, 2017). What was true yesterday, may have become obsolete today, and may be harmful tomorrow. Therefore, Bumblauskas et al. (2017) argue that it requires complex human interaction to augment current findings. This
H3a: The interaction effect between DAC and DCC is positively associated with external search breadth.
Methods Data
To test whether digital analytics capabilities and digital communication capabilities impact a firm’s external search strategy, the Mannheim Innovation Panel (MIP1) database has been used. The MIP is structured as a panel survey and aims to reflect the German economy’s innovation behavior by gathering data from a wide range of German companies on a yearly basis since 1993. The data is collected by the Centre for European Economic Research (ZEW) on behalf of the Federal Ministry of Education and Research (BMBF) and is considered representative for the German firm population.
As part of the European Commission’s Community Innovation Survey (CIS), the MIP follows a strict methodological approach that ensures a high degree of validity and reliability that is consistent with the Oslo Manual (OECD (Organisation for Economic Co-operation and Development) & Eurostat (Statistical Office of the European Union), 2005). The latter has been jointly developed by the OECD and Eurostat to ensure the appropriate and consistent measurement and interpretation of innovation-, technology-, and science-related data. The MIP data is factually anonymized to prevent others from identifying specific firm related information.
When drawing data from the MIP database, a key requirement for data inclusion was the availability of information pertaining to the dependent variables external search breadth and external search depth. While the MIP has already been used in past studies (e.g. Dong & Netten, 2017; Böhringer et al., 2012; Horbach, 2008), a more recent version of the MIP was used in the present paper. However, while the dependent variables were part of the most recent version of the MIP in 2017, the independent variable and moderator were added to a
subset of unique questions focusing on a firm’s application and diffusion of digitalization in 2016. Hence, a one-year time lag design was applied in which the external knowledge search strategies in 2017 were linked to the respective independent, moderator, and control variables in 2016. After merging the 2016 and 2017 responses and matching them based on their firm specific identification number, the final sample size equals 2349 firms. However, due to missing responses for some of the control, independent, and moderator variables, the actual sample size used for the analysis was 2162.
Measures
External Search Breadth
According to Laursen and Salter (2006), external search breadth can be expressed as the aggregated number of external search channels that a firm utilizes in its inbound open-innovation strategy. Hence, to assess whether a firm is following a broadly oriented search strategy, firms needed to indicate whether various external search channels functioned as a knowledge source to a recent innovation project. In the MIP there were 14 different external channels, namely: (1) customers form the private sector and private households, (2)
customers from the public sector, (3) suppliers, (4) competitors or firms in the same industry, (5) consultants and consulting engineers, (6) universities and universities of applied sciences, (7) public research institutions, (8) private research institutions, (9) fairs, conferences and exhibitions, (10) scientific journals, (11) associations, (12) patent specifications, (13)
standardization committees or documents, and (14) crowdsourcing i.e. the public at large. All external search channels were rated on a four-point interval scale (0 = not used, 1 = low importance, 2 = mid-level importance, and 3 = great importance). However, to better
binary value (0 = not used; 1 = used) (Leiponen & Helfat, 2010; Laursen & Salter, 2006). These binary values were subsequently added to create an aggregate measure representing external search breadth. The internal consistency of this measure is considered high (Cronbach’s alpha coefficient = .956). External search breadth ranges from 0 when no
external search channels were used to 14 when all sources were used, larger numbers indicate a higher degree of openness towards a firm’s external environment.
External Search Depth
External search depth represents the degree to which a firm draws intensively from a particular external search channel as a source of innovation related knowledge. Similar to external breadth, external search depth can be constructed using the same 14 external sources mentioned earlier (Laursen and Salter, 2006). However, to distinguish between external channels that are subject to intensive knowledge sourcing and those that are not, each of the external search channels was recoded into a binary value. More specifically, search channels that received a score that indicated that the specific source was of high importance for a recent innovation project (i.e. provided a score = 3) were coded as 1, whereas all other ratings (i.e. provided a score < 3) were scored 0. The sum of these binary values is indicative of a firm’s external search depth strategy. The internal consistency of external search depth is rather low (Cronbach’s alpha coefficient = .675). Again, higher numbers of external search depth are positively associated with a firm’s degree of openness.
Digital Analytics Capabilities (DAC)
low usage, 2 = medium usage, and 3 = high usage). DAC was the independent variable in this study.
Digital Communication Capability (DCC)
Digital communication capabilities refer to the application of tools that facilitate the exchange of information. It is measured on the same four-point interval scale as DAC (i.e. 0 = no usage, 1 = low usage, 2 = medium usage, and 3 = high usage). DCC is the moderating variable in this study. A product term has been computed for both technologies, DAC and DCC (i.e. DACxDCC), to investigate their joint effect on a firm’s external search strategy.
Control Variable
To control for confounding effects, several variables from the 2016 survey that may influence external search breadth and depth were introduced. More specifically, as the degree to which firms have carried out innovation activities in the last three years may be indicative for a firm’s current innovation strategy, a dummy variable distinguishing innovating
companies from non-innovating companies (innovator = 0; non-innovator = 1) was added. Furthermore, larger firms are likely to have greater market power, more resources, or an advantage pertaining to economies of scale and scope that incentivizes firms to innovate (Cassiman & Veugelers, 2006; Jansen, Van Den Bosch, & Volberda, 2005). On the other hand, larger firms often lack the flexibility to integrate newly obtained external knowledge (Jansen, Van Den Bosch, & Volberda, 2005). In contrast, while often missing important resources (Radicic & Pugh, 2017), smaller firms may be more innovative and less
(2) a firm’s total sales was added as a control variable. In the MIP survey, the original sales variable was multiplied by a random number that remained the same across waves to warrant anonymity.
Moreover, because of the historical divide between East and West Germany, and considering that significant firm differences may pertain between regions (Buss & Peukert, 2015), a dummy variable was included (Western Germany = 0; Eastern Germany = 1) to control for potential regional effects (Van de Vrande, Vanhaverbeke, & Duysters, 2011). Lastly, Joshi, Chi, Datta and Han (2010) point out that different industries may affect firm’s innovation activities in distinct ways. Hence, following prior research, an industry control has been included that accounts for the impact of sectorial fixed-effects (e.g. Gómez et al., 2017). Based on the classification used in the ZEW-Indicator reports, the industry control consists of 21 categories, each representing an economic sector. Table 1 provides a distribution overview of firms falling in each sector. Please see Table 2 for the descriptive statistics and correlations of the variables included in this paper.
Table 1
Frequencies – Industry Classification
Industry Classification Frequency Percentage
(10) Machinery 99 4.21
(11) Retail/Automobile 49 2.09
(12) Furniture/Toys/Medical technology/Maintenance 143 6.09
(13) Energy/Water 185 7.88
(14) Wholesale 100 4.26
(15) Transport equipment/Postal Service 191 8.13
(16) Media services 117 4.98
(17) IT/Telecommunications 85 3.62
(18) Banking/Insurance 80 3.41
(19) Technical services/R&D services 186 7.92
(20) Consulting/Advertisement 143 6.09
(21) Firm-related services 108 4.60
Total 2349 100.00
Table 2
Descriptive Statistics and Correlation
Obs. Mean SD (1) (2) (3) (4) (5) (6) (7) (8) (9) (1) Breadth 2349 3.969 4.842 1.000 (2) Depth 2349 .616 1.218 .588*** 1.000 (3) DAC 2188 .441 .739 .268*** .204*** 1.000 (4) DCC 2216 .820 .919 .323*** .270*** .472*** 1.000 (5) Innovator 2349 .539 .499 -.561*** -.370*** -.232*** -.315*** 1.000 (6) Size -Employees 2340 3.234 1.542 .279*** .179*** .197*** .197*** -.272*** 1.000 (7) Size - Sales 2345 1.791 1.825 .290*** .180*** .210*** .218*** -.284*** .929*** 1.000 (8) Region 2349 .330 .470 .001 .028 -.020 -.047* .040 -.116*** -.177*** 1.000 (9) Industry 2349 11.902 5.907 -.078*** -.030 .075*** .098*** .096*** -.155*** -.219*** .049* 1.000
Analysis Strategy
An initial inspection of the data indicated that both dependent variables were non-normally distributed (see Figure 2- Figure 4 for a graphical inspection; in both cases the Skewness-Kurtosis test of normality and Shapiro-Francia test of normality with a log
transformation were significant (p < .001)). Moreover, because the variables were scored on a non-negative discrete level, count data models appear to be a suitable method for analysis (Hilbe, 2014). Typically, it is assumed that count data models follow a Poisson distribution (as cited in Gomez, Salazar, & Vargas, 2017). However, for Poisson regression to be
applicable, the assumption of equidispersion must hold (Cameron & Trivedi, 2010). In other words, checking the equidispersion assumption is instrumental to assess the adequacy of the Poisson regression as opposed to other count models such as negative binominal regression (Greene, 2008). Since there is significant evidence of overdispersion (for the full model with external breadth (G2 = 3769.39, p < .001) and external depth (G2 = 249.57, p < .001) as the dependent variable), an alternative count model other than Poisson should be considered (Hilbe, 2014).
Furthermore, because firms traditionally relied on a closed innovation model in which all R&D was conducted internally (Chesbrough, 2003), it is assumed that a relatively large number of firms does not engage in external knowledge search. Therefore, it is not surprising, that the number of zero counts in the data is excessively high (i.e. external search breadth = 51.34%; external search depth = 69.22%).
Consequently, there are two types of zeros embedded in the present data those related to firms who never engage in external knowledge search (i.e. the “Always-0 Group”) and those who are inclined to search externally but have not done so yet (i.e. the “Not Always-0 Group”) (Long & Freese, 2014). Conceptually, this is plausible because firms may perceive their external environment differently with regards to its innovation knowledge sourcing potential. Firms that are convinced of the closed innovation system or are skeptical of outside ideas (e.g. Not-Invented-Here (NIH) syndrome) may continue to rely on internal R&D (Popa et al., 2017), while others may see the potential of open innovation but may simply lack the appropriate resources to be externally engaged.
To determine which (zero-inflated) count model has the best fit (i.e. Poisson (PRM), negative binominal (NBREG), zero-inflated Poisson (ZIP) or zero-inflated negative
binominal (ZINB)), one can plot and compare the predicted probabilities of each model, and use statistical measures such as the Akaike’s information criterion (AIC) (Akaike 1973) and Schwarz’s Bayesian information criterion (BIC) (Schwarz 1978) to guide model choice (Long & Freese, 2014; Cameron & Trivedi, 2010). Generally, the lower the values for AIC and BIC the better the model fits the data. While both approaches are commonly used (Hilbe, 2014), they differ in their definition of what is considered a good model fit (Joshi et al., 2010). According to Joshi et al. (2010), BIC would point towards the model that would be the closest to a “true” model, given that one of the models tested is actually true. AIC on the other hand, relies on expected predictions of future data, not a hypothetical true model, to assess model fitness. To facilitate the model comparison, the user-written Stata command
countfit was used (Long & Freese, 2014). Based on the results of this model comparison (see Table 3), it appears that the zero-inflated negative binominal regression model has the lowest
pertaining to external search depth. In fact, based on the BIC it seems that the negative binominal regression has better model fit than the zero-inflated version (diff.BIC=
BICNBREG(4258.188) - BICZINB(4323.845) = -65.657). Therefore, as neither of the prediction criteria are superior (Greene, 2008), it is not the aim of this study to identify a true model (Joshi et al. 2010), and BIC tends to penalize complex models more heavily (Greene, 2008), the AIC suggestion for the zero-inflated negative binominal regression model was followed. This is also more consistent with Figure 6 which visualizes the observed against expected count probabilities for each count model, and the Vuong test that shows a preference for ZINB over NBREG (V = 6.52, p < .001) (Vuong, 1989; Hilbe, 2014).
For answering the focal hypotheses, it is less relevant whether a firm falls into the Always-0 or Not-Always-0 group given certain firm characteristics (i.e. the
independent/moderator/control variables). In other words, the focus of this paper is to understand the effect of DAC, DCC, and their joint effect (DACxDCC) on external
knowledge search, given that firms are inclined to search externally (Not-Always-0 group). Hence the logit part of the ZINB is redundant and will not be presented. Moreover, because of overdispersion, standard errors were adjusted using robust variance estimators (e.g. Hilbe, 2014; Hilbe, 2011). Furthermore, following Hilbe’s (2014) suggestion, all count regression coefficients are exponentiated, resulting in incidence rate ratios (IRRs), which are expressed as percentages to facilitate interpretation. For robustness2 Ordinary Least Squares (OLS) regression was run. All data was analyzed using Stata/SE 15.0.
Table 3
Tests and Fit Statistics
DV (1) AIC/BIC (2) AIC/BIC Diff. AIC Diff. BIC
LR G2/Vuong
Breadth PRM 13342.611 (13501.617) NBREG 9575.250 (9739.935) 3767.361 (3761.681) 3769.361*** PRM ZIP 7940.330 (8258.342) 5402.281 (5243.275) 24.02*** PRM ZINB 7827.221 (8150.912) 5515.381 (5350.706) NBREG 9575.250 (9739.935) ZIP 7940.330 (8258.342) 1634.921 (1481.593) NBREG ZINB 7827.221 (8150.912) 1748.030 (1589.024) 26.55*** ZIP 7940.330 (8258.342) ZINB 7827.22 (8150.912) 113.109 (107.430) 115.109*** Depth PRM 4341.073 (4500.079) NBREG 4093.504 (4258.188) 247.569 (241.891) 249.569*** PRM ZIP 4029.040 (4347.052) 312.033 (153.027) 7.77*** PRM ZINB 4000.154 (4323.845) 340.919 (176.234) NBREG 4093.504 (4258.188) ZIP 4029.040 (4347.052) 64.463 (-88.864) NBREG ZINB 4000.154 (4323.845) 93.349 (-65.657) 6.52*** ZIP 4029.040 (4347.052) ZINB 4000.154 (4323.845) 28.886 (23.207) 30.886***
Note. BIC in parentheses. Likelihood-ratio (LR) test (= G2) of alpha = 0, when sig. there is evidence for overdispersion. Vuong test provides insights regarding the preference for the zero-inflated model against its noninflated standard model equivalent (Vuong, 1989; Hilbe, 2014). *** p < .001, ** p < .01, * p < .05
To investigate the degree to which digital analytics capabilities and digital
communication capabilities influence a firm’s external knowledge search, a zero-inflated negative binominal regression was used. The analysis was applied in a step-wise approach. More specifically, first a control model was estimated (model 1), followed by a model that includes the independent variable (model 2), the moderator (model 3) and their interaction term (model 4) in consecutive order. Table 4 shows the regression results for breadth, while
Table 5 shows the results for depth, respectively.
Hypothesis Testing - External Knowledge Breadth
While for breadth all models were statistically significant (model 1: Wald c2(24) = 57.20, p < .001; Pseudo R2 = .132; model 2: Wald c2(25) = 61.47, p < .001; Pseudo R2 = .139; model 3: Wald c2(26) = 64.54, p < .001; Pseudo R2 = .144; model 4: Wald c2(27) = 64.68, p < .001; Pseudo R2 = .147 ), model 4 appeared to have the best fit in terms of AIC. However, in terms of BIC, model 3 is preferred. Due to the inconsistency of AIC and BIC, model 4 will be applied given the arguments mentioned earlier.
Table 4
Results- ZINB - External Knowledge Breadth
Variables Model 1 Model 2 Model 3 Model 4
Location (West†) .070* (.030) .071* (.031) .075* (.032) .075* (.032) Industry (Mining†) Food/Tobacco -.263* (.104) -.236* (.104) -.234* (.102) -.908* (.385) Textiles -.166 (.105) -.179 (.105) -.184 (.104) -1.143** (.399) Wood/Paper -.171 (.110) -.157 (.111) -.160 (.109) -.960* (.443) Chemicals -.075 (.091) -.040 (.091) -.055 (.089) -1.922*** (.468) Plastics -.133 (.096) -.166 (.097) -.177 (.096) -.983* (.426) Glass/Ceramics .017 (.114) .029 (.116) .023 (.117) -1.281** (.410) Metals -.166 (.088) -.171 (.089) -.173* (.086) -.647 (.351) Electrical equipment -.015 (.080) -.027 (.082) -.049 (.080) -1.762*** (.344) Machinery -.135 (.086) -.117 (.086) -.133 (.084) -1.567*** (.397) Retail/Automobile .011 (.093) .010 (.093) -.009 (.092) -.715 (.446) Furniture/Toys/Medical technology/Maintenance -.127 (.088) -.131 (.089) -.146 (.087) -1.026** (.345) Energy/Water -.032 (.095) -.047 (.100) -.049 (.098) .249 (.341) Wholesale -.142 (.108) -.193 (.107) -.225* (.108) -.355 (.384) Transport equipment/Postal Service -.214 (.110) -.254* (.113) -.263* (.112) -.019 (.371) Media services -.134 (.101) -.163 (.104) -.177 (.103) -.506 (.354) IT/Telecommunications -.063 (.092) -.163 (.092) -.203* (.091) -1.112** (.391) Banking/Insurance -.206 (.108) -.232* (.110) -.241* (.106) -.729 (.397) Technical services/R&D services .090
Consulting/Advertisement -.081 (.103) -.133 (.103) -.157 (.102) -.777* (.355) Firm-related services -.253* (.125) -.285* (.128) -.292* (.127) -.659 (.384) Constant 2.079*** (.087) 2.027*** (.089) 2.006*** (.088) 2.004*** (.089) Observation 2337 2177 2162 2162 Log Likelihood -4180.473 -3906.032 -3861.266 -3856.610 Wald chi-square 124.95*** 150.72*** 148.08*** 148.09*** Pseudo (McFadden) R2 .120 .126 .128 .129 AIC 8462.945 7918.064 7832.532 7827.221 BIC 8756.533 8219.406 8144.865 8150.912
Note. Standard error in parentheses. *** p < .001, ** p < .01, * p < .05; † reference category
.595) nor the amount of sales (IRR = (exp(-.027)) = 1.028, 95%CI = .973, 1.086, p = .331) seem to have a significant impact on the expected number of external partners. Hence, with respect to external search breadth, there appears to be no difference with regard to firm size and the expected number of diverse external search channels a firm utilizes. For non-innovators however, the expected number of external partners seems to be 9.9% (IRR = (exp(-.105)) = .901, 95%CI = .832, .975, p = .01) smaller than for innovators, keeping everything else constant. In other words, innovating firms are more likely to attract a diverse set of external search channels from which they can potentially extract innovation related knowledge. Moreover, firms that are situated in the eastern part of Germany appear to have a 7.7% (IRR = (exp(.075)) = 1.077, 95%CI = 1.013, 1.146, p = .018) higher expected number of external knowledge partners than firms located in the western parts of Germany.
Hypothesis Testing - External Knowledge Depth
Following the step-wise approach already mentioned, all models pertaining to the dependent variable depth were statistically significant (model 1: Wald c2(24) = 57.20, p < .001; Pseudo R2 = .132; model 2: Wald c2(25) = 61.47, p < .001; Pseudo R2 = .139; model 3: Wald c2(26) = 64.54, p < .001; Pseudo R2 = .144; model 4: Wald c2(27) = 64.68, p < .001; Pseudo R2 = .147 ). The predicted full model (i.e. model 4) appears to have the best fit in terms of AIC and BIC and will thus be further elaborated.
Table 5
Coefficients - ZINB - External Knowledge Depth
Variables Model 1 Model 2 Model 3 Model 4
Media services .179 (.431) .205 (.461 .210 (.447) .233 (.448) IT/Telecommunications .619 (.428) .447 (.443) .364 (.433) .599 (.458) Banking/Insurance -.019 (.413) -.072 (.441) -.057 (.424) .233 (.448) Technical services/R&D services .702
(.398) .589 (.430) .556 (.417) .797 (.439) Consulting/Advertisement 1.069* (.418) 1.021* (.433) .937* (.421) 1.152** (.443) Firm-related services .425 (.461) .355 (.510) .386 (.504) .606 (.495) Constant -.467 (.410) -.564 (.432) -.633 (.423) .0703 (.424) Observation 2337 2177 2162 2162 Log Likelihood -2118.594 -1977.773 -1951.021 -1943.077 Wald chi-square 57.20*** 61.47*** 64.54*** 64.68*** Pseudo (McFadden) R2 .132 .139 .144 .147 AIC 4339.188 4061.545 4012.043 4000.154 BIC 4632.775 4362.888 4324.376 4323.845
Note. Standard error in parentheses. *** p < .001, ** p < .01, * p < .05; † reference category
The results suggest that the inclusion of the interaction effect from model 3 to model 4 significantly impacted the degree to which DAC and DCC are associated with external search depth. More specifically, while in model 3 both technologies were significant (i.e. DAC (IRR = (exp(.124)) = 1.13, 95%CI = 1.000, 1.282, p = .049)); DCC (IRR = (exp(.111)) = 1.118, 95%CI = 1.002, 1.247, p = .047)), these significant main effects disappeared in model 4 (both technologies p > .05). Hence, hypothesis 1b and hypothesis 2b are not supported. Furthermore, the interaction term in model 4 was found to be highly significant (IRR = (exp(.192)) = 1.212, 95%CI = .1.007, 1.369, p = .002). This points to a fully moderated effect (see Figure 1 for interaction plot).
Figure 1. Interaction Plot Between DAC and DCC on External Search Depth
To be more precise, the effect of both technologies on depth is only jointly significant. Thus, hypothesis 3b is supported. Holding everything else constant, the implementation of both DAC and DCC can together contribute to a 21.2% higher expected number of deeply oriented external search channels among firms that are inclined to engage in external knowledge search. In line with the findings from breadth, for depth, being an non-innovator is associated with a significantly lower expected number of deep partnerships (IRR = (exp(-.448)) = .639, 95%CI = .446, .916, p = .015). In other words, firms that are innovate are 36.1% more likely to have a higher expected number of deep external knowledge sources that are involved in their innovation activities. Apart from some industries (see Table 5), all other control variables appeared to have no significant effect on external search depth.
In the present paper, the effect of digital analytics capabilities, digital communication capabilities, and their interaction on external search breadth and depth were examined. While it was expected that both external knowledge breadth and external knowledge depth would be positively affected by IT, this relationship was more diverse than anticipated. For external knowledge breadth, DAC and DCC both appear to positively affect the number of diverse search channels (i.e. external search breadth). For external search depth, the results point to a different relationship, however. To facilitate a firm’s external search depth strategy, the main effects of both technologies were not significant. Instead, only the joined effect significantly predicted the expected number of intensive relationships. These nuanced findings will be discussed by firstly emphasizing the interaction of DAC and DCC on external search depth, followed by explaining the individual effects on external search breadth.
External Search Depth and the Combined Effect of DAC and DCC
search strategies but to absorb tacit, or otherwise complex knowledge, focal firms likely need to establish deep external relationships (Ferreras-Méndez, 2015). Deep relationships are characterized by intensive interactions between partners and may not only create trust (Fey & Birkinshaw, 2005), but also enhance knowledge sharing, transformation of knowledge, and its exploitation (Teirlinck & Spithoven, 2013). One way through which these deep
relationships are fostered are through frequent communication (Rai, Pavlou, Im, & Du, 2012). Hence, by having digital communication capabilities this effect may be enhanced even further. When the relationship is strong, the context is easier to interpret, and communication will create better awareness to issues that are strategically worth attending to (George & Zhou, 2001; Ocasio, 1997). The information that has come to the focus of decision maker’s attention is now better assimilated given that it receives the adequate attentional resources with the appropriate context to transfer it from one person to another (Gomez, Salazar & Vargas, 2017; Salge et al., 2013). The benefits of analyzing this contextualized information are thus twofold. Firstly, decision makers have access to a richer dataset to draw on.
Secondly, data that has previously been ill understood becomes easier to interpret. In other words, gaining a better understanding of the context may reduce the costs associated with searching deeply because the awareness of decision makers has been channeled appropriately (Flor, Cooper, & Oltra, 2018).
According to Rai et al. (2012), communication may help facilitate contextual
capabilities. When jointly present, the data transformation and exploitation that is enabled by DAC will be better aligned with the context and may thus be better tailored to a focal firm’s strategic needs. Consequently, there is less strain on a person’s attentional capacity pertaining to a specific external partner, given attention was better channeled through greater awareness, prioritization and exploited. Hence, attentional slack resources may be utilized to the
establishment of additional deep relationships. Therefore, the present results regarding the interaction effect between DAC and DCC appears plausible.
Moreover, while the complexity of the information derived from a firm’s external search depth suggest the need for the jointed implementation of DAC and DCC, it also explains why these technologies are not significant when considered in isolation. In the absence of communication capabilities, for example, the information remains rich, but cannot be appropriately channeled and analyzed using DAC. While there is additional data from the depth of the interactions, Bumblauskas et al. (2017, p. 8) argue that this magnitude of
information can stifle action and lead to “paralysis by analysis”. Therefore, without communication channels that guide attention towards important data, or that helps putting data into its context, the bare implementation of digital analytics capabilities may
overstimulate decision-makers attentional capacity and not only be ineffective, but actually detrimental. While barely not significant, this would explain the strong negative effect of DAC on external knowledge depth when DCC is not facilitating these deep external
External Search Breadth and the Individual Main Effects of DAC and DCC
Firms that draw on many different sources (i.e. external search breadth), likely do not have the capacity to establish mutual norms and a reciprocal relationship with their
collaboration partners. The interactions are likely short-lived and instrumental to gain novel insights. Therefore, the information that is accessed from a broad number of external search channels tends to be rather explicit (Ye & Kankanhalli, 2011). Since IT systems are better able to manage explicit data, as stated previously, both DCC and DAC may be effective in channeling a focal firm’s attention towards its strategic objectives. When DAC and DCC are both implemented to cope with a firm’s external search breadth strategy, this does not additionally help recover attentional capacity presumably because the additional attention gained from the data management would be offset by the need to coordinate multiple
technologies. The more these technologies are used in managing external search breadth, the more frequent the interactions may become and the lines between external search breadth and depth become blurry. Hence, when screening the external environment broadly in order to identify novel knowledge, each technology may be sufficiently able to extract the explicit value embedded in the data and may thus allow decision makers to direct their attention to additional sources.
Theoretical Implications
until now been missing. This is an important gap that needed to be filled considering that different technologies may affect a firm’s strategy differently.
To operationalize the technologies under investigation in a way that reflects their empowering effect on external search, Rai et al.’s (2012) definition of IT capabilities has be utilized. Consequently, big data analytics and software-based communication were defined in terms of their digital analytics and digital communication capability respectively. These constructs augment recent investigation on specific IT capabilities (Dubey et al. 2019). While big data has received considerable attention in recent years, theoretical explanations on how it contributes to business value remain scarce. Therefore, the present study complements practitioner’s accounts by putting big data into theoretical perspective with open innovation and external search in particular. While big data may have been described in the context of open innovation before, it was mostly investigated with regard to specific dyads such as customers (Trabucchi et al., 2018). In this paper, the focus shifts from one specific dyad to the entire external search strategy.
Furthermore, while IT capabilities have largely been studied through the lens of absorptive capacity (e.g. Roberts et al., 2012; Joshi et al., 2010) utilizing the attention-based view of the firm is a relatively novel approach (Ocasio, 1997). More specifically, this study contributes to the literature by linking specific IT capabilities to mechanisms that would likely reduce attentional strain. When there is less demand on a decision makers attentional capacity, there are arguably attentional slack resources that decision makers can draw on to expand their external search depth and external search breadth strategy.
Practical Implications
& Weill, 2007). For example, when contemplating the utilization of DAC in an intensive external collaboration context, firms might want to consider the adoption of a complementary technology such as DCC. Otherwise, the information processing may not be utilized fully, and the costs associated with establishing DAC may not be worthwhile. More pragmatically, according to Müller, Fay and Vom Brocke (2018), investing in recent analytics tools such as IBM’s PureData System for Analytics, or a comparable Cloudera Hadoop cluster will amount to a total three-year cost of $39 million and $50 million respectively. Yet in this paper, the results show that implementing these kinds of analytics system by themselves will likely have detrimental effects on the number of external partners with whom the focal firm intends to collaborate intensively. Hence, apart from the financial costs, firms will also incur negative relational costs.
While it is likely that firms have already established some sort of digital
communication channels that could complement a firm’s big data analytics usage, these may need to be upgraded to improve a firm’s digital communication capabilities. Managers need to realize that engaging in external knowledge depth requires building trust and creating a common context which will then facilitate the flow of knowledge. Establishing and utilizing digital communication tools will strengthen the ties between the focal firm and its external sources and thereby help firms uncover the potential value embedded in the external
channel’s knowledge base. This appears to particularly helpful when following a strategy that is focused on intensive external collaborations. On the other hand, when firms seek to utilize a broadly oriented search strategy, both digital analytics capabilities and digital
Limitations and future research
There are a number of limitations that need to be addressed. Firstly, while specific technologies have been selected to explore how IT may facilitate a firm’s external knowledge search, the measures pertaining to these technologies are based on self-reported usage on a 4-point interval scale. However, the actual IT systems under investigation may vary
tremendously and thus might require a more elaborate measure. This is particularly true given that the ratings are relatively ambiguous and can mean different things to different people.
Secondly, the dependent measures remain relatively unchanged since Laursen and Salter (2006) first referred to the concept of external search depth and breadth. Yet, these measures are not without criticism considering that both variables are aggregated terms representing the involvement of certain search channels without actually measuring their idiosyncratic impact on, for example, coordination effort or attentional resources. In the present paper the external search depth scale also appeared to have a relatively low internal consistency and removing a specific item would further impair it. Therefore, it is likely that external search depth would better be represented by a multifactorial construct rather than an aggregated term. An initial attempt to distinguish between vertical, horizontal and societal relationships has been proposed by Dong and Netten (2017) but more research is needed.
Thirdly, the data was drawn from a German firm population which means that it is not necessarily generalizable to other firms in other countries. According to the 2019 European innovation scoreboard (European Commission, 2019), a comparative measure of innovation performance in the EU, German firms score above EU average in terms of innovation performance as measured by 27 innovation indicators. This suggests that comparison
Lastly, while conceptually and methodologically the present analysis appeared suitable, future research should consider testing the effects of specific IT capabilities on the various external search strategies using different model specifications.
Conclusion
This paper makes a unique contribution by showing how different information
technologies may be considered antecedents of external knowledge search depth and breadth. Therefore, unlike prior research, a more nuanced perspective of IT’s enabling effect on external search strategy is presented. More specifically, by drawing from the attention-based view of the firm, the results show that digital analytics capabilities and digital communication capabilities facilitate the channeling of attention which in turn enables firms to collaborate with an increasing number of different external sources. However, the impact of these digital capabilities differs for each external search strategy respectively. Whereas external
knowledge breadth is affected by both DCA and DCC uniquely, for external knowledge depth only their interaction appears to be effective. This finding suggests that knowledge may be more complex when derived from external search depth, rather than external search
Acknowledgement
References
Akaike, H. (1973). Maximum likelihood identification of Gaussian autoregressive moving average models. Biometrika, 60(2), 255-265.
Alavi, M., & Leidner, D. E. (2001). Review: Knowledge Management and Knowledge Management Systems: Conceptual Foundations and Research Issues. MIS Quarterly, 25(1), 107–136.
Bharadwaj, A. S. (2000). A resource-based perspective on information technology capability and firm performance: an empirical investigation. MIS Quarterly, 24(1), 169–196. Buss, P., & Peukert, C. (2015). R&D outsourcing and intellectual property infringement.
Research Policy, 44(4), 977-989.
Cameron, A., & Trivedi, P. (2010). Microeconometrics Using Stata (Kindle. ed.). College Station, TX: Stata Press.
Cassiman, B., & Veugelers, R. (2006). In search of complementarity in innovation strategy: Internal R&D and external knowledge acquisition. Management Science, 52(1), 68-82.
Černe, M., Jaklič, M., & Škerlavaj, M. (2013). Management innovation in focus: The role of knowledge exchange, organizational size, and IT system development and utilization.
European Management Review, 10(3), 153-166.
Chesbrough, H. W. (2003). The Era of Open Innovation. MIT Sloan Management Review,
44(3), 35-41.
Chesbrough, H., Vanhaverbeke, W., & West, J. (2006). Open innovation: Researching a new
paradigm. Oxford: Oxford University Press.
Cohen, W. M., & Levinthal, D. A. (1990). Absorptive capacity: A new perspective on learning and innovation. Administrative Science Quarterly, 35(1), 128-152.
data analytics value in firms. Journal of Business Research, 97, 160-173.
Del Vecchio, P., Di Minin, A., Petruzzelli, A. M., Panniello, U., & Pirri, S. (2018). Big data for open innovation in SMEs and large corporations: Trends, opportunities, and challenges. Creativity and Innovation Management, 27(1), 6-22.
De Mauro, A., Greco, M., & Grimaldi, M. (2016). A formal definition of Big Data based on its essential features. Library Review, 65(3), 122-135.
Dong, J. Q., & Netten, J. (2017). Information technology and external search in the open innovation age: New findings from Germany. Technological Forecasting and Social
Change, 120, 223-231.
Dubey, R., Gunasekaran, A., Childe, S. J., Roubaud, D., Wamba, S. F., Giannakis, M., & Foropon, C. (2019). Big data analytics and organizational culture as complements to swift trust and collaborative performance in the humanitarian supply chain.
International Journal of Production Economics, 210, 120-136.
European Commission. (2019). European innovation scoreboard. Retrieved June 22, 2019, from https://ec.europa.eu/growth/industry/innovation/facts-figures/scoreboards_en Ferreras-Méndez, J. L., Newell, S., Fernández-Mesa, A., & Alegre, J. (2015). Depth and
breadth of external knowledge search and performance: The mediating role of absorptive capacity. Industrial Marketing Management, 47, 86-97
Ferreras-Méndez, J. L., Fernández-Mesa, A., & Alegre, J. (2016). The relationship between knowledge search strategies and absorptive capacity: A deeper look. Technovation,
54, 48-61
Fey, C. F., & Birkinshaw, J. (2005). External sources of knowledge, governance mode, and R&D performance. Journal of Management, 31(4), 597-621.
capacity and radical innovation in high-technology firms. European Management
Journal, 36(2), 183-194.
Gandomi, A., & Haider, M. (2015). Beyond the hype: Big data concepts, methods, and analytics. International Journal of Information Management, 35(2), 137-144. George, J. M., & Zhou, J. (2001). When openness to experience and conscientiousness are
related to creative behavior: an interactional approach. Journal of Applied
Psychology, 86(3), 513-524.
Gómez, J., Salazar, I., & Vargas, P. (2017). Does information technology improve open innovation performance? An examination of manufacturers in Spain. Information
Systems Research, 28(3), 661-675.
Greene, W. (2008). Econometric analysis (6th ed.). Boston, MA: Pearson.
Gupta, M., & George, J. F. (2016). Toward the development of a big data analytics capability. Information & Management, 53(8), 1049-1064.
Hilbe, J. (2011). Negative binomial regression (2nd ed.). Cambridge, UK: Cambridge University Press.
Hilbe, J. (2014). Modeling count data. New York, NY: Cambridge University Press Hitt, M. A., Hoskisson, R. E., Johnson, R. A., & Moesel, D. D. (1996). The market for
corporate control and firm innovation. Academy of Management Journal, 39(5), 1084-1119.
Horbach, J. (2008). Determinants of environmental innovation - New evidence from German panel data sources. Research Policy, 37(1), 163-173.
Huizingh, E. K. (2011). Open innovation: State of the art and future perspectives. Technovation, 31(1), 2-9.
realized absorptive capacity: How do organizational antecedents matter?. Academy of
Management Journal, 48(6), 999-1015.
Johannessen, J. A., Olaisen, J., & Olsen, B. (2001). Mismanagement of tacit knowledge: the importance of tacit knowledge, the danger of information technology, and what to do about it. International Journal of Information Management, 21(1), 3-20.
Joshi, K. D., Chi, L., Datta, A., & Han, S. (2010). Changing the competitive landscape: Continuous innovation through IT-enabled knowledge capabilities. Information
Systems Research, 21(3), 472-495.
Kleis, L., Chwelos, P., Ramirez, R. V., & Cockburn, I. (2012). Information technology and intangible output: The impact of IT investment on innovation productivity.
Information Systems Research, 23(1), 42-59.
Koput, K. W. (1997). A chaotic model of innovative search: some answers, many questions. Organization Science, 8(5), 528-542.
Laursen, K., & Salter, A. (2006). Open for innovation: the role of openness in explaining innovation performance among UK manufacturing firms. Strategic management
journal, 27(2), 131-150.
Leiponen, A., & Helfat, C. E. (2010). Innovation objectives, knowledge sources, and the benefits of breadth. Strategic Management Journal, 31(2), 224-236.
Markus, M. L. (2015). New games, new rules, new scoreboards: the potential consequences of big data. Journal of Information Technology, 30(1), 58-59.
McAfee, A., Brynjolfsson, E., Davenport, T. H., Patil, D. J., & Barton, D. (2012). Big data: the management revolution. Harvard business review, 90(10), 60-68.
Mikalef, P., Boura, M., Lekakos, G., & Krogstie, J. (2019). Big data analytics and firm performance: Findings from a mixed-method approach. Journal of Business Research,
Müller, O., Fay, M., & vom Brocke, J. (2018). The effect of big data and analytics on firm performance: An econometric analysis considering industry characteristics. Journal of
Management Information Systems, 35(2), 488-509.
Nonaka, I. (1994). A dynamic theory of organizational knowledge creation. Organization
science, 5(1), 14-37.
Ocasio, W. (1997). Towards an attention‐based view of the firm. Strategic Management
Journal, 18(S1), 187-206.
Organisation for Economic Co-operation and Development., & Statistical Office of the European Communities. (2005). Oslo manual: Guidelines for collecting and
interpreting innovation data. Paris: Organisation for Economic Co-operation and
Development.
Popa, S., Soto-Acosta, P., & Martinez-Conesa, I. (2017). Antecedents, moderators, and outcomes of innovation climate and open innovation: An empirical study in SMEs.
Technological Forecasting and Social Change, 118, 134-142.
Radicic, D., & Pugh, G. (2017). Performance Effects of External Search Strategies in European Small and Medium‐Sized Enterprises. Journal of Small Business
Management, 55, 76-114.
Rai, A., Pavlou, P. A., Im, G., & Du, S. (2012). Interfirm IT capability profiles and
communications for cocreating relational value: evidence from the logistics industry.
MIS Quarterly, 36(1), 233-262.
Rammer, Peters, & Peters (2013), Innovation Panel Surveys in Germany, in: Gault, F.,
Handbook of Innovation Indicators and Measurement, Cheltenham, UK and
Northampton, MA, USA., 135-177
really matter? A contingency study of health‐care innovation projects. Journal of
Product Innovation Management, 30(4), 659-676.
Schwarz, G. (1978). Estimating the dimension of a model. The Annals of Statistics, 6(2), 461- 464.
Teirlinck, P., & Spithoven, A. (2013). Research collaboration and R&D outsourcing: Different R&D personnel requirements in SMEs. Technovation, 33(4-5), 142-153. Tomlinson, P. R. (2010). Co-operative ties and innovation: Some new evidence for UK
manufacturing. Research Policy, 39(6), 762-775.
Trabucchi, D., Buganza, T., Dell'Era, C., & Pellizzoni, E. (2018). Exploring the inbound and outbound strategies enabled by user generated big data: Evidence from leading
smartphone applications. Creativity and Innovation Management, 27(1), 42-55. Tsai, W. (2002). Social Structure of “Coopetition” Within a Multiunit Organization:
Coordination, Competition, and Intraorganizational Knowledge Sharing.
Organization Science, 13(2), 179-190.
Van de Vrande, V., Vanhaverbeke, W., & Duysters, G. (2011). Technology in‐sourcing and the creation of pioneering technologies. Journal of Product Innovation Management,
28(6), 974-987.
Vuong, Q. H. (1989). Likelihood Ratio Tests for Model Selection and Non-Nested Hypotheses. Econometrica, 57(2), 307.
Wadhwa, A., Bodas Freitas, I. M., & Sarkar, M. B. (2017). The paradox of openness and value protection strategies: Effect of extramural R&D on innovative performance.
Organization Science, 28(5), 873-893.
Ye, H. J., & Kankanhalli, A. (2011). Leveraging Structural Holes For Innovation: The Moderating Effects Of IT-Enabled Absorptive Capacity. PACIS. 1-11.
market knowledge acquisition, and internal knowledge sharing. Strategic
Appendix
Figure 2. Normality Inspection of Dependent Variables - Histogram
Figure 3. Normality Inspection of Breadth Residuals - Histogram, P-P Plot, Q-Q Plot
Figure 5. Count Model Comparison - External Search Breadth
Figure 6. Count Model Comparison - External Search Depth
-. 1 -. 0 5 0 .05 .1 O b se rve d -Pre d ict e d 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
External Search Breadth
PRM NBRM
ZIP ZINB
Note: positive deviations show underpredictions.
Count Model Comparison - External Search Breadth
-. 1 -. 0 5 0 .05 .1 O b se rve d -Pre d ict e d 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
External Search Depth
PRM NBRM
ZIP ZINB
Note: positive deviations show underpredictions.
