KNOWLEDGE: A CURVILINEAR AND CONTINGENCY
PERSPECTIVE
Jork Netten (S2585545)
University of Groningen, Faculty of Economics & Business Duisenberg Building, Nettelbosje 2
9747 AE Groningen, The Netherlands Supervisor: Dr. J.Q. Dong Second supervisor: Dr. R. van der Eijk
Date: 22 June 2015
Word count: 14.738 (references: 2593 tables: 2277)
Master Thesis BA SIM
Abstract - This study explores the effect of information technology (IT) investment, and the moderating role of R&D investment with respect to external knowledge search on firm level. The basic premise is that IT investment has an inverted U-shaped relation with external knowledge search as IT enhances the ability of decision-makers to receive greater and more qualitative information about processes, actors, environment, and (potential) knowledge sources, resulting in lower transaction costs. However, too extensive IT investment will break the equilibrium between the exchange of tacit and explicit knowledge, relegating tacit knowledge to the background. Furthermore, the effect of IT investment on external knowledge search depends on the knowledge source, where tacit knowledge sources will have a lower equilibrium compared to more explicit knowledge sources. R&D investment shifts managerial attention more to tacit knowledge as the investment is more effective when based on tacit knowledge. However this knowledge is not codified and thus more difficult to transfer by IT. R&D investment will, thus, negatively influence the increased amount of focus on explicit knowledge that follows from the increased investment in IT. Subsequently, R&D investment has an overall negative interaction effect when moderating the relation between IT investment and external knowledge search.
JEL codes: M10, M15, M19
Keywords: External Knowledge Search, IT investment, R&D investment, Interaction,
2
I. INTRODUCTION
In today’s highly changing and competitive markets, firms are more often being seen as generators and transformers of various forms of knowledge (von Tunzelmann, 1995; Pan & Leidner, 2003). Firms that are involved in innovation create and define issues, and sequential search for new knowledge to find answers for these issues (Nonaka, 1994; Nonaka & Takeushi, 1995). External knowledge sources are often needed to complement own cognition and competences in order to develop new processes, products or services (Nonaka & Takeushi, 1995; Chesbrough & Crowther, 2006). The trend for increasing the search for new knowledge is indisputable and on global scale (Hagedoorn, 2002). Following past research (e.g. Vega-Jurado, Gutiérrez-Gracia, & Fernández-de-Lucio, 2009), the terminology external knowledge search (EKS) is used to address this practice. EKS can be defined as
‘problem-solving activities that involve the creation and recombination of technological ideas’ (Katila
and Ahuja, 2002, p.1184). The practices of EKS can been divided into two common practices, EKS breadth and depth (Katila & Ahuja, 2002; Laursen & Salter, 2006). The breadth of EKS are the number of distinct external knowledge sources, such as universities, customers and suppliers. EKS breadth can be defined as “how widely a firm explores new
knowledge” (Katila & Ahuja, 2002, p. 1183) . The depth of EKS is the importance of a
certain knowledge source, defined as “how deeply a firm reuses its existing knowledge” (Katila & Ahuja, 2002, p. 1183).
3
Since the early 90’s, the interest in IT to share knowledge increased exponentially as information systems could facilitate the share of valuable information and enhance firms’ competences (Davenport & Prusak, 1998; Hendriks, 1999; Jarvenpaa & Staple, 2000; Kleis, Chwelos, Ramirez, & Cockburn, 2012; Dong & Yang, 2015). Since the beginning of this century, scholars started to question the abilities and negative sides of IT to transfer tacit knowledge (Johannessen, Olaisen & Olsen, 2001). Other scholars (e.g. Panahi, Watson, & Partridge, 2012) argued that IT does have the abilities to transfer tacit knowledge. Although both schools disagree about the role of IT in the exchange of tacit knowledge, they appear to agree on the matter that the exchange of tacit knowledge through the use of IT is less efficient relatively to explicit knowledge (e.g. Griffith, Sawyer, & Neale, 2003).
Despite comprehensive research on the transfer of explicit and tacit knowledge, the effect of IT investment on the breadth and depth of EKS has never taken place. This study aims to fill these literature gaps by pointing out how IT investment affects the breadth and depth of EKS. This is particular fruitful as improving the effectiveness of EKS could lead to more effective decision-making and an improved firm performance (Camisón & Villar-López, 2014; Chesbrough & Crowther, 2006). In order to realize this, the theoretical lenses of attention-based theory and transaction costs (TCE) are used to theorize the paradox that decision-makers face. Furthermore, the interaction between IT and R&D investments is understudied, while this interaction is very important in the context of innovation (Bardhan et al., 2013). Thus the following research question have been conducted; how does IT
investment affects firms’ EKS breadth and depth?
4
investments into account when doing or applying research, and establishing policies regarding EKS. Therefore firms could perform less than their potential as more hazards arise. All industries available of the Mannheim Innovation Panel survey were included in the sample of 2044 firms; both service and manufacturing industries. The analysis of this study was conducted on firm level. The results of this study show four findings. First, IT investment has an inverted U-shaped relation with the breadth of EKS. Second, the interaction of R&D investment and IT investment lowers the overall positive effects of IT investment on EKS breadth. Third, the findings indicate that IT investment have an inverted U-shaped relation with the depth of EKS and that the inverted U-shaped relation differs per knowledge source. Final, R&D investment has an overall negative interaction effect when moderating the relation between IT investment and EKS depth.
This study has multiple important implications. First, this study shows that IT investment is good to enhance the breadth and depth of EKS, but at some point, the firm will mainly focus on explicit knowledge, relegating tacit knowledge to the background. Scholars should take this into consideration when including IT investment in open innovation studies and managers should take this into account when making investment decisions. Second, the effect of IT investment on distinctive sources are different, where the synergy between IT investment and tacit knowledge sources are less effective compared to explicit knowledge sources. Demonstrating that scholars should take a contingency approach when conducting research on the effect of IT on knowledge sources. Third, the interaction between IT and R&D investments could diminish EKS breadth and depth, making it an important implication when doing research on IT and R&D investments as synergies are tried to be theorized (Bardhan, Krishnan, & Lin, 2013) or applied by managers.
5
breadth of EKS. Followed by the last two hypotheses that focuses on the depth of EKS. After that, the methods used in this study will be explained in the methods section (section IV). In section V, the findings of this study will be presented. Furthermore, this paper will discuss the findings in the discussion (section VI) which will be followed by the implications,
limitations and further research directions. Finally ending with the conclusions (section VII).
II. THEORETICAL FOUNDATION
A. Transaction cost economics (TCE)
The question whether a firm should build competences or use the market to gain certain competences has a long history in the extant literature (e.g. Williamson, 1985; Oxley, 1997; Dyer, 2002; Argyres & Zenger, 2012). Prior studies on this question emphasize the hazards when going into the market to gain certain competences or knowledge to create such competences (e.g. Hennart, 1988; Kogut & Zander, 1992; Teece, 1998; Oxley, 1999; Mayer, & Salomon, 2006). An important hazard in market transactions is being exposed to opportunistic behavior. Information asymmetry is often seen as one of the important issues when conducting a transaction on the market across different fields of study as this could enable opportunistic behavior (Croom, 2001; Katila & Mang, 2003; Botosan, 2006). When information asymmetry is present, one party could make use of their information advantage, and act opportunistically.
6
source of competitive advantage, giving firms incentives to pursue activities to lower this costs (Bajari & Tadelis, 2001).
B. Attention based theory
The main concept of this theory, attention, can be defined as the ability to process differential sources of information simultaneous extracting information useful for certain tasks (Van De Laar, Heskes, & Gielen, 1997; Garcia, Oliveira, Grupen, Wheeler, & Fagg, 2000). Within the attention based theory, firms can be seen as systems that structurally distribute attention (Ocasio, 1997; Ocasio, 2011). Similar to the view of the transaction cost theory, the attention based theory also acknowledges that decision-makers have limited cognitive capacity to make rational decisions (Ocasio, 1997). Additional to the line of reasoning of this view is that the behavior and cognition of decisions-makers are not predictable and are primary based on the context and content in which decisions-makers find themselves in (Ocasio, 1997). The main concept of this view, attention, exists of the encoding, interpreting and focusing of effort and time by firms decision-makers on issues and answers (Ocasio, 1997; Ocasio, 2011). Although the decision-makers do the attending, the context and content of the firms’ environment are crucial for the decision and the distribution of the decisions result throughout the firm (Ocasio, 1997). These decision-makers, which are seen as one of the most important assets of a firm, need to allocate their time and effort on an optimal number of issues and answers to achieve an equilibrium (Ocasio, 1997; Ocasio, 2011).
C. Tacitness of Knowledge and IT
7
Watson, & Partridge, 2013). Explicit knowledge, often referred as codified knowledge, is the knowledge that exists in more symbolic or written form, which takes far less attention to exchange (Lee, 2001). Tacit knowledge can often only be exchanged when it can be revealed through its application (Grant, 1996). Thus, when knowledge is not explicit the exchange of knowledge will often be slow, uncertain and costly (Kogut & Zander, 1992; Grant, 1996).
In relation to IT, two schools of thought discuss to what extent IT improves or decreases the exchange of tacit knowledge. The first school of thought explains that IT cannot, or only limited, exchange tacit knowledge (e.g. Johannessen et al., 2001). It can even harm the exchange of tacit knowledge as IT investment can lead to mismanagement relegating tacit knowledge to the background, as sharing knowledge with a more tacit nature can be extremely difficult using IT (Johannessen et al., 2001). The second school of thought has a more nuanced view and argues that IT still can speed up tacit knowledge exchange, although due to its limitations will not fully be a replacement of face-to-face interactions (e.g. Panahi et al., 2012). Although both schools disagree about the role of IT in the exchange of tacit knowledge, they appear to have agreed upon that the exchange of tacit knowledge through the use of IT is less efficient relatively to explicit knowledge (e.g. Griffith, Sawyer, & Neale, 2003).
III. HYPOTHESIS DEVELOPMENT
8 A. Knowledge search breadth and IT investment
IT plays an important role in building a knowledge network as it allows firms to improve their ability to build relations with external parties and acquire knowledge from different sources (Dong & Yang, 2015). Through, but not limited to, establishing relations (Ives & Learmonth, 1984), improvement of communication, coordination, and the reduction of transaction costs (Mukhopadhyay, Kekre, & Kalathur, 1995; Tallon, Kraemer & Gurbaxani, 2000).
From a TCE perspective, IT can influence the decision-makers to use external knowledge sources as the eased knowledge transfer, facilitated by enhanced IT, can substantially lower the uncertainties and information deficiencies (Li, Sikora, Shaw, & Tan, 2006). Furthermore, enhanced IT has been shown to negatively influence the possibility of opportunism occurring, enhancing the willingness of decision-makers to exchange knowledge with external partners (Brynjolfsson, Malone, Gurbaxani & Kambil, 1994). Finally, the use of IT can enhance the ability of decision-makers to receive greater and more qualitative information about processes, actors, environment, and (potential) knowledge sources, resulting in less uncertainty when searching, contracting, controlling and enforcing contracts, increasing EKS (Nooteboom, 1992; Markus, 1992; Cordella, 2001; Cordella, 2006; Laursen & Salter, 2006; Tafti, Mithas, & Krishnan, 2013). This further suggests that enhanced IT will increase the search breadth of the focal firm in question.
9
knowledge (Nonaka and Takeuchi, 1995; Alavi & Leidner, 2001; Griffith et al., 2003). When
“tacit knowledge is high, the need for human expertise in the matching process will become more apparent” (Goodman & Darr, 1998, p. 424) limiting the efficiency and effectiveness of
IT (Johannessen et al., 2001). For example, tacit knowledge first needs to be codified before asynchronous transfer through e-mail, intranet, databases, document management systems are possible (Alavi and Leidner, 2001; Stenmark, 2002). This is a costly and inefficient process compared to explicit knowledge (Johannessen et al., 2001) as explicit knowledge is codified and can be easily shared and controlled through the use of IT (Wasko & Faraj, 2000; Earl, 2001; Johannessen et al., 2001; Stenmark, 2002; Falconer, 2006). Furthermore, the knowledge will be retained even as the contributor is not an active source anymore (Wasko & Faraj, 2000). Due to the lower transaction costs of explicit knowledge, according to TCE view, the transfer of explicit sources is favored to that of tacit knowledge.
Consequently, as IT investment increases, the exchange of explicit knowledge will most likely become more dominant. Overall, this will increase the reliance on explicit knowledge, and could ignore the existence of tacit elements of knowledge (Polanyi, 1966; Ocasio, 1997; Johannessen et al., 2001; Barnett, 2008). This is in accordance with the attention-based arguments, describing that firms have a limited amount of attention that they need to distribute amongst a limited number of options (Barnett, 2008). Combining attention-based view and TCE, it is argued that, due to transaction costs, firms become overly dependent on explicit knowledge and have insufficient attention for tacit knowledge transfer, thereby decreasing the knowledge breadth after a given point.
10
costly tacit knowledge transfer is diminished as a result of transaction cost minimization and limited attention, thereby lowering the external knowledge breadth as the sharing activities with tacit sources will be discontinued. Consequently, the following hypothesis is induced:
H1: The relation between IT investment and external knowledge search breadth has
an inverted U-shape.
B. Moderating role of R&D investment
Previous research has shown that firms need to invest in R&D to be able to develop absorptive capacity in order to utilize knowledge from the environment (e.g. Cohen & Levinthal, 1990; Zahra, & George, 2002; Escribano, Fosfuri, & Tribó, 2009; Lichtenthaler, 2009), as individuals need to understand the underlying assumptions of the other parties before fully assimilating the external knowledge (Cohen & Levinthal, 1990). Within the context of EKS breadth, it is evident that R&D investment itself will increase the knowledge search as absorptive capacity increases the ability of firms to recognize “the value of new,
external information, assimilate it, and apply it to commercial ends” (Cohen & Levinthal,
1990, p. 128) and “increases the ability to make sense of and to assimilate and use new
knowledge”, and to “create new knowledge” (Kim, 1998, p. 507).
11
findings indicate a clear distinction between the preference for the type of knowledge in IT investment and R&D investment. More specifically, a higher investment in R&D will shift managerial attention more to tacit knowledge which is not codified and thus more difficult to transfer by IT, whereas enhanced IT investment will primarily shift attention to more explicit knowledge (Nonaka, 1991; Howells, 1996; Mascitelli, 2000; Alavi, & Leidner, 2001; Johannessen et al., 2001; Berman et al., 2002; Panahi, et al., 2012). This clash is in accordance with the attention-based view, which argues that the focus of a firm on any type of knowledge through enhanced investment in a specific field will automatically lower the attention to the other.
Overall, it is argued that enhanced R&D investment will, thus, negatively influence the increased amount of focus on explicit knowledge that follows from the increased investment in IT, due to the fact that any given firm can only provide sufficient attention to one focus point at any point in time. Therefore, the second hypotheses continues as follows:
H2: R&D investment negatively moderates the relationship between IT investment and
external knowledge search breadth.
C. Towards knowledge depth
12
most efficient knowledge to transfer with that asset. Hence, it will increase the reliance on explicit knowledge, which eventually leads to ignoring the existence of tacit elements of knowledge (Polanyi, 1966; Johannessen et al., 2001), and thus dropping tacit sources as transaction costs relatively to non-tacit sources are too high.
Knowledge that is important for innovation may come from distinct knowledge sources, such as suppliers, competitors, customers, and universities (Darr & Kurtzberg, 2000; Gertler, 2003; Leiponen & Helfat, 2011). Following Cassiman and Veugelers (2002) a distinction is made between three different archetypes of external sources; vertical (suppliers and customers), horizontal sources (competitors) and knowledge intensive (universities, consultancies, and other research institutions). Further mentioned as VKS (vertical knowledge sources), HKS (horizontal knowledge sources), and KIS (knowledge intensive sources).
Various types of knowledge sources react and develop differently in situations, making it most likely the case that these different types of knowledge sources are susceptible for different types of influential actors (Lam, 1997; Paananen, 2009). The nature of knowledge is important as IT is more efficient when transferring explicit knowledge. This study argues, that this does not necessarily mean that tacit knowledge will fully diminish as some schools of thought claim (e.g. Johannessen et al., 2001) or that tacit knowledge is quite easy to transfer (Panahi, et al., 2013). The effects of IT on tacit and more explicit sources depends on the situation, where the effects depend on the knowledge source and where too extensive IT investment can harm EKS depth.
13
knowledge, allowing IT to gain knowledge through but not limited to data-mining software, or web-based surveys (Mukhopadhyay et al., 1995; Su et al., 2006; Soosay et al., 2008). Thus, the tacitness of the knowledge exchange with VKS is likely to be low.
HKS however, is often more important when creating an industry standard where basic or applied research is needed to be conducted, and is more based on tacit knowledge (Audretsch, 1998). For example, the cooperation to establish a new industry standard. Although sharing knowledge with HKS could be beneficial on other occasions and frequently happens. The sharing of this latter knowledge is often more uniform and explicit due to the possibility of standardization that happens within an industry (Wang & Seidmann, 1995; Damsgaard & Lyytinen, 1998). Thus the tacitness of knowledge exchange with HKS is likely to be medium compared to VKS and KIS.
14
Consequently, the following hypothesis is induced:
H3: The relation between IT investment and the external knowledge search depth, (a)
has an inverted U-shape, (b) will have a higher optimum when the knowledge source is more explicit. Vertical EKS Horizontal EKS Knowledge Intensive EKS
Low Tacitness of knowledge sources High
P o si ti v e ef fe ct I T I n v es tm en t L o w
Fig I. Expected relationship between tacitness and effectiveness of IT investment
Relating to the three different knowledge sources, R&D investment will have distinct effects as the nature of the knowledge is different. It is thus likely that firms with a high R&D investment will relatively affect the relation between IT investment and the EKS depth more when the source is explicit. Hence, the following hypothesis is hypothesized.
H4: The inverted U-shaped relation between IT investment and external knowledge
15
The main relations of this study are presented in figure II. The conceptual model consists of one independent variable, one moderator that moderates all relations, and four dependent variables.
R&D
investment
IT
Investment
EKS breadth
VKS
HKS
KIS
EKS depth
∩
∩
∩
∩
-
-
-
-Fig II. Conceptual model
IV. METHODS
A. Data
16
practices German firms conduct represent the practices of firms of other countries with similar innovative capacity such as the United States, Finland, United Kingdom, and Switzerland (Porter & Stern, 2001). This latter explanation results in highly generalizable findings for at least firms that perform practices related to firms in countries with a considerable innovative capacity.
The Mannheim Innovation Panel has been around for more than two decades and was prior part of the broader Community Innovation Survey which were conducted in numerous member states of the European Union since 1993 (Wagner, 2008). The conducted survey is established through the use of the core methodologies and measures published by the Organisation for Economic Cooperation and Development (OECD). The core methodologies and measures used can be found in the Oslo Manual, this manual contributes to the reliability of the collection and interpretation of innovation data as it provides well-grounded and established guidelines (OECD and EUROSTAT, 2005). Besides the Mannheim Innovation Panel, the Community Innovation Survey (CIS), which is conducted by Eurostat, uses the same instruments (Ziegler & Nogareda, 2009). The understandability, reliability, and validity of the surveys questionnaires and collection were extensively tested through the use of pilot-testing (European Commission 2004; Laursen & Salter 2006). Sets with similar data has been used by previous academic scholars within the field of innovation (e.g. Horbach, 2008), knowledge management (e.g. Kaiser, 2002) and other managerial studies (e.g. Poot, Faems, & Vanhaverbeke, 2009).
17
participated in a two year time span (1999 – 2001 and 2003 – 2005) were included in this sample. Time-lagged research design is appropriate for the measurement and testing the effects of IT and R&D investments to avoid tautological correlation issues. Furthermore, this design reduces the threats of causal explanations, common method and omitted variable bias due to that the dependent variables are unlikely to influence explanatory variables of the past (Podsakoff, MacKenzie, & Podsakoff, 2003). Also the data is considered to be panel data as panel data involves at least two dimensions; a time series and a cross sectional dimension (Antweiler, 2001; Davis; 2002). Treating the data as panel data has multiple advantages compared to a more simple time lagged design. Relying on the inter-firm differences of the panel data can reduce the collinearity between the lagged variables due to the possibility to estimate time adjusted design (Pakes & Griliches 1984) which enables studies to create less bias predictions for individual outcomes. In total 2044 firms from different industries were included in the empirical analysis. The original data of the participated of firms was anonymized to ensure confidentiality. As a consequence, extra specific data of the participated firms’ profiles and observations could not be collected. An overview of the sample data on industry level can be found in appendix I.
C. Measures
Dependent variables
18
information sources of each firm (Breadth of EKS). In line with prior research (e.g. Laursen and Salter, 2004; Poot, et al., 2009) that constructed a comparable variable, the sources were coded as binary variables (“0” and “1”). Afterwards, the scores were summed up, and the sum of the total sources was the product, where the score of “0” relates to no external sources used and a score of “9” when all external sources were used. The higher the sum, the more external knowledge sources are used by the firm.
To measure the last two hypotheses, a distinction is made between three previous mentioned different external sources based on the cooperation archetypes of Cassiman & Veugelers (2002); VKS, HKS, and KIS. Finally, adopting from the study Poot et al. (2009) this study will construct a similar indicator which measures the relative importance of external knowledge. This has been done by measuring the value of each archetype by summing up the scores of all variables belonging to that archetype. After the ‘count’ the sum of the total variables was divided by the total variables included. Following Poot et al. (2009), as it was needed to threat the different archetypes equally the score is scaled to a range between 0 and 10. The dependent variables were collected from the years 2001 and 2005. In table I the descriptive statistics of the knowledge sources can be found.
D. Explanatory variables
IT investment (ITINV): In this paper, the IT investment relate to the total investments in IT,
such as hardware and software, whereas IT was scaled by the firms’ total sales. This measure is adopted from prior research (Han & Mithas, 2013). As an inverted U-shaped was tested, one quadratic term was created; ITINV2.The data from the years 1999 and 2003 was used to
19
R&D investment (R&DINV): Following Zahra & George (2002) and Shimizutani &Todo
(2008), the investments of R&D are scaled by total sales of the firm. This measure was used to explain some of the known effects of R&D investment which are related to absorptive capacity (e.g. Cohen & Levinthal, 1990.
Table I: Descriptive statistics
Years Variable Obsa Mean Std. Dev. Min Max
Dependent 2001, 2005 VKS depth 2116 3.539698 341.791 0 10 HKS depth 2091 3.082257 3.486.944 0 10 KIS depth 2111 1.369493 2.102.109 0 10 EKS breadth 2136 3.646536 3.378.409 0 9 Explanatory 1999, 2003 ITINV 1994 .0188104 .0764557 0 .8 R&DINV 2116 .0149034 .0378246 0 .15 Control variables 1999, 2003 East Germany Industry Product development Size Startup 2319 2319 2319 2319 2298 .4294955 .7028892 .4605433 160.1807 .0356832 .4951109 .4570843 .4985482 1234.817 .1855395 0 0 0 0 0 1 1 1 56395.16 1 a
Some survey questions regarding the dependent and explanatory variables were not answered, thus these were not included in the analysis leading to marginal differences between observations.
Consistent with the time lagged design of this study, the data was collected from the years 1999 and 2003. The interaction product was created on basis of R&D investment and IT investment, mainly; ITINV*R&DINV. Similar to IT investment, the data from the years 1999
20 E. Control Variables
In this study five control variables were used which are generally considered as relevant to the breadth and depth of knowledge search.
Table II: Overview of variables
Variable Description Scale References
Dependent variables EKS breadth VKS depth HKS depth KIS depth EKS breadth (the number of distinct external knowledge sources) EKS depth of VKS (The importance of vertical knowledge sources) EKS depth of HKS (The importance of horizontal knowledge sources)
EKS depth of KIS (The importance of knowledge intensive sources) Interval Interval Interval Interval
Laursen and Salter, 2004; Poot et al., 2009
Laursen and Salter, 2004; Poot et al., 2009
Laursen and Salter, 2004; Poot et al., 2009
Laursen and Salter, 2004; Poot et al., 2009
Explanatory variables
ITINV
R&DINV
ITINV*R&DINV
IT Investment (percentage of total sales) R&D investment (percentage of total sales) Interaction term Ratio Ratio Ratio
Han & Mithas, 2013
Zahra & George, 2002; Shimizutani &Todo , 2008 Control variables East Germany Industry Product development Size Startup Geographical region (east or west Germany)
Industry type (Service or manufacturing) Product development Activities
Employees within a firm (firm size) Age of the firm
Nominal Nominal Nominal Ratio Nominal Van de Vrande, Vanhaverbeke, & Duysters, 2011 Jaworski and Kohli, 1993; Lichtenthaler, 2009b
Kessler, Bierly, & Gopalakrishnan, 2000 Laursen & Salter, 2006
Cohen, Nelson, & Walsh, 2002
Variables for robustness
ITINV_FITTED IT investment of firms
in the same industries
21
First, similar to the study of Van de Vrande, Vanhaverbeke, & Duysters (2011) a region dummy was added, which included whether firms’ were located in eastern (1) or western (0) of Germany to account for propensities due to effects that are region-specific. Second, a measure to control whether it is a startup or a more mature firm was added as less mature firms act differently and seem to make other choices compared mature firms (Cohen, Nelson, & Walsh, 2002). Thus, a dummy variable was included which included the classification of the maturity of the firm (1 = younger than three years). Third, in order to account for propensities of firm size, the number of employees was used (Laursen & Salter, 2006). Fourth, one dummy variable was created which consisted of the industry type (service or manufacturing) as industry types seems to be antecedents for the breadth and depth of EKS (Jaworski and Kohli, 1993; Lichtenthaler, 2009b). Final, product development activities was included as this variable has the potential influence to determine the breadth and depth of EKS (Kessler, Bierly, & Gopalakrishnan, 2000).
The overview of the measures can be found in table II, and the description of the measurement of the survey can be found in appendix II.
V. RESULTS
22
some contradicting results as the Skewness and Kurtosis test showed that the data is non-normally distributed and the quantile-quantile plot indicated a normal distribution.
Second, there was also controlled whether explanatory variables within the regression were highly correlated. Controlling for this issue is important as calculations of the explanatory variables can be affected by this issue, often referred to as the multicollinearity. Thus, the full models were tested for multicollinearity issues where – following Bock, Zmud, Kim, & Lee (2005) - the variance inflation factor (VIF) is acceptable (i.e. between 1.00 and 1.80). None of the correlations were substantial enough to require complimentary research of potential multicollinearity problems. Third, the dependent variables take on nonnegative integer values because the amount of knowledge sources or the importance of knowledge sources are counted. As panel data was used, a negative binominal in the case of overdispersion and poisson regression in the case of underdispersion, might seem a justifiable choice (Laursen & Salter, 2014), but due to the dependent variables that can be seen as scale data as they are restricted by a upper bound of 9 and 10 (see table I) this would not be the appropriate test to use (Laursen & Salter, 2014). However, as the normal distribution test did give contradicting findings, first an ordinary least squares (OLS) regression was used as this often is seen as a more robust solution to measure scale data and when residuals can fit the assumption that they are normally distributed. To be sure that I did not include biased results, I also conducted a random- effects of Generalized Least Squares (GLS) regression as the Hausman test showed that (p > 0.05) to ensure the robustness of the results. Both tests were clustered on firm level.
23
whereas the non-normality even can be questioned, thus in this case the F-test is robust to the non-normality distribution (Luh & Guo, 1999).
Final, models were created and sequential implemented in the analysis. First, model 1 only includes all control variables. Second, model 2 included the main explanatory variable, ITINV. Third, in model 3, also the nonlinear term ITinv2 is included, to test whether there is an
inverted U-shaped relationship. Final, model 4 consists of the moderator R&DINV and the
interaction terms. In total this resulted in four regressions that were build up from the model with control variables (model 1) to the full model (model 4). These steps makes it possible to isolate and explain the potentially unique effects on distinct knowledge sources.
In this section, I discuss the results of the analysis. First, the effect of IT investment will be analyzed in terms of the breadth of EKS. Second, I report the findings in relation to the interaction between IT investment and R&D investment on the breadth of EKS. Third, the findings shift from the breadth of knowledge search to the depth of knowledge search and the distinct effects on sources. Final, I address the findings in relation to the interaction between IT investment and R&D investment on the depth of the EKS.
A. Breadth of EKS
24
Table IV presents the OLS regression with the dependent external search breadth. Model 1 is the base model, whereas only the control variables are included. The coefficients of this model show a significant relation with the age of the firm (p < 0.1), the type of industry the firm is established in (p < 0.01), and product development activities (p < 0.01). These findings suggest that within this sample three out of five control variables are significantly predictors of the decision-making to search for new knowledge sources.
These findings are in line with the expectations, suggesting that there are differences between service and manufacturing industries, product development activities, and the age of the firm in the decision to EKS. More specifically, the negative sign of industry indicates that service industries are less likely to have a wide portfolio of external sources. Furthermore, the positive signs of product development activities and startup firms indicate that startup firms and firms that conduct product development are predictors in the external search of a firm.
Hypothesis 1 predicts that the relation between IT investment and the breadth of EKS has an inverted U-shaped relation. First, model 2 is added to the regression, the findings in Table IV show that the IT investment have a negative and insignificant (p > 0.1) linear relation with the breadth of EKS. As this finding does not answer the hypotheses, model 3 which consist of the quadratic term of IT investment is added to the regression. After adding this model, the negative relation shifted to an significant (p < 0.01) and positive relation whereas the quadratic term presents a similar significant (p < 0.01) negative relation. In summary, hypothesis 1 is supported as the relation between IT investment and the breadth of EKS has an inverted U-shape.
25
variable and IT investment. Thus hypothesis 2 can be partially supported as a negative relation was predicted. Overall, these results support the contention that the interaction between IT and R&D investments can hinder the breadth of EKS as investing more in R&D investment leads to a higher need in tacit knowledge. The above findings are graphically represented in figure III.
26 Table III: Descriptive Statistics and Correlations
Variable Mean Std. Dev. Min Max (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
27 Table IV: Breadth of EKS
VARIABLES (1) (2) (3) (4)
East Germany -0.0213 -0.0337 -0.0441 -0.128
(0.133) (0.141) (0.141) (0.140) Industry -1.636*** -1.468*** -1.555*** -1.408***
(0.145) (0.151) (0.152) (0.153)
Size 6.19e-05 5.06e-05 5.43e-05 5.87e-05
(8.79e-05) (7.99e-05) (8.15e-05) (8.66e-05)
Startup 0.580* 0.568 0.486 0.167 (0.345) (0.385) (0.385) (0.375) Product Development 2.631*** 2.717*** 2.630*** 2.224*** (0.137) (0.147) (0.152) (0.162) ITINV 0.376 15.01*** 14.21*** (0.864) (3.699) (4.696) ITINV2 -20.29*** -19.42*** (4.940) (5.991) R&DINV 18.12*** (2.087)
ITINV*R&DINV -81.67*
(43.13)
ITINV2*R&DINV -521.8
(552.7)
Constant 3.526*** 3.478*** 3.433*** 3.324***
(0.157) (0.163) (0.163) (0.164)
N 2,205 1,973 1,973 1,891
R2 0.233 0.231 0.238 0.277
28 B. Depth of EKS
Moving towards the depth of EKS. Table VI presents the OLS regression to show the findings of the EKS depth of VKS, HKS, and KIS. In total the regression consists of 12 models, whereas all dependent variables have 4 models.
Model 1 represents the mode with only control variables added. Similar to knowledge breadth, the findings indicate that industry and product development activities are deeply related to the search breadth but also to the depth of the relations (p < 0.01). Furthermore, relative to startups, this results show that being a startup does not necessarily influence the depth of knowledge search, as the significant relation is limited to a relation with KIS. Towards testing hypothesis 3A, model 2 was included in the regression. The cumulative effect
of IT investment may give an impression that the cumulative value of IT investment give different coefficients for each knowledge source. The F-test, as shown in table V, shows a non-significant result.
Table V: F-test without quadratic term
Equation Obs Parms RMSE F P
VKS 1943 2 3.430.458 .2921106 0.5889
HKS 1943 2 3.469.105 .0018216 0.9660
KIS 1943 2 2.121.198 .3942051 0.5302
However, hypothesis 3A predicted that there is an inverted U-shaped relation between IT
29
archetype. To be sure of this finding, I also conducted a F-test (Table VII) indicating that the inverted U shaped relation indeed differs per knowledge source (p < 0.01).
Table VII: F-test with quadratic term
Equation Obs Parms RMSE F P
VKS 1943 3 3.407.473 137.852 0.0000
HKS 1943 3 3.460.447 5.363.507 0.0048
KIS 1943 3 2.111.785 9.370.364 0.0001
Furthermore, in line with hypothesis 3B, the findings support the theorem that tacitness is an
important determinant concerning the effectiveness of IT investment as the VKS (tacitness = low, coefficient = high) has the highest coefficient, followed by the coefficient of HKS (tacitness = medium, coefficient = medium), and last followed by the KIS sources (tacitness = high, coefficient = low). Indicating that the effectiveness of IT investment depends on the tacitness of the external knowledge source. Thus, hypothesis 3A and 3B are supported.
To answer the final hypotheses, Model 4 was included (table VI), the results indicate that R&D investment significantly (p < 0.05) and negatively moderates the positive effect of IT investment on the depth of all knowledge sources (i.e. VKS, HKS, and KIS). Furthermore, the F-test (table VIII) indicates that the results are significantly heterogeneous (p < 0.01). Due to these findings, hypothesis 4A and hypothesis 4B are supported. Indicating that the inverted
30 Table VIII: F test with moderating effect
Equation Obs Parms RMSE F P
VKS 1863 6 32.576 415.226 0.0000
HKS 1863 6 337.333 2.647.595 0.0000
KIS 1863 6 1.911.559 8.218.752 0.0000
Indicating that this moderating effect will negatively influence the overall positive effects IT investment has on the depth of EKS. The above findings are graphically represented in figure IV (VKS), V (HKS), and VI (KIS).
31
Fig V: Y-axis – EKS Depth; X-axis – IT investment (HKS)
32 Table VI: Depth
of EKS VARIABLES VKS D1 HKS D2 KIS D3 (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) East Germany -0.224 -0.228 -0.239* -0.305** -0.0489 -0.0517 -0.0583 -0.147 0.148 0.158* 0.154 0.0966 (0.139) (0.146) (0.145) (0.145) (0.146) (0.152) (0.152) (0.155) (0.0901) (0.0954) (0.0955) (0.0913) Industry -1.679*** -1.583*** -1.683*** -1.563*** -0.995*** -0.933*** -0.982*** -0.939*** -0.571*** -0.508*** -0.545*** -0.421*** (0.153) (0.158) (0.160) (0.165) (0.159) (0.163) (0.165) (0.171) (0.100) (0.103) (0.104) (0.102) Size 2.39e-05 1.94e-05 2.37e-05 2.56e-05 5.45e-05 4.60e-05 4.81e-05 4.66e-05 6.38e-05 5.96e-05 6.12e-05 6.96e-05
(5.48e-05) (5.18e-05) (5.35e-05) (5.60e-05) (7.69e-05) (7.07e-05) (7.16e-05) (7.09e-05) (6.71e-05) (6.41e-05) (6.48e-05) (7.34e-05) Startup 0.229 0.148 0.0550 -0.116 0.404 0.0413 -0.00746 0.0832 0.615** 0.701** 0.667** 0.332 (0.334) (0.367) (0.364) (0.376) (0.411) (0.412) (0.415) (0.447) (0.271) (0.304) (0.305) (0.269) Product Development 2.583*** 2.653*** 2.554*** 2.311*** 2.566*** 2.558*** 2.510*** 2.347*** 1.376*** 1.421*** 1.385*** 0.901*** (0.141) (0.150) (0.154) (0.168) (0.149) (0.155) (0.158) (0.176) (0.0931) (0.100) (0.103) (0.106) ITINV 0.339 16.99*** 16.42*** 0.678 9.101** 10.50** -0.194 5.899** 4.916** (0.793) (3.718) (4.630) (1.033) (3.893) (4.845) (0.407) (2.613) (2.502) ITINV 2 -23.09*** -22.53*** -11.66** -13.26** -8.446** -7.358** (4.857) (5.843) (5.175) (6.318) (3.409) (3.134) R&DINV 12.61*** 9.078*** 17.16*** (2.295) (2.725) (2.205)
ITINV*R&DINV -107.3** -50.65 -15.04
(48.86) (59.55) (48.47) ITINV 2 *R&DINV 106.2 -1,905** -1,525** (684.5) (828.6) (603.5) Constant 3.579*** 3.539*** 3.489*** 3.403*** 2.581*** 2.591*** 2.564*** 2.541*** 1.033*** 1.003*** 0.985*** 0.900*** (0.165) (0.171) (0.170) (0.175) (0.164) (0.167) (0.167) (0.171) (0.0971) (0.0997) (0.0999) (0.0975) Observations 2,183 1,957 1,957 1,876 2,158 1,937 1,937 1,858 2,179 1,953 1,953 1,874 R-squared 0.225 0.228 0.238 0.255 0.170 0.167 0.169 0.184 0.143 0.145 0.148 0.230
33 C. Robustness checks
Controlling the results on the basis of variables of robustness allows it to control the breath and the depth of EKS in relation with the impacts of IT investment, moderating effect of R&D investment and other variables. The use of instrumental variables are well established within the managerial literature field (e.g. Hall & Reenen, 2000) as is it allows scholars to control for eventually reversed correlation between the dependent variables and endogenous explanatory variables.
Table IX: Instrumental regression breadth of EKS
VARIABLES (1) (2) (3) (4)
East Germany -0.0213 -0.0337 -0.0441 -0.128
(0.133) (0.141) (0.141) (0.140)
Industry -1.636*** -1.468*** -1.555*** -1.408***
(0.145) (0.151) (0.152) (0.153)
Size 6.19e-05 5.06e-05 5.43e-05 5.87e-05
(8.79e-05) (7.99e-05) (8.15e-05) (8.66e-05)
Startup 0.580* 0.568 0.486 0.167 (0.345) (0.385) (0.385) (0.375) Product Development 2.631*** 2.717*** 2.630*** 2.224*** (0.137) (0.147) (0.152) (0.162) ITINV_FITTED 0.00880 0.351*** 0.333*** (0.0202) (0.0866) (0.110) ITINV2_FITTED -0.205*** -0.196*** (0.0498) (0.0605) R&DINV 18.12*** (2.087)
ITINV*R&DINV -81.67*
(43.13)
ITINV2*R&DINV -521.8
(552.7)
Constant 3.526*** 3.452*** 3.035*** 2.954***
(0.157) (0.179) (0.204) (0.230)
Observations 2,205 1,973 1,973 1,891
R2 0.233 0.231 0.238 0.277
34
IT investment in comparable industries could impact the decision to further invest in IT as meso and marcro factors often influence the direction to which decision-making lean. This phenomenon is often referred as idiosyncratic (Lai, Wong, & Cheng, 2006; Lieberman & Asaba, 2006). In this study I first divided the average of IT investment per observation of firms that are situated in the same industry. The total IT investment was summed, and reduced by the individual IT investment of the observed firm. The total sum was divided by the total participated firms within such industry minus 1.
Second, similar to IT investment, R&D investment are often dependent on the type of industry and the situation. The decision makers often hinder at least some form of isomorphism within an industry (Lieberman & Asaba, 2006). Thus the average R&D investment of firms that are situated in a high changing environment and a low changing environment are divided. The total R&D investment was summed, and reduced by the individual R&D investment of the focal firms. Before using the instrumental variables within an OLS regression, the instrument variable was controlled to see whether it was correlated with the explanatory variables (Bound, Jaeger, & Baker, 1995), the instrumental variable of IT investment seems to be significant correlated with IT investment (p < 0.05) while not being significant correlated with the dependent variables. The R&D investment instrument variable based on investments per industry are not significant (p < 0.10). Even when other industry categories are used, such as R&D investment in service relative to manufacturing industries, industry specific R&D investment, or the spending on R&D of multiple years does not seems to correlate with R&D investment or does also correlate with the breadth or depth of EKS. Therefore, this last instrumental variable has not been used in this study.
35
were supported are also supported within this robustness check – see table IX and Table XII. Thus, no endogenous issues did arise.
Third, a second robustness test was conducted, controlling for measurement bias. This is needed as the normal distribution test did resulted in contradicting findings. Thus next to the OLS regressions, I also conducted multiple random-sample GLS regressions – see table X and table XIII. The results of the GLS regressions are in line with the OLS regressions, indicating that the results of all hypotheses are robust.
Table X: GLS regression - breadth of EKS
VARIABLES (1) (2) (3) (4)
East Germany -0.0487 -0.0493 -0.0537 -0.140
(0.135) (0.143) (0.143) (0.141)
Industry -1.672*** -1.475*** -1.564*** -1.422***
(0.143) (0.149) (0.150) (0.151)
Size 5.94e-05 4.74e-05 5.11e-05 5.62e-05
(5.05e-05) (5.11e-05) (5.08e-05) (4.95e-05)
Startup 0.517 0.596 0.512 0.177 (0.347) (0.381) (0.380) (0.390) Product Development 2.438*** 2.545*** 2.456*** 2.102*** (0.130) (0.139) (0.140) (0.151) ITINV -0.136 14.60*** 14.66*** (0.865) (3.467) (4.086) ITINV2 -20.36*** -20.06*** (4.641) (5.437) R&DINV 18.23*** (2.427)
ITINV*R&DINV -86.28
(65.08)
ITINV2*R&DINV -464.9
(1,318)
Constant 3.653*** 3.593*** 3.546*** 3.398***
(0.147) (0.153) (0.152) (0.152)
Observations 2,205 1,973 1,973 1,891
Number of newfirms 1,953 1,763 1,763 1,694
36
Fourth, a third robustness was conducted. Using the quadratic term within a linear regression to test an inverted U-shaped relation is flawed and could be potentially misleading (Lind & Mehlum, 2010). To test an inverted U-shaped relation additional tests are necessary. Lind & Mehlum (2010) propose a method to measure the estimates extreme values of the confidence interval. To control the robustness of the inverted U-shaped relation of this empirical study, the procedures of Lind & Mehlum (2010) are adopted to meet the necessary conditions. The calculations were done using Stata 13.1 where the UTEST ado-file (Lind & Mehlum, 2014) was included. The results of this tests are in line with the results of this study, demonstrating that IT investment has an inverted U-shaped relation with EKS breadth (p < 0,001) and depth (p < 0,01).
Table XI: Overview of hypothesis
Hypothesis Independent variable Relation Dependent
variable
Result Robust
H1 IT investment (ITINV) ∩ EKS breadth Supported Yes
H2 Interaction - IT investment
* R&D investment (ITINV*R&DINV)
∩
-EKS breadth Partially supported
Yes
H3A IT investment (ITINV) ∩ L VKS depth, M
HKS depth, and
H
KIS depth
Supported Yes
H3B IT investment (ITINV) ∩+ L VKS depth, M
HKS depth, and
H
KIS depth
Supported Yes
H4A Interaction –IT investment
* R&D investment (ITINV*R&DINV)
∩- L VKS depth, M HKS depth, and H KIS depth Supported Yes
H4B Interaction –IT investment
* R&D investment (ITINV*R&DINV)
∩E L VKS depth, M HKS depth, and H KIS depth Supported Yes E
negative moderating effect will be stronger when the source is explicit
+
37 Table XII: Instrumental – depth of EKS
VARIABLES VKS D1 HKSD2 KISD3 (1) (2) (3) (4) (1) (2) (3) (4) (1) (2) (3) (4) East Germany -0.224 -0.228 -0.239* -0.305** -0.0489 -0.0517 -0.0583 -0.147 0.148 0.158* 0.154 0.0966 (0.139) (0.146) (0.145) (0.145) (0.146) (0.152) (0.152) (0.155) (0.0901) (0.0954) (0.0955) (0.0913) Industry -1.679*** -1.583*** -1.683*** -1.563*** -0.995*** -0.933*** -0.982*** -0.939*** 0.571*** 0.508*** 0.545*** -0.421*** (0.153) (0.158) (0.160) (0.165) (0.159) (0.163) (0.165) (0.171) (0.100) (0.103) (0.104) (0.102) Size 2.39e-05 1.94e-05 2.37e-05 2.56e-05 5.45e-05 4.60e-05 4.81e-05 4.66e-05 6.38e-05 5.96e-05 6.12e-05 6.96e-05
(5.48e-05) (5.18e-05) (5.35e-05) (5.60e-05) (7.69e-05) (7.07e-05) (7.16e-05) (7.09e-05) (6.71e-05) (6.41e-05) (6.48e-05) (7.34e-05) Startup 0.229 0.148 0.0550 -0.116 0.404 0.0413 -0.00746 0.0832 0.615** 0.701** 0.667** 0.332 (0.334) (0.367) (0.364) (0.376) (0.411) (0.412) (0.415) (0.447) (0.271) (0.304) (0.305) (0.269) Product Dev. 2.583*** 2.653*** 2.554*** 2.311*** 2.566*** 2.558*** 2.510*** 2.347*** 1.376*** 1.421*** 1.385*** 0.901*** (0.141) (0.150) (0.154) (0.168) (0.149) (0.155) (0.158) (0.176) (0.0931) (0.100) (0.103) (0.106) R&DINV 12.61*** 9.078*** 17.16*** (2.295) (2.725) (2.205)
ITINV*R&DINV -107.3** -50.65 -15.04
(48.86) (59.55) (48.47)
ITINV2*R&DINV 106.2 -1,905** -1,525**
(684.5) (828.6) (603.5) ITINV_FITTED 0.0116 0.584*** 0.565*** 0.0543 0.729** 0.840** -0.00896 0.273** 0.228** (0.0273) (0.128) (0.159) (0.0827) (0.312) (0.388) (0.0188) (0.121) (0.116) ITINV2_FITTED -0.361*** -0.352*** -0.434** -0.493** -0.197** -0.172** (0.0758) (0.0912) (0.193) (0.235) (0.0797) (0.0733) Constant 3.579*** 3.503*** 2.860*** 2.807*** 2.581*** 2.429*** 1.717*** 1.541*** 1.033*** 1.013*** 0.935*** 0.867*** Observations 2,183 1,957 1,957 1,876 2,158 1,937 1,937 1,858 2,179 1,953 1,953 1,874 R2 0.225 0.228 0.238 0.255 0.170 0.167 0.169 0.184 0.143 0.145 0.148 0.230 DN dependent variable (depth)
38 *** p<0.01, ** p<0.05, * p<0.1
DN dependent variable (depth)
Table XIII: GLS regression -Depth
VARIABLES VKS D1 HKS D2 KIS D3 (1) (2) (3) (4) (1) (2) (3) (4) (1) (2) (3) (4) East Germany -0.283** -0.282** -0.285** -0.341** -0.0878 -0.0712 -0.0739 -0.156 0.143 0.154 0.153 0.0851 (0.135) (0.143) (0.142) (0.144) (0.144) (0.151) (0.151) (0.154) (0.0899) (0.0958) (0.0956) (0.0915) Industry -1.741*** -1.597*** -1.701*** -1.591*** -1.019*** -0.912*** -0.964*** -0.936*** -0.586*** -0.517*** -0.557*** -0.440*** (0.151) (0.155) (0.157) (0.162) (0.157) (0.161) (0.163) (0.169) (0.0998) (0.103) (0.103) (0.101) Size 2.12e-05 1.46e-05 1.90e-05 2.11e-05 5.35e-05 4.27e-05 4.48e-05 4.36e-05 6.14e-05 5.71e-05 5.88e-05 6.77e-05
(5.39e-05) (4.97e-05) (5.13e-05) (5.34e-05) (7.76e-05) (6.98e-05) (7.07e-05) (6.96e-05) (6.62e-05) (6.33e-05) (6.40e-05) (7.24e-05) Startup 0.133 0.148 0.0513 -0.134 0.531 0.115 0.0631 0.157 0.552** 0.733** 0.695** 0.352 (0.337) (0.361) (0.357) (0.367) (0.413) (0.408) (0.411) (0.442) (0.275) (0.300) (0.301) (0.266) Product Development 2.366*** 2.457*** 2.354*** 2.162*** 2.414*** 2.430*** 2.378*** 2.268*** 1.250*** 1.296*** 1.255*** 0.829*** (0.147) (0.157) (0.160) (0.175) (0.149) (0.155) (0.158) (0.175) (0.0960) (0.103) (0.104) (0.108) ITINV -0.293 16.71*** 17.45*** -0.0375 8.718** 11.28** -0.615** 6.062** 5.288** (0.998) (3.369) (4.496) (1.113) (3.881) (4.868) (0.268) (2.715) (2.570) ITINV2 -23.43*** -23.95*** -12.06** -14.38** -9.212*** -7.875** (4.611) (5.669) (5.227) (6.345) (3.420) (3.219) R&DINV 12.10*** 8.233*** 17.03*** (2.200) (2.765) (2.156)
ITINV*R&DINV -107.2** -53.25 -14.62
39
VI . DISCUSSION
First, I found strong and robust empirical support for the U-shaped relation between IT investment and EKS breadth and depth. In other words, when firms want to complement own knowledge with different sources, firms need to invest in IT, but they do not need to invest too heavily in IT, as this will lead to attention problems. This is consistent with the expectation and elements of research on the role of IT on the exchange of knowledge (Nooteboom, 1992; Johannessen et al., 2001; Panahi et al., 2012). This study demonstrates that while abilities of IT to transfer tacit knowledge are improving (McAfee, 2006), IT and the conditions of the knowledge exchange by IT is an important determinant in the decision-making process.
40
significance of this relation, as some firms could have the ability to allocate the attention of their decision-makers on explicit and tacit sources.
Third, in line what was expected, robust empirical support of an inverted U-shaped relation between IT investment and the depth of EKS was found. More specifically, this relation differs per knowledge source. This is expected as the need for a certain knowledge source dependents on the situation (e.g. Poot et al., 2009). Hence, as theorized, IT investment will increase the depth of EKS as it facilitates relationships while keeping the transaction costs low. However, when IT investment are massive, IT will push decision-makers towards using (more) explicit knowledge, reducing the importance of tacit knowledge sources. Furthermore, as some sources are more based on tacitness compared to other sources, it is supported by the robust findings that more tacit sources will experience hinder relatively faster compared to more explicit sources.
Fourth, the contention is, that the interaction between IT investment and R&D investment can hinder the depth of EKS as R&D investment leads to a higher need in tacit knowledge, which cannot easily be transferred through IT. This findings are empirically supported and robust as all interaction terms are significant. The positive influence (ITINV) of
IT investment is significantly negatively moderated by R&D investment (ITINV*R&DINV) with
the dependent VKS, while the negative influence (ITINV2) of IT investment is significantly
negatively moderated by R&D investment (ITINV2*R&DINV) with the dependents HKS and
41
allocation issue is most likely to occur earlier when investing heavily in IT or R&D, harming the equilibrium and decreasing the overall EKS.
A. Theoretical implications
Literature with respect to the relation between IT investment and EKS is highly underexplored. This study extended the attention-based and transaction costs economics theories, in the context of information systems literature, revealing the importance of IT investment on the EKS breadth and depth. Scholars need to include IT investment when theorizing open innovation as key determinant of firms’ openness.
Another theoretical implication is that the findings of this study can serve both schools of thought on the debate on tacit knowledge. Researchers should acknowledge the differences between the transfer of tacit and explicit knowledge, and that IT could harm the exchange of tacit knowledge if the IT investment is too extensive. Hence, the scale of investment dimension is an important dimension which scholars do not include when theorizing the effect of the transfer of tacit knowledge through the use of IT. Thus, it is important to use a more situational approach considering that different rates of IT investments shift the focus of decision-makers.
42
This paper demonstrates that IT investment is an important driver for the breadth and depth of firms EKS which is crucial for innovation (Chesbrough & Crowther, 2006). Furthermore, the moderating effect of R&D investment on the latter relation affects the need for tacitness when conducting EKS activities. I theorized that IT investment increases the breadth and depth of EKS as more sources are accessible while keeping transaction costs low, and as both explicit and tacit knowledge transfer are facilitated through enhanced IT. However, as firms become more accustomed to transferring explicit knowledge efficiently through enhanced IT, the more costly tacit knowledge transfer will diminish as a result of transaction cost minimization and limited attention, thereby lowering the EKS breadth and depth as the sharing activities with tacit sources will be less important and eventually discontinue. Thus, the relations have an inverted U-shaped relation. Additionally, this affect is negatively influenced by R&D investment as the preferences for the nature of knowledge differs, creating an attention based problem but also increase transaction costs. Furthermore, I demonstrate that more explicit knowledge sources will benefit more from IT investment, giving arguments that a contingency approach is more appropriate when conducting EKS activities.
B. Managerial implications
43
open innovation, it is important to not overspend on IT and to keep in mind that tacit knowledge is still an important element when innovating.
Decision makers should be very careful when investing heavily in IT and R&D as this interaction is not fully compatible. R&D investment will lead to a focus which is based on the tacit knowledge, while IT investment will lead to a focus which is more based on explicit knowledge. Thus, decision-makers of companies that are involved in (open) innovation activities need to take this clash in account when making decisions about IT and R&D investments.
C. Limitations and Future Research
This study has several limitations that provide direction for future research. First, I used the well-established proxy ‘IT investment’ in this study. This proxy is only limited in measuring the effects of IT. There are multiple ways to invest, in this study IT is treated as a black box, while if more data was available on the type of investment (e.g. knowledge systems, infrastructure) more elaborated analysis could be conducted. Thus for future research on this topic, it would be relevant to distinguish different IT investment instead of using one measure.
Second, similar to IT investment, I also treat R&D investment as a black box. R&D investment can be spend in multiple ways, some may impact the nature of knowledge source more than others. Although this proxy is often used in past research, more data on this could give new insights and be a relevant avenue for future research.
44
be comparable to a less innovative context. Thus, researchers could collect data from less innovative countries and refine the findings of this study.
Fourth, the measure of the dependent variables could be improved. The importance of a certain source is scaled from zero to three. Some respondents could respond somewhat differently while being in the same situation. Thus, researchers could replicate and refine our findings but use a more objective measure to determine the breadth and depth of EKS.
Fifth, the data of the explanatory variables within this research are based on data from 1999 and 2003. The IT and R&D landscape is quite different from 10 decades ago, making it relevant to incorporate newer data in further research. Hence, no previous research on the this subject has been conducted. Thus making it an important avenue for researchers to collect newer data and refine the findings of this study.
VII. CONCLUSION
45 Acknowledgement
First, I would like to thank Dr. JQ Dong in particular for the supervision and the advice during the process of this study. His expertise contributed a great deal to the quality of this study. Second, I would like to thank Dr. R. Van Eijk for providing me new insights and relevant feedback on the draft of this thesis, and thereafter. Finally, I would like to thank Esteban Barnhardt, Marc Donders, Timko Ogink, and Irene Warmolts for their feedback and support provided throughout the process of this study.
REFERENCES
Alavi, M., & Leidner, D. E. (2001). Review: Knowledge management and knowledge management systems: Conceptual foundations and research issues. MIS quarterly, 107-136.
Alavi, M., & Leidner, D. E. (2001). Review: Knowledge management and knowledge management systems: Conceptual foundations and research issues. MIS quarterly, 107-136.
Antweiler, W. (2001). Nested random effects estimation in unbalanced panel data. Journal of Econometrics, 101(2), 295-313.
Argyres, N. S., & Zenger, T. R. (2012). Capabilities, transaction costs, and firm boundaries. Organization Science, 23(6), 1643-1657.
Audretsch, B. (1998). Agglomeration and the location of innovative activity. Oxford
review of economic policy, 14(2), 18-29.
46
Bajari, P., & Tadelis, S. (2001). Incentives versus transaction costs: A theory of procurement contracts. RAND Journal of Economics, 387-407.
Bardhan, I., Krishnan, V., & Lin, S. (2013). Research Note-Business Value of Information Technology: Testing the Interaction Effect of IT and R&D on Tobin's Q. Information Systems Research, 24(4), 1147-1161.
Barnett, M. L. (2008). An attention-based view of real options reasoning. Academy of
Management Review, 33(3), 606-628.
Berman, S. L., Down, J., & Hill, C. W. (2002). Tacit knowledge as a source of competitive advantage in the National Basketball Association. Academy of Management
Journal, 45(1), 13-31.
Bock, G. W., Zmud, R. W., Kim, Y. G., & Lee, J. N. (2005). Behavioral intention formation in knowledge sharing: Examining the roles of extrinsic motivators, social-psychological forces, and organizational climate. MIS quarterly, 87-111.
Bonner, D. (2000). The knowledge management challenge: new roles and responsibilities for chief knowledge officers and chief learning officers. Leading Knowledge
Management and Learning, American Society for Training & Development, Alexandria, VA, 3-19.
Botosan, C. A. (2006). Disclosure and the cost of capital: what do we know?. Accounting
and business research, 36(sup1), 31-40.
47
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(4), 597-621.
Brynjolfsson, E., Malone, T. W., Gurbaxani, V., & Kambil, A. (1994). Does information technology lead to smaller firms?. Management Science, 40(12), 1628-1644.
Cassiman, B., & Veugelers, R. (2002). R&D cooperation and spillovers: some empirical evidence from Belgium. American Economic Review, 1169-1184.
Chen, C. J. (2004). The effects of knowledge attribute, alliance characteristics, and absorptive capacity on knowledge transfer performance. R&D Management, 34(3), 311-321.
Chesbrough, H., & Crowther, A. K. (2006). Beyond high tech: early adopters of open innovation in other industries. R&d Management, 36(3), 229-236.
Cohen, W. M., & Levinthal, D. A. (1989). Innovation and learning: the two faces of R & D. The economic journal, 569-596.
Cohen, W. M., & Levinthal, D. A. (1990). Absorptive capacity: a new perspective on learning and innovation. Administrative science quarterly, 128-152.
Cohen, W.M., Nelson, R.R., Walsh, J., 2002. Links and impacts: the 927 influence of public research on industrial R&D. Management Science 48, 1–23.
Cordella, A. (2001). Does information technology always lead to lower transaction costs?. ECIS 2001 Proceedings, 1.
Cordella, A. (2006). Transaction costs and information systems: does IT add up?. Journal
