Is being digital the best approach?
Robin Dekter r.dekter@student.rug.nl
S2563797 University of Groningen Faculty of Economics and Business MSc Strategic Innovation Management
Supervisor
Dr. P. (Pere) Arque Castells
Co-assessor Dr. P.J.O. (Pasi) Kuusela
Abstract
Digital organized firms are able to disrupt entire established industries because of their increased speed, flexibility and adaptability. Nowadays, firms face increasing pressures to innovate successfully in order to create and sustain a competitive advantage. The burden of knowledge implies that when knowledge accumulates as technology advances, successive generations of innovators may face an increasing educational burden. Although the interest of scholars in both digital business and the burden of knowledge increases, the relation between these variables remains unnoticed. This paper addresses this gap by researching the relationship between firm nature and the burden of knowledge, eventually moderated by the Industry 4.0 era (1997-2014). The hypotheses are tested using a longitudinal secondary dataset (1902-2014) consisting of 4,963,225 unique patents. The results of this study imply that digital firms are better able to overcome and encounter the educational burden within the Industry 4.0 era, which eventually results in a decrease of team size.
Keywords: innovation, technological markets, burden of knowledge, team size, firm nature,
1. Introduction
The presence of innovation as a core business more frequently in present day technological industries underlines the importance of a firm’s capability to innovate. According to McKinsey’s research (2019), innovation is a central theme in 84% of the firms. They state that 84% of the firms expect innovation to be the one most important driver to provide them with firm success in the near future (McKinsey, 2019). The increasing attention on innovation and growth has become a central theme in multiple giant consultancy firms in order to help their clients to strive for innovation excellence (KPMG, 2019). The increasing importance of innovation capabilities on the one hand and the increasing attention by giant consultancy firms on the other underlines a phenomenon that some scholars already stated: it has become increasingly difficult to invent and innovate. Consequently, it is harder to create and capture value and growth from innovation in technology markets (Wuchty et al., 2007; Jones, 2009; Bloom et al., 2019).
The phenomenon that ideas are harder to find is called the ‘burden of knowledge’ in today’s literature and research (Wuchty et al., 2007; Jones, 2009). The burden of knowledge implies that when knowledge accumulates as technology advances, successive generations of innovators may face an increasing educational burden (Jones, 2009). As a result, there is an increased need of larger specialized teams to come up with successful inventions. Team size is therefore one of the most important factors explaining the burden of knowledge (Wuchty et al., 2007). The current debate is mainly focused on patent applications and the associated applicants pictured as team size (Jones, 2009). Prior research mainly focuses on team size per patent as the embodiment of the burden of knowledge, since the team size per patent has increased over time in the last decades confirming the existence of the phenomenon (Jones, 2009).
innovative and therefore outperforming traditional firms (BCG, 2019). The digital firms have changed and disrupted multiple markets in the past couple of years and thereby put immense pressures on traditional firms (Verhoef et al., 2019). Since present day literature is more frequently focusing on either digital or non-digital firms, it is highly important to extend this understanding to other fields of research (Verhoef et al., 2019). The current literature on the burden of knowledge is not extensively focused on either digital or non-digital firms and requires a better understanding, which introduces the research gap addressed in this study: the lack of research on firm nature in relationship to the burden of knowledge. In order to battle or minimize the burden of knowledge it is important to understand the implications of being digital. Thus, this paper aims to answer the following central question:
‘To what extent is the impact of the burden of knowledge different to digital firms than it is to non-digital firms?’.
digital era (1997-2014), which means that team size has decreased over time. Correspondingly, the burden of knowledge has a less fierce impact on digital firms than it has on non-digital firms. The following chapter exhibits the theory behind firm nature and the burden of knowledge. Additionally, drawing from this past literature, hypotheses are formed. Next to that, I provide the relevant methods of study and data transformation in order to have the most appropriate empirical setting possible. Thereafter, I provide the most relevant results to answer the research question and show the hypothesized relationship. Lastly, the discussion section contains and discusses the most relevant key findings, limitations and leaves room for future directions of research.
2. Theoretical background
2.1 The increasing importance of innovation
relationship when the internal capital stock is large, concluding that internal capital stock has a positive impact on firm performance.
Not every industry could be labelled as turbulent or rapidly changing. There are some competitive environments which are more stable. In those environments, producers, customers and suppliers try to prevent the industry from changing (Chen, 2017). In these competitive environments the influence of innovation on firm performance, moderated by the external environment, is lower since innovation is unlikely to be useful if the industry is not willing to change the existing state (Chen, 2017). These more stable industries are not innovating in order to change or disrupt the industry, but to serve their customers to such an extent that they do not have to deal with highly changing products and services (Huang and Rice, 2009). Chen (2017) researched the relationship between the structure of firms operating in stable industries and their innovation activities. The results show that the rigidity of the organizational structure, which often relates to the rigidity of the environment, may stifle the innovation activities and therefore affect the whole industry. Moreover, industry leaders rarely innovate first to destroy the existing environment or to strive for first mover advantages, but to cultivate the market in order to provide possibilities for product introductions more gently (Chen, 2017).
Furthermore, prior research shows that the relationship between innovation and firm performance has three different effects: a direct effect, a moderating effect and a mediating effect (Chen, 2017). Consequently, the direct effect could have a positive effect, neutral effect or nonpositive effect. The moderating effect is influenced by environmental variables: the internal capital, the external markets and the environmental factors. Huang and Rice (2009) added that the environment could be either dynamic or stable, which requires a specific approach for the given situation.
and growth by multiple giant consultancy firms could be a logical consequence of a firm’s innovation focus.
2.2 The difficulty of innovation and the existence of the burden of knowledge
The recombination of existing technologies drives a firm’s innovativeness. A strong tenet among prior strategy research is that firms’ recombinant capabilities, defined as “their ability to recombine existing technologies to generate technological innovations”, is a key driver of a firm’s innovative performance and eventually firm performance in general (Carnabuci and Operti, 2013, p.1591). The research by Caranbuci and Operti (2013) resulted in two distinct types of recombinant capabilities that a firm needs to possess in order to lower the difficulty of innovation. First, firms could innovate through a so called ‘recombinant creation’, which is creating innovative technological combinations which are fully new to the firm. Second, firms could innovate through recombinant reuse, which is about creating innovative technological combinations which are already known to the firm. Building on the ideas of Barnett and Clark (1998) the recombinant capabilities are becoming more important since the rapidly changing environment drives firms towards a better understanding of existing and new technologies. Hence, there exists a distinction between established firms and start-ups regarding the recombinant capabilities (Henderson and Clark, 1990). In their seminal study, Henderson and Clark (1990) argued that established firms lose their technological leadership, or market leadership, more frequently due to their inability to link existing technologies in new and more fruitful ways. Consequently, this means that established firms seem to have a lower recombinant reuse capability. Combining the findings of Carnabuci and Operti (2013) and Henderson and Clark (1990), non-established firms, or start-ups, are better able to develop recombinant capacities and deal with rapidly changing external variables. It turns out that innovation is harder to achieve for established firms than it is for start-ups, since they are better able to deal with a changing environment. Multiple scholars have already studied this ‘phenomenon’: it is harder to come up with inventions and innovations and it is therefore harder to capture value and growth from innovation (Wuchty et al., 2007; Jones, 2009).
teams eventually grow larger. Team size is therefore one of the most important factors explaining the burden of knowledge (Wuchty et al., 2007). Prior research already has proven the existence of the burden of knowledge. Wuchty et al. (2007) investigated over 19 million papers combined with over two million patent applications to demonstrate that teams increasingly dominate solo authors in the production of knowledge to invent: research has been done more frequently in specialized teams rather than solo artists in nearly all fields. In addition, technological fields are the most dominant proof of the burden of knowledge (Jones, 2009). Following the arguments as proposed by Jones (2009), technological fields face an increased need and dependence towards innovation relative to less technical fields. Wuchty et al. (2007) discovered that firms often need to combine multiple inventions to come up with new products. It draws upon the findings of Henderson and Clark (1990), Carnabuci and Operti (2013) and Chen (2017): recombinant capabilities are highly important. The burden of knowledge is ‘a shifting process of knowledge creation from broad solo scientists to specialized research teams’ (Wuchty et al., 2007, p. 1037). Successive generations of inventors need to devote an increasing amount of resources to produce inventions (Jones, 2009). Resources are typically defined as either assets or capabilities. Assets could be tangible or intangible and are typically owned and controlled by the firm (Collis, 1994). Capabilities have an intangible nature and are bundles of skills and knowledge exercised through organizational routines (Nelson & Winter 1982; Teece et al., 1997). Hence, assets, skills and knowledge are important drivers of the burden of knowledge.
Taken together, studies are consistent: there is a shift from broad solo scientists to specialized research teams. Bloom et al. (2017) state that the most important driver of the burden of knowledge is the difficulty of coming up with valuable inventions. According to prior studies, the burden of knowledge explains the surge of technology markets since it is a product of team size and inventions, where inventions are considered as a result of patent applications (Wuchty, 2007; Jones, 2009; Bloom et al, 2017).
2.3 Innovation: digital and non-digital firms
can scale their business at an unprecedented pace. These firms compete in a so-called ‘platform economy’, which was established in the early 2010s (Huang et al., 2017). A non-digital firm is a venture which grows by drawing on and adding to a non-digital infrastructure focusing on several products and product lines (Gawer, 2009). Moreover, The European Central Bank investigated traditional firms, or non-digital firms, and defined them as ‘physical assets intensive firms that are centralized and give flat wages to skilled employees’ (ECB, 2001). Since digital firms are able to disrupt established industries due to continuous innovation as one of their core businesses, the impact of digital firms cannot be forgotten (Eisenmann, 2006; Zhu & Iansiti, 2012; Verhoef et al., 2019). How digital firms and non-digital firms are organized and capture value is discussed in the following theoretical sections.
2.3.1 Capturing value: the differences between digital and non-digital firms
The rise of digital businesses has not gone unnoticed. Digital firms are usually based on a digital infrastructure which is fully data driven (Huang et al., 2017). The fact that they are fully data driven makes them a class on their own, which provides firms and enterprises with huge advantages in the shape of speed, adaptability and flexibility (Huang et al., 2017). Consequently, the digital firm generally captures value by offering a widespread of applications while using centralized hubs. They are therefore able to disrupt established industries by continuous innovation (Verhoef et al., 2019; Eisenmann, 2006; Zhu & Iansiti, 2012). Thus, digital firms are more likely to use innovation as their core business resulting in a higher innovation performance and eventually firm performance.
firms innovate, with IT playing an increasingly central role in creating innovation opportunities and outcomes (Hensen & Dong, 2020). It results in the following adjusted value chain:
Figure 1: Digital innovation value chain (Hensen and Dong, 2020)
They found that IT use for knowledge search, meaning that firms increasingly are looking for extensive knowledge in order to increase their own knowledge base, enhances internal innovation benefits. This means that digital firms are capable of increasing their knowledge base more rapidly than non-digital firms, resulting in a higher internal innovation performance. Moreover, the IT use for relational search, meaning that digital firms are more frequently connecting towards other firms, enhances open innovation benefits: the possibility to work on several innovation activities with different companies in order to higher the innovation performance of the firm (Hensen & Dong, 2020). It shows that digital firms are better able to connect with other firms in order to strengthen their position in the modern innovation landscape where innovation seems harder to succeed.
2.3.2 The existence of big data and learning differences
Prior research has shown that innovation is harder to find and team size is one of the most important pillars to prove it, digital firms may have an advantage over non-digital firms because of the existence of big data (Hensen & Dong, 2019; Dong, 2020). Big data is a collection of data from various sources. Over time, research has been conducted frequently to investigate the existence and usage of big data, resulting in a selection of 6V’s: volume, variety, velocity, veracity, value and variability (Chen et al., 2012). The advantage of digital firms is mainly coming from the fact that they are already used to the existence and practices of big data. Non-digital firms are facing difficulties regarding big data gathering and practices. These difficulties are researched
frequently in the past, which results in several distinctive challenges which need to be faced by non-digital firms (Chen, 2017; Dong & Yang, 2019; Mikalef et al., 2019). Firstly, since technology evolves rapidly but work practices evolve slowly, non-digital firms face an unbalanced speed which they cannot follow (Dong & Yang, 2019). Secondly, non-digital firms face a shortage of support inside firm practices to cope with big data. There is a strong conservative culture against change and financial investment of big data combined with a lack of support coming from top management. Hence, there is a lack of qualified skills inside a non-digital firm to deal with big data and to make it valuable to the firm (Mikalef et al., 2019). Thirdly, non-digital firms have a strong trust relationship towards people instead of technology. The most challenging part of today's non-digital business landscape is transforming people, rather than adopting new technologies in order to integrate big data analytics (Chen et al., 2012). This challenge stems mainly from the fact that there is fear among the employees to be replaced by technology resulting in passive, active and aggressive resistance behavior which needs to be battled by top management teams (Chen et
al., 2012). Fourthly, non-digital companies are not able to measure the ideal chase of size of the
data. Bigger does not always mean better when gathering data (Dong & Yang, 2019). Lastly, something that relates to both digital and non-digital firms: privacy. Data gathering and data usage is growing on steroids resulting in a privacy dilemma: data needs to mirror reality as closely as possible in order to avoid privacy issues (Dong & Yang, 2019). Hence, privacy essentially challenges the assumption on accuracy of big data and requires a more sophisticated technical and analytical approaches than most of the non-digital firms are able to (Chen et al., 2012).
(Alom et al., 2019). This ‘deep learning’ capability ensures digital firms to increase performance measures while simultaneously increasing the amount of data, as pictured in figure 2.
Figure 2: deep learning (digital) versus old machine learning (non-digital) (Alom et al., 2019)
Being capable of deep learning and the usage of big data helps firms to increase their
internal knowledge base, resulting in a higher innovation performance. Moreover, combining the resources of digital organized firms in a deep learning network creates more output than the resources of non-digital firms (Alom et al., 2019).
2.3.3 Operating models: digital versus non-digital
The traditional operating model The digital operating model
Organizational level
• Several products and product lines; • Focused and dedicated teams in
heavyweight project management
Organizational level
• One integrated software and data platform based on many applications; • A functional structure with ad hoc
created teams in a centralized hub
Process level
• A top down approach
• Focused on the internal development
Process level
• Iterative, agile application development process;
• Core platform focus with top down approach
• Focused on open innovation
KPIs
• Profit and cost per product (line); • Quality valued per product (line) • Product features as innovation
KPIs
• Total user base, audience; • Multi-sided platforms;
• Data quality and completeness
Table 1: adapted from Cennamo & Santalo (2013), Gawer (2009) and McIntyre & Subramaniam (2009)
2.3.4 Users as innovators – a possibility to lower team size
Digital businesses are more frequently involving users or customers to participate in their innovation process than non-digital businesses (Chang & Taylor, 2016). Hence, Chang and Taylor (2016) empirically proved that involving customers has a positive relationship towards new product innovativeness and the speed-to-the-market. This eventually leads to a higher new product financial performance. Because team size increases when looking at the burden of knowledge, users could be of importance when trying to decrease it. Moreover, user ideas are proven to score higher on novelty and customer benefit. In other words: users are better able to come up with ideas that are more novel and represent a higher customer benefit than professionals can (Poetz & Schreier, 2012). Based on the same empirical research, users are also better able to come up with a so-called ‘top idea’: an idea that is significantly better than ideas that are currently active in these industries. It can be concluded that involving your users or customers could help battle the burden of knowledge, since there is an increased input figure coming from users without having to deal with a bigger team. Hence, digital kick-started projects normally show better financial performance figures than non-digital projects where the help of users or customers was asked (Chang & Taylor, 2016).
2.3.5 The introduction of Industry 4.0
and develop high-tech solutions such as Artificial Intelligence and Blockchain. Since the debate on industry 4.0 is currently been held, we can conclude that we are still part of this fourth revolution. Companies that are competing in platform economies are fully build on the abilities of the industry 4.0 revolution. This revolution is of importance since it helps them create and sustain competitive advantages over their non-digital competitors (Huang & Rice, 2017).
3. Battling the burden of knowledge and hypotheses building
The burden of knowledge has to do with a trend where firms face increased team sizes to come up with new inventions and innovations (Jones, 2009). Encountering the burden of knowledge, or decreasing team size to some extent, clearly differs across either digital or non-digital firms, following the insights of the theoretical section. Several aspects of firm nature could cause the burden of knowledge to evolve differently over time.
First, due to the increasing importance of innovation, literature is consistent: in order to successfully innovate, firms need to be highly flexible in today’s rapidly changing industries. Non-digital firms competing in more stable environments may stifle the innovation activities due to the rigidity of the organizational structure (Chen, 2017). This means that established firms within industries with a low willingness to change are less able to innovate.
Secondly, in order to be a successful innovator, a firm needs to be able to recombine existing technologies to drive a firm’s innovativeness (Carnabuci & Operti, 2013). Following the insights by Henderson and Clark (1990), established firms more frequently lose their technological leadership due to their inability to recombine these existing technologies. This actually means that non-digital firms are less able to develop recombinant capabilities, which causes digital firms to take the lead in technologically innovative leadership (Carnabuci & Operti, 2013).
Thirdly, digital firms are able to react towards a rapidly changing environment more easily (Hensen & Dong, 2020). Since digital firms are competing on a platform economy, speed, flexibility and adaptability take on different levels with the arrival of IT (Mikalef et al., 201). Due to the existence of big data, digital firms are creating their own operating model, which differs from the traditional operating model in terms of structure, processes and KPIs (Hensen & Dong, 2020; Cennamo & Santalo, 2013; Gawer, 2009; McIntyre & Subramaniam, 2009).
non-digital firms to change in order to keep the company running (Pouspourika, 2019). The arrival of industry 4.0 combined with the ability of users to innovate themselves these days, made it possible for digital firms to invent and innovate at a much faster rate compared to non-digital firms (Chang & Taylor, 2016; Poetz & Schreier, 2012).
Taken together, this theoretical overview implies that firm nature has a different effect on the burden of knowledge, taken into account that this study considers a firm to be either digital or non-digital. In this case, it means that the effects of the burden of knowledge are minimized by being digital, while they are maximized by being non-digital. In order to test these statements, I hypothesize:
Hypothesis 1a (H1a): Digital firms have a negative effect on the burden of knowledge
Hypothesis 1b (H1b): Non-digital firms have a positive effect on the burden of knowledge
Since the literature on digital business is more extensively conducted in the Industry 4.0 era, it is important to keep the digital era in mind (Hensen & Dong, 2020; Huang & Rice, 2017; Pouspourika, 2019). The industry 4.0 revolution could have different effects on the burden of knowledge pictured within a timeframe. The evolution over time is measurable using data prior to the revolution and data after the evolution had started. However, prior research is not united on the exact starting year of the last industrial, digital evolution. Given the dataset and research conducted earlier, I have chosen to set the 1997-2014 timeframe as the Industry 4.0 revolution. In order to test the evolution of the burden of knowledge over time, I hypothesize:
Hypothesis 2a (H2a): Digital firms have a negative effect on the burden of knowledge
moderated by the Industry 4.0 era (1997-2014)
Hypothesis 2b (H2b): Non-digital firms have a positive effect on the burden of knowledge
moderated by the Industry 4.0 era (1997-2014)
4. Methodology
4.1 Data sources and empirical setting
To test the hypotheses, a quantitative analysis is conducted using secondary patent application data (1902-2014) of U.S.-based firms. The data used originates from two distinct sources, both provided by the USPTO (United States Patent and Trade Office). The first dataset, to investigate patent applications and the time frame, is available through the BASIC_BIB-dataset from the USPTO (2017). The second dataset on team size is available through the INVENTOR-dataset provided by the USPTO (2017). Both INVENTOR-datasets contained the same patent applications which makes it applicable for all variables. Hence, the datasets and sources used were available through the University of Groningen.
4.2 Data sample and transformation
Firstly, the original USPTO BASIC-dataset was used, consisting of 5,842,509 patents issued over the 1902-2014 timeframe. This sample provides us with insights on the patent number, the date it was applied and the technology class it was issued in as provided by the National Bure of Economic Research (NBER) USPTO working paper (Hall et al., 2001). Following the insights of the NBER working paper, the technology class is of an important manner since it shows us the nature of the patent: either digital or non-digital. Using a dummy variable, I linked a patent application to either digital firms or non-digital firms. Appendix B shows the different technology classes used to determine whether a firm is considered to be either digital or non-digital. Moreover, the time frame is of an important manner since it shows the evolution over time. Lastly, the patent number, a unique code attached to every patent issued, shows the information on each patent. This unique patent number is important since it is used to match several variables to the same patent.
prior to applying a patent, which eventually shows every unique ‘adjustment’ to the patent before application. This variable is, just as team size is, useful to research the burden of knowledge.
Merging these two datasets into one new dataset gives us the overview we need to test the hypotheses and answer the research question. Doing so, this study is able to research the burden of knowledge and how it has evolved over time for both digital and non-digital firms. Since it is only relevant to draw conclusions upon the same unique patent identifier, the finalized dataset contains 4,963,225 observations containing all the necessary variables. Taken together, the analysis is conducted on the patent level, using the unique patent identifier attached to every patent applied, which labels the data as cross-sectional data. It is important to understand the different variables and the corresponding differences. An overview of the variables is displayed in table 2.
Variable Description
1. Patent ID 7-digit unique patent number
2. Class Technology class of patent applied
3. Aclass Sub-category technology class of patent applied
4. Team Size Team size per patent applied
5. Ayear Patent application year
6. Digital Dummy variable to determine patent nature
7. Last years Dummy variable to show evolution over time (t) starting in 1997
8. Citations Sum of backward citations per patent applied
Table 2: overview of the variables in the finalized dataset
4.3 Sample statistics
The descriptive figures on the different variables used can be found in Table 3. In order to test the hypotheses and answer the research question, 4,963,225 observations were used and considered. In order to determine whether the data is normally distributed or not, a skewness test was conducted. All variables showed skewness to that extent that P < 0.001. Given that the researched data was panel data, this skewness result is very common.
Variable Observations Mean St. Dev. Min Max
1. Class 4,963,225 359.397 184.982 1 850 2. Digital 4,963,225 0.150 0.357 0 1 3. Team Size 4,963,225 2.326 1.692 1 76 4. Year 4,963,225 1997.226 10.520 1902 2014 5. Last years 4,963,225 0.602 0.490 0 1 6. Citations 4,963,225 16.044 44.370 1 7691
Table 3: descriptive statistics
The correlation matrix, Table 4, shows the correlation of variables. Since this research is fully based on regression tests, it is of an important manner to test for multicollinearity. The rule of thumb is that a correlation between two variables should not exceed the maximum threshold of 0.7. Given the results, multicollinearity only exists between the variables ‘year’ and ‘last years’. Given the fact that last years is a dummy variable resulting from the year variable, this multicollinearity is accepted and therefore not problematic.
1 2 3 4 5 1. Digital 1.000 (0.00) 2. Team Size 0.054* (0.00) 1.000 (0.00) 3. Year 0.195* (0.00) 0.187* (0.00) 1.000 4. Citations 0.043* (0.00) 0.086* (0.00) 0.174* (0.00) 1.000 5. Last years 0.180* (0.00) 0.161* (0.00) 0.843* (0.00) 0.149* (0.00) 1.000
4.4 Method of analysis
The final dataset contains cross-sectional data through the categorization by unique patent identifiers and time series data with the usage of the application year. In order to choose for either random effects or fixed effects, a Hausmann (1978) test was conducted. The results of this test showed significance (p < 0.001) to that extent that running fixed effects regressions appeared to be the most suitable. Since the dataset does not contain panel data, a basic regression is conducted in order to test the hypotheses. The basic regression is explained by using year fixed effects in the regression formulae. In order to test the relationship between the variables in the dataset, the following regression formulae are proposed:
(1) 𝑇𝑒𝑎𝑚 𝑆𝑖𝑧𝑒𝑡 (𝑇𝑆) = 𝛽1 𝐷𝑖𝑔𝑖𝑡𝑎𝑙𝑡 + 𝜀𝑡
(2) 𝑇𝑒𝑎𝑚 𝑆𝑖𝑧𝑒𝑡 (𝑇𝑆) = 𝛽1 𝐷𝑖𝑔𝑖𝑡𝑎𝑙𝑡 + 𝛽2 𝐵𝑎𝑐𝑘𝑤𝑎𝑟𝑑 𝐶𝑖𝑡𝑎𝑡𝑖𝑜𝑛𝑠𝑡+ 𝜀𝑡
(3) 𝑇𝑒𝑎𝑚 𝑆𝑖𝑧𝑒𝑡 (𝑇𝑆) = 𝛽1 𝐷𝑖𝑔𝑖𝑡𝑎𝑙𝑡 + 𝛽2 𝑃𝑜𝑠𝑡 − 1997 𝑑𝑢𝑚𝑚𝑦 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑡+ 𝜀𝑡
(4) 𝑇𝑒𝑎𝑚 𝑆𝑖𝑧𝑒𝑡 (𝑇𝑆) = 𝛽1 𝐷𝑖𝑔𝑖𝑡𝑎𝑙𝑡 + + 𝛽2 𝑃𝑜𝑠𝑡 − 1997 𝑑𝑢𝑚𝑚𝑦 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑡+
𝛽3 𝐵𝑎𝑐𝑘𝑤𝑎𝑟𝑑 𝐶𝑖𝑡𝑎𝑡𝑖𝑜𝑛𝑠𝑡 + 𝜀𝑡
A commonly used standard error term is the robust standard error terms. Regarding these commonalities, during this research I made use of the robust standard error in order to resolve serial correlation errors in the data. Moreover, a Breusch-Pagan LM test for random versus OLS effects was conducted to reconsider using individual specific effects. Hence, the Breusch-Pagan LM test showed individual specific effects which needed to be recovered running an alphafehat test.
Lastly, it is important to know under which circumstances the hypotheses 1 and 2 could be confirmed. Hypotheses 1a holds true if the results show a significant negative result on the relationship between digital firms and the burden of knowledge, displayed as team size with a negative beta parameter. Hypothesis 1b holds true if the results show a significant positive result on the relationship between non-digital firms and the burden of knowledge, displayed as team size with a positive beta parameter. Furthermore, hypothesis 2a holds true if the results show a significant negative result on the relationship between digital firms and the burden of knowledge, displayed as team size while taking into account the Industry 4.0 era, or 1997-2014 timeframe. Hypothesis 2b holds true if the results show a significant positive result on the relationship between non-digital firms and the burden of knowledge, displayed as team size while taking into account the Industry 4.0 era, or 1997-2014 timeframe.
5. Results
5.1 Hypothesis testing: H1a and H1b
backward citation control variable proves to be significant while it shows a small, positive beta-parameter (β = 0.00212, p < 0.01). This indicates that an increase of one citation adds up 0.00212 onto team size per patent. While model 2 has an explanatory factor which is three times higher compared to model 1, it seems that model 2 is the most appropriate one to derive results from. Hence, based on the regression results, the following can be concluded regarding hypothesis 1a and 1b.
Hypothesis 1a states that being digital has a negative effect on the burden of knowledge, which means that team size would decrease while being digital, cannot be confirmed. The beta-parameters show a significant increase with respect to the relationship between digital firms and the burden of knowledge. Hypothesis 1b states that being non-digital has a positive effect on the burden of knowledge, which means that team size would increase while being non-digital, is confirmed following the regression results. Both digital and non-digital firms show a significant increase in team size given the whole timeframe 1902-2014.
Variables Model 1 Model 2
Digital 0.0849*** (0.0104) 0.0828*** (0.0105) Backward Citations 0.00212*** (0.0002) Constant 2.313*** (0.00156) 2.279*** (0.00266) Observations F-statistic 4,963,225 1583.80*** 4,963,225 153.08*** R-squared 0.000 0.003
Year FE Yes Yes
Dependent variable: team size per patent Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 5: models 1 and 2 for hypothesis testing (1a and 1b)
5.2 Hypothesis testing: H2a and H2b
displays a regression with the effect of being digital or non-digital on team size, while taking into account the Industry 4.0 era, meaning that the regression is conducted on the 1997-2014 timeframe. Taking the results of model 3 into account, there is a significant decrease of team size while being digital, since β = -0.146, p < 0.01. This means that, although team size has increased over time given the results of hypothesis 1a and hypothesis 1b, digital firms are experiencing a less educational burden explained as team size in the digital era compared to non-digital firms. The beta-parameter shows that digital firms have a team size that is 0.146 smaller compared to the constant than non-digital firms do. Model 4 displays the addition of the commonly used control variable backward citations. Hence, the beta-parameter does change and shows a significant decrease in team size with respect to digital firms, since β = -0.148, p < 0.01. The impact of backward citations remained unchanged compared to model 1 and 2, showing a significant increase since β = 0.00212, p < 0.01. This confirms that one citation adds up 0.00212 onto the team size per patent. Given the fourth model, it remains clear that digital firms experience a less educational burden expressed as team size compared to non-digital firms.
Variables Model 3 Model 4 Digital 0.200*** (0.0181) 0.200*** (0.0185) Industry 4.0 dummy variable -0.146*** (0.0193) -0.148*** (0.0197) Backward Citations 0.00212*** (0.0002) Constant 2.313*** (0.0010) 2.280*** (0.0026) Observations F-statistic 4,963,225 93.42*** 4,963,225 127.54*** R-squared 0.006 0.008
Year FE Yes Yes
Dependent variable: team size per patent Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 6: models 3 and 4 for hypothesis testing (2a and 2b)
All four models were tested using the robust standard error terms since heteroscedasticity could have caused less efficient results. These robust standard errors are found in the parentheses below each beta parameter compared to the values below the model. The R-squared values increased while adding variables to the last model, model 4, which appeared to be the most appropriate model to test the hypothesis. Hence, F-statistics in all four models show a good model fit for model 1 (F = 1583.80, p < 0.01), model 2 (F = 153.08, p < 0.01), model 3 (F = 93.42, p < 0.01) and model 4 (127.54, p <0.01).
6. Discussion
6.1 Key findings
et al., 2017). As aforementioned, the burden of knowledge implies that if knowledge accumulates
as technology advances, successive generations of innovators may face an increasing educational burden (Jones, 2009). With team size used as a proxy to mirror the burden of knowledge, this research shows an increase of team size regarding the 1902-2014 timeframe. Hence, research on the burden of knowledge conducted in the past was fully focused on the existence of the burden of knowledge without considering different firm natures. This study is conducted in order to show differences between digital firms on the one hand and non-digital firms on the other. Since digital firms are able to disrupt entire established industries, the burden of knowledge in relation to digital firms needed clarification. Regarding the results of this study, considering the timeframe 1902-2014, there is no empirical evidence found that digital firms do indeed have a lower increase in team size than non-digital firms do. On the contrary, the entire dataset considered, digital firms do have a larger increase in team size than non-digital firms do. Clarification for the larger increase in team size for digital firms compared to non-digital firms could have several reasons. First, summarizing the dataset shows that just 14% of the patents in the dataset can be considered digital. Second, the rise of digital firms has started in the last decade (1990-2000), which can cause biased results when investigating the whole dataset. Lastly, digital firms are nearly started to act as a common competitor in present-day industry, since they are still at the initial stages of business. However, since digital firms are more present in the last years of the business landscape (Chen, 2012; Huang et al., 2017; Dong, 2019) and in the USPTO dataset, throughout the study named as the Industry 4.0 era, this study shows an empirical evidence that digital firms do indeed have a less educational burden compared to non-digital firms, given the fact that team size has decreased in the 1997-2014 timeframe. Based on these findings, hypothesis 1a cannot be confirmed and does not hold true. Hypothesis 1b, 2a and 2b however are fully confirmed following the results and findings of this study.
firms and their ability to innovate, it has never been linked to the burden of knowledge phenomenon. This study connects those two highly important facets of current innovation and strategy literature to create a better understanding of the innovation capabilities of digital firms.
6.2 Implications
This study has implications for both practitioners and academics. First, regarding practitioners, digital natured firms tend to face a lower educational burden in the digital era. This study shows that it seems indeed challenging to overcome the burden of knowledge in general since team size is increasing, but that it could be somewhat easier if your firm is digital based. The extremely fast speed and rate of flexibility and adaptability could help firms battle the burden of knowledge and eventually lower team size. Next to that, despite the fact that team size has increased over time regarding the full 1902-2014 timeframe, being digital could help firms to increase their learning capabilities. In that case, a small increase in team size could outweigh the benefits of doing business digitally. Hence, based on the results, established non-digital firms could change the scope of their business, since they still face an increase in team size despite the timeframe. The results indicate that there exists an increased importance for non-digital firms to do business more frequently on a digital basis. Next to that, the results show that regarding to digital firms, it is important to be aware of the advantages platform economies provide firms with. Speaking of a (sustainable) competitive advantage, as mentioned earlier in this study, digital firms tend to create capabilities that are more suitable to strive for innovational excellence. The advantages digital firms have over non-digital firms are therefore twofold. First, digital firms are able to combine capabilities more easily relative to non-digital firms which puts them in the lead when aiming for innovational improvement. Second, digital firms are competing faster and experience higher levels of flexibility and adaptability to that extent that they are always one step ahead relative to non-digital firms. It seems clear that speed, flexibility and adaptability are three important pillars which non-digital firms should strive for in order to overcome the burden of knowledge more easily. Moreover, it could help established firms to improve their innovation performance and eventually their firm performance. The fact that we can no longer mis out on digital firms was already stated, but the results prove that this statement is fully correct.
phenomenon (Hall et al. 2001; Jones, 2009; Wuchty et al., 2017). It can therefore be concluded that this study enriches the literature on the burden of knowledge. Second, although digital business faces an increasing amount of research nowadays, the firm nature in relation to the burden of knowledge has not been researched before. This means that this study fills the research gap on firm nature in relation to the phenomenon. Next to that, by filling this research gap, it becomes clear which differences between digital firms and non-digital firms are the most important while researching the burden of knowledge, but also when researching innovation. Moreover, it creates possibilities for future research to investigate new insights on the burden of knowledge on the one hand and doing business digitally on the other.
7. Limitations and future research
Despite the fact that this research has enriched current literature on the burden of knowledge, just as any research it suffers from some limitations. First, since the dataset was provided by the USPTO, fruitful insights are probably left out of the secondary data. For example, digitalization in U.S.-based firms could be concentrated among specific places within the country (e.g. Sillicon Valley). Since these locations are not taken into account investigating the data, it results in a probability of construct validity of the variables which eventually could have had a certain relationship towards team size. Consequently, it results in less used control variables since the dataset is not collected specifically for this study.
Secondly, due to the lack of data, team size on its own was used as a proxy of the burden of knowledge. In order to understand the phenomenon, apart from firm nature, multiple variables could be taken into account to enrich literature on the burden of knowledge. One can enrich it by using a moderating analysis for example. Variables as firm size, cultural factors, (technological) distance might be very interesting to investigate.
Thirdly, since there has not been any research done in the field of digital firms and the burden of knowledge, I needed to combine separate studies in order to find the best combination possible. This made it hard to find the right direction in the literature review and to provide the research with all the necessary information on the phenomenon.
this phenomenon. This means that there is still a lot to explore on this research theme. As frequently stated during this study, the business landscape is changing due to the presence of digital natured businesses. Since this study is limited to the patent-level, it could be very interesting to choose the scope of digital business on its own and enrich the literature on digital business and the educational burden it experiences.
Next, this study has focused on a large timeframe at first, 1902-2014. The results are clear: digital business has most of its impact in the Industry 4.0 era of the dataset since this timeframe mirrors the digital era. Future research should be based on this digital era to provide the literature with clear suggestions on the burden of knowledge in relation to the digital era we currently work in. Furthermore, since the Industry 4.0 era is not yet established, it could be highly relevant to study the phenomenon in relation to digital firms from 2014 onwards.
8. Conclusion
to digital firms than it is to non-digital firms?’. The results suggest that digital firms do have,
9. References
Alom, Z., Tarek, M.T., Yakopcic, C., & Westberg, S. 2019. A State-of-The-Art Survey on Deep Learning Theory and Architectures. Electronics, 8(3): 1-67.
Barnett, T., & Clarke, S.R. 1998. Combining Player Statistics to Predict Outcomes of Tennis Matches. IMA Journal of Management Mathematics, 16(2): 113-120.
BCG (2019). Innovation in 2019: The most innovative companies in 2019. doi: https://www.bcg.com/publications/2019/most-innovative-companies-innovation.aspx
Bloom, N., Jones, C.I., van Reenen, J. & Webb, M. (2017). Are Ideas Getting Harder to Find? NBER Working Paper, No. 23782.
Carnabuci, G., & Operti, E. 2013. Where do firms’ recombinant capabilities come from? Intraorganizational networks, knowledge, and firms’ ability to innovate through technological recombination. Strategic Management Journal. 34(13): 1591-1613
Cennamo, C., & Santalo, J. 2013. Platform Competition: Strategic Trade-Offs in Platform Markets. Strategic Management Journal, 34: 1331–1350.
Chang, W., & Taylor, S.A. 2016. The Effectiveness of Customer Participation in New Product Development: A Meta-Analysis. Journal of Marketing, 80(1): 47-64.
Chen, H., Chiang, R.H.L., & Storey, V.C. 2012. Business Intelligence and Analytics: From Big Data to Big Impact. MIS Quarterly, 36(4): 1165-1188.
Chen, S. 2017. The Relationship Between Innovation and Firm Performance: A Literature Review. Computer Science Research. 82: 648-653.
Collis, D.J. 1994. How Valuable Are Organizational Capabilities? Strategic Management
Dong, J. Q. 2019. Moving a Mountain with a Teaspoon: Toward a Theory of Digital
Entrepreneurship in the Regulatory Environment. Technological Forecasting and Social Change, 146(1), 923-930.
Dong, J. Q., & Yang, C.H. (2019). Information Technology and Innovation Outcomes: Is Knowledge Recombination the Missing Link? European Journal of Information Systems, 28(6): 612-626.
Eisenmann, T. R., Parker, G. G., & Van Alstyne, M. W. 2006. Strategies for TwoSided Markets. Harvard Business Review, 84: 1–11.
European Central Bank, 2001. Sources of Economic Renewal: From the Traditional Firm to the Knowledge Firm, Working Paper Series, no. 43
Gawer, A. 2009. Platform Dynamics and Strategies: From Products to Services. In A. Gawer (Ed.), Platforms, Markets and Innovation: 45–76. Cheltenham, U.K, Northampton, MA: Edward Elgar.
Hall, B.H., Jaffe, A.B. & Trajtenberg, M. 2001. THE NBER PATENT CITATIONS DATA FILE: LESSONS, INSIGHTS AND METHODOLOGICAL TOOLS, NBER Working Paper no. 8498.
Henderson, R.M., & Clark, K.B. 1990. Architectural Innovation: The Reconfiguration of Existing Product Technologies and the Failure of Established Firms. Administrative Science
Quarterly. 35(1): 9-30.
Hensen, A. H. R., & Dong, J. Q. (Accepted/In press). Hierarchical Business Value of Information Technology: Toward a Digital Innovation Value Chain. Information and
Huang, J., Henfridsson, O., Liu, M.J. & Newell, S. 2017. Growing on Steroids: Rapidly Scaling the User Base of Digital Ventures Through Digital Innovation, MIS Quarterly, 41(1): 301-314.
Huang, J., Henfridsson, O., Liu, M.J., & Newell, S. 2017. Growing on Steroids: Rapidly Changing the User Base of Digital Ventures Through Digital Innovation. MIS Quarterly, 41(1): 301-314.
Huang, F., & Rice, J.L. 2009. The Role of Absorptive Capacity in Facilitating ‘Open
Innovation’ Outcomes: A Study of Australian SMEs in the Manufacturing Sector. International
Journal of Innovation Management, 13(2): 201-220.
Jones, B.F. 2009. The Burden of Knowledge and the “Death of the Renaissance Man”: Is Innovation Getting Harder?, Review of Economic Studies. 76 (1): 283-317.
KPMG. 2019. The KMPG Innovation Factory: Smart Tech Solutions. doi: http://www.smarttech.kpmg.nl
McIntyre, D. P., & Subramaniam, M. 2009. Strategy in Network Industries: A Review and Research Agenda. Journal of Management, 35: 1494–1517.
McKinsey & Company. 2019. Growth and Innovation. Strategy and Corporate Finance. doi: https://www.mckinsey.com/business-functions/strategy-and-corporate-finance/how-we-help-clients/growth-and-innovation
Mikalef, P., Pappas, I.O., Krogstie, J., & Giannakos, M. 2017. Big Data Analytics
Capabilities: A Systematic Literature Review and Research Agenda. Information Systems and
e-Business Management, 16: 547-578.
Nelson, R.R., & Winter, S.G. 1982. An Evolutionary Theory of Economic Change. The
Nieto, M.J., & Santamaria, L. 2007. The Importance of Diverse Collaborative Networks for the Novelty of Product Innovation. Technovation, 27(6): 367-377.
Poetz, M.K., & Schreier, M. 2012. The Value of Crowdsourcing: Can Users Really Compete with Professionals in Generating New Product Ideas? Journal of Product Innovation
Management, 29(2): 245-256.
Porter, M. 1985. The Competitive Advantage: Creating and Sustaining Superior Performance. NY: Free Press.
Teece, D.J., Pisano, G., & Shuen, A. 1997. Dynamic Capabilities and Strategic Management.
Strategic Management Journal, 18(7): 509-533.
Verhoef, P.C., Broekhuizen, T., Bart, Y., Bhattacharya, A., Dong, J.Q., Fabian, N. & Haenlein, M. 2019. Digital transformation: A multidisciplinary reflection and research agenda.
Journal of Business Research, in press.
Wuchty, S., Jones, B.F & Uzzi, B. 2007. The increasing dominance of teams in production of knowledge. Science 316, (5827): 1036-1039.
Zhu, F., & Iansiti, M. 2012. Entry Into Platform-Based Markets. Strategic Management
10. Appendices
Appendix B. Technology classes USPTO overview – Digital determination
Category code Category name Sub-cat. code Sub-cat. name Patent class nr.
2 Computer & Communications 22 Computer Hardware and Software 341, 380, 382,700-710, 712-725, 726 23 Computer peripherals 345, 347 24 Information storage 360, 365, 369, 711
• 341 = coded data generation or conversion • 380 = cryptography
• 382 = image analysis
• 700 = data processing: generic control systems or specific applications • 701 = data processing: vehicles, navigation, and relative location • 702 = data processing: measuring, calibrating, or testing
• 703 = data processing: structural design, modeling, simulation, and emulation
• 704 = data processing: speech signal processing, linguistics, language translation, and audio compression/decompression
• 705 = data processing: financial, business practice, management, or cost/price determination
• 706 = data processing: artificial intelligence
• 707 = data processing: database and file management or data structures • 708 = electrical computers: arithmetic processing and calculating
• 709 = electrical computers and digital processing systems: multicomputer data transferring
• 710 = electrical computers and digital data processing systems: input/output
• 712 = electrical computers and digital processing systems: processing architectures and instruction processing (e.g., processors)
• 713 = electrical computers and digital processing systems: support • 714 = error detection / correction and fault detection / recovery
• 715 = data processing: presentation processing of document, operator interface processing, and screen saver display processing
• 716 = computer-aided design and analysis of circuits and semiconductor masks • 717 = data processing: software development, installation, and management • 718 = electrical computers and digital processing systems: virtual machine task or
process management or task management/control
• 719 = electrical computers and digital processing systems: interprogram communication or interprocess communication (IPC)
