1 Master Thesis – Text Analysis for consultancy firms
“To what extent can information problems be reduced for
external funding in innovation project descriptions”
A quantitative empirical studyfocusing on information asymmetry
AUTHOR
Willem Lieuwe Hoekstra | S3538745
MSc BA -Strategic Innovation Management
University of Groningen | Faculty of Economics and Business
Thesis Supervisor: Dr. F. Noseleit
Co-Assessor: Prof. Dr. P.M.M. de Faria
2
Abstract
Building up on information asymmetry, which states that quality is hard to asses beforehand as one party has more or better information than the other to make the decision. This paper tries to explain the relationship of innovativeness of project descriptions on the amount of funding received and how previous experience and additional information influences this outcome. Three hypotheses are created and tested on a sample of 106 project descriptions from firms located in the Netherlands. Text analysis, by experts from a Dutch consultancy firm, is performed on these project descriptions to grade the innovativeness of these projects. A positive significant relationship has been found for innovativeness on the amount of funding received. Also, for the moderating effects: additional information, and previous experience, a positive significant relationship has been found on the relationship innovativeness and the amount of funding received. Thereby, confirming all three hypotheses. This research contributes by exploring how innovativeness influences external funding, and how information problems can be reduced.
Keywords:
3
1. Introduction
A new revolution can be seen within the world, the revolution of big data management. Large amounts of data are created every day and the amount is constantly increasing. At this moment there is more data created and spread on the internet every second than was stored 20 years ago on the entire internet (McAfee, Brynjolfsson, Davenport, Patil, & Barton, 2012). For example, Walmart collects more than 2.5 petabytes of data to increase performance of the firm and exploit new opportunities (Marr, 2019). McAfee et al. (2012) state that data helps managers with their decisions and that data management in itself has the potential to directly revolutionize management and the performance of a firm.
“Data-driven decisions are better decisions, it helps managers to create
decisions based on evidences rather than intuition”
(Andrew McAfee and Erik Brynjolfsson, 2012, p.5)
Exploiting data is an important necessity and requirement for firms to use the data to their advantage (McAfee et al., 2012). Linoff and Berry (2002) describe three different information technologies used to discover knowledge in text: (i) information retrieval, (ii) document clustering, and (iii) data-mining. The structure of the data is used by Hui and Jha (2000) to determine to which stream it includes, structured data to data-mining and unstructured data to text-mining. The unstructured data has different pieces of information with varying lengths, written in a free form, which are particular interesting for this research. Humans can easily interpret and derive conclusions from unstructured datasets, computers however have a greater challenge in deriving conclusions from unstructured data (Stevens, 2014). Using computers to discover new opportunities do have several challenges but also offer create opportunities. Computers are able to find patterns within large datasets faster and better compared to humans (Sebastiani, 2002). Studies have shown that text-mining can create deep insights which then can be further exploited to the firms’ advantage (Hui, 2017; Jia, 2018; Kim & Park, 2019). This study will explore the possibilities of text-analysis on innovative project descriptions used for subsidies.
1.1 Problem Statement
4 innovativeness measures to decide which projects are innovative. However, there is a lack of clarity on which innovativeness measures are suitable to determine the innovativeness of projects (Möller, Steinmann, & Calabretta, 2014). Within the literature two types of innovativeness measures can be found: (i) input measures, and (ii) output measures (Arvanitis, 2008). It can be said that both these measures are not representative on how innovative the project/idea is – a lack of i.e. R&D employees does not unequal innovativeness (Brouwer & Kleinknecht, 1999). Furthermore, the necessary input and output statistics required for these measures might be missing as the firm is too young or small (Hoffman, Parejo, Bessant, & Perren, 1998). Therefore, it can be said that current measures have several major drawbacks when judging the innovativeness of a project or idea (Brouwer & Kleinknecht, 1999).
Even though subsidy providers have large sets of data, they cannot apply traditional innovation measures to judge the innovativeness of the projects due to i.e. missing statistics. Therefore, a different measure for innovativeness is required to better understand which projects are innovative and tackle the information asymmetry. Text analysis on the project descriptions could provide subsidy providers with the required measure and tackle the information problems. Hence, the question occurs wheter project descriptions can be used as a measure for innovativeness to judge the projects. The current literature field has researched whether text-based can be used as an innovation measure (Bellstam, Bhagat, & Cookson, 2017). However, limited their scope to the S&P 500 leading companies within the United States. This research will include firms from different industries, age and size.
1.2 Research Question
Based on the identified literature gap, research questions have been derived. The research question focuses on whether text-based descriptions can predict innovativeness and thereby lead to a new measure of innovation and further reduce information asymmetries. Furthermore, it will be researched whether innovativeness is linked to project funding. This results in the following research questions:
RQ.1: Can text analysis be used to assess the potential innovativeness of an innovation project description?
RQ.2: Is the innovativeness of a project description positively related to funding success in terms of received amount of project funding?
RQ.3: Can information problems be reduced in project descriptions.
5 • This paper empirically tests if expert rating of the innovativeness of project descriptions can
explain the success of external funding in terms of amount of funding received.
• By assessing the possibilities to utilize text analysis to assess the innovativeness of a project, this paper will contribute to the agency theory by minimizing information asymmetries.
Moreover, the empirical research conducted has several managerial implications specifically for a consultancy firm. They possess large datasets and are looking for new ways to analyze their applications on innovativeness and thereby limit the information asymmetry. By having a more effective and efficient way of analyzing subsidy applications, the overall project performance could be improved.
This paper is structured as follows: in the first chapter after the introduction, the theoretical framework is explained which further elaborates on the research question and includes the hypothesis. Then the institutional setting of the WBSO is explained. In the next chapter the data and methodology of this research is discussed. Then the results will be presented created by the analysis. This paper concludes with a discussion and conclusion with the key findings regarding the results. At last the managerial implications, the limitations of this research, and future research will be described.
2. Theoretical Framework
This theoretical chapter further elaborates on the research questions drawn in the previous chapter and cover the literature field focusing on information asymmetry and innovativeness. Firstly, I will describe the current literature about the three main aspects – information asymmetry, innovation measures, and the contextual setting – and discuss them from a theoretical perspective. Secondly, the hypotheses are drawn. At last the conceptual model created from the drawn hypothesis can be found within this chapter.
2.1 Information Asymmetry
6 and therefore is obliged to hire people with specialized skills to perform the task. The central concern for the firm is how to have the agent behave in the interest of the principle. The agency theory deals with the difficulties which arise of incomplete information and information asymmetries between the principle and agent (Eisenhardt, 1989).
Within the subsidy industry, both of the two potential relationships of the agency theory can be found; adverse selection and moral hazard (Mishra, Heide, & Cort, 1998). Recently, firms committed fraud to be able to receive subsidies from the RVO (the subsidy provider of the Netherlands).1 The firms received over 1 million euro in subsidy money meant for innovation research, but was used for machines and salaries. By making these false claims, the subsidy providers inability to determine the applicant’s goal, which results in the adverse selection as described by Akerlof (1970). The applicant might know for example that the subsidy project will fail but still applies for the subsidy as it can be used for signaling for private investors or quality of the project (Kleer, 2010). Furthermore, this is also in line with moral hazard in which the applicant changes its behavior after the deal is setand using the received subsidy for other purposes (Pauly, 1968). Mishra et al. (1998) state that the relationship is often characterized by information asymmetry in which one party has more information than the other. As the previous mentioned example shows, also subsidy providers do not fully understand the goals of the applicant and thereby lack information to have the agent (applicant) behave in the best interest of the principle (provider). Information problems create concerns for both parties (Mishra et al., 1998): the subsidy provider who cannot correctly evaluate the project, and for the subsidy applicants whose projects are based on quality but whose offerings are difficult to differentiate from the lower-quality ones.
Possibly, information asymmetry can lead to the breakdown of a market (Akerlof, 1970). This same principle could also be applied to the subsidy system in which the system crashes due to too many information failures (Hall, 2002). Healy & Palepu (2001) explain it as a situation where investors can choose from ideas from which half are good and the other half are bad. When investors cannot distinguish the bad ideas from the good ones, the owner of a bad idea will try to sell his idea as a valuable good idea. This lemon problem as explained by Akerlof (1970), will cause that investors will average the ideas to minimalize the risk. According to Peneder (2008), the accuracy of the distribution of resources depends on two critical factors: (i) information availability, and (ii) ability to interpret information. When the lemon problem is not resolved, bad ideas will be overvalued while good ideas will be undervalued resulting in less accurate distribution of resources. This whole principle can be directly linked towards the subsidy system. When the subsidy provider is not capable in distinguishing high and low innovative ideas, innovative projects will be undervalued and non-innovative project will be overvalued. This causes that funding and resources will be inefficient distributed over the projects resulting that not the maximum output can be achieved with the funding. According to the literature,
7 there are several well-known solutions to prevent information asymmetry: optimal contracts to provide incentives for full disclosure (Healy & Palepu, 2001), or restructuring the transaction to have the party making the actions perform in the interest of the second party for its own benefit (Mishra et al., 1998). However, the best solution would be for both parties, to fully disclose the information, and thus preventing information asymmetry and possible breakdown of the system (Healy & Palepu, 2001).
2.2 Measure innovativeness
Many studies have shown that innovativeness is directly positively linked to the performance of a firm (Hult, Hurley, & Knight, 2004; Kleinschmidt & Cooper, 1991; Rhee, Park, & Lee, 2010) making it an interesting factor for firms to measure. By increasing the innovativeness, firms can directly extend and improve their survival, profitability, and growth (Avlonitis & Salavou, 2007). Existing research shows that a U-shaped relationship can be found (Avlonitis & Salavou, 2007; Kleinschmidt & Cooper, 1991). Furthermore, it can be traced back to the famous old aphorism: “What gets measured gets done” (Behn, 2003). When setting a specific target and by measuring the progress: (i) firms will be more focused, (ii) base their decisions on previous results, and (iii) will put more effort in achieving the set goal (Locke, Shaw, Saari, & Latham, 1981). This makes it interesting to measure innovativeness and set goals to further improve the performance of the firm. According to Van Dijk, Den Hertog, Menkveld, & Thurik (1997), a large number of studies has concluded that small firms can be just as innovative as larger firms, making it interesting for both firm types to measure innovativeness.
8 & Traoré, 2008; Harabi, 1995) and that patents are more important for discrete products such as pharmaceutical and chemicals (Hall et al., 2014). These missing statistics directly influence how innovativeness can be measured as traditional innovation measures depend on factors such as firm-size and industries and that traditional measures cannot be applied to all firms due to firm-characteristics.
Brouwer & Kleinknecht (1999) found that the two most commonly used indicators for innovativeness – R&D and patenting – have even more weaknesses than often assumed and find little correlation between the five indicators researched by them. Acs et al. (2002) did find empirical evidence suggesting that patents could provide a reliable measure for innovativeness but do claim it is not perfect. Patents as an indicator for innovativeness have several weaknesses as not all innovations can be patented, are not patented in general (secrecy), or are used for defensive reasons (Bellstam et al., 2017). Not only for patenting weaknesses can be found, but also for other innovativeness measures: R&D – it is an input and therefore can be used more or less efficient, it has a manufacturing bias, and undercounting of small firm R&D (Brouwer & Kleinknecht, 1999), Total Innovation Expenditures – none R&D related activities are difficult to judge (Kleinknecht, 1993), New product Announcements – depending on the journals selected, process innovations not equally covered, and some firms choose not to publish in journals (Kleinknecht, 1993), and significant innovations – depend on the quality of the experts, statistical procedures cannot be applied, and best assessment is ex-post (Brouwer & Kleinknecht, 1999). Therefore, it can be said that current indicators with their biases do not reflect the innovativeness of a project and are limited to firm characteristics. Nonetheless, the empirical research has often used these measures as an indicator for innovativeness (Archibugi, 1992; Pavitt, 1982). Bellstam et al. (2017) say that by using a text analysis innovation measure, firms which lack statistics for traditional innovation measures such as patents or R&D could also be included which would expand the scope of innovativeness that can be researched.
2.3 Governmental funding
9 funding (Rijksdienst, 2019). Leyden, Link, & Bozeman (1989) found that governmental R&D funding positively influenced private R&D innovation, 5 years before the WBSO started in the Netherlands. Furthermore, research shows that public funded firms have higher R&D expenditures and no external financial constraints (Czarnitzki, 2006). Additionally, by receiving an R&D subsidy, it creates a positive signal about the quality which results in better access to long-term finance possibilities (Meuleman, & De Maeseneire, 2012). Lerner (1996) looked at firms which did and did not receive funds and found that firms that did receive funding grew faster. It can therefore be said that external financing is ideal for firm’s which do not possess the required financial resources internally itself but do want to exploit new opportunities created by R&D research (Peneder, 2008).
Firms requiring investments for intangible assets tends to be more difficult than physical assets, as they are riskier and harder to collateralize (Czarnitzki, 2006). Governments therefore provide two different methods of incentives to increase innovation: (i) fiscal incentives, and (ii) direct funding of targeted expenditures (Peneder, 2008). Especially in the early stage of the firm cycle, bank loans and government support play an important role for financing new projects (Nofsinger & Wang, 2011). However, many firms may not even know tax incentives exist or do not apply for them due to the costs involved or are inexperienced with dealing with authorities (Baghana & Mohnen, 2009). Fiscal incentives are mainly interesting for firms which are profitable as they are related to corporate income taxes therefore less interesting for smaller firms (Baghana & Mohnen, 2009). Lokshin and Mohnen (2007) say that small firms are more sensitive towards direct funding than large firms as small firms have difficulty in gathering finances for their projects. According to Peneder (2008): “Direct funding instruments give governments more scope to make deliberate choices about which projects they want to support. In contrast, fiscal incentives generally leave these decisions to the firms themselves” (p. 13). Showing that direct funding has some major benefits when boosting specific areas, firm types, or industries and therefore important for innovation policies to target (Peneder, 2018).
10 appropriation of the returns. Many countries providing governmental funding mainly focus on SMEs, as SMEs are more dealing with market failures and information (Baghana & Mohnen, 2009). Furthermore, as large firms rely more on internal finance than on external finance, small firms are more constraint than large firms (Fazzari, Hubbard, & Petersen, 1987). These tighter constraints are due to information barriers and fixed costs related to the access of the financial systems (Beck, Demirguc‐Kunt,
Laeven, & Levine, 2008). Governmental funding improves these financial gaps within the market which have especially a positive effect on smaller firms.
2.4 Hypothesis & Conceptual model
The conceptual model, reported in the end of this section, is constituted from factors which could be relevant in explaining the difference in received amount of project funding across firms. As the application process for subsidies is standardized with specific rules, firm characteristics; i.e. size, age or earnings, are not taken into account or influence the assessment of the subsidy provider. Translated from the Dutch RVO (2019a) website: “It does not matter what the size is of the firm or in which industry it operates, all firms can apply”. Therefore, it is likely that the only factor that can influence the difference of the received amount of project funding across firms is the project description provided by the applicant. Project descriptions could be influenced by several factors explaining the difference. However, it is important to know that all these factors explaining the difference can be linked back to the innovativeness of the project. This information is used to provide the subsidy, as the main goal of the applicant is to provide the subsidy provider the necessary innovative information (RVO, 2019b). As mentioned before, asymmetric information between the subsidy provider and the applicant, could positively or negatively influence the decision-making process for the applicants. Therefore, the hypothesis development will focus on the aspect of asymmetric information.
Innovativeness - Expert Rating
It can be assumed that the experts of a consultancy firm have gained experience and understanding which projects are innovative as they have encountered and worked with a large number of projects. As innovativeness is the most important criteria for a project to receive the subsidy (RVO, 2019a), the assumption is created that high-innovative projects will receive more funding than low-innovative projects. When experts would be able to predict whether projects are innovative, it could save consultants time and work more efficient by allocating resources to more promising projects.
11 products focusing on incremental innovations are more understandable and thereby generate more funding compared to products with radical innovations. They argue that this is caused by higher risks and harder understanding of the radical innovation and that customers prefer incremental innovations because these are more familiar and beneficial to the user. Marom & Sade (2013) found that projects on Kickstarter – an online crowdfunding platform – with descriptions had higher rates of success compared to projects without the descriptions. This can be linked back to the information problems in which project quality cannot be signaled and thereby creating a failure (Baghana & Mohnen, 2009). Scitovsky (1954) argued that non-market institutions such as the government arise when market transactions are not mediated by economic agents to fill the gap left by them. Subsidy providers should therefore be trained in understanding which projects would not get funded by market institutions. According to Kortum & Lerner (2001) there is a difference between different external funders, equity financiers and crowd-funders focus on different factors of importance. They say that crowdfunding is complementary to the traditional equity financing. Early stage investors prefer radical innovative projects with the possibility for extraordinary profits while crowd-funders mainly focus on incremental projects as they lack knowledge which projects have high outcomes and search for products closer related (Chan & Parhankangas, 2017). Even though the experts are not looking for returns on investment, the goal of the experts and early stage investors are similar. While early stage investors want innovative projects and receive a high ROI, subsidy experts also want innovative projects and thereby receive subsidies. Research shows, that experts are different compared to novices and are better in their tasks. (Brand-Gruwel, Wopereis & Vermetten, 2005; Cleary & Zimmerman, 2001; Mandel & Johnson, 2002; Murray, 1999). For example, the results of Murray (1999) showed that experts had significant lower investments failures and homerun outcomes compared to non-experts. Furthermore, Chan & Parhankangas (2017) argue that experts are capable in finding promising projects with extraordinary returns.
It is likely that with the experience gained from working on a large number of subsidy projects the experts of the consultancy firm are capable in finding high innovative projects which would increase the amount of received project funding. Therefore, the following is hypothesized:
Hypothesis 1. The innovativeness of the project description is positively related to the project funding in terms of amount approved.
Information asymmetry
12 Furthermore, the balance of arguments favoring for receiving the subsidy are likely to be strengthened or changed. According to Stasser & Titus (1985), there is a spectrum of knowledge sharing, on one side “shared information” where all members have the same information, and on the other side “unshared information” where information is held by one or a small group of the members. They show in their example that this spectrum can have a large influence and bias on the outcome. As there is only one fair possible outcome for the subsidy system, if applicants commit fraud they have to pay back the subsidy plus interest and risk consequences as a recent example showed 2, it is in best interest for both parties to fully disclose and share the information to improve the quality of the decision making (Healy & Palepu, 2001). However, Stasser & Titus (1985) found that providing all information available created an information overload decreasing the overall result. In this situation there can be found an abundance of information but it is difficult to obtain the useful and relevant information needed (Edmunds & Morris, 2000). This could also be the case for applicants who provide too much irrelevant information which makes the judgement harder for the subsidy provider resulting in a question letter from the subsidy provider. In total this results in three possible outcomes when applying for a subsidy:
i. Low innovativeness: The project description does not show enough innovativeness, based on this information the subsidy provider has two options, denying subsidy or requesting additional information to provide the innovativeness information and draw a better conclusion.
ii. High innovativeness: The project description shows all the required innovativeness, based on this information the subsidy provider approves the subsidy request.
iii. High innovativeness – Complex/Asymmetry: The project description shows high innovativeness but it is not (fully) understood by the subsidy provider and needs more information to draw a better conclusion and will therefore request additional information. If the applicant receives a question letter from the RVO it failed to provide the necessary information within the project description and therefore has to provide detailed information about for example the technical complexity of the project, materials used, investments costs, and risk analysis to convince the RVO that the project is innovative (Innovencio, 2019). The following is assumed: providing additional information positively moderates the relationship innovativeness and received amount of funding (H1) as additional and specific information is added to the pool of information from which the subsidy provider (RVO) can draw conclusions resulting in the following hypothesis:
Hypothesis 2. Additional information has a positive moderating effect on the relationship innovativeness of the project description and project funding in terms of amount.
The accuracy on how resources are distributed is not only depended on the information available but also on the ability to interpret information (Peneder, 2008). Several factors influence the efficiency on
13 how firms understand and interpret information (Schlemmer & Mash, 2006). Knoblich, Ohlsson, & Raney (2001) argue that research shows a relationship between previous experience and problem solving in which previous experience appears to be positive to problem solving. According to Huber, G. P. (1991), there are five constructs of knowledge acquisitions from which: learning from experience and building on previous information one is. According to March (1991), organizations collect information by storing knowledge in their procedures, norms, rules, and forms over time and have to make choices between using new information or previous information for future projects. Furthermore, as results over time will affect the lessons learned, organizations will learn from experience what will be best for the firm to use (March, 1991). By creating routines which are history dependent and target oriented, firms can learn from previous experiences which will guide the behavior of a firm (Levitt & March, 1988). Slow learning improves the overall knowledge found in the organization as it can be accumulated over a longer time period. Slow learning maintains diversity longer, thereby providing the exploration that allows the knowledge found in the organizational code to improve (March, 1991). This implies that firms which have learned longer, experienced more and encountered problems in the past can use this knowledge to their advantage. The application process for the subsidy application can be seen as a routine as they are based on past actions (Levitt & March, 1988). Furthermore, as the subsidy application process is standardized, previous experience can gain an advantage as firms could learn from the process and thereby improve the quality of the application (Adler & Clark, 1991). Therefore, as firms gain experience it is likely that they improve their knowledge about subsidy applications and have an understanding what is required from firms, thereby better able to deliver the required information for project funding. Therefore, the following is hypothesized:
Hypothesis 3. Previous experience has a positive moderating effect on the relationship innovativeness of the project description and project funding in terms of amount.
2.6 Conceptual Framework
The following conceptual model is created from the hypotheses. It shows the relationship between the independent, dependent, moderators and four control variables:
14
3. Institutional Setting - WBSO
This research paper focuses on the subsidies provided in the Netherlands called: “Wet Bevordering Speur- en Ontwikkelingswerk” (WBSO) for R&D at companies and institutions, and is provided by the Netherlands Enterprise Agency (RVO). The WBSO, together with the “Innovatiebox”, are the most important tax/subsidy incentives in the Netherlands (Rijksdienst, 2019). According to Keijzers & Bos-Brouwers (2006), the application for the tax incentives is relatively easy. The WBSO’s main purposes is improving and increasing the R&D of firms as well as the climate in which R&D is created (RVO, 2019a). By doing so, the Dutch government wants to attract foreign firms, stimulate R&D firms, and maintain and if possible, expand existing companies with R&D (Rijksdienst, 2019). To be able to apply, the innovations must contribute to the improvement of the profitability and competitiveness of firms. Projects that qualify for the WBSO include technical scientific research and the development of new technical products, process, and software/hardware. The WBSO is a generic R&D incentive that mainly serves small and medium firms – in 2017, 97% were SMEs (Rijksdienst, 2019). While 97% of the users are SMEs, 37% of the WBSO budget goes to large firms. However, when looking at budgetary terms, large companies carry out almost 60% of all R&D performed. Furthermore, the number of users for the WBSO have grown as well in the last couple of years from 20,533 users in 2012 to 21,263 in 2017 (Rijksdienst, 2019). According to the CBS (2017, p.199), 31% of the firms with 10 or more employees performing R&D do not use the WBSO. The main reason for not using them, is that the R&D activities are not covered by the R&D definition within the WBSO or because the burden of administrative activities do not outweigh the benefits gained from the WBSO. Overall there are a large number of companies which apply for the WBSO to further increase their innovative activities.
15
4. Methodology
In this chapter, the dataset used for this research will described, including the transformations made, and analyses performed. The dataset needs to be transformed and adapted to make it suitable for the text-analysis and the regression analyses performed.
4.1 Project Description Dataset
The dataset used for this research is accumulated by a consultancy firm by storing the project information from the subsidy requests received over the last couple of years. The dataset contains a large set of text-based information from all the subsidy projects where the consultants worked on between 01-01-2018 and 06-05-2019 with a total number of 6536 projects throughout the whole Netherlands. This time period was chosen to prevent time-related biases concerning innovativeness of the projects and improve the quality of the grading. For this research, other fields i.e. company name, project type and period were not included within the dataset as the project descriptions were only used for the text analysis. Even though most projects are located within the Netherlands, some project descriptions are in Dutch while others are in English. According to Schlemmer & Mash (2006), language barriers interfere with the working efficiency and create uncertainty about interpretation and understanding. Therefore, to create a higher qualitative result, this research focused only on the projects which were described in Dutch. This would make the judgements of the experts better as language barriers would be prevented. As last, the sample for the experts to grade consist of a total of 120 projects. This time frame has been chosen to make sure the experts are grading up-to-date information making it easier to judge the innovativeness of the projects. A second dataset is used for other variables about the project such as industry of the firm, requested funding, and received funding.
4.2 Approach
16 Quantum computing, which should have shown in project descriptions in the year 2019, showed already
in project descriptions of 2015. Other trends also showed large gaps: AI Foundation – 6-year gap,
immersive experience – 7-year gap, blockchain – 6-year gap, and AI & Advanced Machine Learning –
7-year gap. These gaps showed that the trends in the codebook would not be representative and could not predict the innovativeness of recent projects as future project trends were missing. Furthermore, a quick manual check of a small sample of project descriptions showed that the highly diverse projects would make it even harder using trends to grade the innovativeness of project. It was discovered that the created code-book for grading the innovativeness did not create the sufficient results. Therefore, this method was not chosen as the main method for analyzing the project descriptions.
Expert approach - The first research showed that the credibility of the codebook was not sufficient. Therefore the expert-based method was chosen as the main method to grade the innovativeness of the project descriptions. A sample of 120 project descriptions from the total 6536 project descriptions was created for the experts to grade. As the 6536 projects were in one single .CSV file, it was opened in Microsoft Office Word which resulted in a document with a total of 7415 pages. An online random number generator is used to create a random sample of 120 numbers. 3 The following data was entered into the generator: not sorted, the number of projects looking for: 120, range: 1 to 7415. These 120 numbers would then be used to find the specific page in the document, the project on top of that page is used for the sample. Two conditions when selecting the project on the random page: (i) when the first project description on the random page is in English the first following Dutch project has been taken for the sample, and (ii) when an incomplete project was found on the random generated page (starting on a previous page), the first following Dutch project was taken. In total eight experts were approached to each grade 15 projects on the innovativeness, the response rate was 87.5% resulting in 106 projects descriptions used for this research.
4.3 Measurements
Dependent variables - In this study the dependent variable consist of how much funding a firm has received for a specific project. The variable is indicated as: Approved_Funding. The natural logarithm +1 has been taken to make the data less skewed and create a normal distribution. See figures 2 and 3 for how the natural logarithm influences the skewness between with and without the log +1 function.
17
Independent variable - The independent variable indicates whether a firm received more funding for a specific project compared to others, based on the innovativeness of the project description (Grade_Innovative). This ordinal variable is calculated by a score given by experts who judged the innovativeness of the projects. The experts judged, based on their experience, how innovative the project was according to them and ranked them on a scale from 1-not innovative, up to 5-very innovative without knowing and having other knowledge (i.e. firm size, previous experience, or age) except the project descriptions. This likert-scale has been used as it improves validity and reliability (Dawes, 2008).
1 2 3 4 5
Niet innovatief Matig innovatief Neutraal Innovatief Erg innovatief Table 1. Scale grading experts
Moderators – The moderator variable Additional_Info is a nominal variable indicating two options no [0] or yes [1] for having provided extra information to the subsidy provider. The second moderating variable used is: Previous_Experience. This indicates how long the applying firm is already working together with the consultants for applying for subsidies and is measured in amount of years. The median (8 years) has been taken to divide the group into two categories of low and high experience: [0] - 51 projects with low experience (<8), and [1] - 55 projects with high experience (>= 8). The variable has been recoded as the continuous variable did not show a linear relationship and a threshold at 9 years of experience. See appendix A for more information.
Control Variables - To test the relationship between the different variables, other variables are controlled for as these could influence the results. The following variables are used to control.
Industry – According to Cohen & Klepper (1996), the total R&D dedicated to innovation differs greatly across industries and that these differences are caused by the exogenous conditions. Thornhill
18 (2006) found that the high-tech sector introduced more than double the number of new products compared to the low-tech sector indicating that industries with higher levels of R&D are home to firms with higher innovativeness. Castellacci and Lie (2015) found that manufacturing firms perform more R&D and apply for more R&D subsidies compared to other industries. Dummy variables for all industries were included, however only for the manufacturing industry a difference was shown. Furthermore, the model with all industries dummies included showed no large differences compared to the model with fewer variables. Therefore, manufacturing industry will be used a dummy variable, as fewer variables is preferred. The second dataset contains the SBI codes of the projects which will be used to create the variable (KvK, 2019). All projects starting with SBI-codes 10 up to 33 are coded as [1] manufacturing, all others will be coded as [0] other. This resulted into the following distribution: [0] - 72 projects with other as industry, and [1] - 34 projects with manufacturing as industry.
Project type – Utterback & Abernathy (1975) observed difference the types of innovations: process and product innovations. Porter (1981) elaborates that product innovation is the dominant mode and aims on improving the performance while process innovations focus on lowering costs. Van Auken, Madrid-Guijarro, & Garcia-Perez-de-Lema (2008) conclude that the different types of innovation have different kind results on performance. Therefore, it is important to have a good understanding of the nature of the innovations (Gunday, Ulusoy, Kilic, & Alpkan, 2011). The project descriptions are classified according three different types of innovation: (i) process, (ii) product, and (iii) programming – resulting in two dummy variables with process as baseline.
Size – Strategic resources are uneven distributed across firms in which larger firms have more resources compared to smaller firms (Barney, 1991). This could make it possible to have differences, in terms of quality, providing the required project descriptions for the subsidy. According to several studies, firm-size affect the innovation activity of a firm (Acs & Audretsch, 1988; Shefer & Frenkel, 2005; Thornhill, 2006). According to Shefer & Frenkel (2005) firm-size has a positive impact on the R&D expenditure, and that R&D expenditures are directly linked to the firm’s innovativeness. Furthermore, they argue that larger firms are more likely to have the necessary resources needed for R&D. Therefore, number of employees has been used to control and the natural logarithm has been taken to make the data less skewed.
19
4.4 Poisson Regression
The Poisson regression is a form of regression analysis to model count data and is appropriate for low frequencies (n) and deals as well with the skewness related with these low frequencies. The regression models expected frequencies and is ideal for regressing a count variable: Approved_Funding with a categorical variable: Grade_Innovative. Furthermore, Poisson regression also specifies how the dependent variable Approved_Funding, relates to any of the explanatory variables. The Poisson regression models the log of the dependent variable as a function of the explanatory variables resulting in the coefficient. According to UCLA (2019), the coefficient can be interpreted as follows: “for every unit change in the explanatory variable, the difference in the logs of the expected counts is expected to change by the coefficient, given the other explanatory variables in the model are held constant”. The Poisson model is assumed to be the appropriate model as the dependent variable is not over-dispersed and does not have a large number of zeros (UCLA, 2019). For this research the vce (robust) options is used to obtain robust standard errors to control for heteroskedasticity (Cameron and Trivedi, 2009). Before starting the regression analysis, multicollinearity was checked for to prevent if two or more explanatory variables in the regression model are strongly correlated. Therefore, the data is tested for the Variance Inflation Factors (VIF) after running a linear regression analysis. According to Miles & Shevlin (2001) multicollinearity can be assumed when the VIF-values are larger than 4. The results for the VIF test range from VIF = 1.02 (Grade_Innovative) up to VIF = 1.79 (Firm_Age), therefore it is assumed that multicollinearity is not a problem for this research.
5.
Results
The following section will start with the descriptive statistics of the sample used for this research. This will be followed by the Poisson regressions which will test the direct effect of Grade_innovativeness on
Approved_funding. Furthermore, for each of the moderating variables: Previous_Experience and Additional_Information, additional Poisson regressions will be tested. Finally, interaction models
between the moderating variables and the full model will be created by Poisson regressions. With these analyses this paper hopes to explain the different hypothesis.
5.1 Descriptive Statistics
The total sample in the analysis consist of 106 project descriptions. As the independent variable
Project_Descriptions are graded for the purpose of this research, all observations were graded and no
missing values were found. All other variables used for this research did not miss any values.
20 maximum, a highly right-skewed graph, which can also be seen in figure 2. A boxplot from the approved funding – see appendix B – shows that 50% of the projects applying for subsidies receive 27,187.50 euro or less per project. Furthermore, the summary details – see appendix C – show that the top 1% received a total amount of 2,002,980.00 euro subsidy funding, receiving more than the bottom 50%. On average the project descriptions were graded 3.113 out of 5 on innovativeness, meaning that on average the projects are just above “neutral innovative”. Also, the variable Previous_Experience is highly right-skewed – see appendix D – showing that this sample mainly consists of firms which have not worked longer than 10 years together with the consultants. Approximately 30% of the project descriptions do not possess the required information, therefore these firms had to provide additional information to the RVO for subsidy approval. However, these 30% of applications receive on average 101.932.13 euro more funding than firms that did not provide additional information – see appendix G. Both Firm_Size and Firm_Age are highly right-skewed, size is more skewed compared to age. This is also more likely as the range of the two variables differ highly: firm size (1-3336) is higher compared to firm age (0-96). The high right-skewed graphs of size and age show that mainly younger and smaller firms are represented in this sample – see appendix E and F for more information. According to the European Commission (2017), firms with less than 250 employees can be classified as SMEs. In this sample, 92 out of the 106 projects (87%) can be classified as a SME as they have less than 250 employees. Showing that mainly in this sample, SMEs focus on applying for subsidies for external funding on R&D projects. See table 6 for more descriptive statistics.
Descriptive Statistics
Variable Obs Mean Std.Dev. Min Max
Approved Funding 106 102464.5 208609 0 1032080
Grade Innovativeness 106 3.113 1.054 1 5
Previous Experience 106 9.028 6.437 1 26
Additional Information (1=Yes) 106 .311 .465 0 1
Firm Size 106 220.160 513.675 1 3336
Firm Age 106 23.028 20.727 0 94
Industry (1=Manufacturing) 106 .321 .469 0 1
Table 2. Descriptive statistics
5.2 Correlation Matrix
21 The variable Approved_Funding shows a strong postive correlation five variables: Firm_Size, Grade_Innovativenes, Additional_Information, Firm_Age, and Industry, this could imply that the amount of funding is dependent on several factors. Also, the variables Additional_Information and Firm_Size are strong positively correlated, in combination with the variable Approved_Funding, this could mean that larger firms apply for larger amounts of funding but do have to provide additional information to justify the amount requested. Furthermore, Firm_Age and Previous_Experience are strong positively correlated, his is likely as older firms have more previous experience as they have been around longer than younger firms. Only the variables Project_Type and Industry are strong negatively correlated. As the industry variable is categorized according to manufacturing and others, it seems logic that manufactures mainly apply for process and product subsidies as they produce products and have process for which subsidies can be applied, therefore probably the negative relationship. Approved_Funding and Previous_Experience are not correlated according to matrix. This could imply that previous experience does not directly influence the amount of received funding. Furthermore, Additional_Information and Previous_Experience, the two moderating variables, are not correlated, implying that they independently influence the amount of funding. At last, the variable Grade_Innovativeness is only strong positively correlated with the variable Approved_Funding. As the experts only received the information about the project descriptions, the correlation matrix indeed implies that variables such as Firm_Size, Industry, and Previous_Experience were not considered when grading the projects.
Correlation Matrix Variables Appr. funding Grade inno. Add.info Previous Ex.
Size (ln) Age (ln) Industry Project type Appr. fund. 1.000 Grade inno. 0.191** 1.000 Add. Info. 0.227** 0.083 1.000 Previous exp. 0.013 -0.022 0.077 1.000 Size (ln) 0.453*** -0.007 0.363*** 0.146 1.000 Age (ln) 0.254*** 0.035 0.184* 0.392*** 0.536*** 1.000 Industry 0.344*** 0.080 0.105 0.095 0.349*** 0.416*** 1.000 Project type -0.042 0.053 0.101 -0.119 -0.052 -0.077 -0.210** 1.000 *** p<0.01, ** p<0.05, * p<0.1
Table 3. Correlation Matrix
5.3 Direct effect of innovativeness on funding
22 innovativeness tend to be associated with a higher amount of received funding. Model 2 shows the same effect with the added independent variable Additional_Information to the regression. Model 3 shows an even slightly stronger positive coefficient with the added independent variable Previous_Experience. From the three models, it shows that higher innovativeness is more likely to receive more amount of project funding as the variable Grade_Innovative is significant positively related to Approved_Funding. This is in line with the RVO (2019a) which states that innovativeness is the most important criteria for a project to receive the subsidy. Furthermore, confirming the assumption of Chan & Parhankangas (2017) that experts are capable in finding promising projects. Therefore, supporting hypothesis one. The margins are included for the baseline regression to get a better understanding on how innovativeness influences approved funding, also showing the positive significant relationship. With every increase unit of innovativeness, the approved funding increases with approximately 18,500.00 euro. Continuing with the margins for the baseline regression, all other variables except Firm_Age show a positive increase of amount of product funding. The positive significant relationship of Industry shows that manufacturers are likely to receive 59,643.10 euro more funding compared to other industries. Finally, a strong positive effect can also be seen for Firm_Size in which for every unit increase result in 25,062.67 euro more amount of funding. See table 4 for more information.
Margins for Approved Funding
dy/dx Std.Err. z P>z 95%Conf. Interval] Grade_Innovative 18521.490 6409.328 2.890 0.004 5959.438 31083.540 Firm_Size (ln) 25062.670 5537.943 4.530 0.000 14208.500 35916.840 Firm_Age (ln) -2432.047 8569.024 -0.280 0.777 -1.92e+04 14362.930 Industry 59643.100 25826.530 2.310 0.021 9024.037 1.10e+05 Project_Type 2 14220.690 12660.410 1.120 0.261 -1.06e+04 39034.650 3 21420.870 32042.360 0.670 0.504 -4.14e+04 84222.750 Note: dy/dx for factor levels is the discrete change from the base level
Table 4: Margins Model 1 – Approved Funding
5.4 Moderating Effects
23 For hypothesis three it, is hypothesized that previous experience has a positive moderating effect on the direct relationship between innovativeness and amount of received funding. Model 5 shows a non-significant coefficient for low previous experience and positive coefficient for high previous experience on innovativeness and approved funding. Knoblich, Ohlsson, & Raney (2001) which state that previous experience improves problem solving and thereby the performance. Furthermore, in line with the research of Adler & Clark (1991) which state that as firms could learn from the process, they gain an advantage. Therefore, having more previous experience with applying for subsidies therefore has a positive moderating effect on the relationship between innovativeness and amount of received funding, supporting the third hypothesis.
5.5 Interaction Effects
Model 6 shows four significant interaction effects between additional information and previous experience. As there are significant interaction effects, the moderating effects should not be interpreted without considering the interaction effects (Frost, 2017). Two interaction effects show a lower coefficient, one a negative coefficient, and one a stronger coefficient than without the interaction effects.
The variables: Additional_Information (model 4) and Previous_Experience (model 5) show both a significant positive moderating effect on the relationship innovativeness and amount of approved funding. However, the interaction effects in model 6 shows that the decision cannot be based on either of the main effects alone (model 4 or 5). When a firm has no previous experience and has not provided additional information, according to model 4 and 5, this would have a non-significant positive coefficient. However, with the interaction effect from model 6 a strong negative coefficient can be found. The interaction between providing no additional information and having previous experiences shows significant positive coefficient, meaning a positive influence on amount of funding, but smaller coefficient than without the interaction effect (model 4). When providing additional information but having no previous experience a positive significant coefficient is shown implying a positive influence on the amount of funding, but also smaller than without interaction effect (model 5). The interaction between having previous experience and providing additional information is positive significant with the highest coefficient (0.750, p<0.01) compared to the other 3 interactions and moderating effects in models 4 and 5. Meaning that if a firm has both previous experience and provided additional information, it has the strongest effect on receiving more funding.
24 Full Model – Approved Funding
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Independent Grade_Innovative 0.339*** 0.339*** 0.357*** (-0.114) (-0.111) (-0.115) Moderators Additional_Info (1=Yes) 0.009 -1.284 -1.713** (-0.329) (-0.791) (-0.743) Previous_Experience (1=High) 0.240 -1.148 -1.859** (-0.330) (-0.793) (-0.750) Control Variables Firm_Size (ln) 0.459*** 0.458*** 0.475*** 0.440*** 0.478*** 0.432*** (-0.081) (-0.101) (-0.093) (-0.102) (-0.098) (-0.118) Firm_Age (ln) -0.045 -0.044 -0.070 -0.038 -0.028 -0.131 (-0.158) (-0.152) (-0.157) (-0.136) (-0.137) (-0.144) Industry (1=Manufacturing) 0.899*** 0.897*** 0.971*** 0.928*** 1.052*** 1.202*** (-0.330) (-0.304) (-0.315) (-0.292) (-0.309) (-0.307) Project_Type (1=Product) 0.299 0.297 0.337 0.316 0.400 0.344 (-0.293) (-0.311) (-0.316) (-0.299) (-0.336) (-0.339) Project_Type (1=Software) 0.422 0.420 0.518 0.474 0.651 0.687 (-0.581) (-0.625) (-0.643) (-0.611) (-0.651) (-0.692) Interactions 0.Additional_Info# Grade_Innovative 0.121 (-0.152) 1.Additional_Info# Grade_Innovative 0.498*** (-0.149) 0.Previous_Experiencep# Grade_Innovative 0.134 (-0.127) 1.Previous_Experiencep# Grade_Innovative 0.534*** (-0.155) 0.Additional_Info#0.Previous_ Experiencep#Grade_Innovative -0.398** (-0.183) 0.Additional_Info#1.Previous_ Experiencep#Grade_Innovative 0.407** (-0.186) 1.Additional_Info#0.Previous_ Experiencep#Grade_Innovative 0.305** (-0.125) 1.Additional_Info#1.Previous_ Experiencep#Grade_Innovative 0.750*** (-0.171) Constant 7.697*** 7.701*** 7.416*** 8.435*** 7.939*** 9.764*** (-0.520) (-0.586) (-0.711) (-0.738) (-0.727) (-0.792)
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
25
6. Discussion
The outcomes of the Poisson analyses have provided interesting insights concerning the relationship between project innovativeness and amount of project funding. The expertise of the experts to understand which projects are innovative and the amount of funding received is significant positive related. This is in line with Chan & Parhankangas (2017) which assumed that experts are capable in finding promising projects with extraordinary returns. Furthermore, innovativeness of the project is indeed an important criterion for the RVO to provide funding as innovativeness is significant positive related to the amount of funding received (RVO, 2019a). Besides innovativeness, firm size and industry also significant positively influences the amount of received funding. Combining these results, it can be said that innovativeness of the project is not the only predictor on how much funding a project receives and that a large firm, in the manufacturing industry with an innovative project is likely to receive the highest amount of funding.
26
7. Conclusion
This paper has focused on the direct relationship between project innovativeness and external funding, and how information symmetry influences this relationship. In the current literature there is ambiguity about which innovativeness measures can be used for a firm and industry wide approach as current measures are not representative on how innovative the project or idea is. Therefore, this paper researched if expert ratings can be used as an innovativeness measurement and thereby understand the differences in amount of funding between project descriptions. Furthermore, to examine why some firms with similar innovativeness receives less funding this paper looked at the information asymmetries by looking at the previous experience of the firm and additional information provided. As discussed in the results section, significant support is found for all three hypotheses. It is found that the amount of funding received is positively impacted by the innovativeness of the project based on the gradings of experts. It can therefore be concluded that higher innovativeness explains why some firms receive more funding compared to others. Furthermore, as the expert’s ratings innovativeness measure shows a significant relationship with the amount of funding, which is based on the innovativeness of the project, it can be concluded that manual text analysis can be used to measure the innovativeness of a project. It is found that additional information positively influences the amount of funding received, this same effect is found for previous experience. Hypothesis two and three therefore show that information asymmetry can be reduced and increase the amount of received funding. However, it is important to take into consideration the interaction effects between these two variables when applying the research. The interaction effects can have a large impact on the overall outcome of the two variables. Hence, it can be concluded that innovativeness has a positive effect on approved funding, but it depends if previous experience and additional information has a positive effect.
8. Managerial implications
The results of this paper are especially interesting for consultancy firms operating in the funding industry, governments, and institutions. By implementing the findings, the overall subsidy system could be improved resulting in higher returns.
27 request at least on average 70,000 euro less, consultancy firms should provide this group with more expertise and information to justify larger amount of requested funds.
For the governments, it can be seen from the results that the subsidy funds are unequally distributed over the number of firms. With the current application process, all firms have to put in the same amount of effort for highly different amounts of requested funds. As the additional information is positively related to larger amount of funding, a more stepwise subsidy process should be considered. This could reduce the number of requested letters to provide additional information and thereby reduce information asymmetry. Furthermore, this would decrease the bureaucracy as projects do not have to be reconsidered and letters to be send.
9. Limitations & Future Research
One of the limitations of this study is the small sample size. Even though the dataset consists out of over 6000 projects, this research was limited by the number of experts and time of the experts available. As the smaller sample led to a reliable manual text-analysis, it limits the generalizability of the findings. Furthermore, due to the small sample and wide research approach of including all firms form all industries, some firm types and industries are thereby are underrepresented in the sample. Furthermore, as the firms from this sample are only from the Netherlands, firms and governments should be careful about generalizability of the findings. A second limitation is the knowledge of the experts, it is unknown if the experts are biased towards specific information or project descriptions.
28
References
Aboody, D., & Lev, B. (2000). Information asymmetry, R&D, and insider gains. The journal of Finance, 55(6), 2747-2766.
Acs, Z. J., & Audretsch, D. B. (1988). Innovation in large and small firms: an empirical analysis. The American economic review, 678-690.
Acs, Z. J., Anselin, L., & Varga, A. (2002). Patents and innovation counts as measures of regional production of new knowledge. Research policy, 31(7), 1069-1085.
Adler, P. S., & Clark, K. B. (1991). Behind the learning curve: A sketch of the learning process. Management Science, 37(3), 267-281.
Akerlof, G. A. (1970). 'The Market for 'Lemons': Quality Under Uncertainty and the Market Mechanism," Quarterly Journal of Economics, 84 (August), 488-500.
Amara, N., Landry, R., & Traoré, N. (2008). Managing the protection of innovations in knowledge-intensive business services. Research policy, 37(9), 1530-1547.
Andrew McAfee & Brynjolfsson, E., Davenport, T. H., Patil, D. J., & Barton, D. (2012). Big data: the management revolution. Harvard business review, 90(10), 60-68.
Archibugi, D. (1992). Patenting as an indicator of technological innovation: a review. Science and public policy, 19(6), 357-368.
Atalay, M., Anafarta, N., & Sarvan, F. (2013). The relationship between innovation and firm performance: An empirical evidence from Turkish automotive supplier industry. Procedia-Social and Behavioral Sciences, 75, 226-235.
Avlonitis, G. J., & Salavou, H. E. (2007). Entrepreneurial orientation of SMEs, product innovativeness, and performance. Journal of Business Research, 60(5), 566-575.
Barney, J. (1991). Firm resources and sustained competitive advantage. Journal of management, 17(1), 99-120.
Beck, T., Demirguc‐Kunt, A. S. L. I., Laeven, L., & Levine, R. (2008). Finance, firm size, and growth. Journal of Money, Credit and Banking, 40(7), 1379-1405.
Behn, R. D. (2003). Why measure performance? Different purposes require different measures. Public administration review, 63(5), 586-606.
Bellstam, G., Bhagat, S., & Cookson, J. A. (2017). A text-based analysis of corporate innovation (No. 2803232). SSRN working paper.
Bond, S., Harhoff, D., & Van Reenen, J. (1999). Investment, R & D and financial constraints in Britain and Germany (No. W99/05). London: Institute for Fiscal Studies.
Brand-Gruwel, S., Wopereis, I., & Vermetten, Y. (2005). Information problem solving by experts and novices: Analysis of a complex cognitive skill. Computers in Human Behavior, 21(3), 487-508. Brouwer, E., & Kleinknecht, A. (1999). Innovative output, and a firm's propensity to patent.: An
exploration of CIS micro data. Research policy, 28(6), 615-624.
29
Castellacci, F., & Lie, C. M. (2015). Do the effects of R&D tax credits vary across industries? A meta-regression analysis. Research Policy, 44(4), 819-832.
CBS. (2017). ICT, kennis en economie 2017. Retrieved May 16, 2019, from: https://www.cbs.nl/-/media/_pdf/2017/26/ike2017-web.pdf
Chan, C. R., & Parhankangas, A. (2017). Crowdfunding innovative ideas: How incremental and radical innovativeness influence funding outcomes. Entrepreneurship Theory and Practice, 41(2), 237-263.
Cleary, T. J., & Zimmerman, B. J. (2001). Self-regulation differences during athletic practice by experts, non-experts, and novices. Journal of Applied Sport Psychology, 13(2), 185-206.
Cohen, W. M., & Klepper, S. (1996). Firm size and the nature of innovation within industries: the case of process and product R&D. Review of Economics and Statistics, 78(2), 232-243.
Czarnitzki, D. (2006). Research and development in small and medium‐sized enterprises: The role of financial constraints and public funding. Scottish journal of political economy, 53(3), 335-357. Dawes, J. (2008). Do data characteristics change according to the number of scale points used? An
experiment using 5-point, 7-point and 10-point scales. International journal of market research, 50(1), 61-104.
Edmunds, A., & Morris, A. (2000). The problem of information overload in business organisations: a review of the literature. International journal of information management, 20(1), 17-28.
Eisenhardt, K. M. (1989). Agency theory: An assessment and review. Academy of management review, 14(1), 57-74.
European Commission. (2017). What is an SME?. Retrieved June 6, 2019, from https://ec.europa.eu/growth/smes/business-friendly-environment/sme-definition_nl
Fazzari, S., Hubbard, R. G., & Petersen, B. C. (1987). Financing constraints and corporate investment. Brookings Papers on Economic Activity, 1988(1), 141-206.
Frost, J. (2019). Understanding interaction effects in statistics. Retrieved June 21, 2019, from: https://statisticsbyjim.com/regression/interaction-effects/
Gunday, G., Ulusoy, G., Kilic, K., & Alpkan, L. (2011). Effects of innovation types on firm performance. International Journal of production economics, 133(2), 662-676.
Hall, B. H. (2002). The financing of research and development. Oxford review of economic policy, 18(1), 35-51.
Harabi, N. (1995). Appropriability of technical innovations an empirical analysis. Research policy, 24(6), 981-992.
Harhoff, D. (2000). Are there financing constraints for R&D and investment in German manufacturing firms?. In The economics and econometrics of innovation (pp. 399-434). Springer, Boston, MA. Healy, P. M., & Palepu, K. G. (2001). Information asymmetry, corporate disclosure, and the capital
30 Hoffman, K., Parejo, M., Bessant, J., & Perren, L. (1998). Small firms, R&D, technology and innovation
in the UK: a literature review. Technovation, 18(1), 39-55.
Huber, G. P. (1991). Organizational learning: The contributing processes and the literatures Organization science, 2(1), 88-115.
Huergo, E., & Jaumandreu, J. (2004). How does probability of innovation change with firm age?. Small Business Economics, 22(3-4), 193-207.
Hui, I. (2017). Shaping the Coast with Permits: Making the State Regulatory Permitting Process Transparent with Text Mining. Coastal Management, 45(3), 179–198.
Hui, S. C., & Jha, G. (2000). Data mining for customer service support. Information & Management, 38(1), 1-13.
Hult, G. T. M., Hurley, R. F., & Knight, G. A. (2004). Innovativeness: Its antecedents and impact on business performance. Industrial marketing management, 33(5), 429-438.
Innovencio. (2017). Vragenbrief WBSO. Retrieved June 18, 2019, from https://innovencio.nl/vragenbrief-wbso/
Jia, S. (Sixue). (2018). Behind the ratings: Text mining of restaurant customers’ online reviews. International Journal of Market Research, 60(6), 561–572.
Keijzers, G., & Bos-Brouwers, H. E. J. (2006). Innovaties voor duurzaamheid in het MKB; naar nieuwe
aangrijpingspunten voor beleid. (Ministerie van VROM Publicatiereeks). Den Haag: Ministerie
van VROM.
Kim, J., & Park, Y. (2019). Leveraging ideas from user innovation communities: using text‐mining and case‐based reasoning. R&D Management, 49(2), 155–167.
Kleer, R. (2010). Government R&D subsidies as a signal for private investors. Research Policy, 39(10), 1361-1374.
Kleinknecht, A. (1993). Testing innovation indicators for postal surveys. In: Kleinknecht, A., Bain, D. (eds). New Concepts in Innovation Output Measurement. Macmillan: London, pp. 153-188. Kleinschmidt, E. J., & Cooper, R. G. (1991). The impact of product innovativeness on
performance. Journal of product innovation management, 8(4), 240-251.
Knoblich, G., Ohlsson, S., & Raney, G. E. (2001). An eye movement study of insight problem solving. Memory & cognition, 29(7), 1000-1009.
Kortum, S., & Lerner, J. (2001). Does venture capital spur innovation?. In Entrepreneurial inputs and outcomes: New studies of entrepreneurship in the United States (pp. 1-44). Emerald Group Publishing Limited.
Kvk. (2019). Overzicht standaard bedrijfsinstelling Retrieved June 18, 2019, from:
https://www.kvk.nl/over-kvk/over-het-handelsregister/wat-staat-er-in-het-handelsregister/overzicht-standaard-bedrijfsindeling/
