Examination of Stage-Based National Key Performance Proxy Variables Using Neural Networks
University of Portsmouth, United Kingdom
Decrypting the secret code for innovation performance improvement enables the strengthening of countries’ national innovation system (NIS) by governments’ effective interventions in relation to innovation policies. Evaluating governments’ effectiveness led to innovation indices (Mahroum et al., 2013) which in turn enable to predict the effect of policy makers’ decisions.
In the last two years, studies have used fuzzy logic and machine learning techniques such as neural networks to identify determinants that lead to high innovation performance. The overriding objective is to predict the effect of innovation policy related decisions on countries’ positions in a global innovation performance environment. The study of Proksch et al. (2017) and Crespo et al. (2016) uses fuzzy-set qualitative comparative analysis (fsQCA) to investigate the determinants for high innovation performance. Both outlined that multiple innovation strategies lead in combination to an improvement. Hajek and Henriques (2017) used machine learning to investigate the determinants for regional innovation performance by applying multi-output neural networks. They confirmed the conclusions about the combination of strategies and stated entrepreneurship support as the effective intervention of innovation policy. Pençe et al. (2018) also used neural networks based on index data from 2016 and identified 27 features for innovation performance improvement.
Studies using QCA have limitations in terms of feature selection, sample size (Cooper et al., 2011) and perform incorrectly for causal inference (Baumgartner et al., 2017). Cross-sectional studies evaluate determinants empirically on small data although machine learning enables unsupervised feature selection and examines longitudinal effects on countries' innovation performance.
- Which machine learning model is most accurate in feature selection and predicting the effect of innovation performance-related decisions?
- What proxy variables do exist and how can they be identified?
- What implications result for policymakers, how would innovation performance improve and what is the forecasted performance improvement?
This study applies a quantitative approach based on index data. At first, a five-year GII dataset from 2014 to 2018 is retrieved. Data is cleaned and missing values are imputed using multiple imputations with chained equations. Countries are grouped using k-means clustering and similar profiles are identified by merging path plot. Subsequently, machine learning techniques are cross-validated by analysing RMSE values and learning curves. Variable importance plots are analysed after validating the model performance and model selection. In a next step, explanations are compared and breakdown plots evaluate proxy variables. Conclusions in terms of their stage-based effect are drawn.
Our results show four groups of countries cluster whereas the positioning aligns with the GII ranking. Countries with similar innovation profiles as for instance Sweden, Netherlands, Luxembourg and Great Britain have been categorized. The cross-validation revealed that Bayesian Regularized Neural Networks (BRNN) fit better than other methods. Further, the learning curve of BRNN is almost congruent which indicates a very good accuracy in prediction. Proxy variables have been identified for countries with similar profiles and predictions have been decomposed. As one result the proxy variables ‘Cost of redundancy dismissal’ and ‘Scientific and technical publications’ are notable, which both have a strong positive impact on the top countries’ innovation performance (CHE, SWE, NLD, GBR, SGP). The future significance of the identified proxy variables and their effects have been forecasted by the assumption of Ceteris Paribus.
Contribution to Scholarship
This study contributes to scholarship by suggesting a structured methodology to evaluate machine learning techniques to identify relevant features. Furthermore, it outlines how the effect of key performance proxy variables can be used to predict innovation performance in future.
Contribution to Practice
This study contributes to practice by offering policymakers implications about key performance determinants for effective interventions in innovation policy. Decisions can thus be validated in advance and their influence on the future position of a country in the global innovation environment can be assessed.
The issue of this research is crossing the boundaries of systems of innovation with its different actors and interrelations. Since the national innovation system contains various actors from academia, industry and society, our research bridges the gap with regard to collaboration and the common objective to foster economic growth.
Baumgartner, M., & Thiem, A. (2017). Often Trusted but Never (Properly) Tested: Evaluating Qualitative Comparative Analysis. Sociological Methods and Research, 1–33.
Cooper, B., & Glaesser, J. (2011). Paradoxes and pitfalls in using fuzzy set QCA: Illustrations from a critical review of a study of educational inequality. Sociological Research Online, 16(3).
Crespo, N. F., & Crespo, C. F. (2016). Global innovation index: Moving beyond the absolute value of ranking with a fuzzy-set analysis. Journal of Business Research.
Hajek, P., & Henriques, R. (2017). Modelling innovation performance of European regions using multi-output neural networks. PLoS ONE, 12(10), 1–21.
Kafadar, K., Kotz, S., Read, C., & Banks, D. (2006). Encyclopedia of Statistical Sciences. Journal of the American Statistical Association. American Statistical Association.
Mahroum, S., & Al-Saleh, Y. (2013). Towards a functional framework for measuring national innovation efficacy. Technovation, 33(10–11), 320–332.
Pençe, I., Kalkan, A., & ÇeSmeli, M. S. (2018). Estimation of the Country Ranking Scores on the Global Innovation Index 2016 Using the Artificial Neural Network Method. International Journal of Innovation and Technology Management.
Proksch, D., Haberstroh, M. M., & Pinkwart, A. (2017). Increasing the national innovative capacity: Identifying the pathways to success using a comparative method. Technological Forecasting and Social Change.
Which Policy Intervention Is Better to Encourage for Open Innovation? - Comparison of Direct and Indirect Government Financial Support on Innovation Collaboration
1Sogang University, Graduate School of Management of Technology, Republic of (South Korea); 2Institute for Manufacturing, University of Cambridge, UK
Open innovation (OI) is increasingly recognised as a successful approach for businesses. However, whilst policy makers are increasingly interested in supporting its adoption, it is not yet clear whether innovation policy tools have effectively evolved to encourage it (de Jong et al., 2010).
To date, research has mainly focused on firm level OI adoption and implication rather than on policy level (Greco et al., 2017). Direct support (such as R&D grants) release funds upfront, potentially reducing the efforts that firms need to invest to start developing an innovation, while indirect support (e.g., tax credit) is an ex post incentive which is easy to claim (Radas et al., 2015). Also, direct R&D grants often include binding clauses for the applicants, such as the formation of a certain type of collaboration, and innovation vouchers promote interactions between different innovation actors (Chapman and Hewitt-Dundas, 2018). However, indirect support, such as patent boxes and tax credit, is measured via traditional innovation indicators, such as the number of intellectual property (IP) or the amount of internal R&D investment, which does not reflect the nature of OI – organisational interaction and resource swap.
What is the optimal policy for OI? The answer to this question may not be simple, because policy tools include both direct and indirect support which are differently designed. Although indirect support may have effectively contributed to competence enhancement from traditional innovation perspective, it may have not adequately encouraged OI.
In this respect, the current study attempts to explore the influence of the central government’s direct and indirect financial support on various types of innovation collaborations.
Propensity score matching (PSM) has been regarded as one of the most effective analysis approaches estimating treatment effect (i.e., policy impact) using non-experimental or observational data, while also addressing selection bias from recipient firms and policy programmes. After matching between government support recipient and non-recipient firms, the average treatment effect on the treated (ATET) for five different types of innovation collaborations were estimated.
For an analysis, this paper used the UK Innovation Survey (UKIS) data sets which follow the OECD Oslo Manual. The two recent waves of the UKIS, UKIS 2015 (wave 9) and 2017 (wave 10) were used to see the longitudinal impact of public funding on openness. UKIS 2015 and 2017 were collected during the period between 2012 and 2014 and between 2014 and 2016 respectively, and they have almost identical question structure. The survey was voluntary, and was conducted through both a postal questionnaire and telephone interview for businesses that had not yet completed a postal response. To make a panel data set, we aggregated sample firms which appear both in UKIS 2015 and 2017. Consequently, 10,100 observations of 5,050 firms were selected for analysis.
The analysis results are very consistent and findings are summarised as follow.
First, direct financial support positively affected innovation collaboration but mainly in SMEs and only for specific types of collaboration. Direct financial support was not helpful for large firms, but it encouraged SMEs to expand its collaboration boundary by stimulating them to work with geographically distant partners or partners who are not in their value chains.
Second, indirect financial support did not significantly affect innovation collaboration to any types of examined OI and in any types of sub-sample groups. This suggests that the current indirect financial support, such as tax credit, does not reflect the nature of OI. As usually indirect financial support, such as tax credit, attempts to measure observable and easily quantifiable variables, such as revenue or the number of new product/IP, it may fail to promote innovation collaboration in firms.
Contribution to Scholarship
The paper adds to the literature on OI and policy by examining the impact of direct and indirect government support on innovation collaboration. Government financial support could encourage firms to invest in internal R&D, which in turn will help firms to develop their absorptive capacity for the integration of external knowledge (Spithoven et al., 2011). Also, policy intervention may result in the accumulation of knowledge in an industry (e.g. high tech), which may lead to knowledge spill-overs (Roper et al., 2013). However, despite these potential positive impact, the empirical analysis on the relationship between public policy and OI has been rarely explored, and this study attempts to address this issue. To date, OI studies has focused on firm level analysis. However, in order to establish vibrant innovation ecosystem, OI research must target higher-level OI initiatives, and, in this regards, this study expands the boundary of OI research.
Contribution to Practice
The findings from this study provide some policy implications. First, as shown in our results, the government’s direct financial support enhances open innovation in SMEs and helps them to overcome challenges in collaboration with distant or new partners. Therefore, bearing this in mind, policy focus must have been given to this direction. Second, policy for direct financial support must be entirely re-designed to reflect the nature of OI. The current tax incentive system may not be optimised for OI promotion, so more complicated design should be made to nudge firms’ collaborative behaviour.
This study attempts to cover one of the recent research themes in the open innovation literature, which may provide good food for thought for both scholars and practitioners.
Chapman, G. & Hewitt-Dundas, N. (2018). The effect of public support on senior manager attitudes to innovation. Technovation, 69, 28-39.
de Jong, J. P., Kalvet, T. & Vanhaverbeke, W. (2010). Exploring a theoretical framework to structure the public policy implications of open innovation. Technology Analysis & Strategic Management, 22, 877-896.
Greco, M., Grimaldi, M. & Cricelli, L. (2017). Hitting the nail on the head: Exploring the relationship between public subsidies and open innovation efficiency. Technological Forecasting and Social Change, 118, 213-225.
Radas, S., Anić, I.-D., Tafro, A. & Wagner, V. (2015). The effects of public support schemes on small and medium enterprises. Technovation, 38, 15-30.
Roper, S., Vahter, P. & Love, J. H. (2013). Externalities of openness in innovation. Research Policy, 42, 1544-1554.
Spithoven, A., Clarysse, B. & Knockaert, M. (2011). Building absorptive capacity to organise inbound open innovation in traditional industries. Technovation, 31, 10-21.
The Role of Governmental Intermediary in Cross-Country Entrepreneurial Resources Transfer: A Four Countries Comparison
1National Formosa University, Taiwan; 2Yuan Ze University, Taiwan
With the aim to spur economic prosperity by encouraging entrepreneurship, governments (i.e. China, Korea, Singapore, and Taiwan) initiate entrepreneurial programs that focus on connecting Silicon Valley (SV) resources and establishing local executive offices. These governments' work through ecosystem intermediaries that provides resources from SV to strengthen their local entrepreneurial ecosystems.
Start-ups develop their competitive advantages from resources within the entrepreneurial ecosystem rather than residing within the ﬁrms themselves (Spigel, 2017). Resources (e.g. capital, technologies, talents, and markets) flow frequently through key actors (e.g. entrepreneurs, mentors, advisors, investors, and so on) who can facilitate the availability of those resources that are crucial for start-ups to develop their businesses. A strong entrepreneurial ecosystem (Spigel, & Harrison, 2018) consists of not only a comprehensive set of resources and actors, but a dense network of actors. Intermediation is a process that connects entrepreneurs with key actors within an entrepreneurial ecosystem (Dutt, Hawn, Vidal, Chatterji, McGahan, Mitchell, 2016; Goswami, Mitchell, & Bhagavatula, 2018). Examples of intermediaries include accelerators, incubators, co-working spaces, and development organizations. Armanios, Eesley, & Eisenhardt (2017) highlight government use of institutional intermediaries to bridge gaps between the public resources and entrepreneurs and to spur entrepreneurial activity.
Despite significant understanding of the entrepreneurial ecosystem (Spigel, & Harrison, 2018; Goswami et al., 2018), there remains a limited understanding of the government’s role in strengthening the entrepreneurial ecosystem. There is an especially limited understanding of how governments shape regional entrepreneurial ecosystems at the macro level with entrepreneurial ecosystem boundary‐spanning.
In this study, we seek to understand how governments use institutional intermediaries to shape regional entrepreneurial ecosystems of home countries (i.e. China, Korea, Singapore, and Taiwan), and integrate distant, cross-boundary, entrepreneurial actors and resources simultaneously.
Case study is adopted as the research method of this paper. The activities of the research objects are studied over a window of time (2013~2017) to determine the cause or the process that led to their current state. Analytic generalization (Yin, 1994, pp.10, 31) is conducted to infer the constructive empirical results. The selection of cases is based on the “find the cross-case patterns” method proposed by Eisenhardt (1989, pp. 540–541). More profound meanings can be induced through the selection of cases and by finding similarities and differences between them.
The sources we use in this paper are categorized as primary and secondary information. Primary information includes in-depth interviews and the field notes taken while visiting the selected countries’ office (i.e. China, Korea, Singapore, and Taiwan) in SV. Secondary information consists of government websites, official and related websites and the media coverage of the research objects. Besides, the first author of this paper was a member of the Taiwan Innovation and Entrepreneurship Centre (TIEC) and served as an expert on government entrepreneurial programs for three years. As such, she knows the selected cases well. China, Korea, Singapore and Taiwan have been selected as the research objects for studying the role of intermediation in macro-level entrepreneurial ecosystem. The four cases are all initiated and supported by the government and all aim to bridge gaps in resources between SV and the home countries.
China, Korea, Singapore, and Taiwan devote themselves to shaping strong regional entrepreneurial ecosystems. Their intermediation processes can be categorized as broker, hub, or disciple.
Broker: China uses the broker intermediation process executed by Zhongguancun Silicon Valley Innovation Center, an entity funded by the Beijing local government. The process involves key actors from both SV and China entrepreneurial ecosystems forming relationships to collaborate on investments, markets development, or technology integration.
Hub: Singapore uses the hub intermediation process executed by BLOCK71 (BLK71), a University-Industry Alliance supported by the Ministry of Trade and Industry. This process pulls together the dispersed resources of Southeast Asia and SV and provides start-ups points of access into a new and unfamiliar entrepreneurial ecosystem.
Disciple: Taiwan uses the disciple intermediation process executed by the TIEC, supported by the Ministry of Science and Technology, while Korea uses the disciple intermediation process executed by Korea Innovation Center-Silicon Valley, supported by Ministry of Science, ICT and Future. Korea and Taiwan governments are dedicated followers and try to learn SV’s secrets to a successful entrepreneurial ecosystem. They arrange start-ups as a pilgrimage to SV and aim to instil a SV mindset and replicate the SV ecosystem style in their home countries.
Contribution to Scholarship
This study contributes to the entrepreneurial ecosystem literature by bringing the role of governmental institutions to attention. Governmental institutions can foster the developing process of domestic entrepreneurial ecosystems by bridging foreign, entrepreneurial successful ecosystem. This study also contributes to the intermediary literature by identifying three types of connective roles. While previous studies identified functional roles of intermediation, such as incubation, raising awareness, training (Howells, 2006; Küçüksayraç, E., Keskin, D., & Brezet, H., 2015), this study focuses on the connective roles that engage required actors into the network and arrange required resource flow within this network.
Contribution to Practice
Our findings provide a practical way of shaping regional entrepreneurial ecosystems. We present our results and organize them into three distinct roles of government that align with the entrepreneurial policy goals and original entrepreneurial ecosystems. For government policy makers, when considering connections to foreign ecosystems, three distinct roles are useful for identifying the main tasks they should devote themselves to.
This study is relevant to both the theme of “Innovation Policy” and “Ecosystem” of the 2019 R&D Management conference.
1. Armanios, D. E., Eesley, C. E., Li, J., & Eisenhardt, K. M. (2017). How entrepreneurs leverage institutional intermediaries in emerging economies to acquire public resources. Strategic Management Journal, 38(7): 1373-1390.
2. Dutt N, Hawn O, Vidal E, Chatterji A, McGahan A, Mitchell W. 2016. How Open System Intermediaries Fill Voids in Market-Based Institutions: The Case of Business Incubators in Emerging Markets. Academy of Management Journal, 59(3): 818-840.
3. Eisenhardt, K. M. (1989). Building theories from case study research. Academy of Management Review, 14(4): 532-550.
4. Goswami, K., Mitchell, J. R., & Bhagavatula, S. (2018). Accelerator expertise: Understanding the intermediary role of accelerators in the development of the Bangalore entrepreneurial ecosystem. Strategic Entrepreneurship Journal, 12(1): 117-150.
5. Howells, J. (2006). Intermediation and the role of intermediaries in innovation. Research policy, 35(5): 715-728.
6. Küçüksayraç, E., Keskin, D., & Brezet, H. (2015). Intermediaries and innovation support in the design for sustainability field: cases from the Netherlands, Turkey and the United Kingdom. Journal of Cleaner Production, 101: 38-48.
7. Spigel, B. (2017). The relational organization of entrepreneurial ecosystems. Entrepreneurship Theory and Practice, 41(1): 49-72.
8.Spigel, B., & Harrison, R. (2018). Toward a process theory of entrepreneurial ecosystems. Strategic Entrepreneurship Journal, 12(1): 151-168.
9. Yin, R. (1994). Case Study Research: Design and Methods. Beverly Hills.