مدل سازی بهره وری نسبی سیستم های ملی نوآوری

Although a large amount of past research has theorized about the character of national innovation systems (NISs), there has been limited process-oriented empirical investigation of this matter, possibly for methodological reasons. In this paper, we first propose a relational network data envelopment analysis (DEA) model for measuring the innovation efficiency of the NIS by decomposing the innovation process into a network with a two-stage innovation production framework, an upstream knowledge production process (KPP) and a downstream knowledge commercialization process (KCP). Although the concept of innovation efficiency is a simplification of the innovation process, it may be a useful tool for guiding policy decisions. We subsequently use a second-step partial least squares regression (PLSR) to examine the effects of policy-based institutional environment on innovation efficiency, considering statistical problems such as multicollinearity, small datasets and a small number of distribution assumptions. The hybrid two-step analytical procedure is used to consider 22 OECD (Organisation for Economic Co-operation and Development) countries. A significant rank difference, which indicates a non-coordinated relationship between upstream R&D (research and development) efficiency and downstream commercialization efficiency, is identified for most countries. The evidence also indicates that the overall innovation efficiency of an NIS is mainly subject to downstream commercial efficiency performance and that improving commercial efficiency should thus be a primary consideration in future innovation policy-making in most OECD countries. Finally, the results obtained using PLSR show that the various factors chosen to represent the embedded policy-based institutional environment have a significant influence on the efficiency performance of the two individual component processes, confirming the impact of public policy interventions undertaken by the government on the innovation performance of NISs. Based on these key findings, some country-specific and process-specific innovation policies have been suggested.

The national innovation system (NIS) approach was introduced in the late 1980s (Freeman, 1987 and Dosi et al., 1988) and further developed in the years that followed (Lundvall, 1992, Nelson, 1993 and Edquist, 1997). It enjoys wide currency in both academic and policy-making contexts (Sharif, 2006) and is considered a useful and promising analytical tool for academic study and for the development of innovation policy-making, fostering an understanding of innovation processes and determinants (Edquist, 1997, Furman et al., 2002 and Lundvall, 2007). Although no single definition of NISs has yet been adopted, a semantic core is common to most of the definitions used (Sharif, 2006). From a general perspective, an NIS results from the interaction between the knowledge innovation process (KIP) and the embedded2 innovation environment represented by framework conditions and infrastructure related to government intervention (Furman et al., 2002, Faber and Hesen, 2004 and OECD and EUROSTAT, 2005). As Edquist has pointed out, an institutional set-up geared toward innovation and an underlying production system are the basic characteristics of an NIS (Edquist, 1997). In terms of its physical composition, an NIS is a set of interacting institutions/actors (e.g., universities, industries and governments) that produce and implement knowledge innovation. These actors provide the national innovation production framework within which governments form and implement policies to influence the innovation process. Through interface structures (Molas-Gallart et al., 2008) or intermediate organizations (Howells, 2006), actors in different cultural and organizational contexts across an NIS are connected, and these connections tighten the institutionally embedded relationship between innovation production and the innovation environment. The contributions of the extant literature regarding the NIS approach lead policy-makers to employ systematic thinking rather than linear thinking about innovation at the national level (Edquist, 1997, Edquist and Hommen, 1999 and Groenewegen and van der Steen, 2006). The system thinking approach that supports a demand-side orientation in innovation policy (Edquist and Hommen, 1999) is a more holistic system perspective on innovation that focuses on the interdependencies among various agents, organizations and institutions (Groenewegen and van der Steen, 2006). In contrast to the traditional linear thinking approach, which supports a supply-side orientation in innovation policy (Edquist and Hommen, 1999), this alternative approach can better take into account the factors influencing innovation processes besides those shaping innovation processes, thus inspiring innovation policy-making. From a systems perspective, the NIS approach reminds policy-makers of the need to improve the collaboration among interacting institutions participating in the KIP and the influence of the innovation environment on the KIP. As national innovation policy-makers, governments mostly concern themselves with innovation efficiency as closely related to the innovation input/output ratio and emphasize the effect of public policy intervention on the innovation efficiency. Innovation efficiency is related to the concept of productivity, which is improved when the same amount of innovation input generates more innovation output or when less innovation input is needed to produce the same innovation output. This concept involves comparing the observed output to the maximum potential output obtainable from the input or, alternately, comparing the observed input to the minimum potential input required to produce the output. In this context, in the two comparisons, the optimum is defined in terms of production possibilities, and efficiency is technical (Fried et al., 2008, pp. 8). In this sense, efficient NISs are operating at their production possibility frontier (PPF) or “transformation curve”, which indicates the maximum amount of innovation output that can be produced with a given input. Clearly, the innovation efficiency of an NIS is measured by the latter's ability to transform innovation input into output and generate profits. Assessing innovation efficiency helps both to identify the best innovation practitioners for benchmarking and to shed light on ways to improve efficiency by highlighting areas of weakness. In empirical management, in countries seeking to enhance policy learning and thus develop more appropriate policy recommendations, examples of “best practices” are currently employed. Additionally, the effect of innovation environment on the innovation process is related to the effectiveness of the innovation policy instruments formed by governments. If the aim is to foster effective innovation policy-making, it is advisable to further investigate the effect of factors embedded in the innovation environment on the efficiency of the KIP based on system thinking. Prior studies (Freeman, 1987, Dosi et al., 1988, Furman et al., 2002, Lundvall, 1992, Nelson, 1993, Edquist, 1997, Fernández-de-Lucio et al., 2001, Fernández-de-Lucio et al., 2003, Fritsch and Slavtchev, 2007 and Fritsch and Slavtchev, 2009) have indicated or empirically demonstrated that the differences in the innovation performance of geographic units are closely related to variation in the innovation environment embedding the innovation process. To obtain effective information for innovation policy-makers, it is important to choose an appropriate modeling/mythological framework to accommodate the production structure of the KIP and its embeddedness in the institutional environment. As an emerging current of thought in the economics of innovation, the innovation systems approach provides a useful analytical tool for the development of innovation policy-making for geographic units (e.g., nation or regions) (Edquist and Hommen, 1999). However, simply grasping the conceptual structure of innovation systems does not allow one to control the operational quality of innovation systems via specific empirical management, which depends on measuring innovation performance and exploring its determinants. This means that the innovation systems approach mainly promotes our understanding of what relevant factors innovation policy-makers should consider. However, to understand what to do to improve innovation performance and how to do it requires (1) the construction of a new measurement framework for benchmarking the innovation performance of geographic units in comparison to “best innovation practitioners” and (2) the creation of an examination framework for exploring the determinants of cross-unit differentials. It becomes necessary to use a two-step integrated analytical framework for this purpose. Possibly for methodological reasons, there has been little formal empirical investigation of these two issues. As Balzat and Hanusch (2004) have argued, the NIS approach still lacks coherent theoretical backing and the methodology necessary to allow for more systemic, empirical comparisons among countries. Using novel modeling tools, this study aims to construct an integrated analytical framework for quantitatively investigating the NIS, taking into account the internal physical transformation structure of the KIP and the embeddedness of the KIP in the external institutional environment. The rest of this paper is organized as follows. Section 2 defines the conceptual production framework of a typical NIS within which the KIP operates under the influence of the innovation environment. Section 3 is devoted to constructing our modeling methodology. Section 4 provides an empirical analysis based on the innovation activities of 22 OECD countries. Finally, Section 5 offers concluding remarks.

Increasing attention should be paid to the efficiency performance of national innovation activities in the development of knowledge-based economies. It is true that if innovative resources are not utilized or commercialized effectively, additional innovative investment may be of little help in stimulating economy progress. The existing literature about the NIS has made considerable theoretical progress, but it fails to empirically investigate the internal transformation process for innovative investment and the participating effects of the policy-based innovation environment based on systems thinking. This study addresses this gap using a novel two-step analytical framework composed of a first-step measurement procedure based on network DEA modeling and a second-step examination procedure using PLSR modeling. 5.1. Theoretical contributions The primary purpose of this paper is to present an integrated process-oriented analytical framework for modeling national innovation systems with interactions between the KIP and the embedded policy-based innovation environment. To improve our measurements, we have constructed a coherent two-step analytical framework including a network DEA measurement model and a PLSR examination model. Our hybrid analytical framework not only describes the internal physical innovation production structure embedded in the innovation environment within an NIS but also produces reliable measurement and examination results. Specifically, we developed a flexible relational network two-stage DEA measurement model for calculating the innovation efficiency of the NIS along with a network two-stage innovation production framework under both CRS and VRS assumptions. Then, we used the PLSR examination model to explore the environmental determinants of efficiency variation across peer NISs for the two individual sub-processes. In the hybrid analytical framework, the proposed relational network two-stage DEA model can account for the interactive relationship between the two component innovation processes and the inefficiency embedded in them. Additionally, using the PLSR model can help to resolve multiple problems: multicollinearity, the strong distribution assumption and the small sample size. Our measurement results help to shed light on the efficiency performance of innovation activities within NISs from a production perspective. Compared with the extant “black-box” measurements obtained using the traditional one-stage DEA model, the dual-role intermediate technologically innovative products (with patents as proxy measures in our study) in innovation measurement are explicitly and fully considered in our analytical framework. The explicit analytical framework provides an in-depth measurement of innovation performance, which should facilitate the development of process-focused innovation policies. Furthermore, compared with the independent efficiency measures, our calculations via network two-stage DEA provide interdependent efficiency measures, which are useful in helping us to compare the two component efficiency scores for each country under the CRS and VRS assumptions. 5.2. Empirical applications In the empirical analysis, the proposed network DEA modeling framework has been novelly introduced into the innovation evaluation field oriented to knowledge innovation process for measuring empirical innovation performance and improving innovation policy-making based on the “best innovation practitioners”. The empirical study benchmarked the relative efficiency of the KIP and the two internal sub-processes (KPP and KCP) of 22 OECD nations. It also explored the determinants of variations in efficiency across those nations in the two individual sub-processes. We have found a great difference in their rankings, and these results imply that there exists a non-coordinated relationship between upstream KPP efficiency and downstream KCP efficiency in most countries. Furthermore, we have found that the overall innovation efficiency of an NIS is mainly subject to the operational quality of the downstream KCP on average. This reminds us that the KCP efficiency improvement should be a primary consideration for most OECD countries as they embark on future innovation policy-making. Furthermore, the results of a second-step examination procedure with PLRS show that various factors related to external innovation policy-based environments had a strong or moderate influence on the efficiency of both the KPP and the KCP, confirming the contribution of government intervention to the innovation performance of the NIS. Our empirical results indicate that it is advisable to take into account the operation of internal interdependent sub-processes in modeling the NIS. 5.3. Limitations and future research The limitations of our conceptual and analytical frameworks provide potential starting points for future work. First, although the proposed conceptual framework of the NIS in this study explicitly characterizes the knowledge innovation process, it is still a simplified structure within which some practical complexities (e.g., dynamics) are not taken into account. As mentioned previously, the input/output measure correspondence in the knowledge innovation process is complicated by inter-temporal dependencies because of time lags and the multiple-period influence of one-time innovation investment on outputs. Secondly, our two-step analytical procedure involves potential problems related to both measurement and statistical inference. The first-step efficiency calculation may be flawed due to possible statistical noise and outliers in the data (Simar and Wilson, 1998 and Simar and Wilson, 2000). The second-step statistic inference results are likely to be biased because they do not necessarily follow from the underlying process generating the data (Simar and Wilson, 2007 and Simar and Wilson, 2008). Thirdly, there are some common issues with our efficiency estimation related to indicator measurements in empirical analysis, partially due to restrictions on statistics. These issues might include the problem of double counting (or vice versa) in indicators’ measurements (Brown and Svenson, 1998). In future empirical analyses on this subject, our hybrid two-step analytical framework can be used to investigate the NIS based on panel data associated with the Malmqusit Index (Guan and Chen, 2010) to measure changes in innovation efficiency over time. Of course, it can also be applied to other geographic units (e.g., regions or sectors). Moreover, our analytical framework is not subject to the internal structure framework, so it can be easily extended to fit more complex production structures. To improve the estimations, bootstrapping procedures (Simar and Wilson, 1998 and Simar and Wilson, 2000) or a robust non-parametric approach (Daraio and Simar, 2007 for order-m or order-α frontier estimation) may be used to reduce outliers, extreme points and statistical noise in the data. Thirdly, to obtain a “pure” estimation of efficiency, it may help to use Fried et al., 1999 and Fried et al., 2002 multi-stage approach in our analytical context, which can help to remove/isolate the influences of the environmental factors from the efficiency measures. Finally, our analytical framework can be used as an alternative approach to investigating the determinants of national process-specific innovative capacity because innovation efficiency embodies the ability of a nation to produce and commercialize a flow of innovative technology.