On the efficiency of national innovation systems

https://doi.org/10.1016/S0038-0121(02)00046-0Get rights and content

Abstract

This paper presents a non-parametric approach to the disentanglement of the related effects of efficiency and productivity of a country's technological effort. The ultimate purpose is to identify the extent to which the alleged decrease in the productivity growth of many countries can be explained by differences in efficiency and by differences in its components, namely scale and congestion. The role of moderators in explaining productivity fluctuations if also assessed. The database consists of the countries included in the World Competitiveness Report.

Introduction

In his 1994 Presidential inaugural to the American Economic Association, Griliches [1, Chapter 14] addressed the issue of the effect of research and development (R&D) investment on productivity and problems associated with the quality of existing data needed to study this relationship. One of the more salient issues of the presentation is the contention that the impact of R&D on productivity growth has declined over time in most of the countries studied. In fact, Griliches [1], as well as many of the references listed therein, present a thorough analysis of the R&D/productivity relationship. The basic conclusion is that, whereas R&D policies are certainly an important contributor to productivity growth, variations in national productivity growth cannot be entirely explained by dissimilarities of the various countries’ R&D policies. Market distortions also account for an important portion of these variations e.g. [1], [2], [3], [4]. Some may lead to under-investment in R&D, at least compared to what are considered socially acceptable levels. Examples include the surplus appropriability problem, knowledge spillovers, monopolistic pricing and the long-term nature of the relationship between R&D and productivity. Some, such as imitation, creative destruction and duplication externalities, induce over-investment in R&D. Still others may go either way, depending upon the intended purpose of the distortion.

This paper examines the proposition that, even if data considerations and market distortions are indeed important reasons in explaining the decline in the impact of R&D upon productivity growth, the modelling used for testing the hypothesis requires substantial modification. This is needed to disentangle the related effects of efficiency and productivity of a country's technological effort. The basic rationale for this position is threefold. First, the overwhelming majority of studies on the subject assume that existing technology exhibits constant returns to scale e.g. [2]. However, in their productivity analysis of the Services, Total Industry and Manufacturing sectors of 14 Organization for Economic Co-operation and Development (OECD) countries, Arcelus and Arocena [5], [6], and others, found strong evidence of variable returns to scale for many of these countries across many years. Such evidence has clear efficiency implications, independent of productivity considerations e.g. [7], suggesting the need to differentiate between the two concepts.

Second, there is a problem related to the existence of two types of efficiency and their interpretations: R&D efficiency and R&D productivity. R&D efficiency is related to usage of its input/output ratio, while R&D productivity considers the contribution of R&D effort to the national economy. These distinctions have been frequently neglected in the literature [8], yet are considered explicitly in the present study. Thus, this paper considers the incidence and degree of a country's R&D effort within the context of its National Innovation System (NIS) e.g. [9], [10], [11], [12], [13], [14]. An NIS may be defined as the “network of agents and set policies and institutions that affect the introduction of technology that is new to the economy. The key aspects of NIS are the extent to which the economy acquires technology from abroad, the intensity of domestic technological effort it undertakes, and the level of technical human capital” [9, pp. 541–542]. NISs are formed in order to foster development, application, and diffusion of technology, thereby improving productivity. As a result, investment in technology can be considered one of the inputs of a sector of an economy. Within this context, efficiency relates to an NIS's ability to transform R&D inputs into R&D outputs. The presence or absence of changes of national productivity attributable to that transformation process provides a way to test whether the said process is carried out with constant or with variable returns to scale relative to that of other countries.

Third, an important distinction not generally made is whether an R&D effort is assessed for the country's ability to generate or to consume technology. The latter is the purview of most countries; the former is practiced by very few. Hence, differences in productivity and efficiency between the two types of economic activity are expected to abound. As a result, the current analysis is carried out with the variables weighted by POPU, the country's population, and then repeated with the variables weighted by GDP, the gross domestic product. POPU is used as a proxy for a the size of a country's market for technology, as no other index of market power is available for the entire data set. It is intended to account for a country's potential absorptive capacity as a receptor of technology, be it developed in the country or acquired abroad. GDP, as a measure of wealth, is a proxy for each country's ability to generate technology and, hence, provides a measure of a country's potential as a producer/generator of technology.

The paper is organized as follows. Section 2 describes the research framework. Included here are (i) identification of the variables embedded in an NIS; and (ii) the two-step model designed to evaluate the efficiency of each country's NIS. Section 3 develops the first step of the model, namely, measurement of the efficiency with which the inputs are transformed into outputs using data envelopment analysis (DEA). The resulting efficiency indexes are then decomposed to purge the influences of congestion and scale. The second step, subject of Section 4, analyzes the extent to which the moderating elements of the various NIS’ can explain the fluctuations in efficiency indexes and in those measures resulting from the decomposition. Conclusions are provided in the last section of the paper.

Section snippets

The research framework

This section outlines the two-step model designed to measure the efficiency of each country's NIS. Initially, however, it is necessary to identify those variables used in the description of a country's NIS.

Step 1: measuring the efficiency of the input/output transformation process

The first step of the efficiency model computes the DEA-based efficiency measures using the inputs and outputs identified previously. This section presents first the DEA formulation and its decomposition into various sources of efficiency/inefficiency. The issue here is to determine the extent to which these sources are responsible for the alleged decrease in productivity gains of the countries in the sample. This is followed by an analysis of the empirical results. The data with the two

Step 2: the effect of moderators on efficiency

As stated in Section 2.1, the moderators cannot be classified, within the DEA context, as either inputs or outputs. Hence, their inclusion into the DEA formulation is questionable, even if their directional effect upon efficiency may resemble that of an input or that of an output. This section considers the second step of the efficiency model subject of this paper; namely, the effect of moderators or environmental variables on the NIS efficiencies computed in the first step.

In principle, the

Some concluding comments

This paper has endeavored to assess the role of R&D on a country's productivity. It has presented substantial evidence in favor of the proposition that most countries operate under VRS. This accounts for an important portion of the inefficiency embedded in the results, as well as of the diminishing returns evidence reported in the literature. It has also illustrated the importance of differentiating between R&D's contribution to national productivity and its ability to transform inputs into

Acknowledgements

Financial assistance to the second-listed author for the completion of this research from the Natural Sciences and Engineering Research Council of Canada is gratefully acknowledged, as is the computational assistance of Tricia Maksymnuk. We would also like to thank Barnett R. Parker, Editor-in-Chief, and the anonymous referees for their comments in improving the paper.

References (32)

  • G. Dosi et al.

    Technical change and economic theory

    (1988)
  • C. Edquist

    Systems of innovation approachestheir emergence and characteristics

  • C. Freeman

    The national system of innovation in historical perspective

    Cambridge Journal of Economics

    (1995)
  • Lundwall BA, editor. National systems of innovation: towards a theory of innovation and interactive learning. London:...
  • Nelson, RR, editor. National innovation systems: a comparative analysis. Oxford: Oxford University...
  • W. Nasierowski et al.

    On the stability of countries’ national technological systems

  • Cited by (129)

    • Impacts of environmental constraint target on green innovation efficiency: Evidence from China

      2022, Sustainable Cities and Society
      Citation Excerpt :

      The comparison of the number of green patents cannot reflect the difference of green innovation capability between different regions. The latter mainly constructs comprehensive indicators such as green innovation efficiency, but uses the total patent data, which cannot highlight the "green" concept of green innovation efficiency (Nasierowski and Arcelus, 2003; Feng et al., 2018; Luo and Zhang, 2021). In other words, using comprehensive index just measures innovation efficiency instead of green innovation efficiency.

    View all citing articles on Scopus
    View full text