An empirical investigation of the National Innovation System (NIS) using Data Envelopment Analysis (DEA) and the TOBIT model

Abstract

The goal of this paper is to investigate national innovation systems’ input–output components and to model a robust efficiency measurement using the DEA Bootstrap technique. Most of the previous NIS studies are descriptive and little emphasis is given to complex analysis. In our previous study, we evaluated the innovation performance of 20 emerging and developed countries, from the point of view of technical efficiency. This study makes an important contribution using the DEA Bootstrap technique, whereby we rank the countries based on bias-corrected estimation parallel to conventional DEA efficiency. The efficiency scores obtained from this technique show which countries are considered to be innovation leaders because their innovation performance is efficient under both constant and variable returns to scale in the process of transforming innovation inputs into innovation outputs. We suggest some key policy implications that can be learned from these innovation leaders. Subsequently, we apply the Tobit model to explain inefficiency. Based on the Tobit regression model, the DEA CRS technical efficient score of inefficient countries could be improved through three main variables: the secondary school enrolment ratio; the labour force (ages 15–65), as a percentage of the total population; and domestic credit expansion by the business sector, as a percentage of GDP.

Keywords:

• national innovation system
• knowledge economy
• DEA bootstrap technique
• Tobit model
• inefficiency

JEL Classifications:

• O20
• O30
• O57

Notes

1. Distance usually relates to all of the attributes and assumes that they all have the same effects on distance. This incorrect classification, due to the presence of many irrelevant attributes, is often termed the problem of dimensionality.

2. Korean term for a conglomerate of many companies clustered around one parent company. The companies usually hold shares in each other and are often run by one family.

3. DEA Bootstrapping needs a larger data set to improve the quality of estimation. This is a limitation of our research. Using a larger data set, we may have more robust results to draw inferences. Nevertheless, we try to demonstrate how to obtain bias-corrected results, which is missing in conventional DEA analysis.