This paper analyzes the impact of Research and Development (R&D) on the productivity of China's high technology industry. In order to capture important differences in the effect of R&D on output that arise from geographic and socioeconomic differences across three major regions in China, we use a novel semiparametric approach that allows us to model heterogeneities across provinces and time. Using a unique provincial level panel dataset spanning the period 2000-2007, we find that the impact of R&D on output varies substantially in terms of magnitude and significance across different regions. Results show that the eastern region benefits the most from R&D investments, however it benefits the least from technical progress, while the western region benefits the least from R&D investments, but enjoys the highest benefits from technical progress. The central region benefits from R&D investments more than the western region and benefits from technical progress more than the eastern region. Our results suggest that R&D investments would significantly increase output in both the eastern and central regions, however technical progress in the central region may further compound the effects of R&D on output within the region.

No abstract

In efficiency studies using the stochastic frontier approach the main focus is to explain inefficiency in terms of some exogenous variables and computation of marginal effects of each of these determinants. Although inefficiency is estimated by its mean conditional on the composed error term (the Jondrow et al. (1982) estimator), the marginal effects are computed from the unconditional mean of inefficiency (Wang (2002)). In this paper we derive the marginal effects based on the Jondrow et al. estimator and use the bootstrap method to compute confidence intervals of the marginal effects.

This paper uses a semiparametric smooth coefficient model (SPSCM) to estimate TFP growth and its components (scale and technical change). The SPSCM is derived from a nonparametric specification of the production technology represented by an input distance function, using a growth formulation. The functional coefficients of the SPSCM come naturally from the model and are fully flexible in the sense that no functional form of the underlying production technology is used to derive them. Another advantage of the SPSCM is that it can estimate bias (input and scale) in technical change in a fully flexible manner. We also used a translog input distance function framework to estimate TFP growth components. A panel of U.S. electricity generating plants for the period 1986 − 1998 is used for this purpose. Comparing estimated TFP growth results from both parametric and semiparametric models against the Divisia TFP growth, we conclude that the SPSCM performs the best in tracking the temporal behavior of TFP growth.

No abstract