Inference on the Sources of Innovative Inefficiency Using Data Envelopment Analysis and Semi-parametric Bootstrapping Inference on the Sources of Innovative Inefficiency Using Data Envelopment Analysis and Semi-parametric Bootstrapping

Using the US states plus the District of Columbia, output-oriented innovative efficiency scores are computed using data envelopment analysis (DEA) where total patent counts serve as the measure of innovative output and industry and academic research and development expenditures, and numbers of research scientists, graduate students, and postdoctoral positions are inputs. In the second stage the extent to which state and local government fiscal policy explain the estimated inefficiency is considered. Using a truncated regression model where parameter inference is based on semi-parametric bootstrapping, the results indicate that reducing government administration expenditure share, increasing total education expenditure share, and reducing the share of revenues collected from institutions of higher learning correlate with greater efficiency. INTRODUCTION Technology plays a critical role in economic development and it is precisely for this reason that so much research has been devoted to explaining the innovation process . However, it is interesting that so few studies have addressed issues related to innovative efficiency. Roughly speaking, a scientist conducting research will require a lab, lab equipment, research assistants and graduate students, etc. The idea a scientist in their lab would be consciously wasteful, i.e. inefficient, is of course absurd. However, given that innovative activity is region specific, it is conceivable that factors beyond the control of scientists themselves may give rise to inefficiency in the short-run (i.e. administrative and bureaucratic red tape). Unfortunately existing research is scant regarding the relationship between regional and other location-specific characteristics and innovative inefficiency. The purpose of this paper is to explore the relationship between innovative inefficiency and government policy. To evaluate this relationship, two important issues need to be addressed. First, given the complexities of the knowledge production function, how can we quantify innovative inefficiency it is present? Secondly, what type of regression framework is appropriate to test for and assess the significance of any potential interrelationships between this measure of innovative inefficiency and a collection of government policy variables? To explore the link between innovative efficiency and government policy, this paper adopts a two-step procedure. In the first stage, data envelop analysis (DEA) is used to estimate measures of innovative inefficiency where innovative output is measured by total patents granted (utility patents, design patents, defensive publications, etc) and inputs include R&D expenditures and research-related labor using data for the contiguous United States plus the District of Columbia. The implicit assumption is that all innovative output and inputs are homogeneous and given an equal weight when estimating measures of innovative inefficiency via DEA. That all states should strive to be efficient in this sense is reasonable since under the DEA methodology, each state is compared only to its peers. The innovative efficiency of states with a large patent output is thus compared only to other states with similar large patent outputs, while states reporting fewer patents are compared to states that are less patent prolific.