Department of Statistics Unitmark
Dietrich College of Humanities and Social Sciences

Measuring Appropriability in Research and Development with Item Response Models

Publication Date

February, 1999

Publication Type

Tech Report

Author(s)

Matthew Johnson, Wesley Cohen and Brian Junker

Abstract

The extent to which firms are able to capture, or appropriate, the profits created by their innovations is thought to be a key determinant of the amount of Research and Development (R&D) they do. Firms lose the incentive to innovate if they are unable to capture the returns on their innovations, and may perform R&D at a less than socially optimal level. For this reason, for example, the National Institute of Standards and Technology's (NIST) Advanced Technology Program (ATP) invests in such innovations, and by cost-sharing research to foster new, innovative technologies, benefits the U.S. economy (NIST 1998).

In this paper we use the results of the 1994 Carnegie Mellon Survey of Industrial R&D in the U.S. Manufacturing Sector (CMS; Cohen, 1996) to model and examine the effectiveness of six appropriability mechanism on product and process innovations in 72 U.S. industries. The survey questions address subjects such as information flows, time until competitors imitate innovations, the amount of R&D focused on product versus process innovations, and the effectiveness of various appropriability mechanisms. R&D unit directors from 1489 units responded to the survey, and self-reported the industries they do work in. Our goal is to determine how well the various appropriability mechanisms work in protecting firms' profits; and to determine how the effectiveness of these appropriability mechanisms varies across industries, and varies according to the focus of the R&D unit: product or process innovations.

We develop a mixed effects generalized linear model (Stiratelli, Laird, and Ware 1984) to help characterize the responses to the CMS survey questions on appropriability mechanisms. This class of statistical models contains item response theory (IRT; van der Linden and Hambleton, 1996) models which are commonly used in educational testing. Using our statistical model, we investigate whether industries differ in their ability to appropriate returns created by innovations, the effectiveness of each mechanism relative to others, and the extent to which response behavior is a function of the focus, product versus process innovations, of the R&D.