Working Paper: NBER ID: w0438
Authors: Ariel Pakes; Mark Schankerman
Abstract: This paper explores the economic factors which determine the variation of research effort across firms. The intra-industry coefficient of variation of research intensity is much larger than those of traditional factors. We show that this important fact is consistent with the theoretical argument that knowledge possesses unique economic characteristics, and that the demand for research depends both on the parameters of the production function for knowledge and on the ability of the firm to appropriate the benefits from the knowledge it produces. We propose and implement a framework for decomposing the observed intra-industry variance In research intensity into three components: demand inducement, a firm-specific structural parameter, and errors in the observed variables. The main empirical findings are that errors in the variables (especially research) are important, that very little of the structural variance in research intensity is accounted for by demand inducement, and that the bulk of the variance is related to differences in the firm-specific parameter. Both the theoretical and empirical analysis indicate that it is not reasonable to treat the demand for research in a manner analogous to the demand for traditional inputs, including capital. Substantially richer models are required to provide insight into the structure of incentives driving the demand for research.
Keywords: Research intensity; Economic factors; Knowledge production; Private returns
JEL Codes: O31; O32
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| intraindustry coefficient of variation of research intensity (O32) | research demand (R22) |
| differences in firms' ability to appropriate benefits from knowledge + expected growth rates of output (F12) | intensity of research efforts (O32) |
| measurement errors in research intensity (C83) | variance in observed research intensity (D29) |
| firm-specific parameters (L20) | variance in research intensity (I23) |