Review of Industrial Organization (2006) 28:145–164 DOI 10.1007/s11151-006-0010-z

© Springer 2006

The Determinants of Pharmaceutical R&D Expenditures: Evidence from Japan
¨ JORG C. MAHLICHa and THOMAS ROEDIGER-SCHLUGAb,
Economic Policy Department, Austrian Federal Economic Chamber, Wiedner Hauptstrasse 63, A-1045 Vienna, Austria. b Department of Technology Policy, ARC systems research, Donau-City-Strasse 1, A-1220 Vienna, Austria. Author for correspondence. E-mail: thomas.roediger@arcs.ac.at a Abstract. During the past 20 years, the world pharmaceutical industry has experienced a dramatic increase in R&D intensity. We apply and extend a model developed by Grabowski and Vernon (2000, Journal of Evolutionary Economics, 10, 201–215) with a pooled data sample of the 15 publicly listed Japanese drug firms for the period 1987– 1998. As in the original study, we find expected returns to be an important determinant of R&D spending in the Japanese drug industry, albeit considerably smaller than in the U.S., which is particularly obvious in the case of returns from newly introduced drugs. However, our results are sensitive to econometric model specification, in particular to controlling for serial correlation and to a dynamic specification of the baseline model. Likewise, estimates on financial constraints are sensitive to model specification, indicating that Japanese drug firms face small or no financial constraints. Our results are consistent with the general literature on R&D investment behaviour, yet raise some methodological questions with regard to the original study. Key words: investment, Japan, panel data estimation, pharmaceuticals, R&D. JEL Classifications: L65, O31, O33.

I. Introduction During the past 20 years, the world pharmaceutical industry has experienced a dramatic increase in R&D intensity. In a recent paper, Grabowski and Vernon (2000) explore the determinants underlying this trend in R&D investment behaviour for a pooled sample of 11 major U.S. drug firms over the period 1974–1994. Estimating a static panel data model, they find that expected returns and cash flow are important explanatory variables of firm research intensities during this period. In this paper, we explore whether similar conclusions also hold for the Japanese pharmaceutical industry. Given institutional differences and the differences in the way how firms are funded in the U.S. and Japan,

146

¨ JORG C. MAHLICH AND THOMAS ROEDIGER-SCHLUGA

there is good reason to expect divergent results. In particular, the U.S. is the only drug market in the world where prices are unregulated, and financial systems are quite different in the two countries. Using a pooled data sample of the 15 publicly listed (and largest) Japanese drug firms for the period 1987–1998, we start out with the baseline model estimated by Grabowski and Vernon (2000). We then extend their model to accommodate for special features of the Japanese drug market and for econometric issues not dealt with in the reference paper. This aligns our results with the existing general literature on the determinants of investment in R&D. We find expected returns to be an important determinant of R&D spending in the Japanese pharmaceutical industry. In absolute terms, however, R&D spending is considerably less responsive than in U.S., in particular in a dynamic panel data model, which is a theoretically more appropriate specification. Likewise, sensitivity to cash flow is considerably smaller in Japan and disappears in the dynamic model formulation, confirming the result of other studies that Japanese firms face fewer or even no financial constraints. The paper is structured as follows. In Section II, we briefly highlight the particular features of R&D, which impact the R&D investment decision. In Section III, we describe recent trends in pharmaceutical R&D in Japan and discuss the particularities of the Japanese drug market. In Section IV, we present the data and discuss our econometric model specifications. In Section V, we list and interpret our estimation results. Section VI summarises the paper and concludes with directions for future research. II. The R&D Investment Decision According to basic economic theory, a firm’s R&D investment decision is determined by the intersection of the marginal rate of return on investment schedule and the marginal cost of capital schedule. These are obtained by arranging potential investment projects in order of decreasing rates of return and the opportunity costs of alternative investment in increasing order. However, investment in R&D differs from ordinary investment in a number of fundamental ways, which impact the way investment in R&D is modelled. First, and perhaps most importantly, a significant share of R&D spending are wages and salaries of scientists and engineers, who create ideally profitable new knowledge. This knowledge is specialized, sunk and intangible, and its value hence very difficult to determine. Second, even to the extent that R&D produces tangible output, its ultimate value is uncertain and difficult to appraise ex-ante. Time horizons may be quite long, as in the pharmaceutical industry, where it takes

THE DETERMINANTS OF PHARMACEUTICAL R&D EXPENDITURES

147

14 years on average at present to develop and introduce a new chemical entity (NCE) (see PhRMA, 2003). Not only is the variance of the distribution of expected returns much larger than for other types of investment. Rather, returns to R&D are highly skewed (Scherer et al., 2000) and appear to be fat-tailed, which may affect the computation of moments. In a recent paper, Silverberg and Verspagen (2005) estimate Pareto exponents