Article contents
Size of Firm, Oligopoly, and Research: A Comment*
Published online by Cambridge University Press: 07 November 2014
Abstract
![Image of the first page of this content. For PDF version, please use the ‘Save PDF’ preceeding this image.'](https://static.cambridge.org/content/id/urn%3Acambridge.org%3Aid%3Aarticle%3AS0315489000002115/resource/name/firstPage-S0315489000002115a.jpg)
- Type
- Notes
- Information
- Canadian Journal of Economics and Political Science/Revue canadienne de economiques et science politique , Volume 31 , Issue 2 , May 1965 , pp. 256 - 266
- Copyright
- Copyright © Canadian Political Science Association 1965
Footnotes
The research underlying this paper was supported by a grant from the Inter-University Committee on the Microeconomics of Technological Change, sponsored in turn by the Ford Foundation. Use was also made of computer facilities supported in part by National Science Foundation grant NSF-GP579. I am indebted to Mrs. Noah Meltz for help in collecting data.
References
1 “Size of Firm, Oligopoly, and Research: The Evidence,” this Journal, 02 1964, 62–75.Google Scholar
2 Worley, J. S., “Industrial Research and the New Competition,” Journal of Political Economy, 04 1961, 183–6.CrossRefGoogle Scholar
3 US National Academy of Sciences, National Research Council, Industrial Research Laboratories of the United States, 10th ed., 1956.Google Scholar The 11th edition of this publication was the source of Hamberg's R&D data.
4 See Johnston, J., Econometric Methods (New York, 1963), 148–50.Google Scholar
5 An even more violent change occurred in the correlation coefficients. In patenting, as in R&D employment, the highest correlations for the observations of all industries together were obtained with employment as the scale variable. But when the electrical industry firm observations were deleted, assets turned out to be more strongly correlated with patenting than employment. This was true when the variables were correlated in their natural form as well as when logarithms were taken.
6 Cf. Habakkuk, H. J., American and British Technology in the Nineteenth Century (Cambridge, 1962), 163.Google Scholar
7 Scherer, F. M., Herzstein, S. E., et al., Patents and the Corporation (rev. ed., Boston, 1959), 114.Google Scholar
8 “Research and Development, New Products, and Productivity Change,” American Economie Review, 05 1962, 179.Google Scholar
9 Scherer, Herzstein, et al., Patents and the Corporation.
10 Cf. US National Science Foundation, Science and Engineering in American Industry (1956), 46–7Google Scholar; and Mansfield, Edwin, “Industrial Research and Development Expenditures,” Journal of Political Economy, 08 1964, 319–32.CrossRefGoogle Scholar
11 Although Hamberg used 1960 data, the base year 1955 is accepted here for two reasons. One is a practical one: I had data for that year only in usable form. But in addition, 1955 data give a better picture of private R&D incentives, since in that year the US federal government supported only (?) 47 per cent of all R&D performed by industry, compared to 58 per cent in 1960. See US National Science Foundation, Reviews of Data on Research and Development, 04 1962, p. 2.Google Scholar
National Science Foundation expenditures survey data provide some insight into the possibility of structural changes in the period separating our data. Between 1955 and 1960 total R&D expenditure intensity, and perhaps therefore employment intensity, increased more for firms with more than 5,000 employees than it did for smaller firms. But this was due almost entirely to an increase in the concentration of government R&D contract expenditures. See National Science Foundation, Research and Development in Industry: 1961, 39–45.Google Scholar
12 On the other hand, regressions of this form are biased in a less serious way by errors in measuring the scale variable, as discussed earlier. The greater such errors are, the more the evidence of either increasing or decreasing intensity will be obscured, since the non-linear coefficients will be biased towards zero. With the logarithmic approach of equation (2), the bias is consistently in the direction of showing decreasing intensity.
13 See Mansfield, Edwin, “Size of Firm, Market Structure, and Innovation,” Journal of Political Economy, 12 1963, 565–7.Google Scholar It should be noted that both of the industry-by-industry analyses undertaken by Hamberg—a rank correlation of R&D employment per thousand employees with scale and the equation (1) test discussed earlier-give maximum relative weight to the smaller firms. The rank correlation approach, for example, implies that the difference of 128,000 employees between US Steel and Bethlehem Steel is equivalent to the difference of 200 employees between Lukens Steel and Granite City Steel. If, as I conclude, research intensity increases with firm size for small but not for large firms, the emphasis on small firms in Hamberg's equation (1) tests could explain the predominance of b coefficients exceeding 1.0. In this respect it should also be noted that among the 448 firms in my full sample, the 73 firms with sales exceeding $500 million accounted for 63 per cent of the 1955 sales of all the sampled firms. This concentration of research-supporting potential is good reason for placing statistical emphasis on the behaviour of the largest firms.
14 All coefficients in these two equations were statistically significant at the .05 level or higher.
15 The great contrast between General Motors and Ford is clearly not due to an imbalance of defence orders, since General Motors' defence sales were only about three times Ford's in 1955. It seems to have two main grounds: the greater diversification of General Motors into such dynamic fields as electronics, and the tradition of technical conservatism inherited by Ford from the last years of the senior Henry Ford's reign. Financial and styling reforms came much more quickly during the administration of Henry Ford II than did changes in research and development policy.
16 See also Mansfield, “Size of Firm, Market Structure, and Innovation,” who uses a similar methodology and reaches similar conclusions in his analysis of important innovations in four industries. Precise location of the scale conducive to maximum research intensity by this method is hazardous, since estimates for the smallest firms may be biased by small errors in the R&D function's intercept term estimate. The intercept estimates are subject to error not only because of random disturbances and curve-fitting problems, but also because no 1955 sales observations under $55 million were included in my sample.
17 No attempt is made here to analyse eight other industry groups defined in connection with my study of corporate patenting. In most of these cases there were too few observations for a meaningful non-unear analysis. The aircraft industry is also excluded because private incentives have so little to do with a contractor's military R&D employment.
18 Mansfield ako found the chemicals industry to be the only one with increasing intensity in his analysis of R&D spending in five industries. “Industrial Research and Development Expenditures,” 334.
19 Because of its inflectionless character equation (1) also does not fit the 1955 R&D employment data as well as equations (3) and (4). The standard errors of estimate with all 352 available observations were 1,081 (employees) with equation (1), 999 with equation (4), and 927 with equation (3). The differences between (1) and (3) are significant at the 1 per cent point in an F–ratio test.
20 All of the 67 firms not listed in the 1955 National Research Council survey and which received four or fewer patents in 1959 had 1955 sales of less than $400 million. Thirty-seven of them had sales of less than $100 million, and 17 sales between $100 million and $199 million.
21 Some hypotheses concerning these exceptions are presented in my paper, “Firm Size, Market Structure, Opportunity, and the Output of Patented Inventions,” American Economie Review (forthcoming).
- 46
- Cited by