Firm size and types of innovation

Abstract

We propose a general theory of innovation that illustrates the relative benefits of performing process versus product R&D when firm size is endogenous. A firm's size, scope, and R&D portfolio are shown to reflect the same underlying characteristic of the firm, namely manufacturing efficiency. We demonstrate that efficient firms become larger, have greater scope, and perform more of both process and product R&D. In light of decreasing returns to R&D, this implies small firms obtain more product innovations per dollar of R&D than large firms, which is consistent with evidence we present that small firms are more innovative than large firms as they obtain more patent counts and citations per dollar of R&D.

Keywords:

• firm size distribution
• technological change
• R&D portfolio

JEL Classification :

• L11
• L25
• O33

Acknowledgements

I would like to thank the editor Cristiano Antonelli, three anonymous referees, Michael Gort and workshop participants at the University at Buffalo for helpful comments, Bronwyn Hall for making available the NBER patent citations dataset, Dunli Li for research assistance, and the Center of Excellence on Human Capital, Technology Transfer, and Economic Growth and Development (HEAD) for financial support.

Notes

The exception here is Miravete and Pernias (2006): innovations are classified via direct interviews with managers.

In a 1982 sample of 247 four-digit industries from the US Small Business Administration, Acs and Audretsch (1987) find that large firms perform more innovations per employee than small firms in markets characterized by imperfect competition, specifically industries that are advertising- and capital-intensive, concentrated, highly unionized, and produce a differentiated good; while small firms perform more innovations per employee than large firms in industries in the early stages of the life cycle, where innovation and the use of skilled labor play a large role.

To see this, let S denote firm size, and let R(S) denote the product R&D expenditures of a firm with size S. By hypothesis, small firms perform more product R&D, so . Let DR(S)^β denote the number of product innovations performed by a firm incurring product R&D expenditures R(S), where D>0 is a constant. By hypothesis, there are decreasing returns to product R&D, so β∊(0, 1). Let denote the number of product innovations per dollar of product R&D. We find since and β∊(0, 1). In other words, large firms are more innovative per dollar of product R&D than small firms if small firms perform more product R&D and there are decreasing returns to product R&D.

Negative binomial MLE takes into account the fact that the dependent variable is a count variable.

Jensen (1987) analyzes 28 firms in the US pharmaceutical industry over the period 1969–79. For 21 of the 28 firms studied, the elasticity of the expected number of discoveries with respect to R&D is not significantly different from one, but it is significantly less than one for the five smallest firms. Once the effect of firm size on R&D is controlled for, there are no advantages or disadvantages to locating a given R&D program in a smaller firm.

In studying about 2000 significant innovations in Britain since 1945, Pavitt (1984) finds that innovating firms in electronics and chemicals are relatively big and innovate over a wide range of product groups; firms in mechanical and instrument engineering are relatively small and specialized; large firms in scale-intensive sectors like metal manufacture and vehicles make contributions to their own process technology; and for textile firms, most process innovations come from suppliers.

The qualitative results of the model do not change if a patent lasts more than one period.

Process R&D is assumed to solely have a contemporaneous effect on production costs. The qualitative results would not change if current process R&D efforts affect current and future production costs.

Cohen et al. (1987) find that overall firm size has a very small, statistically insignificant effect on business unit R&D intensity when either industry fixed effects or measured industry characteristics are taken into account. Business unit size has no effect on the R&D intensity of business units that perform R&D, but it affects the probability of conducting R&D. Business unit and firm size jointly explain less than 1% of the variance in R&D intensity; while industry effects explain nearly half the variance.

In traditional Cournot theory, a monopoly produces more than an oligopolist. Though we cannot prove this property holds in our model in general, it can be shown for specific examples.

In Nesta and Saviotti (2005), scope is measured as the number of technology classes in which a firm has applied for patents. In Henderson and Cockburn (1996), scope is measured as the number of research programs run by the firm with a budget exceeding $500,000. In Lunn (1987), diversification is measured as the number of industries of origin for which the firm has obtained its patents.

That is, if we envision R&D as a Schumpeterian engine of growth, then a larger market generates higher growth.