Do firms face a trade-off between the quantity and the quality of their inventions?
Highlights
► We show that firms face a trade-off between invention quality and quantity. ► We address the identification problem caused by differences in patent propensity. ► Invention quality must be taken into account when assessing research productivity.
Introduction
An important topic in the economic and management of innovation literature is the study of firms’ research productivity (e.g. Henderson and Cockburn, 1996, Penner-Hahn and Shaver, 2005, Girotra et al., 2010). Research productivity is the quality-constant measure of the efficiency at which inputs to the innovation process are converted into output. The quantity of output is the number of inventions created and it is often proxied with the number of patents, as imperfect a proxy as it is (Griliches, 1990). The quality of output is more difficult to define and, a fortiori, to measure. Following Lanjouw and Schankerman (2004), the quality of output encompasses both the technological and economic value of inventions.
The existing research has mostly focused on the determinants of the quantity of inventions created, with quality considerations usually relegated to the background. Yet, the quality dimension is just as critical to the understanding of research productivity as the quantity dimension. Little is known about the relationship between the quantity of inventions created and their quality, in particular with respect to a potential trade-off between these two dimensions. For any given level of research inputs, it seems obvious that an increase in the number of inventions created would be associated with inventions of lower average value. This is not necessarily the case, however. For one thing, some otherwise comparable firms may be more productive than others due to a better use of IT resources, more appropriate contracting and management practices, or more skilled researchers. For another, it is difficult to target a quality level due to the uncertain nature of the innovation process. A firm investing all its resources in a risky but promising project may end up with a limited output of low value. And further, dynamics of the invention process itself may affect the quantity/quality trade-off. For instance, Fleming (2001) shows that inventors’ experimentation with new components and combinations leads to less success on average, but it also increases the variability of success that can lead to breakthrough inventions. To the best of our knowledge, the existence of a trade-off has yet to be shown.
Only a handful of studies have looked at the hypothesis of a trade-off and none has come up with conclusive evidence. A first group of studies has tested the hypothesis using firm-level patent data. Lanjouw and Schankerman (2004) regress the number of patents per dollar invested in R&D on an index that captures the mean quality of patent applications for a panel of U.S. firms. Within-firm regressions provide no support for the hypothesis, while between-firm regressions provide only weak support. Sørensen and Stuart (2000) provide an indirect test of the hypothesis. Applying organizational theory, they argue that firm age should be positively associated with the rate of innovation, but negatively associated with how influential the innovations are. They find strong evidence that firm age increases the rate of (patented) inventions, but only weak evidence on the impact of firm age on patent quality. A second group of studies uses patent data at the inventor level. Mariani and Romanelli (2007) use data from the PatVal-EU survey of inventors to test whether the quantity of patents produced affects their average value as measured by patent indicators. They find a positive effect of the quantity of patents on the average quality using forward citations but no effect using a composite value indicator. Gambardella et al. (2011) propose a related test of the hypothesis. The authors also use data from the PatVal-EU survey of inventors but rely on a self-assessed measure of value. They find a negative relationship between the number of inventions and the average value of the patents in the portfolio, but the effect is not statistically significant.1
A limitation of existing studies is that they rely on patent data but do not address the confounding effect of the propensity to patent. Were the decision to patent an invention independent of its value, the trade-off would easily be estimated with patent data. One would simply regress the number of patents against average patent value. A negative coefficient would signal a trade-off. However, since the marginal value of inventions patented is likely to decrease with the propensity to patent (defined as the proportion of inventions patented), a negative correlation between patent quantity and patent quality would not be evidence of a trade-off.
The objective of this paper is to test whether innovative firms face a trade-off between the quantity and the quality of their inventions (holding research inputs constant). An attempt to address the identification problem caused by heterogeneous patent propensities is a distinguishing feature of this study. We put forward an empirical model that links the average quality of inventions with the average quality of patent applications and adopt an instrumental variable approach to account for differences in the propensity to patent. The econometric analysis uses cross-sectional survey data on patent applicants at the European Patent Office (EPO) and finds evidence of a trade-off between invention quantity and quality, as measured by the patent family size. The existence of a trade-off has profound implications for the economics of science and the management of innovation. It stresses the need to take the quality of inventions into account to properly assess – and to study the determinants of – the productivity of research spending. It also raises questions regarding the optimal quantity–quality mix that firms should target.
The paper is organized as follows. Section 2 introduces the empirical framework and the data is presented in Section 3. Section 4 presents the econometric results and the final section discusses the implications of the findings.
Section snippets
Empirical framework
A careful econometric analysis is needed to control for the potentially confounding effect of patent propensity. The next section introduces the building blocks that are necessary to test the existence of a trade-off, while the following section presents the econometric implementation.
The data
The data come from the Applicant Panel Survey carried out from June to September 2006 by the EPO. The main purpose of the survey is to provide information on filing intentions for the EPO's forecasting exercise for budgetary planning purposes. The population is composed of all applicants that have filed at least one patent application at the EPO in the year 2005. A sample of 2098 applicants was selected, partly from among the largest applicants and partly at random, covering overall about 31
Results
The first step, the estimation of the quantity parameter from Eq. (4), is only an intermediary step and is not reported. The dependent variable is the number of inventions (inv) and the covariates are the number of researchers (res), the number of employees (emp), a dummy representing whether the firm is part of a group (group), as well as country and technology dummies. Since the number of inventions is a count variable and exhibits overdispersion (the observed variance is larger than the
Discussion and conclusion
This paper tests whether firms face a trade-off between the quantity and the quality of their inventions. The empirical test is not trivial because inventions are usually observed at the point of patent application. Since the marginal value of inventions patented is likely to decrease with the propensity to patent, a negative correlation between patent quantity and patent quality would not necessarily be evidence of a trade-off. The empirical analysis addresses the identification problem caused
Acknowledgements
The author is grateful to John Haisken-DeNew, Paul Jensen, Anne Leahy, Christian Lebas, Kwanghui Lim, Alfons Palangkaraya, Peter Sivey, Russell Thomson, and Beth Webster for helpful comments and discussions as well as to Peter Hingley for having provided access to the data. Dominique Guellec and Bruno van Pottelsberghe provided valuable comments on an earlier version of the paper. The author also would like to thank 4 anonymous referees for helpful comments.
References (42)
- et al.
R&D and the patent premium
International Journal of Industrial Organization
(2008) - et al.
Patent systems for encouraging innovation: lessons from economic analysis
Research Policy
(2006) The decision to patent, cumulative innovation, and optimal policy
International Journal of Industrial Organization
(2005)Industrial research as a source of important patents
Research Policy
(1998)- et al.
Applications, grants and the value of patent
Economic Letters
(2000) - et al.
Citations, family size, opposition and the value of patent rights
Research Policy
(2003) - et al.
What is behind the recent surge in patenting?
Research Policy
(1999) - et al.
Stacking and picking inventions: the patenting behavior of European inventors
Research Policy
(2007) - et al.
Faster, better, cheaper: a study of NPD project efficiency and performance tradeoffs
Journal of Operations Management
(2006) - et al.
Little patents and big secrets: managing intellectual property
RAND Journal of Economics
(2004)
Determinants of international patenting for nine agricultural biotechnology firms
Journal of Industrial Economics
Absorptive capacity: a new perspective on learning and innovation
Administrative Science Quarterly
A new industrial ecology
Science
How SMEs exploit their intellectual property assets: evidence from survey data
Small Business Economics
On the price elasticity of demand for patents
Oxford Bulletin of Economics and Statistics
Recombinant uncertainty in technological search
Management Science
The Determinants of the Private Value of Patented Inventions
Idea generation and the quality of the best idea
Management Science
Pseudo maximum likelihood methods: applications to Poisson models
Econometrica
Issues in assessing the contribution of research and development to productivity growth
Bell Journal of Economics
Patent statistics as economic indicators: a survey
Journal of Economic Literature
Cited by (22)
Financial market integration and the effects of financing constraints on innovation
2024, Research PolicyVolatile organic compounds pollution control technologies: Past, current and future analysis based on patent text mining and technology life cycle analysis
2022, Journal of Cleaner ProductionCitation Excerpt :We compared the different application characteristics in different regions and also compared the transfer characteristics of different technologies between countries. Countries with a relatively large number of published patents are usually the main technology holders in this field (De-Rassenfosse, 2013). For VOCTs, China, the United States, Japan, Korea, and major European countries were the major patent holders, with 48%, 19%, 13%, 12%, and 4%, respectively (Fig. 6), and these dynamic markets attracted more companies and research institutes to propose better devices or methods for VOCs treatment.
Technology status and trends of industrial wastewater treatment: A patent analysis
2022, ChemosphereCitation Excerpt :With China's efforts on optimizing patent mechanism, we believe that the quality of patents will gradually improve in the foreseeable future. A time-series analysis of patents by countries shows the development pathway of patents for each country over the same period of time (Rassenfosse, 2013). As shown in Fig. 4, the growth trend of the six countries was relatively different.
Competition and invention quality: Evidence from Swiss firms
2020, Technological Forecasting and Social ChangeCitation Excerpt :Addressing this innovation vs. patenting selection issue has been a perennial problem in the innovation literature. Either being ignored entirely or all results bearing the caveat “amongst patenting firms”, dealt with using instrumental variables (de Rassenfosse, 2013), or using a sophisticated selection model with detailed micro-data, exemplified by Arora et al. (2008). Our approach is a more generalizable to datasets without a valid instrument or a way to estimate an endogenous system.