ZEW Discussion Papers
Selection Bias in Innovation Studies: A Simple Test
de Rassenfosse, Gaétan and Annelies Wastyn (2012), Selection Bias in Innovation Studies: A Simple Test, ZEW Discussion Paper No. 12-012, Mannheim. Download
The study of the innovative output of firms often relies on a count of patents filed at one single patent office, although companies have the option to file patents at multiple offices such as their national patent office, the European Patent office (EPO), the World Intellectual Property Office, the US Patent and Trademark Office, or in any other jurisdiction. In Europe, the office of reference used is often the EPO. Yet, not all firms file their patents at the EPO, raising the specter of a selection bias.
Motivated by the tension between the popularity of the single office count and the threat of a selection bias, the objective of this paper is twofold. First, we study whether the single-office count biases econometric estimates of innovation production functions. Innovation production functions, which relate a firm’s inventive input to its output, are a key object of analysis in the innovation literature. Second, we propose a simple way to test the existence of bias when the econometrician observes patents at only one patent office.
The analysis is performed on a panel of Belgian patenting firms. Three databases are combined: The panel of R&D surveys providing annual data on R&D-related variables for the period 2002–2008 is merged with the Belfirst database which provides yearly information on balance sheets and income statements. The Patstat database by the EPO (April 2009 edition) was used to collect a novel dataset of the whole population of patents by Belgian firms.
We show that the single-office count results in a selection bias that affects econometric estimates of innovation production functions. In addition, we demonstrate the usefulness of our test to evaluate whether estimates that rely on the single-office count are affected by a selection bias. The test, which uses information that is readily available to most researchers, successfully spots variables that are subject to a selection bias. It should be of interest to a wide audience given its ease of use and the popularity of the single-office count among innovation scholars.
Keywords: innovation production function, patent, R&D, selection bias