Working Paper: NBER ID: w15851
Authors: Grid Thoma; Salvatore Torrisi; Alfonso Gambardella; Dominique Guellec; Bronwyn H. Hall; Dietmar Harhoff
Abstract: This paper discusses methods for the harmonization and combination of large-scale patent and trademark datasets with each other and other sources of data. Dictionary- and rule-based approaches to the consolidation of applicant names in patent data are presented and shown to have both benefits and drawbacks in isolation. We combine the two methods and develop a set of rules and dictionaries to consolidate European, Patent Cooperation Treaty (PCT) and US patent data with firm accounting data. The resulting data encompass about 131,000 patent applicant names from 46 countries, covering 58.8 percent of EPO applications and 50.6 percent of PCT applications by business organizations during the time period from 1979 to 2008. For US data, the resulting dataset includes around 54,000 assignee names and 51.3 percent of US granted patents during approximately the same time period.
Keywords: Data harmonization; Patent data; Accounting data; Innovation research
JEL Codes: C81; O34
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| traditional manual matching approaches (C78) | significant errors and biases in research outcomes (C90) |
| employing a combination of dictionary-based and rule-based methods (C45) | improve the matching process (C78) |
| their approach (B50) | significantly increase the quality of harmonization (L15) |
| their proposed methods (C90) | improved data accuracy (Y10) |