Determinants of locational patenting behavior of Canadian firms

ABSTRACT

Using a unique data set combining Canadian and U.S patent data with firm-level data, we analyze Canadian firms' locational patenting decisions during the period 2000–2008. We find that the likelihood of Canadian firms' patenting in both the U.S. and Canada is associated with past patenting experience, firm size, profitability and patent scope. While manufacturing firms in export intensive industries are more likely to patent in both countries, firms in Foreign Direct Investment intensive industries are more likely to patent domestically. Finally, Canadian Intellectual Property Office's role as an International Search Authority under the Patent Cooperation Treaty (PCT) is associated with an increase in the use of PCT by Canadian firms.

KEYWORDS:

• Canadian Firms
• Locational Patenting Behavior
• Patent Cooperation Treaty

JEL CODES:

• O12
• O34

Acknowledgments

The statistical analysis was conducted at Statistics Canada, under arrangements that maintain legal confidentiality requirements. Views expressed are those of the authors and do not necessarily reflect those of Statistics Canada. Funding from Industry Canada is gratefully acknowledged. We thank Cristiano Antonelli (the Editor) and two anonymous referees for valuable comments and suggestions We thank Norman Chalk and Nan Zhou for their help with the data. We would like to thank participants of IPSDM 2017, IRPP Symposium 2018, and CEA 2018 as well as participants of Brown Bag seminar at University of Alberta for their comments. Special thanks are given to Ryan Kelly for his help with the data construction and project coordination. All remaining errors are ours.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 For example, in 2018, 55% of patent applications by Japanese applicants were submitted in Japan, 57% of patent applications by U.S. applicants were submitted in the U.S., and 70% of patent applications by Korean applicants were submitted in Korea, respectively.

2 Theoretical models have been proposed to understand why firms patent. For instance, Horstmann, MacDonald, and Slivinski (1985) argue that innovating firms possess private information about the profitability of their inventions and their decisions to patent can influence rivals' behavior. Langinier (2005) suggests that when the leader in a patent race has more information about the improvability of an innovation, he can choose to patent strategically. In the context of sequential innovations, Bessen and Maskin (2009) argue that patents can be used to block potential rivals in their subsequent research.

3 As described later, we only consider patent applications that are ultimately granted because we only have access to USPTO data that have information on granted applications.

4 The NALMF dataset only contains data on Canadian firms. Therefore, we exclude patents applied for or held by individuals and universities. In addition, some patents cannot be matched to any firm in NALMF. Overall, approximately 20 percent of observations in patent data cannot be matched to NALMF.

5 PATSTAT Data are combined with USPTO data internally by Statistics Canada so that the matching was done by Statistics Canada.

6 Data on exports were downloaded from the website of the Government of Canada: Search by industry (NAICS codes) - Trade Data Online. https://www.ic.gc.ca/app/scr/tdst/tdo/crtr.html?productType=NAICSlang=eng. Last accessed January 2, 2018.

7 Data on direct investment were downloaded from the website of Statistics Canada: Table 36-10-0009-01 (formerly CANSIM 376-0052). http://www5.statcan.gc.ca/cansim/a26?lang=engretrLang=engid=3760052tabMode=dataTablep1=1p2=-1srchLan=-1customizeTab. Last accessed January 2, 2018. which is classified using the North American Industry Classification System (NAICS) 2012. The industries in NALMF were also classified using NAICS 2012. As a result, data on industrial business opportunities were merged with NALMF by NAICS industry and year.

8 Note that IPC is a hierarchical system for the classification of patents according to the different areas of technology to which they pertain. As IPC is assigned when a patent application is examined, it cannot reflect the actual influence of the patent in the future. In this respect, citation based measures could better capture the value of patents.

9 We have explored if the effect of firm age is monotonic by including a squared term in the regressions. It turns out that the squared term is not significant. In what follows, we keep firm age in its logarithm form.

10 We have explored if the effect of firm size is monotonic by including a squared term in the regressions. It turns out that the squared term is not significant. In what follows, we keep firm size in its logarithm form.

11 While reporting R&D is not mandatory, staff at Statistics Canada told us that if a Canadian firm invests in R&D, it usually would report it. As a result, if a firm does not report any R&D expenses, it is reliable to assume that this firm has not done any R&D.

12 The logarithm of R&D expenditure is also used to indicate R&D intensity. As the results are consistent, in all regressions, we use the current definition.

13 The ratio of total current liability and total current assets is alternatively used. The results will remain, so we use the current definition in all regressions.

14 CIPO's application to be an ISA was approved by WIPO in 2002 but started its service as an ISA in July 2004.

15 CIPO's role as an ISA may have different effects on firms' patenting decisions. We will not explore this issue in depth in this paper and leave it to another paper.

16 Note that it could be the same firm in different years.

17 We carry out a robustness check in the subsequent section regarding the truncation issue.

18 The decline after 2006 in Figure 2 is clearly related to the truncation problem. To check how much this truncation problem affects our findings, in the robustness check section, we included two regressions. First, we considered only patents that were granted within 5 years from the application date to the granting date. Second, we considered only the period 2001–2007. The results suggest that truncation issues do not affect our main conclusions.

19 As the number of observations in the provinces Newfoundland and Labrador (NL), Prince Edward Island (PE), Yukon (YT) and Nunavut (NU) is very small, we exclude observations in these provinces. In the empirical analysis, we control for two digits NAICS fixed effects. However, as a few industries have a very small number of observations, we combine a few industries according to their similarities. In particular, we combine NAICS 44 and 45 together to represent retail trade, NAICS 48 and 49 to represent transportation and warehousing and NAICS 61 and 62 to represent educational and other social assistance services.

20 We have run regressions by excluding the largest 5% and 10% of firms in our sample and results are consistent with current findings. In other words, our findings are not driven by extremely large firms.

21 To solve the truncation problem, Hall, Jaffe, and Trajtenberg (2000) weight the number of patent applications, where the weights are constructed from the average lag distributions. However, this approach cannot be applied to our paper., because our patenting location variable is discrete, and indicates whether all of a firm's successful patent applications in a year were in Canada only, the U.S. only, or in both countries. Thus, we do not count the number of patents per firm in each country, but rather we create a dummy variable that takes a value 1 whenever a firm has a patent,and zero otherwise. As a result, if the firm has zero application, the approach of Hall, Jaffe, and Trajtenberg (2000) will still give zero application to that firm, which will not change the patenting location variable in our paper.