Patent portfolio analysis of cities: statistics and maps of technological inventiveness

ABSTRACT

Cities can be considered as engines of the knowledge-based economy, because they are the primary sites of knowledge production activities that subsequently shape the rate and direction of technological change and economic growth. Patents provide rich information to analyse the knowledge specialization of specific places, such as technological details and information on inventors and entities involved. The technology codes attributed at the level of individual patent documents can be used to indicate the diversity and scope of the knowledge claims underlying a specific invention. In this study we introduce tools for portfolio analysis in terms of patents that provide insights into the technological specialization of cities. The mapping and analysis of patent portfolios of cities exploits data derived from the Unites States Patent and Trademark Office (USPTO) and dedicated tools (at https://leydesdorff.net/software/patents/). The results allow policy makers and other stakeholders to identify promising areas of further knowledge development, including smart specialization strategies.

KEYWORDS:

• Knowledge production
• technological change
• knowledge space
• patent analysis
• invention
• agglomeration
• specialization
• diversity
• cities

JEL CODES:

• O11
• O31
• O34

Acknowledgments

The authors would like to thank two anonymous referees for comments on an earlier draft. The usual disclaimer applies.

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Dieter Franz Kogler http://orcid.org/0000-0002-6744-5632

Gaston Heimeriks http://orcid.org/0000-0002-0577-6938

Loet Leydesdorff http://orcid.org/0000-0002-7835-3098

Notes

1 CPC contains new categories classified under ‘Y’ that span different sections of the IPC in order to indicate new technological developments (Scheu et al., 2006; Veefkind, Hurtado-Albir, Angelucci, Karachalios, & Thumm, 2012).

2 The cosine is similar to the Pearson correlation except that the distributions are not z-normalized to the mean. Since the patent distributions are non-normal (but skewed), this measure is more appropriate (Ahlgren et al., 2003).

3 Because of the diacritical characters searching with these names is difficult in the USPTO search interface; we found one patent with ‘Essone’ in the address field, three with ‘Val-de-Marne’, and seven with ‘Yvelines’, granted in 2014.

4 UCInet enables the user to generate these network statistics in a single pass.

5 We use the transposed matrix because factor scores are more difficult to read, while factor scores do not vary between −1 and +1.

6 (p_i * p_j) is the Gini-Simpson index. The Gini-Simpson is equal to the complement to one of the Herfindahl–Hirsch index or equivalently the Simpson index (Stirling, 2007).

7 We use Zhang et al.’s (2016) diversity measure (² D^S) for estimating this correlation since measures ‘true diversity’ with which one is allowed to calculate as a variable at the ratio scale.

Additional information

Funding

Dieter F. Kogler would like to acknowledge funding from the European Research Council under the European Union’s Horizon 2020 research and innovation programme (grant agreement No 715631, TechEvo).