Updates and Extension of Measuring Technological Innovation Over the Long Run. 2021.

The main data file provides an updated series of KPST indicators for patent importance and an indicator of whether the patent constitutes a breakthrough patent. The extended data is constructed based on updating the methodology of Kelly, B., Papanikolaou, D., Seru, A. and Taddy, M., 2021.. The other datasets provide the linkage file between the patent number and its CPC class, and patent-patent pairwise citations.

Data

Data Version:

The version released on Sept 29, 2023 is the latest data that updates until the end of 2022.

Data Description:

We provide two updated data sets constructed from the paper here:

PatentSimilarityImportanceBreakthrough_forPost2022.csv: Patent level panel data with patent forward (impact) and backward (novelty) similarities, importance, and breakthrough indicators from 1836 to the present
PatentFullCPC_forPost2022.csv: a dataset containing the assigned CPC technology class to all patents issued from 1836 to 2022
PatentPairwiseCitations_forPost2022.csv: A dataset that contains updated citation pairs from the original paper augmented with data from the USPTO until the end of 2022

For the patent-level panel data, the variable definitions are:

Variable name	Definition
patent_num	Patent ID number
issue_year	Year when a patent is issued (yyyy)
filed_year	Year when a patent is filed (yyyy)
fsim01	Forward 1-year aggregate patent similarity
fsim25	Forward 2-5 years aggregate patent similarity
fsim610	Forward 6-10 years aggregate patent similarity
bsim5	Backward 5-year aggregate patent similarity
lqsim05	5-year importance (log ratio of fsim5/bsim5)
lqsim010	10-year importance (log ratio of fsim10/bsim5)
fcitALL	Total Forward citations till 2022
bk_p90_alqsim05	Breakthrough Indicator (based on 5-yr importance)
bk_p90_alqsim010	Breakthrough Indicator (based on 10-yr importance)

For the patent-CPC class match data, the variable definitions are:

Variable name	Definition
patent_num	Patent ID number
CPC	Full CPC class of a patent

For the patent citations data, the variable definitions are:

Variable name	Definition
citing_patent_num	ID number of the citer patents
cited_patent_num	ID number of the patents being cited

Notes

Our patent indicators are affected by truncation lags (patents only appear in the dataset when they are issued). As a result, we compute patent importance, and breakthrough indicators until 2011 (for the 10-year forward measure) and until 2016 (for the 5-year forward measure).
The patent's importance measure and breakthrough indicators may not precisely match the original version of the data due to the following reasons:
- Changes in the raw patent text data: The extended data is constructed using a new scrape of Google Patents database. The text of patents in Google Patent does change for some patents, likely due to improvements in OCR technology.
- Errors in Filing Years: 294 patents, primarily from year 1882-1884, had the wrong filing year in the original data. These have been corrected.
- Improvements in Text Cleaning: We have improved the text cleaning process by excluding a few more frequently occurring but irrelevant phrases, such as "United States Patent Office," "sheets-sheet," etc., in the raw text files. Those phrases predominantly appear in the patents from earlier years (before 1926).
The patent pairwise citation data is constructed from KPST pairwise citations by augmenting it with newer pairwise citation records from USPTO. Using this data, researchers can create analogous measures as our baseline indicators, by restricting the horizon over which forward citations are constructed.

Contact:

Please contact Dimitris Papanikolaou (d-papanikolaou@kellogg.northwestern.edu) or Amit Seru (aseru@stanford.edu) for any questions regarding the data.