Database Information
Data current through
{date_database}
Saturday - November 18, 2017
McGuire Center for Entrepreneurship
The University of Arizona | Eller College of Management The University of Arizona Eller College of Management
Eller College Home > McGuire Center for Entrepreneurship > CRIE > Patent Data > Patent Rank Documentation
Commercialization Research on Innovation and Entrepreneurship


Patent Rank Documentation

Patent Rank defines the intrinsic value of a patent within the USPTO-patent network based on the graph-theory mathematics known as "eigenvector" centrality.

As of now, the index tables for Patent Rank and forward-citation counts contain annual scores from 1975-2012.

Details about Patent Rank

What is Patent Rank?

Patent Rank is a recursive-weighting methodology to define a patent's importance within the network by considering all links within the network. At a given point in time, a network is formed, all Patent Rank scores are computed simulataneously using this recursive algorithm. Based on Patent Rank1 we can define a patent's value objectively over time and even estimate its lifetime value. This new measure and data can be utilized to help academics better understand innovation and entrepreneurship.

1. Shaffer, Monte J. 2011. Entrepreneurial Innovation: Patent Rank and Marketing Science. Dissertation: Washington State University.

How does it compare to forward-citation counts?

Below you will find details comparing Patent Rank to Trajtenberg's forward-citation counts:

Comparisons

 1 Trajtenberg (1990)2 Shaffer (2011)
Weighting SchemaForward CitationsPatent Rank
Weighting LogicNot all patents are equal, so let's weight them by their subsequently cited patents.If not all patents are equal, why would we equally weight subsequently cited patents?
WeightingNonrecursiveRecursive
Network MathematicsIndegree CentralityEigenvector Centrality
Network EffectsImportance EffectsTotal Effects


Because Patent Rank represents a more robust mathematical model for network centrality, we posit that the forward-citation metric creates systematic bias in determining value which has deleteriously influenced our study of innovation and entrepreneurship. The recursive weighting used in Patent Rank provides a more precise measure of value than forward-citation counts.

1. Trajtenberg, Manual. 1990. A Penny for Your Quotes: Patent Citations and the Value of Innovations. The RAND Journal of Economics 21(1) pp. 172—187.

2. Shaffer, Monte J. 2011. Entrepreneurial Innovation: Patent Rank and Marketing Science. Dissertation: Washington State University.

Patent Rank model specifications

As outlined above, Patent Rank recursively weights all forward and backward citations to determine the patent's intrinsic value within the patent network. Annually, we can compute these scores cumulatively (total diffusion since introduction in the network) or marginally (what has been the influence in a recent network [e.g., the last 5 years]). Additionally, we can compute the scores just using the citations (structure) or we can utilize the citations and their strength (combined1).

 Structure-OnlyCombined
Cumulative(cs)(cc)
Marginal(ms)(mc)

1. Each link is defined by its presence and strength: the strength is defined by a ClassMatch algorithm which is a softmatch of technological-classification overlap between the two patents that define the link More details can be found in Shaffer (2011), see above.

Patent Rank version (0.2)

Patent Rank scores are computed annually and stored within the patent data repository. Currently, all known link structures are included in the computation. Since link structures started in 1975 and became standardized in 1976, data collection to compute Patent Rank begins in 1975.

Patent Rank scores (a mathematical vector of numbers) are normalized so the minimum score is equal to 1. If a patent is not found in the network it will not have a score; if it is in the network, but has no forward citations it will have a score of 1; if it is in the network, and has at least one forward citation it will have a score greater than 1. Generally, the distribution of such scores are highly skewed (following a power-law distribution). Since these links are controlled and governed by a legal prosecution procedure and represent the value of innovation, we note potentially a double-log normal distribution, see Shaffer (2011) for more details.


For additional information, please contact us.

hosted by Mergent

powered by Patent Rank

Patent Data Repository
* Email:
Patent Data Repository
* Password:
Patent Data Repository
Lost Password Reset Password Activate Account
* Email:
This will be your user name (EDU email)
* Create Password:
* Repeat Password:

  Title:
* Name:      
William        H.       Gates       III 
* Preferred:
Bill Gates or "Bill"      

* Service:
By checking this box, you agree to our terms of service.
If you check this box, we will send you a monthly newsletter.
If you check this box, we will send you promotional information about the Patent Data Repository, etc.

- Reference:
Monte  

http://crie.patent-rank.com/ [user] [crie-sandbox]