Assigning Patents to Industries: Tests of the Yale Technology Concordance

The Yale Technology Concordance is tested on several subsets of Canadian patents by comparing out-of-sample predictions with industry assignments made by the Canadian Patent Office, and it is found that the predictions of patents by industry are quite accurate for the subset of patents form US inventors.

Abstract

We describe a method to predict patent counts disaggregated by industry, using available data on patenting by technology field. This method—the Yale Technology Concordance (YTC)—exploits a data set of patents that have been individually assigned by the Canadian Patent Office to both an industry and a technology field. The procedure for predicting patents by industry is developed as a statistical model so that the standard errors of the predictions can be estimated. The YTC is tested on several subsets of Canadian patents by comparing out-of-sample predictions with industry assignments made by the Canadian Patent Office. We find that the predictions of patents by industry are quite accurate for the subset of patents form US inventors. The prediction errors are much greater for the subset of patents granted or published after 1989. This suggests that the relationship between the technology fields and industries has shifted in a way that the procedure does not capture. Nonetheless, predictions from the YTC do appear to give a reasonably accurate picture of the pattern of patenting by industry.