Long abstract

About Statistics Publications & Documents Information by Country

Indicators based on patents provide a good measure of the innovative performance and technology outputs of countries. However, because of legal rules imposed by the patent application process, information on patents is generally publicly disclosed after 18 months. Patent indicators are consequently faced with a timeliness issue, which can extend to more than five years depending on the computational method used to develop indicators.

This study aims at designing simple but robust methods that would enable to "nowcast" patent indicators -- forecast the present (or the recent past) -- in order to mitigate the timeliness issue. The nowcasting exercise is conducted here on two separate sets of patent indicators: the number of patents applied to the European Patent Office (EPO) and the number of Triadic Patent Families (patents taken at the EPO, the Japan Patent Office (JPO) and the United States Patent and Trademarks Office (USPTO)).

The nowcasting method developed in the present document is based on estimates of the transfer rate of patents filed under Patent Coooperation Treaty (PCT) into the EPO regional phase, given that information on PCT patents at international phase is disclosed before reaching the regional/national phase. This method provides robust estimates up to year t-2 (instead of year t-4), even though patenting activity of small patenting countries or emerging economies are difficult to predict, in terms of both level and growth.

Another method is implemented to nowcast triadic patent families’ indicators, as these are based on three different datasets (EPO patent applications, JPO patent applications and USPTO patent grants). Timeliness of triadic patent families strongly depends on the disclosure of USPTO grants, which sometimes may occur more than 5 years after the patents' first filing date. Therefore, a two-step method is implemented, extending triadic patent families' coverage up to year t-3, possibly year t-2. First, a model based on the relationship between triadic and biadic patent families (patents taken at the EPO and JPO only) enables to gain at least three additional years of data. Then, a second model built on the relationship between EPO patent applications and triadic patent families is used to add two to three years of data. However, as for EPO nowcasts, the method shows weaknesses for predicting small patenting countries/emerging economies.

The new two-step method leads to significant revisions of the time series on triadic patent families that were previously presented in various OECD publications (which were derived from a less robust nowcasting method based on estimates of USPTO grants). Emphasis on differences in nowcasting methods (old and new) should be made while publishing revised figures on triadic patent families.