Patent Statistics Manual

What a patent is

The manual starts from the legal definition in Article 28 of the WTO's TRIPS Agreement. A patent gives its owner the right to stop others from making, using, offering for sale, selling or importing the protected product, or from using the protected process (section 2.2). To be patentable, an invention must be new, involve an inventive step and be capable of industrial application. What exactly is protected is set out in the patent's “claims”.

Three properties matter for statistics:

• Patents are temporary. Protection usually lasts at most 20 years from the filing date; then the invention passes into the public domain.
• Patents are territorial. A patent applies only in the country, or countries, that granted it. A US patent gives no exclusive rights in Japan.
• Not everything can be patented. Laws of nature, natural phenomena and abstract ideas are generally excluded. As the manual notes, “there is of course debate about the boundaries of the system”, for instance whether software is an “abstract idea” or a patentable invention.

The disclosure that the patent system requires is what makes the statistics possible. A patent document typically lists the inventors and applicants with their addresses, a technical description, the claims, classification codes, a series of dates and references to earlier patents and scientific literature (section 2.3).

How a patent is obtained

The first application anywhere in the world for a given invention is the priority application, and its date is the priority date. From there an applicant can extend protection in three ways (sections 2.3 and 3):

• National route: the first filing, usually at the national office of the applicant's own country, which the office then searches and examines.
• International route: under the Paris Convention of 1883, the applicant has 12 months from the priority date to file in other Convention countries and keep the original date. Alternatively, a single filing under the Patent Cooperation Treaty (PCT), in force since 1978 and run by the World Intellectual Property Organization (WIPO), lets the applicant delay national procedures until 30 months after the priority date.
• Regional route: an application to a regional office such as the European Patent Office (EPO), whose European patents then have to be validated in each country where protection is wanted.

Applications are generally published 18 months after filing. Examination takes between two and eight years, depending on the office.

What patent data are good for

The manual lists the main advantages (section 2.6):

• They cover a wide range of technologies, including new ones for which there are few other data, such as nanotechnology.
• They are closely, though imperfectly, linked to invention.
• Each document is detailed, and the data are public, unlike survey data protected by statistical confidentiality.
• Coverage is almost global and, in most OECD countries, goes back to the 19th century.
• They are cheap for statisticians and impose no extra reporting burden on firms, because patent offices collect the data anyway.

Patents also sit between R&D and innovation. The manual describes them as both an output of R&D and an input to innovation, which makes patent data a bridge between R&D surveys and innovation surveys.

Where patent counts mislead

The same section is just as clear about the drawbacks:

• Not all inventions are patented. Some are not worth the cost, some do not meet the legal criteria, and some firms prefer secrecy.
• The propensity to patent differs. In some fields, such as semiconductors, firms surround one invention with many applications on small variations. Small firms find patenting costlier than large ones.
• Value is highly skewed. Many patents are worth little or nothing, a few are worth a great deal, so simple counts can mislead.
• Laws and practices differ between countries and change over time. Protection was strengthened from the early 1980s, and some countries extended it to software and genetic sequences.
• A patent is not an innovation. A patented invention need not ever be used. The manual cites the PATVAL survey (2005): about 40% of patents in its sample were not used industrially or commercially, including 18.7% kept to block competitors and 17.4% “sleeping patents” not used at all. Whether something was actually brought to market is the subject of the Oslo Manual.

Most of these problems can be reduced by method: weighting by citations, restricting the analysis to high-value patents, comparing within one industry, or combining patent data with R&D and innovation surveys.

Which date to count

A patent carries several dates: priority, application, publication and grant (section 4.2). Counting by grant date means counting only granted patents, reporting the past (grants take on average three years at the USPTO and five at the EPO) and mixing inventions from many years. Counting by application date mixes domestic applicants, who file at home immediately, with foreign ones, who usually file a year later.

“[I]n order to reflect inventive performance, it is recommended to use the priority date to compile patent statistics.”

OECD Patent Statistics Manual (2009), section 4.2

The choice is not cosmetic. For EPO patents of OECD countries in 2000, counts by priority date and by application date differed by 9% on average for applications and by 28% for grants. For granted patents, Germany led on priority date, while the United States led on grant date.

Which country to count

A patent can be attributed to the inventor's country, the applicant's (owner's) country or the country of first filing. Each answers a different question (section 4.3):

Inventor's country

Where the research was done: the inventiveness of a country's laboratories and workforce. Recommended for measuring inventive activity.

Applicant's country

Who owns or controls the invention. Useful for analysing the strategies of companies, especially multinationals. Distorted when patents are filed by affiliates set up for that purpose, sometimes for tax reasons.

Priority office

Where inventors choose to file first: a measure of how attractive a country's patent system is.

The differences are large. In EPO applications, small countries such as Belgium, the Czech Republic, Hungary and Mexico have higher shares as inventor countries than as applicant countries; the Netherlands, Switzerland and Finland the reverse, and Luxembourg's share as applicant is double its share as inventor.

When inventors come from several countries, a patent can be split between them (fractional counting) or counted in full for each (whole counting). Fractional counts avoid double counting in regional or world totals; whole counts are sometimes preferable, for example for measuring how international a country's technological activity is (section 4.3).

Comparing countries: PCT and patent families

As a general rule, the manual advises against mixing indicators from different patent offices (section 4.1). Each office has its own procedures, and applicants show a home bias: in proportion to their inventive activity, they file more at home than foreigners do. Two kinds of indicator largely avoid this:

• PCT applications, which are international by design. Since 2004 every PCT filing designates all PCT member states, so it can be seen as a “worldwide” application. They are timely, as they are published 18 months after priority.
• Patent families, the set of applications in different countries linked by common priority filings.

“[A] triadic patent family is a set of patent applications filed at the EPO and the JPO, and granted by the USPTO, sharing one or more priority applications.”

OECD Patent Statistics Manual (2009), section 4.5

Triadic families remove home advantage, because every family is filed in the same three offices, and select higher-value inventions, because few applicants pay to protect a minor invention in Europe, Japan and the United States. The manual recommends counting them by earliest priority date, inventor's country and fractional counts. Their weakness is timeliness: complete counts are available only about three years after the year of interest, which the OECD bridges with estimates (“nowcasting”). For small inventions aimed at a local market, a broader family definition is more suitable.

Raw counts also need scaling. The manual shows triadic families per unit of GDP and per million inhabitants, and notes that country rankings change significantly once indicators are normalised (section 4.6).

Technologies, sectors and regions

Patents are classified by technical field, mainly under the International Patent Classification (IPC), established by the Strasbourg Agreement of 1971 and used in more than 100 countries (section 5.2). A patent can carry several IPC codes. Because the IPC is revised and, before 2006, revisions were not applied retroactively, long time series need care. Chapter 5 also covers converting technology classes into industries, attributing patents to regions from inventors' addresses, identifying the sector of the owner (company, university, government body, individual) and matching patents to companies and inventors.

For emerging technologies, patents are often the only reliable source of data; the manual's example is the growth of fuel cell patenting.

Citations

Patent documents cite earlier patents and non-patent literature, such as scientific articles, standards and technical manuals. These references define what the new patent cannot claim (chapter 6). Statisticians use them in three ways:

• Backward citations, to earlier work, trace flows of knowledge between firms, regions and fields.
• Forward citations, received later, are a common indicator of an invention's importance.
• Citations of scientific literature show links between science and technology.

The signal is noisy. Many citations are added by patent examiners or attorneys, not by inventors, so the inventor may never have known the cited work. Citation practices also differ between offices; at the USPTO, for example, applicants have a “duty of candour” to disclose prior art they know of.

Patent value

“Value” can mean the private value to the owner or the social value of the knowledge the patent adds; and the value of the patent is not the same as the value of the invention (section 8.1). According to the manual, all studies of patent value find a quite skewed distribution, and surveys of patent holders show a peak at zero. Indicators can therefore either weight patents by a measure of value or count only patents above a value threshold. The measures discussed include whether the patent was granted, forward citations, family size (the number of jurisdictions where protection is sought), renewals, opposition and litigation (chapter 8, Table 8.1), as well as the number of technical classes (section 8.4.2).

Internationalisation of research

When inventor and applicant live in different countries, the patent records cross-border ownership. From this the manual builds two indicators: foreign ownership of domestic inventions and domestic ownership of inventions made abroad (section 7.2). Both mostly reflect the research laboratories of multinational companies, and they complement data on R&D by foreign affiliates. Patents listing inventors from several countries measure international co-invention.

Polish figures and how to read them

The Patent Office of the Republic of Poland (UPRP) reports these figures for 2025 in its annual report:

• 3,453 applications for inventions; by first applicant, 3,345 came from Polish applicants and 108 from foreign ones;
• 2,283 patents granted for inventions;
• 10,580 European patents validated in Poland.

This is the single-office problem the manual warns about (section 4.1). Counts from one national office mix two effects: domestic applicants file more at home in proportion to their inventive activity (home bias), and foreign firms that want protection in Poland largely choose the European route instead of the Polish office. Note also that the two figures measure different stages: applications at the UPRP, and European patents already granted and then validated in Poland. Neither number on its own says much about Poland's place in world technology. For international comparisons the manual recommends PCT applications or patent families counted by inventor's country and priority date.

Editions and status

The OECD's interest in patent statistics dates from the late 1970s. The first manual, Using Patent Data as Science and Technology Indicators – Patent Manual, appeared in 1994; the current text was published in 2009 and prepared with the OECD's working party of national experts on science and technology indicators (NESTI). It is available in English and French; there is no Polish edition, so the Polish version of this page uses our own translations.

Some details reflect the time of writing. The OECD then had 30 members and the EPO 32 member states (section 2.3), and its examples use data from the early 2000s. The counting rules themselves do not depend on those figures.