Patent Quality

Measuring the quality of patents is among the most consequential — and most contested — challenges in the economics of innovation. Unlike patent counts, which capture the quantity of inventive activity, quality indicators attempt to measure how significant, broad, and influential individual inventions are. No single metric suffices: claim structures reveal the legal scope of protection, classification breadth captures technological interdisciplinarity, forward citations measure downstream impact, and knowledge flow indicators such as originality and generality illuminate the diversity of inputs and outputs in the inventive process.

This chapter synthesizes eight complementary dimensions of patent quality, drawing together evidence from claim analysis, scope measurement, citation dynamics, and knowledge flow indicators. The patterns that emerge are not always consistent — quality trends in one dimension do not necessarily parallel trends in another — but together they provide a multifaceted portrait of how the nature and significance of patented inventions has evolved from the late 1970s through the present.

Beyond the growth in patent volume and the shifting composition across types, the internal structure of patents themselves has evolved. Patent complexity, as measured by claim counts, provides a window into how applicants have sought to define and protect their inventions over time.

Each claim defines a specific element of the invention that receives legal protection. The initial upward trend in average claims per patent reflected both the growing technical sophistication of inventions and the strategic incentive for applicants to secure broad coverage, though the gap between average and median claims initially widened but has recently reversed, with the median surpassing the average by the mid-2010s.

Examining the distribution more closely — particularly at the extremes — reveals how the upper tail of claim counts has shifted relative to the median over the same period.

Patent scope has broadened as technologies become more interdisciplinary. The number of distinct CPCCooperative Patent Classification — a hierarchical system jointly managed by the USPTO and EPO that categorizes patents by technology area (e.g., H = Electricity, G = Physics). subclasses assigned to each patent has increased steadily, indicating the convergence of once-separate technology domains. Contemporary inventions in areas such as IoT, biotechnology, and AI inherently span multiple classification categories.

Forward citations — the frequency with which a patent is cited by subsequent inventions — constitute the most widely employed indicator of patent quality in the economics of innovation literature. A patent that receives numerous forward citations has produced knowledge that subsequent inventors deemed sufficiently valuable to build upon.

Raw citation counts are misleading for cross-field and cross-time comparisons because citation norms differ substantially by technology field and grant year. Cohort normalization divides each patent's 5-year forward citations by the average for its grant-year × CPC section cohort, yielding a field-adjusted measure of relative impact.

The average number of backward citations per patent has grown substantially over the decades, reflecting the expanding body of prior art that new inventions must acknowledge. Patents increasingly build on larger bodies of prior art, with the average rising from 4.9 in 1976 to 21.3 by 2023. The gap between average and median suggests a long tail of patents citing an exceptionally large number of prior works, a pattern consistent with both the expanding universe of prior art and more thorough examination and disclosure requirements.

Self-citation rates — the fraction of backward citations directed to patents held by the same assignee — provide a window into how firms balance internal knowledge accumulation with external knowledge absorption. Organizations that consistently cite their own prior patents are building on internal knowledge stocks, a characteristic of cumulative innovation within technological trajectories.

The citation lag — the temporal distance over which patents cite prior art — has increased steadily, indicating that the useful life of patented knowledge continues to extend. Contemporary patents draw on an increasingly expansive base of prior art, reaching further back in time than patents of earlier decades.

The originality index (Trajtenberg, Henderson, and Jaffe, 1997) measures the diversity of a patent's technology sources — a patent that cites prior art across many different CPC sections is more "original" than one citing only its own field. The generality index measures the converse: the diversity of citing patents, capturing whether an invention has broad applicability across fields.

Certain patents receive minimal attention for years, only to be "rediscovered" when technology develops sufficiently. These "sleeping beauties" received fewer than 2 citations per year during their first 10 years, then experienced a citation burst of 10 or more citations within a 3-year window. Such patents represent inventions that preceded the technology necessary for their practical application.

Non-patent literature (NPL) citations — references to journal articles, conference proceedings, technical reports, and other scientific publications — provide a window into how closely patent activity is linked to the underlying science base. Rising NPL citation counts indicate a tightening relationship between scientific discovery and commercial invention.

The share of backward citations directed to foreign (non-US) patents reveals the growing internationalization of the knowledge base upon which US patents are built. A rising foreign citation share indicates that US inventors are increasingly building on innovations from abroad.

The number of figures per patent provides an indicator of visual and technical complexity. More figures typically correspond to more elaborate inventions that require extensive illustration to describe, including circuit diagrams, flowcharts, molecular structures, and mechanical drawings.

Having examined patent quality across multiple complementary dimensions — from claim complexity and scope breadth through citation dynamics, self-citation patterns, originality, generality, sleeping beauty patents, and non-patent literature citations — the next chapter turns to the technological composition of patent activity. Where this chapter asked how good patents are, the next asks what fields they cover and how the distribution of inventive effort across technology domains has shifted over five decades.