Abstract
Why are there so many measures of patent novelty? Novelty is a central concept in innovation studies but the way it is measured is in a state of confusion. We examine patent-based novelty measures and find that they are rarely grounded in theory or empirically verified. Even though all measures are based on the same data, they use different parts of patents, often constrain which patents they include in comparisons, and frequently impose conditions with unknown effects. Not surprisingly, some measures are highly similar and correlated, while others are not at all. However, we presently do not know exactly why or how these measures differ. We need a better model of how patent data captures novelty to produce robust research with reliable results. Besides calling for external validation of measures, we propose a framework that makes explicit what reference data and decision model are used, and why, with the objective to guide the design of more valid novelty measures.