The problem of using patent applications as trend indicator
In statistics it is known as "the law of large numbers": the larger the reference quantity, the smaller the differences, individual outliers disappear in the mass, deviations disappear in a background noise. A new study examines the relationships between the number of living patents in a patent portfolio and the values. It is obvious that there is a global correlation, but on a closer view it is surprising how heterogeneous this correlation is and to what wrong insights the mere counting of patents can lead to.
Two recently published studies of InTraCoM GmbH impressively prove how deceptive correlation analyses on large data sets are. With aa appropriate segmentation, e.g. by company size, sector or country, correlations between the number of living patents and their values become less and less significant and larger in dispersion.
Deviations from the average are greater the more specific the segment is
In plain language this means: If someone wants to find anomalies, e.g. which industry has the most valuable patents, which countries or which company sizes are the most innovative ones, then the number of patents and their values deviate more and more from each other and from the average. This means that the counting of patents would lead to completely different results than the consideration of the respective patent values. This is impressively shown by the study in comparisons.
Particularly impressive in this context is the examination of patents on "sustainable" technologies. Here, the counting of living patents would have led to the result of a negative relative trend, although the calculation of relative patent values shows an increasing or stagnating trend at a higher value level. This has already been reported on in the last blog post.