Tracing technological shifts: time-series analysis of correlations between patent classes

This is the authors' abstract. We don't add key points for this paper.

Patents are used as a reliable indicator for the study of technological evolution in specific fields. Patent citation networks can further enlighten the relation between individual classes of patents that are used to categorize innovation. The tightening or loosening of bonds between a pair of them can point to a changing landscape in either of the two, or in both. It does, however, clearly signal one or more changes. Thus, it is important to point out pairs of classes that undergo processes of this kind, and try to provide plausible explanations for them. We use patent citation data from the European Patent Office to create the time series of all IPC classes. We then examine all pairs of patent classes for correlations, and discuss those which show the greatest increase, or decrease, over time. We identify classes which show both a significant decrease in their correlation with one class and simultaneously an increase with another. We further proceed to check the cross correlations of all pairs in order to identify pairs which show a time lag in following one another. By implementing specific criteria for the selection of the most promising pairs we distinguish some cases which exhibit strong correlation values with time lags of several months (3–10), and for which we can provide a plausible explanation.