The Value of Failures in Pharmaceutical R&D

We build a cumulative innovation model in which both success and failure provide valuable information for future research. To test this learning mechanism, we use a dataset covering outcomes of world-wide R&D projects in the pharmaceutical industry, and proxy knowledge flows with forward citations received by patents associated with each project. Empirical results confirm theoretical predictions that patents associated with successfully completed projects (i.e., leading to drug launch on the market) receive more citations than those associated to failed (terminated) projects, which in turn are cited more often than patents lacking clinical or preclinical information. We therefore offer evidence of the value of failures as research inputs in (pharmaceutical) innovation