How Licensing Resolves Hold-Up: Evidence from a Dynamic Panel Data Model with Unobserved Heterogeneity

We study licensing in a high technology industry affected by a patent thicket. Licens-ing contracts signed with rival innovators allow firms to contain the threat of hold-up that arises in the patent thicket. Licensing contracts will have very different effects on firms’ R&D incentives depending on when they are signed. Therefore we distinguish between ex-ante and ex-post licensing contracts which are signed before or after R&D invest-ments are made. We find that there is considerable variation in firms ’ choices between these forms of licensing. To explain this variation we develop an empirical model that nests transactions costs, technology and product market determinants of the choice of licensing. This model is tested employing a dynamic binary choice model with unob-served heterogeneity, strictly exogenous variables as well as a lagged dependent vari-able. The initial conditions problem is dealt with according to a method suggested by Wooldridge (2005). We estimate a random effects probit model using conditional ML and find evidence of state dependence in our model. The predictions of our empirical model are confirmed. In particular we find that ex-ante and ex-post licensing resolve different forms of hold-up and derive implications for the regulation of licensing in the industry we study