Does licensing resolve hold up in the patent thicket?

In a patent thicket licensing provides a mechanism to either avoid or resolve hold up. We study the choice between ex ante licensing to avoid hold up and ex post licensing to resolve it. Firms' choice of licensing contract is studied in the context of a patent portfolio race. We show that high expected blocking leads to ex ante licensing while ex post licensing arises if expected blocking is low but realized blocking is high. Also, ex ante licensing reduces firms' R&D incentives. A sample selection model of licensing is derived from the theoretical model. In this framework theoretical predictions on effects of blocking are tested with data from the semiconductor industry. We show that licensing helps firms to resolve blocking. However, licensing is not a cure all: it decreases as fragmentation of property rights increases and arises mainly between large firms with similar market shares. Using a treatment effects model we also confirm the prediction that ex ante licensing reduces the level of R&D investment.