This paper analyzes selection biases in the project choice of complementary technologies that are used in combination to produce a final product. In the presence of complementary technologies, patents allow innovating firms to hold up rivals who succeed in developing other system components. This hold-up makes innovation rewards independent of project difficulties and firms excessively cluster their R&D efforts on a relatively easier technology in order to preemptively claim stakes on component property rights. This selection bias is persistent and robust to several model extensions. Implications for the optimal design of intellectual property rights are discussed.