A consistent estimator for the binomial distribution in the presence of “incidental parameters”: an application to patent data