Maximum Simulated Likelihood Estimation: Techniques and Applications in Economics

Abstract This chapter discusses maximum simulated likelihood estimation when construction of the likelihood function is carried out by recently proposed Markov chain Monte Carlo (MCMC) methods. The techniques are applicable to parameter estimation and Bayesian and frequentist model choice in a large class of multivariate econometric models for binary, ordinal, count, and censored data. We implement the methodology in a study of the joint behavior of four categories of U.S. technology patents using a copula model for multivariate count data. The results reveal inter-esting complementarities among several patent categories and support the case for joint modeling and estimation. Additionally, we find that the simulated likelihood algorithm performs well. Even with few MCMC draws, the precision of the likeli-hood estimate is sufficient for producing reliable parameter estimates and carrying out hypothesis tests.