A limited information estimator for the multivariate ordinal probit model

A limited information estimator for the multivariate ordinal probit model i s developed. The main advantage of the estimator is that even for high dime nsional models, the estimation procedure requires the evaluation of bivaria te normal integrals only. The proposed estimator also avoids the potential problem of encountering local maxima in the estimation process, which is lo oming using maximum likelihood. The performance of the limited information estimator is shown by Monte Carlo experiments to be excellent and it is com parable to that of the maximum likelihood estimator. Finally, an applicatio n of the limited information multivariate ordinal probit to model the consu mption level of cigarette, alcohol and betel nut is presented.