A choice experiment (CE) has been popularly applied to various fields, including environmental economics and agricultural economics. Concerning farm animal welfare (FAW), there have been several valuation studies, yet, the CE has rarely been employed to estimate the willingness to pay (WTP) for individual attributes of FAW. This study attempts to apply the CE in order to estimate WTP for FAW, which uses a mixed logit model, allowing for correlations among random coefficients of attributes of FAW. A panel logit and rank ordered logit models were employed to compare the estimates with those based on the mixed logit model. The estimation results show that mixed logit with allowing for correlations among random parameters is best fitted based on the AIC test and direct use value (banning usage of hormone) was evaluated more highly than indirect use value (providing balanced feed or humane breeding space) associated with the attributes of FAW. Furthermore, positive correlations exist among the random coefficients of the attributes.