In decision making processes, people often anticipate future. Future anticipation can take place in medium-term choice scenarios such as vehicle holding and transaction decisions and longterm ones such as residential and job location choices. Most of the discrete choice models developed in the area of travel behavior are incapable of accounting for expectation of future utility. It remains an open question in terms of how the existing modeling frameworks, such as the popular mixed logit models, can be modified to account for the expectation of future utility. This paper explores this question by constructing two theoretical models: a binary model and multinomial (3-alternative) model. For each model, a series of monte carlo simulations are conducted to explore the properties of the parameter estimates. In particular, two main questions are answered. First, what consequences will result if a without-future-utility model is applied to a dataset generated with the future utility component? Second, how well the existing modeling frameworks can be modified to account for the expectation of future utility in different choice scenarios? The simulation results illustrate three points mainly. First, if a without-future model is applied to a dataset generated with a future utility component, the estimated parameters attenuate to zero. In more complicated choice scenarios involving more than 2 alternatives and correlation between alternatives, the estimated parameters have wrong signs. Second, if a with-future model can be correctly specified on a dataset generated with the future utility component, both the probit model and the mixed logit model can recover true parameters quite well. Third, the use of model forms (probit vs. mixed logit) should be cautious.