Both in theoretical frameworks and empirical applications, proximity model is most frequently used in its basic version. This version assumes symmetric utility with the same shape for all voters and perfect information. In this paper I evaluate whether it is justifiable to use proximity model with such restrictive assumption, or if some generalized version of proximity model is more appropriate. Three different types of generalized proximity models are tested on Parliamentary Elections Survey Data from the Netherlands. Using multinomial logit estimation, the performances of generalized and basic proximity models are compared. The results show that only some of the generalized versions can outperform the basic version, and even in these cases the differences are relatively small.