Within Political Science, there is significant concern about the type of discrete choice model best suited to represent the vote. Looking for an answer, most scholars have opted to evaluate the performance of a wide variety of probit and logit models, including more flexible logit versions such as the mixed logit model, in specific elections in relatively stable pool of party’s contexts. Despite the limited evidence resulting from the studies conducted, there is growing enthusiasm in favor of probit models since this type of model relaxes the independence of irrelevant alternatives property (IIA) imposed by logit alternatives. The IIA is especially detrimental in the case of studies interested in changes in the probability of voting when political parties enter, leave, coalition or split in a partisan system, and it is precisely in these contexts where, as Dow and Endersby (2004, p.112) point out, empirical researchers have paid less attention in the arena of democratic elections. Thus, this paper aims to empirically evaluate the performance of logit and probit models in such contexts. Specifically, it takes advantage of the changes produced in the Spanish political party system since the emergence of Podemos, Ciudadanos and Vox.