An Advantage of the Linear Probability Model over Probit or Logit.

Linear probability models, logit models, and probit models have been used to estimate dichotomous choice models in the past, but recently, the linear probability model has fallen into disfavor because it can yield predicted probabilities outside the 0-1 interval. However, there are some parameters of interest that can be estimated in the linear probability model, but not in either logit or probit models. If the model contains a dummy variable for membership in some group, and every member of the group has the same value for the dependent variable, the coefficient of the group dummy variable cannot be estimated in logit or probit models, but can be estimated in the linear probability model. Copyright 1988 by Blackwell Publishing Ltd