I am starting to work on an IO model and I am puzzled in front of the choice of the appropriate utility function. Since the concept of utility is so abstract, and one cannot directly refer to empirical articles on what factors should the decision be based on? Is it exclusively on the basis of what the assumptions on marginal effects of the variables taken into account are? Isn't this prone to subjective reasoning that may bring researchers to forge their result based on a particular specification? An example of this may be the results one obtain in Hotelling models depending on the transportation cost functions, how do you choose? Thanks, and sorry if the question seems to vague.