Article ID: | iaor19901149 |
Country: | Germany |
Volume: | 34 |
Start Page Number: | 143 |
End Page Number: | 156 |
Publication Date: | May 1990 |
Journal: | Mathematical Methods of Operations Research (Heidelberg) |
Authors: | Morawitz B., Tutz G. |
In business surveys, the response categories most often are of the type ‘increase’, ‘no increase’ and ‘decrease’. In modelling how the response variable depends on exogeneous variables one has to decide whether the response categories can be treated as ordered or unordered categories. Though the response categories are constructed as ordered categories it is not so clear that the respondent actually uses them as ordered. The middle category may be used as a ‘don’t know’ category if the respondent has insufficient information. Here categorical regression models are considered for the ordered and the unordered case. Three types of models are investigated: unordered models resulting from utility maximization, threshold models and two-step models which treat the middle category explicitly as a separate category. The relationship between models is considered. The models are applied to a sample of the IFO-Konjunkturtest and are compared to each other.