Architectural Configuration and Probability Calculation of the Sequential Logit Model over a Household Budget Survey Data in Turkey

Abstract

The aim of this study is to determine the architectural configuration and probability structure of the sequential logit model. In situations where there are more than two alternatives and where the choice between the alternatives are made sequentially, the situation turns into the estimation of sequential models with fewer alternatives and this reduces the number of calculations that need to be done. In multiple-choice models, individuals make a choice between more than two alternatives and the probability of this choice is calculated. In sequential models, dependent variable levels have a multi-staged sequence and response level on each stage contains the answer level on the previous stage. Success of each step depends on the success of the previous step. An important point regarding sequential models is the necessity that choice probability in each stage must be independent from the choice probability in other stages. In this study, the most important factors that affect an individuals indebtedness status were estimated using sequential logit model and interpretations were made by calculating odds ratios and marginal effect values related to this model. The architecture configuration and detailed interpretations of the sequential logit model were discussed over a real data set on indebtedness.