Hide/Show Apps

Micro-level analysis of unregistered employment in Turkey with group comparisons

İner, Mehmet
Group comparison of logistic regression models in a similar way with OLS is manipulating depending on the unobserved heterogeneity in logistic regression. In this sense, this study focuses on the group comparison problem in logistic regression. In order to get to the root of the comparison problem in logistic regression, the theoretical background of the logistic regression is explained with the latent propensity interpretation in which the extent of the dependent variable’s closeness to success is taken into consideration. In this respect, the discussions on the diagnosis and the remediation of the problem in the literature are revealed and analyzed. The application of group comparison in logistic regression is made by means of unregistered employment data in Turkey. In this respect, comparisons among genders, regions and years are made in terms of unregistered employment in this thesis. For this aim, Long’s (2009) and Long & Mustillo’s (2018) methods to conduct comparisons among groups by means of predicted probabilities and marginal effects are utilized since the test of difference in predicted probabilities based on the models and the marginal effects are not scaled by unobserved heterogeneity. Various important socio-economical results and implications including gender differences and regional differences are reached through the comparisons. Moreover, the differences in marginal effects in 10 years period are analyzed and the changes over time are associated with the measures taken in the field.