Application of the Binary Logistic Regression Model in a Social Security Scheme in Albania

Abstract

Social security system continues to become more complex and difficult to understand. Countries vary considerably in the extent to which their social security apparatus is centralized and unified. The social security scheme in Albania is based on the system "pay as you go" according to which current payments to pensioners are funded by current contributions within the system. This creates a direct link between the number of persons contributing to the scheme, the amount of money they pay, and the benefits they receive. As a statistical modeling technique that can be applied to evaluate the simultaneous effect of a prediction group on the result of a specific variable that one or one of the two possible values, we use a binary logistic regression model in a social security scheme. We apply this method to predict the probability of non-inclusion in the social security scheme of individuals employed in Albania. From the database, we use a dependent variable and five independent variables. The dependent variable is categorical which takes two values ​​and indicates whether an individual is part of a social security scheme or not, independent variables are: years of school, type of employment, age, sex, and zone. As a result of the application, we conclude that adding independent variables to the model increases the total percentage of the correct classification.