A Statistical Analysis Of American Income Distribution in the 1990s

Income inequality in the United States is characterized by a combination of demographic, educational, and occupational factors. This paper investigates which variables are most influential in predicting whether or not the annual income of a person exceeds $50,000 annually and evaluates the characteristics of individuals who exceed their predicted income classification. Using an income distribution dataset from Kaggle, statistical models such as logistic regression, random forest, decision tree, and XGBoost were implemented and the results of each model was compared. The analysis of the results examined model performance, variable coefficients, and false negative cases. Results indicate that logistic regression with balanced class weights provide the most accurate analysis of the data. The model found that the factors that generally influenced income the most were factors that people could control such as hours per week they worked or the occupation that they work in. The analysis also found that false negative cases where the person was predicted to make less than $50,000 but exceeded this threshold were generally attributed to capital gain and loss through investments and assets. Overall, the findings demonstrate that income predictions are mostly influenced by variables that the subject can control rather than assigned variables that one has no control over such as race and gender.