Details

Autor(en) / Beteiligte

• Rock, Elizabeth

Titel

Predicting Financial Well-Being Score in America Through Perceived Financial Well-Being and Financial Vulnerability Using CatBoost and XGBoost

Ort / Verlag

ProQuest Dissertations & Theses

Erscheinungsjahr

2022

Quelle

ProQuest Dissertations & Theses A&I

Beschreibungen/Notizen

• Given the high housing prices in some United States cities, increasing housing and food costs, and limited savings, the question of where to live can profoundly affect financial well-being. Consumer financial education tools could benefit from research into predicting financial well-being using location. This quantitative, non-experimental dissertation expands upon limited research into predicting the financial well-being score using measurable financial vulnerability, perceptions of financial well-being, the U.S. Census Geographic Division, and home-ownership status. This dissertation utilized a multitrait-multimethod approach to analyze the National Financial Well-Being in America Survey using gradient boosting methods with XGBoost and CatBoost. Features were selected using Recursive Feature Elimination and a RandomForestRegressor. The ability to predict financial well-being scores using perceived financial well-being with XGBoost (RMSE = 2.223) and CatBoost (RMSE = 2.225) was supported with an Adjusted R2 score of 0.975. Predicting financial well-being scores using measurable financial vulnerability was also supported, with XGBoost (RMSE = 8.882 and Adjusted R2 = 0.590) and CatBoost (RMSE = 8.872 and Adjusted R2 = 0.591). Predicting the financial well-being score with the U.S. Census Geographic Division and home-ownership status was not supported. The study also found that renters had median financial well-being scores that were lower than homeowners' first quartile financial well-being scores in all U.S. Census Geographic Divisions.