<p>Household income plays a vital role in socioeconomic well-being and inequality. Due to its critical role, this indicator has been extensively analyzed in policy studies. However, modeling household income while accounting for hierarchical structure remains a challenge. In this paper, we propose the Hierarchical Additive Mixed-Effects Model, namely the 3Trees-CART model, as a semi-parametric framework for examining variable influences and interaction effects of household incomes. This model integrates linear mixed-effects components with three regression trees to capture non-linear patterns and household-level, district-level, and cross-level interactions. We employed eight empirical datasets to identify the best depth of trees for each tree component in our proposed model. Among the various depths of trees evaluated, we found that the depth of three demonstrated the best performance based on predictive evaluation criteria. Our proposed model also outperformed the Linear Mixed Model and the REEMTree model. We set the tree depth to three in the 3Trees-CART model for analyzing household income. The results revealed that the explanatory variables on the household level, such as household members, age, education level, employment status, job search duration, working hours, health insurance, and pre-employment card, had significant effects. On the district level, expected years of schooling, human development index, number of micro and small enterprises, and regional expenditures had significant impacts. Interaction effects were also successfully identified in each regression tree. Specifically, the first tree captured interactions among job search duration, education level, and employment status. The second tree detected interaction effects between expected years of schooling, the number of micro and small enterprises, and regional expenditures. In contrast, no cross-level interactions were identified in the third tree.</p>