<p>In order to develop a robust framework for analyzing household income and expenditure data, this paper proposes flexible regression models capable of capturing skewness, heterogeneity, and heavy-tailed behavior commonly observed in socio-economic data. Specifically, the study introduces the Type II Half Logistic Exponentiated Exponential (TIIHLEtE) distribution and extends it into a regression structure within the T–X family of distributions. In addition, a regression model based on the Pareto IV distribution, referred to as the Type II Half Logistic Pareto IV (TIHLPIV) model, is formulated as an alternative approach for modeling heavy-tailed household budget data. The main methodological contribution of this paper is the development of these flexible distribution-based regression models for household budget analysis, where conventional models may fail to adequately represent extreme values and complex distributional characteristics. Model parameters are estimated using the Maximum Likelihood Estimation (MLE) method, and model performance is evaluated through goodness-of-fit measures, residual diagnostics, and information criteria, including AIC and BIC. A simulation study is further conducted to assess the consistency and efficiency of the estimators. The proposed models are applied to data from the Tanzania Household Budget Survey (HBS) 2017/2018 to examine patterns of household income and expenditure. The empirical results reveal that the TIIHLEtE and TIHLPIV regression models outperform conventional alternatives in capturing the complex distributional features of household budget data. These findings provide valuable insights into income inequality, consumption behavior, and poverty-related disparities, with important implications for social policy formulation and targeted poverty reduction strategies in Tanzania.</p>