Modeling of Per Capital Household Expenditure in Tanzania using Bayesian Two Levels Hierarchical Log-Logistic Approach: The Case Study of Dodoma Region
In Tanzania, the fight against poverty is a long-standing agenda, various efforts and initiatives designed to eradicate poverty and increase economic growth among the citizen. However, there is evidence that the real growth over the past decade have not been reflecting in a rapid reduction in poverty rates. In this regard, the government needs analysis of household welfare or poverty. The objective of this study is to get the best model and determine the factors that explain household expenditure. Household expenditure has a hierarchical structure therefore modeled using the two-level hierarchical linear model with the characteristic of households in the first level and district characteristics at the second level. The modeling is set based on Log-logistic with three parameters (LL3) and the estimation process then accomplished by using Bayesian approach with Markov chain Monte Carlo (MCMC), and Gibbs sampling algorithms. We found that three predictors in micro model among all are statistical insignificant. These factors are age of household head, level of education and gender of head. Furthermore, in macro model all estimated parameter of the district predictors was significant at 95% credible interval. It means that the four districts predictor effected on per capita household expenditure.
Keywords: Bayesian Hierarchical Linear Model, Log-logistic Approach, MCMC, and Per Capital HE