This paper uses a multilevel logit model and a multilevel mixed linear model to simultaneously analyze the micro-level (household) and macro-level (governorate) factors that might affect the nature and social patterning of poverty in Tunisia. We find convincing evidence that the likelihood of a household being poor is positively and significantly related to household size, number of children per family and education level of household head. Macro-level analysis indicates that a greater neighborhood unemployment rate is associated with higher odds of poverty, while greater industrial agglomeration and migration balance are associated with reduced odds of poverty.