Drawing on two nationally representative surveys, this paper analyzes how and why determinants of household welfare changed in Azerbaijan during the transition in the period from 1995 to 2002. Household welfare is measured by quartiles of total household consumption per capita. Ordered probit regression was used to estimate the probability that a household belongs to one of the quartiles as a vector of individual, household, and community characteristics. The empirical findings demonstrate that some factors significantly affecting welfare distribution in 1995, such as attachment to the formal labor market and renting land, became insignificant in 2002, while some factors that played no significant role in determining household welfare in 1995, such as having a large number of women in the household and a younger head of the household, become important predictors in 2002. Finally, strong effects of spatial inequalities, large household size, and a large number of dependents in the households were consistent over time. Designing and implementing an effective social protection system, ensuring the accessibility and affordability of university education, gender mainstreaming of poverty reduction policies, and reducing spatial inequalities should be the priorities for Azerbaijan.