The alleviation of poverty is one of the most important objectives of the Australian social security system. Estimates of poverty across different groups and over time are thus in principle a significant indicator of the effectiveness of social security programs. This paper investigates the reliability of Australian poverty estimates for 1981-82 and 1985-86 by assessing their sensitivity to changes in the assumptions used to construct a poverty line. The exercise utilises data from the 1981-82 Income and Housing Survey and the 1986 Income Distribution Survey. The results indicate that many broad conclusions about poverty incidence and trends are robust over the period studied. However, poverty estimates exhibit considerable sensitivity for those groups heavily reliant on the social security system for their income.