This study complements existing poverty information in Sri Lanka by presenting a comprehensive poverty profile that examines bi-variate relationships between poverty and its covariates over a 17 year period from 1985 to 2002 using comparable data and consistent measures of poverty. The poverty profile includes poverty measures by location characteristics, such as sector, district and province, and by demographic, human capital and labour market characteristics of the household head, such as gender, educational attainment, employment status, industry, occupation, ethnicity and religion and composition of household income, including the distribution of welfare benefits, disability payments and remittances. The study also decomposes changes in poverty in Sri Lanka over time into their growth and redistribution components, at the national, sectoral, provincial and district level and provides estimates of relative poverty changes during this period, based on several scenarios for a relative poverty line. The data used is from the 2002, 1995/96, 1990/91 Household Income and Expenditure Surveys (HIES) and the 1985/86 Labour Force and Socio-economic Survey (LFSES). The analysis shows that poverty in Sri Lanka has declined over the last 20 years, but that the decline has neither been large, nor steady. In fact, a relative definition of poverty based on a value close to the absolute value of the poverty line in 1995/96 would indicate that poverty has risen. The only sustained decline observed was in urban areas, particularly in the relatively affluent Western province, and within this province, in the Colombo district. The relative regional distribution of poverty at the end of the 17 year period appears to have changed little from its initial picture; the Western province has the lowest poverty, the next cluster of provinces with moderate poverty comprised the Central, North Central and North Western provinces, and the highest poverty was observed in the Southern, Sabaragamuwa and Uva provinces. However, during this period, striking reductions in poverty were observed in the North Central province and Southern province. Trends in absolute numbers and contribution to poverty moved parallel to trends in poverty incidence, with the situation in the least poor province, Western province, improving, and the situation in the poorest province, Uva province, deteriorating. At the beginning of the period, one in five persons in the former, and one in 10 persons in the latter, was poor. By the end of the period, these numbers had converged with a 15% contribution to poverty of the more populous Western province, and a 12% contribution to poverty of the sparsely populated Uva province. Similarly, trends in poverty incidence and contribution to poverty moved parallel among all three measures of poverty. Thus, the depth and severity of poverty are generally greatest where the incidence of poverty is highest. Decomposition of poverty changes highlighted the importance of both growth and redistribution, depending on the period under consideration. Decompositions at the disaggregated level indicated that adverse redistribution, especially in the last period, occurred not simply between districts, but also within districts. Adverse redistribution and low consumption growth led to increases in the depth and severity of poverty, indicating that the inequality experienced in the last period was not limited to the upper part of the distribution but had an effect on the poor. Simulation exercises with different growth rates indicated the potential of growth to reduce poverty reduction. However, the high inequality experienced in the last period (1995/96-2002) indicated that the predictive power of these simulations is limited in the face of adverse redistribution, especially around and below the poverty line. Occupational and income profiles of the poor indicated that poverty was associated more strongly with households whose head was engaged in agriculture, and least associated with households with only non-agricultural sources of income. Trends in poverty over the 17 year period indicate that the least decline in poverty was experienced by households with only agricultural income or a combination of agricultural and wage income. The situation of female headed households, and households headed by those with low or moderate educational attainment has also worsened over the period, while poverty rates disaggregated by ethnicity and race have not changed very much. The latter may be due to the fact that the analysis is conducted only for the seven districts outside the North Eastern province. A ‘quick and dirty’ simulation exercise with relative poverty lines suggests that an appropriate relative poverty line for Sri Lanka is that of 75% of median income. This was close to the value of the absolute poverty line in 1995/96. The exercise also demonstrated that relative poverty lines will always overstate the extent of poverty (relative to absolute poverty) during periods of rising (overall) living standards and will always understate the extent of poverty during periods of falling living standards. Thus, they should be used in conjunction with, rather than instead of, absolute poverty lines. While the lesson from the regional patterns in poverty reduction appears to be that economic growth is essential for poverty to decline (as was the case in the Western province), growth will need to occur in the outlying provinces and in the agricultural sector in order for it to have an impact on poverty. Alternatively, sufficient alternatives away from agriculture need to be developed in order to enable the vast majority of the rural poor to escape poverty, especially in periods of low growth. Analyses of the distribution of welfare benefits, disability and relief payments and remittances indicated that the role of transfers - and by inference, of public and private redistribution mechanisms - in reducing poverty is limited.