Abstract: The analysis of poverty dynamics yields important insights about the nature of poverty and the expected effectiveness of alternative social policies in order to fight against it. The recent literature on poverty dynamics proposes various approaches to the measurement of the effects of poverty spell length and the experience of multiple-spells on poverty exit or re-entry rates. However, none of the proposals in the literature considers the expected effect on transition probabilities of the accumulation of poverty spells in an individual’s poverty history. This paper proposes the estimation of a hazard regression model where the individual probability of leaving and entering poverty is affected by the length of the current poverty spell, the fact that the individual experiences repeated spells which accumulate in time. Results indicate that poverty transitions still show some negative duration dependence even if we introduce controls for unobserved heterogeneity and lagged durations. The duration of previous poverty spells reduces the exit and increases the re-entry hazard. Finally, estimating separate hazards by spell order allows for some control for the relevant impact of left-truncation on results and shows the significant differences in the covariates that turn out to promote transitions for individuals that often fluctuate into and out of poverty (transitory poor) in comparison with those that suffer a rather more persistent poverty experience (chronic poor).