Using survival models, we test whether short-term changes in the labour market affect poverty duration. Data are from the Brazilian Monthly Employment Survey. Such a monthly dataset permits more accurate estimations of events than using annual data, but its panel follows households for a short period. Then methods that control for both right- and left-censoring should be used. The results are as follows: households with zero income are not those with the lowest chances of exiting; changes in aggregate unemployment do not affect poverty duration; and increasing wages in the informal sector has a negative effect on poverty duration.