Using linked employer-employee data from the Canadian Workplace and Employee Survey 1999-2004, we provide new evidence on how the cost of absence affects labor supply decisions. We use a particular feature of the data by which total absences are divided into three separate categories: sick paid days, other paid days and unpaid days. This division introduces variations in the way workers are compensated for absence (the cost of absence) and allows us to estimate more precisely how variations in such costs affect absenteeism decisions. We find an absence elasticity of -0.37. We also find unobserved heterogeneity to play different roles for workers and workplaces: some workers are more frequently absent whatever the reason, but paid and unpaid leaves are negatively correlated at the workplace level.