We quantify labour market effects of changes in the potential benefit duration (PBD) in Poland. Individual workers' PBD depends on the county unemployment rate relative to the national average---12 months of PBD above a cut-off of 125 per cent and 6 months below. This cut-off shifted from 125 to 150 per cent in a 2009 reform. We utilize (i) the natural experiment of the reform and (ii) the sharp discontinuity generated by the cut-offs to estimate effects of shortening the PBD. Our administrative data cover unemployment spells for prime age workers during the years 2004-2019. A one-month shorter PBD decreases average benefit duration by 0.5 months and average unemployment duration by 0.4 months. The PBD reduction by six months increased the job finding rate within the first 9 months by 6 percentage points. We next investigate whether those micro-level effects spill over to the macro-level. Using information on monthly unemployment rates at the county level, we find evidence that a longer PBD increases a county's unemployment rate by 0.3 percentage points after 12 months.