We use Norwegian micro-data to identify the driving forces behind unemployment spells following temporary- and permanent dismissals. The duration of unemployment spells for permanently dismissed workers is primarily explained by individual resources and economic incentives, while spell-duration for temporary dismissed workers is explained by firm incentives. Higher benefits reduce the employment hazard for permanently- but increase it for temporary dismissed workers. Structural duration dependence is nonparametrically identified and sharply estimated with the aid of lagged hazard rate variation and repeated spells. The employment hazards for both spell-types exhibit positive duration dependence during the first months and negative duration dependence thereafter.