Using Norwegian individual register data of young workers, from the period 1986-2008, we analyse whether there are large and persistent negative relationships between unemployment and the risk of repeated unemployment and being out of labour force. A nearest-neighbour propensity score matching method is applied to make the treatment group (the unemployed) and the control group (the employed) as similar as possible. By tracking workers over a 10-year follow-up period, we find that unemployment has a negative effect on later labour market attachment. This is consistent with existing findings in the literature. The negative effects decrease over time. Using the bounding approach proposed by Rosenbaum (2002) to analyse the importance of unobserved variables, our results indicate that a relatively high level of unobserved selection bias could be present in the data before changing the inference. Thus, unemployment leaves young workers with long-term scars.