as absolute levels in explaining the education-health gradient. We show that relative education impacts smoking, when direct utility is relative, or when there is signalling in the labour market. We use data from the "Enquête sur les Conditions de Vie des Ménages 2001" and a major reform of...