This study reconsiders the empirical question of whether men's earnings increase because of children. Large Norwegian register data are used for brother and twin pairs who are followed over their life cycle from their first entry into the labour market. The data permit family-fixed effects to be modeled in various ways, as well as observing earnings growth before and after having children. The simple conditional correlation between children and earnings is positive. When only variation from between-sibling differences is used, the earnings effect post entry into first-fatherhood declines. The effect becomes small and non-significant when we use twins.