A large representative sample of twins is used to estimate the causal effect of schooling on earnings in Norway, for both men and women. The within-twin-pair estimates reveal that standard cross section (OLS) estimates of the effect on male hourly earnings are biased upwards. For women, no such...