In this paper I investigate the causal returns to education for different educational groups in Germany by employing a new method by Klein and Vella (2010) that bases identification on the presence of conditional heteroskedasticity. Compared to IV methods, key advantages of this approach are...