The problem is subtler and worse than simple censorship. A person who concludes the "wrong" casual effect from the data is branded a racist or at least under suspicion. A person who concludes the "right" causal effect might face criticism but at least is not in danger of being labeled an evil person. Such a lopsided set of incentives creates a very strong bias. In my opinion this bias is much stronger than the bias ("racism") that it evolved in response to.