We provide evidence on the least biased ways to identify causal effects in situations where there are multiple outcomes that all depend on the same endogenous regressor and a reasonable but potentially contaminated instrumental variable that is available. Simulations provide suggestive evidence...