Falsification before Extrapolation in Causal Effect Estimation