Finally, differences in sample implementation can also lead to overestimation or underestimation of fertility in both surveys [7, 23]. For instance, if educated women were accidentally oversampled in the first survey, fertility would be biased downward.