Extended field efforts to reduce the risk of non-response bias: have the effects changed over time? Can weighting achieve the same effects?