A survey oversamples one region, and respondents there have a higher average outcome than respondents elsewhere. The weighted overall mean is lower than the unweighted sample mean. Which explanation is most plausible?