Survey statistics

A population total for education level is known, and education level is strongly related to the survey outcome. Why might calibrating weights to education totals help estimate the outcome?