In other words, shouldn't they call 3,000 people and then take the 1,000 that meet the demographic measures, instead of calling 1,000 people and then weight the responses?
Otherwise, their weighting factors only assume how a respondent will reply, as if they are automatons displaying group herd mentalities.
-PJ
Methodologies vary, but in this case they random sample, and then weight the responses to make them match the expected demographics.