Census demographer Martin O’Connell explained that the survey results are weighted by age, race and gender. If a 26-year-old black man is married to a 35-year-old white woman, and if 26-year-olds and blacks are underrepresented in the survey sample, then his interview may be weighted more heavily than hers. "There's no final step to make sure the number of husbands and wives is exactly equal," Mr. O’Connell told me. "If we did that, we would have to reweight the data again, and the data would endlessly spin around.” Nonetheless, he says of the American Community Survey numbers, "They are the best numbers you will ever get, any place."They are already spinning the data.
There was some controversy over whether the 2000 Census should weight the data for Congressional apportionment. As you can see, spinning the data has some adverse consequences that are subject to political manipulation.