@_cingraham wait, just 1000 ppl interviewed? I assume the 3.5 pct error is just for total, and errors for individual groups are *way* higher