TIL polling averages hide more than they show until you dig into the crosstabs

I used to just look at the top-line numbers in political surveys and call it a day. Last month I got curious about a state poll in Ohio that showed a 4 point lead, but when I broke it down by age group, the under 35 sample was only 120 people out of 800. That tiny slice can swing the whole average if it misses. I started checking the margin of error on each subgroup before trusting any headline. It took me about 20 minutes per poll, but now I spot bad takes before they go viral. Has anyone else found a simple way to quickly verify sample sizes without reading the full methodology PDF?

1 comments

1 Comment

stellat466d ago

The "4 point lead" part got me. That number means nothing alone. I saw the same thing with a poll here. Under 30s were like 15% of the sample. Whole thing flipped when you looked at folks over 50. People just read the headline and move on. Takes five minutes to spot the real story.