Political Polling: Sample Size and Margin of Error


New Member

My country have a proportional/majority system.

I wonder how can a poll have a margin of error of 2.0%, if all the regional samples used in the poll have a bigger margin of error?

- Southwest Region: N~64MM, n=880 and E~3.5%;
- Northwest Region: N~39MM, n=553 and E~4.5%;
- Central Region: N~22MM, n=303 and E~5.5%;
- South Region: N~22MM, n=294 and E~6.0%;
Total: N~147MM, n=2030 and E~2.5%;

I kind of get the formula, but not the intuition behind. How a subset of "less precise" surveys generate a better one?

Thank you!
Last edited: