[mlo]

Every poll we hear about "overweighting". Just like four years ago, when everyone thought the polls needed correcting.

The problem then, as it is now, is that many of these polls aren't "weighting" for party at all, so they can't be "overweighted". If you pick your random sample by criteria other than party, and 41% of the people answer the party ID question as "Democrat", that's not weighted, it's just reporting the results.

[Alberta's Child]

The point is that they should be corrected/calibrated by various metrics in order to ensure their accuracy. The term "weighting" doesn't refer to a process that the pollster does; it refers to a breakdown of the voters by these various metrics.

If you conduct a poll of a "random sample" of American adults and you end up with a 95%-5% male/female breakdown for whatever reason, then the weighting is way off and it needs to be adjusted to reflect the reality of this split.

[Steve Van Doorn]

Just to be clear most polls sampling Democrats as it was 2012. When in reality they're not going to the polls as they did in 2012.

Most people in freerepublic believed the democrats knew who Obama was in 2012 and fewer would go too the polls. Obama did get fewer numbers then in 2008 but it wasn't that far off from 2008 for the Rino to win.