Buyers (and campaigns) beware. Not all polls are equal.

As of yesterday, some 1.12m Floridians had already cast their ballots, including 600.2k Republicans and 486.9k Democrats.

As opposed making calls from a registration-based sample in which the voter’s party affiliation and vote history is known (or should be known) from information pulled from public records, some pollsters random digit dial Floridians, asking to speak to a registered voter in the household and then asking them to self-identify their political party.

If the survey’s sample frame (among other factors) is wrong, biased results can result.

For example, a pollster may only want to survey voters who respond that they cast ballots in the past two general elections (say, 2012 and 2014), thinking that this will generate a sample of voters who are most likely to turn out in the 2016 presidential primaries.

But this might be a very misguided assumption in the pollster’s likely turnout model.

Indeed, if we look at the vote history of the 1.12m votes cast thus far in the Florida PPP, according to the voter files 18% of those who have voted did not vote in the 2012 and the 2014 General Elections.

On balance, we’re seeing both Republicans (17.3%) and Democrats (17.5%) coming to the polls early and voting absentee ballots who didn’t bother voting in the last two elections.

Hard to imagine that these are voters — Republicans and Democrats alike — who are supporting their party’s establishment.