Posted 10:51 PM by Robert

Not A Fully Satisfactory Metric

This election was characterized by an unusual number of controversies regarding public opinion polls. The point at which the polls may be least significant is the very end of the campaign when the final polls are taken. Yes, there is some speculation that final polls may affect turnout a bit, but that surely is a very minor effect. Further, by the final days preceding the election the campaign has been run, the candidates and their supporters have advanced their arguments, deployed their strategies and tactics and marshaled their positions and forces - all informed by public opinion polls taken much earlier in the campaign. Even beyond that, the early dynamics and psychology of a campaign are largely determined by reports on polling results. Those factors can seriously affect voting patterns in primaries in which strategic voting can be very significant. In contrast, the final polls are mostly curiosities, surrogates for the real thing soon to come - but not influencing that real thing much.

So it strikes me as a very odd thing indeed to measure which pollster was "most accurate" on the basis of whether the pollster's final poll "called" the election. But that's what Real Clear Politics is doing. And they are very smart guys. But consider, for example, Zogby's bizarre, outlying polling reports the week before the New Hampshire primary, which he pulled into line with the other polls in his "final" poll. Was Zogby "accurate" in new Hampshire? I don't think so, but the RCP metric says he was accurate ... and Zogby has used this trick on other occasions.