If that were actually true then we would see wild differentials between poles and election results. I admit the possibility that results may be shifted a few percent either way, enough to swing an election, but this will only ever be at the margins. Should the population be truly set on one candidate or the other (more than 5% difference) then rigging would differ from poling very significantly and be detected.