Click to enlarge
"The chart above shows how much the polling average at each point of the election cycle has differed from the final result. Each gray line represents a presidential election since 1980. The bright green line represents the average difference. At this point – 167 days before the election – a simple polling average has differed from the final result by about nine percentage points. We expect this average to become more meaningful by the week, until the national party conventions temporarily make it less so, as shown in the bump about 100 days before the election. The average difference begins to flatten about two months before the election. The day before the voting, an unadjusted polling average has been about 3.5 points off the final result."
'via Blog this'
The same phenomenon can be observed in Australia. The further away from election day an opinion poll is taken the less reliable a guide to what the actual result will be.
On this blog I only have historical data for Newspoll but by my rough calculation the 50:50 result of the recently published polls will end up some 3.5 to 4 percentage points wrong when the real poll takes place.
Only with a week or so to go will I be taking much notice of what the opinion pollsters excite the media with.