Originally Posted by A Tyrant Named Miltiades!
I find it difficult to believe that pollsters haven't thought of this, given that this phenomenon has been in play for well over a decade.
Opinion polling is closer to Astrology than Statistics.
The telephone polling is largely a US import but the problem is that the Irish phone system is not laid out along US geographical zones where the area code is more accurate. The mobile phones don't have any geographical indicators. The phone numbers are supposed to be randomly generated but there are some anecdotal reports of people who had cooperated in one poll being repolled again.
The response rate (those who when concated are not willing to be polled) is also important but it rarely seems to be published. This means that rather than it being the poll of 1,000 people, it may actually be a poll of more than 1,000.
The pollsters also apply their own "secret sauce" adjustments to some of these polls to derive other indications. The problem is that it can lead to complete bolloxology like the "Burton Bounce" which had to remove SF voters along with other voters from the poll to get the "bounce".
There's also a cyclical effect in the main Irish polls in that each poll seems to favour specific party. Thus a party might do well in one poll but will not do well in the other polls that month. It is visible in the poll of poll type graphs. This kind of cyclical effect is very unusual and odd for supposedly random samples from the electorate.
They don't just gather random samples, not if they have the most elementary notion of how to do their job.
That changes a poll from being a random sample to something completely different.