Fascinated with why the polls were so wrong in New Hampshire

Every so often we delve into "Politics 2.0, ie the impact on the 'Net on politics and government.

By now everyone will be aware that the predictive polls in the US in New Hampshire were wrong - but not completely, they got Obama and Edwards roughly right, but were nearly 30% out on Clinton. The question is why were they so wrong on one person?

Pollsters don't know in truth - they argue that this is a less predictable field, or huge numbers of voters lied, or positioning of names on ballots, or a large lump of undecideds all dcided on one candidate - but to be that far out is not really explainable those ways, especially as the polls were being taken until fairly close before the actual voting.

There are only really three possible explanations:

(i) There has been some form of foul play

(ii) Some late action influenced a mass of people independently

(iii) The poll mechanisms didn't pick up a tipping point in the underlying trends (the "undecided swing" or somesuch)

Attractive as (i) seems to be for the conspiracy theorist in us all, I am more curious about whether (iii) occurred, and if so, why wasn't it picked up?

The usual argument is that polls are not actually representative - not truly random, truly universal, and they usually have over-arching historical metadata that "smoothes" the raw results which can and does go wrong.....but even so, this is a huge shift considering the typical accuracy range of polls. Some of those levers must have been way, way out.

I wonder if internet media has something to do with it, as polls are typically artefacts of mainstream media and are unlikely to have any views of the digital backchannels around today. Looking at it from the outside, this looks like it might be a social network driven tipping point - a "flash mob" sort of thing.

If it was, it means that elections going forward are going to be very interesting, as these backchannels will be acting as feedback loops, commenting on the candidates ' broadcast, controlled media output in ways that just cannot be tracked or predicted.

The implications for the democratic process in the "Politics 2.0" world are both interesting and scary. Scary in that these new media are probably still far more open to manipulation, interesting in that it is showing that the electorate can be given newer, faster tools to vote - and if that is possible, it can move from voting for candidates today, to voting for policies tomorrow.