Quote:
Originally Posted by Damien
Well this is why we have elections rather than let polls decide everything but they don't just get a random sample of 2,500. They balance out the sample with different demographics and then weight it to reflect society. It's not like an internet poll or asking random people in the street.
It does tend to work. They are quite good at this kind of thing and it's in use everyday in all sorts of industries.
Even the polling errors people like to cite are still roughly correct, at least in terms of sheer numbers if not when they try to project seats.
|
Well, I don't think 2,500 is sufficient when you factor in the fact that elections are won by constituencies. You need to get a bigger sample in all like constituencies to get the accuracy required. It was the failure to do this adequately at the last election that led to the result being such a surprise.