Quote:
Originally Posted by Uhuniau
I hope you're being facetious.
|
I was being serious. Given all the sampling bias errors that pollsters accumulate, 1505 is a pretty small sample size for 28 million electors.
If the samples they gathered were purely representative, then 1505 is plenty. But the problem is they're not. All forms of gathering data create sampling bias:
1) Calling people on landlines--thus excluding the large number of people who don't have landline telephones. These people are disproportionately likely to vote NDP.
2) Internet polling--thus excluding people who don't go on the internet. These people are disproportionately likely to vote Conservative.
3) Calling people on mobile phones--excludes those without them, who are more likely to vote Conservative.
Most pollsters these days include a mix of all 3 but getting the balance exactly right is almost impossible. So you need LARGE sample sizes (5000+) to get an accurate picture--especially when you're breaking the data down to get numbers by region, sex, age, etc.