Will, also frustrated with the apparent disconnection of most poll results with the rest of the news and given the vast variety of information distribution and uptake, I have come to seriously doubt the ability of ANY organization to effectively poll the populace. How would they accomplish contact with enough different people on enough d…
Will, also frustrated with the apparent disconnection of most poll results with the rest of the news and given the vast variety of information distribution and uptake, I have come to seriously doubt the ability of ANY organization to effectively poll the populace. How would they accomplish contact with enough different people on enough different platforms who are actually willing to respond to represent the populace in general?!
Oh, it can be done; there are decades upon decades of solid mathematics behind the methods used by reliable practicioners in the field. The thing people tend to struggle with is understanding that even the most perfectly-weighted, gold-star poll has a margin of error and confidence interval, and since things like voter attitudes towards candidates are fluid, you are aiming your dart at a moving target. If you can be 95% sure that your poll results have a margin of error of 3%, that would be considered a job well done, and useful to a campaign. Yet if you are a reader looking to know who is "ahead," and the poll says Candidate A is ahead by 3%, all that actually means is that at the very second you're reading the article Candidate A is somewhere in the range of losing by a fraction of a percent to winning handily by 6, except for the 5% chance where either a win or a loss is even bigger, and who the heck knows in three weeks?
So... always look at averages and trends, not a single poll. Ever. And preferably don't look at them much at all. Just make some phone calls for who you want to win and then go read something fun instead.
Will, also frustrated with the apparent disconnection of most poll results with the rest of the news and given the vast variety of information distribution and uptake, I have come to seriously doubt the ability of ANY organization to effectively poll the populace. How would they accomplish contact with enough different people on enough different platforms who are actually willing to respond to represent the populace in general?!
Oh, it can be done; there are decades upon decades of solid mathematics behind the methods used by reliable practicioners in the field. The thing people tend to struggle with is understanding that even the most perfectly-weighted, gold-star poll has a margin of error and confidence interval, and since things like voter attitudes towards candidates are fluid, you are aiming your dart at a moving target. If you can be 95% sure that your poll results have a margin of error of 3%, that would be considered a job well done, and useful to a campaign. Yet if you are a reader looking to know who is "ahead," and the poll says Candidate A is ahead by 3%, all that actually means is that at the very second you're reading the article Candidate A is somewhere in the range of losing by a fraction of a percent to winning handily by 6, except for the 5% chance where either a win or a loss is even bigger, and who the heck knows in three weeks?
So... always look at averages and trends, not a single poll. Ever. And preferably don't look at them much at all. Just make some phone calls for who you want to win and then go read something fun instead.