3 ms·

FiveThirtyEight: When to Freak Out About Shocking New Polls

10 points by dirtyaura 10y ago

marcoperaza 10y ago
Not hating on FiveThirtyEight for the interesting work they do--and I understand that covering this angle of things is their main purpose--but I can't help but feel that the intense focus on poll numbers by the media throughout the entire race is upside-down. The poll that matters is on November 8th and we'll all get the chance to participate in it. Until then, maybe we should focus on informing ourselves about the candidates and the issues, and less on who's winning the rat race each morning.
- toyg 10y ago
  But it's not a poll, it's a race with weird rules. Polls are a huge part of that race: if you are winning state X by double-digit margins, there is no point in flying there to campaign; whereas the swing state where you're down by 1% needs ad-hoc policy announcements and lots of time spent on the ground. Paying attention to the polls will often tell you where a candidate will stand on certain issues, so it's part of the whole "understanding" bit.
marcoperaza 10y ago
>There’s less to fight about when polls show similar results. But that doesn’t necessarily mean they’ll turn out to be more accurate. Instead, that consensus may reflect herding — pollsters suppressing results that they deem to be outliers, out of fear of embarrassment. And there goes my entire faith in the process. Deciding which polls to release based on the results is absurd. If the media is going to focus so much on polls (and see my other comment for what I think about that) then they should only source them from pollsters that share all results, not just the ones the pollster deemed "correct" because it fit with their expectations. Nate Silver's strategy for dealing with it isn't encouraging either: >In fact, you should trust a pollster more if it’s willing to publish the occasional “outlier.” Clinton probably isn’t winning Colorado by 13 percentage points right now or losing Pennsylvania by 6 points. But the fact that Monmouth and Quinnipiac are willing to publish such results are a sign that they’re letting their data speak for itself. It sounds like he's admitting he doesn't even know what the practices of each pollster are. For all we know, a pollster could be so sure in a Hillary or Trump lead that they don't publish the outliers in the other direction. How can Nate Silver aggregate these results with any certainty whatsoever?

10 points by dirtyaura 10y ago

Not hating on FiveThirtyEight for the interesting work they do--and I understand that covering this angle of things is their main purpose--but I can't help but feel that the intense focus on poll numbers by the media throughout the entire race is upside-down. The poll that matters is on November 8th and we'll all get the chance to participate in it. Until then, maybe we should focus on informing ourselves about the candidates and the issues, and less on who's winning the rat race each morning.

toyg 10y ago
But it's not a poll, it's a race with weird rules. Polls are a huge part of that race: if you are winning state X by double-digit margins, there is no point in flying there to campaign; whereas the swing state where you're down by 1% needs ad-hoc policy announcements and lots of time spent on the ground. Paying attention to the polls will often tell you where a candidate will stand on certain issues, so it's part of the whole "understanding" bit.

marcoperaza 10y ago

>There’s less to fight about when polls show similar results. But that doesn’t necessarily mean they’ll turn out to be more accurate. Instead, that consensus may reflect herding — pollsters suppressing results that they deem to be outliers, out of fear of embarrassment. And there goes my entire faith in the process. Deciding which polls to release based on the results is absurd. If the media is going to focus so much on polls (and see my other comment for what I think about that) then they should only source them from pollsters that share all results, not just the ones the pollster deemed "correct" because it fit with their expectations. Nate Silver's strategy for dealing with it isn't encouraging either: >In fact, you should trust a pollster more if it’s willing to publish the occasional “outlier.” Clinton probably isn’t winning Colorado by 13 percentage points right now or losing Pennsylvania by 6 points. But the fact that Monmouth and Quinnipiac are willing to publish such results are a sign that they’re letting their data speak for itself. It sounds like he's admitting he doesn't even know what the practices of each pollster are. For all we know, a pollster could be so sure in a Hillary or Trump lead that they don't publish the outliers in the other direction. How can Nate Silver aggregate these results with any certainty whatsoever?