2016 Postmortem
Related: About this forumNate Silver: Who’s Winning Iowa And New Hampshire?
Nate silver's two different results as measured by two different methods.
http://fivethirtyeight.com/features/whos-winning-iowa-and-new-hampshire/
(snip)
The first model, which we call polls-only, is based only on polls from one particular state. (Iowa polls in the case of Iowa, for example.) Its basically an updated version of the model we used for the primaries four years ago.
The second model, polls-plus, also considers endorsements and national polls, in addition to state polls, and tries to consider the effect that Iowa and New Hampshire could have on subsequent state contests. (National polls arent necessarily a positive for a candidate in the polls-plus model; instead, its a bearish indicator when a candidates state polls trail his national numbers.)
Historically, polls-plus would have been somewhat more accurate, but its pretty close so we think the models are most useful when looked at together. Indeed, they present different perspectives on the races this year, mostly because of that endorsements variable, which helps Hillary Clinton and Marco Rubio but hurts Bernie Sanders and Donald Trump. If you take a strictly empirical view of the primaries, accounting for the establishments historical tendency to win out in the end, youll probably prefer the polls-plus model. If you think this time is different or youre a Trump or a Bernie fan youll probably like polls-only instead.
(snip)
Iowa: Because public opinion can shift rapidly in the primaries, our models put a lot of emphasis on the most recent polls. Thats good news for Sanders, who has been neck and neck with Clinton in Iowa polls published this month after trailing her for most of last year. In fact, the race is nearly a tossup: He now has a 45 percent chance of winning Iowa according to polls-only, although the polls-plus model, noting Clintons dominance in endorsements, is more skeptical of Sanders, giving him a 27 percent chance instead.
New Hampshire: Here, theres a split between the models. Sanders is a 73 percent favorite according to polls-only, while polls-plus noting Clintons advantage in endorsements and that shes favored in Iowa gives Clinton the slightest edge, with a 53 percent chance to Sanderss 47 percent. Essentially, shed be following the path that Al Gore took over Bill Bradley in 2000, when an Iowa victory propelled him to a narrow victory in the Granite State. But the polls-plus model is designed to lower the effect of the endorsements variable to zero by election day in each state. So if Clinton keeps falling in New Hampshire and Iowa polls instead of rising, the establishment may not be able to bail her out, and shell have to contemplate the possibility of being swept in both states.
litlbilly
(2,227 posts)highprincipleswork
(3,111 posts)jfern
(5,204 posts)Will be trying to unbias themselves for their final prediction if they don't want to lose all credibility. So predictable.
Matariki
(18,775 posts)catnhatnh
(8,976 posts)In New Hampshire it costs Bernie 26 points!!! You know-enough to give Hillary a small lead. I wonder why it only gives an 18 point penalty in Iowa???
Maybe it smells less when Nate pulls less out of his ass?
wilsonbooks
(972 posts)The natester's credibility has hit a wall.
jfern
(5,204 posts)And that Trump would immediately crash and burn.
Indepatriot
(1,253 posts)Camp Weather Vain must be in full meltdown mode.
BlueMTexpat
(15,365 posts)It's hardly "hedging" his bets. He explains the differences quite clearly.
- Polls only = all polls from one state only
- Polls plus = all polls + endorsements + other factors
Of course, they will give different results. It is more likely that the polls-plus prediction will be closer to the voting results.
But we'll only know that for sure on the respective caucus/election days.
In the meantime, people will see what they want to see and will believe what they want to believe. There's certainly nothing wrong with that.