# When does FiveThirtyEight start projecting the Presidential race based on polls? **URL:** \ **Category:** Politics & Elections\ **Created:** [June 16, 2016, 10:42pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566 "2016-06-16T22:42:16Z")\ **Posts on this page:** 18\ **Page:** 2
**Author:** ![DSeid](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/dseid/32/20194_2.png) [@DSeid](https://boards.straightdope.com/u/DSeid)\ **Post date:** [June 23, 2016, 1:26am UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/21 "2016-06-23T01:26:57Z")
> [@septimus](#): > > … I did call PA as a key tipping state in earlier threads and was soundly ridiculed: “PA is lock Blue, not swing.” … [Per 538](http://fivethirtyeight.com/features/the-clinton-campaign-seems-to-think-pennsylvania-is-in-the-bag/) Team Clinton is apparently of the belief that PA is a lock Blue and 538 is with you that it may be more a tipping point state. > [@538](#): > > … when it comes to spending that cash, her campaign may be making a strategic miscalculation. The campaign and an allied super PAC have reserved $137 million of ads across eight states — yet they’ve conspicuously left out the state that might be likeliest to tip the 2016 election: Pennsylvania. > > Michigan and Wisconsin were absent from the list, as well, but the Keystone State is the most curious Rust Belt omission. In May, we laid out the case for why the Keystone State could be on pace to decide a close national race. Evidently, Democratic ad strategists don’t share that view. Last week, former Obama advisor David Plouffe tweeted that the buy “shows real discipline” because “PA is not a true battleground.” That’s a bit of bravado, considering it was tied for Obama’s fourth-closest win in 2012. … ---
**Author:** ![DSeid](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/dseid/32/20194_2.png) [@DSeid](https://boards.straightdope.com/u/DSeid)\ **Post date:** [June 29, 2016, 6:38pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/22 "2016-06-29T18:38:21Z")
> [@Roderick\_Femm](#): > > After the conventions, presumably, but how long after? I wasn’t paying enough attention last time until the last month or so before the election. How soon do they start this? … And the answer is … [today](http://fivethirtyeight.com/features/donald-trump-has-a-20-percent-chance-of-becoming-president/). > [@Nate Silver](#): > > Our polls-only model has Trump with a 19 percent chance of beating Clinton as of early Wednesday afternoon. (The forecasts will continually update as new polls are added.) Our polls-plus model, which considers economic conditions along with the polls, is more optimistic about Trump, giving him a 26 percent chance. > > Still, Trump faces longer odds and a bigger polling deficit than John McCain and Mitt Romney did at the same point in their respective races. He needs to look back to 1988 for comfort, when George H.W. Bush overcame a similar deficit against Michael Dukakis to win. Wang is currently at 70% Clinton with random drift and 85% “Bayesian.” Betting markets are also in that same range. ---
**Author:** ![Derleth](https://avatars.discourse-cdn.com/v4/letter/d/b9e5f3/32.png) [@Derleth](https://boards.straightdope.com/u/Derleth)\ **Post date:** [June 29, 2016, 7:01pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/23 "2016-06-29T19:01:02Z")
As long as we can keep Hillary out of a tank we’ll be fine. ---
**Author:** ![DSeid](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/dseid/32/20194_2.png) [@DSeid](https://boards.straightdope.com/u/DSeid)\ **Post date:** [June 29, 2016, 7:19pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/24 "2016-06-29T19:19:22Z")
It is interesting that the different models all pretty much converge: 538 gives Trump 19 to 26 depending on method; Wang 15 to 30 also depending on model variation; betting markets (per Silver) 20 to 25. That’s all averaging 22.5. Meanwhile Silver notes the same thing I noted a few days ago in another thread: at this point at least Trump performs _better_ in swing states than in the nation overall. He offers a potential explanation though, that such is a function of his _relatively_ weak performance in some traditionally solidly Red states bringing his national numbers down even though he will still win most of those states. Of course it could just be that state polling is still relatively sparse. Still it raises a question … IF Trump wins places like Texas but by smallish margins does that have any impact on down ticket races? IOW is it worth competing in places like Texas (especially if you have a bigger war chest) not with an expectation of winning the state (excepting in such an overwhelming national win that it would not matter) but rather in order to make some down ticket headway and build for future elections as the demographics shift further? ---
**Author:** ![Derleth](https://avatars.discourse-cdn.com/v4/letter/d/b9e5f3/32.png) [@Derleth](https://boards.straightdope.com/u/Derleth)\ **Post date:** [June 29, 2016, 7:31pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/25 "2016-06-29T19:31:21Z")
> [@DSeid](#): > > It is interesting that the different models all pretty much converge: > > 538 gives Trump 19 to 26 depending on method; Wang 15 to 30 also depending on model variation; betting markets (per Silver) 20 to 25. That’s all averaging 22.5. That’s what happens when you have multiple honest ways of looking at the same reality, yes. It would be _shocking_ if they _didn’t_ all pretty much converge, especially Wang and Silver, who are both applying statistical methods to the same polling data. (Betting markets are another story, because the market can stay irrational for longer than you can stay solvent. That is, bettors are not necessarily basing their actions on anything rational, even though they ostensibly have money on the line; they can over-value and under-value all they want based on gut feelings, especially if individual bettors don’t feel they have _that much_ riding on the outcome.) > [@](#): > > Still it raises a question … IF Trump wins places like Texas but by smallish margins does that have any impact on down ticket races? IOW is it worth competing in places like Texas (especially if you have a bigger war chest) not with an expectation of winning the state (excepting in such an overwhelming national win that it would not matter) but rather in order to make some down ticket headway and build for future elections as the demographics shift further? I think so, yes. This could be a big year for the Senate if Red-State Democrats are energized to get out and vote and are primed with that downticket strategy. ---
**Author:** ![Uncle\_Jocko](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/uncle_jocko/32/2932_2.png) [@Uncle\_Jocko](https://boards.straightdope.com/u/Uncle_Jocko)\ **Post date:** [June 29, 2016, 7:44pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/26 "2016-06-29T19:44:13Z")
Just for grins, I took today’s 538 polls-plus numbers and fired up the 270 To Win app. I distributed all but the eight closest states (as determined by Silver) to Clinton or Trump, leaving Colorado, Arizona, Iowa, Missouri, Ohio, New Hampshire, North Carolina and Florida undecided. That map gave Clinton 266 electoral votes. Trump would need to win _ **every single one** _ of the eight closest states to take the presidency. If Clinton could only manage to swing just NH, those 4 EVs put her at 270. (Of course this doesn’t account for the odd split EVs in Maine and Nebraska, though.) Trump’s got a yooge challenge with the map. Although it’s not even July yet, I know lots can change before November. ---
**Author:** ![Ulf\_the\_Unwashed](https://avatars.discourse-cdn.com/v4/letter/u/b4bc9f/32.png) [@Ulf\_the\_Unwashed](https://boards.straightdope.com/u/Ulf_the_Unwashed)\ **Post date:** [June 29, 2016, 8:36pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/27 "2016-06-29T20:36:03Z")
> [@DSeid](#): > > [Per 538](http://fivethirtyeight.com/features/the-clinton-campaign-seems-to-think-pennsylvania-is-in-the-bag/) Team Clinton is apparently of the belief that PA is a lock Blue and 538 is with you that it may be more a tipping point state. You probably know this by now, but that article was jumping the gun a little–about two days after this article appeared a PAC supporting Clinton announced plans to spend $10.5 million on TV ads in Pennsylvania. IOW, right in line with what they had already earmarked for Virginia. I suppose you could argue that the Clinton campaign was pushed to change their minds by the fivethirtyeight article 🙂 but I kind of doubt it… ---
**Author:** ![DSeid](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/dseid/32/20194_2.png) [@DSeid](https://boards.straightdope.com/u/DSeid)\ **Post date:** [June 29, 2016, 10:33pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/28 "2016-06-29T22:33:04Z")
> [@Uncle\_Jocko](#): > > Just for grins, I took today’s 538 polls-plus numbers and fired up the 270 To Win app. I distributed all but the eight closest states (as determined by Silver) to Clinton or Trump, leaving Colorado, Arizona, Iowa, Missouri, Ohio, New Hampshire, North Carolina and Florida undecided. > > That map gave Clinton 266 electoral votes. Trump would need to win _ **every single one** _ of the eight closest states to take the presidency. If Clinton could only manage to swing just NH, those 4 EVs put her at 270. (Of course this doesn’t account for the odd split EVs in Maine and Nebraska, though.) > Trump’s got a yooge challenge with the map. Although it’s not even July yet, I know lots can change before November. > [@Ulf\_the\_Unwashed](#): > > You probably know this by now, but that article was jumping the gun a little–about two days after this article appeared a PAC supporting Clinton announced plans to spend $10.5 million on TV ads in Pennsylvania. IOW, right in line with what they had already earmarked for Virginia. I suppose you could argue that the Clinton campaign was pushed to change their minds by the fivethirtyeight article 🙂 but I kind of doubt it… **Ulf** I had missed that! Thanks. I must admit I don’t quite get 538 polls-only currently giving PA a Clinton +8 with an 81.6% probability of winning. June polls are Clinton +1, +1, and +4. To say that current polling (polling only) supports Clinton winning PA by more than the national margin seems odd. So **Uncle Jocko** that also presumes that all states Silver calls not in the eight closest do indeed go that way. Remember that list has the number 9 state being called at 70% probability and you have to go 13 more states until the certainty gets to over 80%. 80% certainty implies being wrong one in five times and they are not likely independent events. ---
**Author:** ![Roderick\_Femm](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/roderick_femm/32/14875_2.png) [@Roderick\_Femm](https://boards.straightdope.com/u/Roderick_Femm)\ **Post date:** [June 29, 2016, 11:16pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/29 "2016-06-29T23:16:05Z")
> [@DSeid](#): > > And the answer is … [today](http://fivethirtyeight.com/features/donald-trump-has-a-20-percent-chance-of-becoming-president/).Wang is currently at 70% Clinton with random drift and 85% “Bayesian.” > > Betting markets are also in that same range. Cool. Thanks. ---
**Author:** ![DSeid](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/dseid/32/20194_2.png) [@DSeid](https://boards.straightdope.com/u/DSeid)\ **Post date:** [June 30, 2016, 3:06am UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/30 "2016-06-30T03:06:12Z")
You are welcome. So as to how 538 comes up with PA so much more strongly Clinton than recent polls state … the method uses all the polls and uses recency as only one factor, along with size and track record of the house. An older Marist poll, two months old (!), is weighted most strongly because it was large and the house is highly regarded. That two month old poll was Clinton +15 and offsets more recent data showing a closer race in the 538 model. Wang’s model is simpler: last three or last week’s polls for the state, whichever is greater, use the median. PA last three are now: tie; Clinton +1; Clinton +4. Median is +1. So Wang rates PA as current Clinton +1, a virtual toss-up, and Silver as Clinton +8, while RCP rolling average is Clinton +2.5 Again, impressive that the different techniques or 538 and PEC come up with the average of their two variations for the national results (and the sauce added by each to create their value added approaches is very different as well) as exactly the same. ---
**Author:** ![Ulf\_the\_Unwashed](https://avatars.discourse-cdn.com/v4/letter/u/b4bc9f/32.png) [@Ulf\_the\_Unwashed](https://boards.straightdope.com/u/Ulf_the_Unwashed)\ **Post date:** [June 30, 2016, 3:37pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/31 "2016-06-30T15:37:35Z")
Anybody else notice the way fivethirtyeight phrased the headline on the article presenting their first prediction? For background, note on the side of the site the baseball predictions, which specify for example that Cleveland has a 52% chance of defeating the Blue Jays tonight. Or the predictions for the primaries, earlier, which would state things like “Bernie Sanders has a 72% chance of winning the Wisconsin Democratic primary.” That seems to be the standard way of expressing these things: “[Favorite] has a [percentage] chance of winning [event].” Yet the headline is “Donald Trump has a 20 percent chance of becoming president.” NOT the more standard formulation “Hillary Clinton has an 80 percent chance of becoming president.” I am not accusing fivethirtyeight of any particular bias toward Trump or against Clinton, or anything of the sort…but I find it fascinating. It’s just always all about Trump! ---
**Author:** ![Uncle\_Jocko](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/uncle_jocko/32/2932_2.png) [@Uncle\_Jocko](https://boards.straightdope.com/u/Uncle_Jocko)\ **Post date:** [June 30, 2016, 3:41pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/32 "2016-06-30T15:41:43Z")
> [@DSeid](#): > > **Ulf** I had missed that! Thanks. > > I must admit I don’t quite get 538 polls-only currently giving PA a Clinton +8 with an 81.6% probability of winning. June polls are Clinton +1, +1, and +4. To say that current polling (polling only) supports Clinton winning PA by more than the national margin seems odd. So **Uncle Jocko** that also presumes that all states Silver calls not in the eight closest do indeed go that way. Remember that list has the number 9 state being called at 70% probability and you have to go 13 more states until the certainty gets to over 80%. 80% certainty implies being wrong one in five times and they are not likely independent events. Oh, I know, absolutely. I wouldn’t presume to say anything based on polling in the first half of the year equates to a lead-pipe lock in November. So throw PA in the mix. With a snapshot of today’s situation, Trump would need to make practically a clean sweep of the nine states polling the closest. As of today, that’s a big hill to climb. ---
**Author:** ![Morgenstern](https://avatars.discourse-cdn.com/v4/letter/m/f9ae1b/32.png) [@Morgenstern](https://boards.straightdope.com/u/Morgenstern)\ **Post date:** [June 30, 2016, 4:00pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/33 "2016-06-30T16:00:37Z")
It’s as if Trump gets a “get out of jail” card on everything he does from his supporters. All you really need to ask the population is whether they watch Fox News, CNN or MSNBC. That’s going to give you a better count of the eventual vote. I couldn’t live without Rachel Maddow’s reassuring voice each night. Trump getting more than 200 electoral votes is Freddy Krueger on the loose to me. ---
**Author:** ![DSeid](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/dseid/32/20194_2.png) [@DSeid](https://boards.straightdope.com/u/DSeid)\ **Post date:** [June 30, 2016, 4:25pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/34 "2016-06-30T16:25:08Z")
> [@Uncle\_Jocko](#): > > Oh, I know, absolutely. I wouldn’t presume to say anything based on polling in the first half of the year equates to a lead-pipe lock in November. > So throw PA in the mix. With a snapshot of today’s situation, Trump would need to make practically a clean sweep of the nine states polling the closest. As of today, that’s a big hill to climb. Yes, but again, not all independent events. Even if somewhere north of 80% probable not. Take [Wang’s current status with Trump just +2 map](http://www.270towin.com/maps/princeton-election-consortium-trump-outperforms-polls). Give Trump the three toss-ups of IA, NH, and FL. Then have him just flip ME as an any given Sunday (Clinton is relatively underperforming there) and it is 269 to 269 with the House deciding. Have him instead flip VA (in four-way currently only +3) and it’s a Trump EV win even with a popular vote loss. The agregated popular vote seems (at this point) sticky at 5 to 7 Clinton in the face of Trump being just about as awful as can be; is it just as sticky in the other direction if he starts campaigning better, gets funding, and Clinton has some significant gaffes? I hope so. (And even more hope to not find out.) ---
**Author:** ![Morgenstern](https://avatars.discourse-cdn.com/v4/letter/m/f9ae1b/32.png) [@Morgenstern](https://boards.straightdope.com/u/Morgenstern)\ **Post date:** [June 30, 2016, 4:50pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/35 "2016-06-30T16:50:10Z")
> [@DSeid](#): > > Yes, but again, not all independent events. > > Even if somewhere north of 80% probable not. > > Take [Wang’s current status with Trump just +2 map](http://www.270towin.com/maps/princeton-election-consortium-trump-outperforms-polls). Give Trump the three toss-ups of IA, NH, and FL. Then have him just flip ME as an any given Sunday (Clinton is relatively underperforming there) and it is 269 to 269 with the House deciding. Have him instead flip VA (in four-way currently only +3) and it’s a Trump EV win even with a popular vote loss. > > The agregated popular vote seems (at this point) sticky at 5 to 7 Clinton in the face of Trump being just about as awful as can be; is it just as sticky in the other direction if he starts campaigning better, gets funding, and Clinton has some significant gaffes? I hope so. (And even more hope to not find out.) Are we seeing the same thing here? According to that cite, Clinton has 3 more EVs than she needs to win POTUS. (273) ETA, elsewhere on that site; State polls: Clinton 330 EV, Trump 208 EV ---
**Author:** ![Pleonast](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/pleonast/32/1183_2.png) [@Pleonast](https://boards.straightdope.com/u/Pleonast)\ **Post date:** [June 30, 2016, 6:22pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/36 "2016-06-30T18:22:57Z")
When will 538 publish its Congressional predictions? ---
**Author:** ![DSeid](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/dseid/32/20194_2.png) [@DSeid](https://boards.straightdope.com/u/DSeid)\ **Post date:** [June 30, 2016, 9:50pm UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/37 "2016-06-30T21:50:24Z")
> [@Morgenstern](#): > > Are we seeing the same thing here? According to that cite, Clinton has 3 more EVs than she needs to win POTUS. (273) … Not sure what is confusing you. Yes Trump just +2 over current polling by itself gets Clinton down to 273 if she does not win any of the three toss-ups, and if VA flips too (it is currently Clinton +3 and would be Clinton +1 in that scenario) then Trump wins. Make sense? ---
**Author:** ![Lantern](https://avatars.discourse-cdn.com/v4/letter/l/d26b3c/32.png) [@Lantern](https://boards.straightdope.com/u/Lantern)\ **Post date:** [July 1, 2016, 6:22am UTC](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566/38 "2016-07-01T06:22:04Z")
> [@DSeid](#): > > So as to how 538 comes up with PA so much more strongly Clinton than recent polls state … the method uses all the polls and uses recency as only one factor, along with size and track record of the house. An older Marist poll, two months old (!), is weighted most strongly because it was large and the house is highly regarded. That two month old poll was Clinton +15 and offsets more recent data showing a closer race in the 538 model. Interesting. I am guessing that timeliness is given a higher weight the closer you get to the election. Perhaps they have data indicating that June polls don’t have significantly more predictive value than April polls and that poll quality matters more. [Previous page](https://boards.straightdope.com/t/when-does-fivethirtyeight-start-projecting-the-presidential-race-based-on-polls/757566.md?page=1)