It’s been a week since my first post on this, so it’s tie to check back in on what the polls — and all those polling aggregators with their algorithms and indicators and ‘leans’ and ‘likely’ and ‘tossup’ states — are telling us.
And, because each has their own unique approach to integrating and summarizing the now-ceaseless stream of political polls (otherwise, all their calculations would look the same), it’s probably also time to take a closer look under the hood and see how they differ. One lesson to emerge from this should be that no one should take a single poll as a clear picture of anything, whether it suggests imminent doom or glorious triumph or a nail-biting dead heat. But polls can be very good indicators of the current state of affairs, and the more of them we have to integrate and contemplate, and the more methods we test for doing so — the more confident we should feel about what they are telling us. So, without further ado:
Electoral-vote.com (EV): This site uses only polls from nonpartisan pollsters. It always uses the most recent poll and, if there are others completed within the last week, includes those and averages them all together. So only the most recent polls are used, but in a rarely-polled state the polls may be weeks or months out of date. If no polling is available for a race, result from the most recent election are used as placeholders.
Pollster (Huffington Post): This site uses every available poll that claims to have a representative sample, but will exclude those that do not disclose their sampling techniques, sample size, and other key methodological information. A trendline is calculated through the polls that shows the current percentages for each candidate and how they have changed over time.
RealClearPolitics (RCP): This site takes the most recent available polls and creates an average or ‘spread’ showing the difference between the two candidates. If a state is polled infrequently, this average will be affected by polls that are weeks or months out of date, though.
Fivethirtyeight (538): Here I just want to consider the polls-only state-by-state forecast of vote percentages (538 also offers a nowcast, polls-only, and ‘polls-plus’ version of the national vote, and a % likelihood of winning as well as an estimated % of the vote). The site calculates a weighted polling average for each state, adjusted for things such as likely voters, the trend line, and ‘house effects' (systematic biased found for particular polling organizations, and then also adjust for demographics to predict the election day outcome.
Sabato's Crystal Ball: To be honest, I couldn’t find any general info on methods for this site. If someone has the link, please let me know.
There are more aggregators, such as Princeton Election Consortium, and I may get around to adding more into the table. Also, I may bet around to adding individual senate races from key states, as control of the Senate will greatly impact what the next President is able to achieve.
Now for the juicy part:
state |
|
EV |
538 |
pollster |
RCP |
sabato |
The Presidential Race in Swing States
co |
|
C, 41-29 |
C 49.3-38.8 |
C, 44.1-36.8 |
C +11.0 |
Likely D |
fl |
|
C, 44-41 |
C, 49.1-42.8 |
C, 45.6-42.7 |
C, +3.6 |
Leans D |
ia |
|
tie, 36-36 |
C, 48.0-42.2 |
C, 39.7-37.4 |
C, +0.4 |
Leans D |
nv |
|
C, 42-41 |
C, 48.8-41.8 |
T, 42.7-42.1 |
C, +2.3 |
Leans D |
nh |
|
C, 48-37 |
C, 49.4-40.4 |
C, 42.5-35.7 |
C, +8.2 |
Leans D |
nc |
|
C, 43-41 |
C, 48.9-43.6 |
C, 44.8-41.2 |
C, +2.0 |
Leans D |
oh |
|
C, 42-39 |
C, 48.0-41.8 |
C, 44.1-42.4 |
C, +2.6 |
Leans D |
va |
|
C, 46-34 |
C, 49.6-40.0 |
C, 43.7-37.4 |
C, +8.0 |
Likely D |
mi |
|
C, 42-32 |
C, 51.6-39.3 |
C, 42.1-34.2 |
C ,+6.6 |
Likely D |
pa |
|
C, 47-39 |
C, 50.0-40.7 |
C, 46.7-39.7 |
C, +9.2 |
Likely D |
wi |
|
C, 42-33 |
C, 51.3-38.9 |
C, 47.1-36.3 |
C, +9.4 |
Likely D |
az |
|
C, 44-43 |
C, 46.6-45.0 |
T, 45.6-45.1 |
T, +0.3 |
Leans R |
ga |
|
T, 42-41 |
C, 46.6-45.9 |
C, 44.4-42.5 |
C, +0.3 |
Leans R |
Here’s the story the numbers tell us now:
(1) In the first 8 states (consensus ‘swing’ states), Clinton leads by every measure in 7 of 8 and four of five in NV. Her position has strengthened in the crucial states (for Trump) of Ohio and Florida.
(2) In the 3 states (MI, PA, WI) identified by the Trump team as part of a ‘winning’ strategy to take the Rust Belt, the situation has gotten worse for Trump. Clinton has consolidated her lead to 8-10 points in all 3.
(3) In the new ‘battleground' states that used to be reliably Republican, AZ and GA, Trump also continues to struggle. Georgia in particular now appears to be moving toward Clinton.
On some sites, aggregate polling suggests that the race is within 3 points or less in additional ‘red' states, like South Carolina (yes, South Carolina). If this persists for another week or two, I will expand the list of battleground states. The mere fact that we are even discussing this shows how far this election has gone off the rails for the Republicans.
National polling: The national numbers, although less important, also show that Clinton’s lead remains solid at 7 — 8% over Trump. HuffPo’s Pollster has Clinton up 47.5 — 40.4; RCP has her up 47.7 — 41.0; and 538 puts her at 44.4 — 36.4 (with 9% for Johnson).
The Senate: Sabato puts the Senate split at 50-50. RCP does as well, if you use the “no tossups" option. EV still puts it at D 49, R 51. However, EV puts Indiana in the 'Safe R' because it refers back to the 2010 results, due to a lack of outside nonpartisan polling of this race. With Evan Bayh as the Democratic nominee, other sites now regard this as a tossup. Notably, everyone sees WI, IL, PA, and NH as Democratic pickups now. Nevada remains the only D seat in play. A 50-50 split produces a Democratic majority because VP Kaine would be the tiebreaker vote. But outright control is in reach with just one more seat flipping to the Democrats.