STATE POLLING
New polls today from Pennsylvania (Clinton +1), Utah (Trump +6, Trump +12, Trump +11), Georgia (Trump +1), New Hampshire (Trump +5, tied), Florida (Clinton +4), Colorado (Clinton +6), and Texas (Trump +9, Trump +14).
Election Day is Tuesday. We know this. So do the vast majority of voters, 98% or so of whom have apparently made up their mind. So why are the polls jumping all over the place? Why are we seeing such swings day-to-day when there realisiticaly just aren't enough undecided voters out there to affect such changes?
Steven Shepard at Politico has a theory: when one candidate dominates headlines in a negative way, the supporters of that candidate are more likely to decline to take part in polls. It's an interesting theory, but my own take is harder to solve: I believe polling itself is broken.
The sharp decline in landlines, the less likely people are to even answer unknown numbers, the often sharp differences in how pollsters set up their preferential calculations (percent of each group as represented in the overall sample), and so forth, have all led to increasingly inaccurate polling.
One of the reasons sites like this use cumulative polling is to try and average out the swings. And many sites, like this one as well, use weighted averaging, which gives more "strength" to recent polls and less to older ones, under the theory the newer the poll, the more representative it us of the current status of the state. Yet we've seen often wild swings even with our cumulative weighted modeling.
How can a candidate in three polls over just two days be up five, tied, and up one? Or up six and twelve on the same day? Those are statistically huge swings (and those were just from today's polls). I sincerely doubt there are that many voters in New Hampshire who in less than two weeks went from a +12 for Clinton to a +5 for Trump. A seventeen point swing in less than two weeks isn't a sign of a huge change of heart, it's a sign (at least to me) of a major flaw in polling. And I never bought in to Texas being a swing state, at least not this year.
So, what can be done about it? I'll be honest: I have no idea. I'm not a pollster (though I've learned an awful lot since I started this blog a few months ago); I'm a statistics geek with a dusty political science degree who wanted to see if I could develop a better algorithm than was being used by other sites. Arrogant? A bit. Is the algorithm working? I won't know until next Wednesday when all the results are in. But as to fixing the problem...that's simply out of my area of expertise.
However, one thing I do know: there will be an awful lot of experts (both actual and armchair) who will be spending and lot of time (and money) trying to solve the polling problem. Until then...who really knows if any of the numbers I've been crunching for months have any basis in reality.
As for the predictions themselves, I just have to go with the data I am presented. And, as such, there are for changes to map today: Texas moves from "likely" to "solid" Republican, Utah moves from "leaning" to "likely" Republican, Georgia moves from "likely"to "leaning" Republican, and New Hampshire moves from "likely" to "leaning" Democratic.
Currently, Clinton has a projected overall lead of 316 to 222. Here are the current averages from the battleground states:
Likely Republican
Nebraska (CD2): Trump up by 6%
Utah: Trump up by 5.9%
Alaska: Trump up by 4.6%
Leaning Republican
Georgia: Trump up by 3.4%
Arizona: Trump up by 2.8%
Maine (CD2): Trump up by 2.4%
Iowa: Trump up by 2%
Nevada: Trump up by 0.8%
Ohio: Trump up by 0.7%
Leaning Democrat
New Hampshire: Clinton up by 2.1%
North Carolina: Clinton up by 2.5%
Florida: Clinton up by 2.7%
Pennsylvania: Clinton up by 3.1%
Colorado: Clinton up by 3.9%
Likely Democrat
Virginia: Clinton up by 4.3%
Michigan: Clinton up by 4.9%
Wisconsin: Clinton up by 5.1%
New Mexico: Clinton up by 6.4%
Here are the State Polling Averages for all 50 states and the District of Columbia.
TOP POLITICAL HEADLINES
Associated Press -- Voter intimidation arguments head to federal court
Democrats will argue before a federal judge in New Jersey on Friday that Republicans are coordinating with GOP presidential nominee Donald Trump to intimidate voters, charges that the Republican Party says are not true in either the federal case or in the four other states where Democrats are waging similar battles.
CNN -- Democrats, Trump lawyers going to court early -- and often
As Democrats warn their supporters of voter suppression efforts and Donald Trump urges his followers to watch polling areas, lawyers from both sides are filing fiery legal briefs and laying the groundwork for potential challenges on Election Day.
Fox News -- Philly transit strike could cause Clinton, Democrats headaches
An ongoing transit strike in Philadelphia is raising concerns among some Democrats that the stand-off could depress voter turnout in the battleground state's deep-blue stronghold – but union officials say the agency is using Election Day as a scare tactic and unfair bargaining chip.
Politico -- Trump has a path, and if the polls are wrong, it’s wider than thought
Hillary Clinton leads in most national polls, and in enough battleground states to put her on pace to surpass the 270 electoral votes she needs Tuesday to become the next president. But not far beneath the surface, as Donald Trump has narrowed the gap following the late-breaking FBI announcement of a renewed review of emails related to her private server, lurks a question making Democrats squirm in these frenzied final days.
The Hill -- Four days to go: What to watch for
The presidential campaigns are sprinting for the finish amid tightening polls that show new opportunities for Donald Trump as he seeks a late comeback against Hillary Clinton. "If the Cubs can do it, he can too," one Republican told The Hill.
No comments:
Post a Comment