A fun post-mortem of the polls..


Or rather, the pollsters.. Specifically, I’ll focus on 2, one from “each side of the aisle.” The assumed “left-leaning” Public Policy Polling (actually, in looking at the last 3 elections, “left-leaning” really means “accurate” in PPP’s case) and our “right-leaning” old buddy Scott Rasmussen, the holy grail of right-wing poll cherry-pickers. I’ll stick to national and swing-state polls, because it’s irrelevant to see how close they were in Kansas and Connecticut.

I’ll also add Nate Silver’s projections as proof of his surreal god damn sorcery.

**PLEASE NOTE** As of my writing this blog, there may be some ballots yet to be counted in Washington State, Oregon, California, and a few other states, which could nudge Obama’s margin up slightly.

Anyhoo…. not even going to mention Gallup, who as I previously displayed effectively removed huge swaths of voters under 30 from their “LV” sample. So let’s have a gander, shall we?

I’m going to hold off excessive commentary and just focus on the numbers. As (hopefully) we’ve learned, the numbers speak for themselves, no matter how my right-wing friends bravely and futily fight them with unyielding spirit.

Another note: As of writing this, Rasmussen’s site has made their final presidential tracker nearly impossible to find. LOL.. good move Scotty. 

USA!! USA!! Time for the voice of all Americans to be heard to pick the next President.. unless you live in one of the 39 states where the race won’t be close enough to even bother polling..

Final projections/actual tallies:

PPP National projection: Obama 50/Romney 47

Rasmussen National projection: Romney 49/Obama 48

Nate Silver projection: Obama 50.8/Romney 48.3

Actual result:   Obama 50.4/48.1

PPP Verdict:  CORRECT

Rasmussen Verdict: INCORRECT 

Notes: PPP didn’t publish fractions on this poll, rounding either up or down, so I can’t say for sure how close they were to the tenth like with Nate Silver.. but on first look they may have under-estimated Romney’s national support by ~ 1 point. And then there’s good ol’ Rass.. like Gallup, giving Romney supporters some false hope, since EV/PV splits are rare.

Eagles-Fans

Pennsylvania: Home to FUCK YOU, SANTA…

Pennsylvania:

PPP: Obama 52/Romney 46

Rasmussen: Obama 51/Romney 46

Nate Silver: Obama 52.5/Romney 46.6

Actual result: Obama 52/Romney 46.8

PPP Verdict:  CORRECT

Rasmussen Verdict: CORRECT

Notes: Can’t get much closer to “nailing it” than this for both sides. Perhaps another slight underrating of Romney support, irrelevant to final tally. And good job Rass!!.. Maybe he did better than ’08 and ’10…. maybe..

Ohio: Did Romney remember this was a critical swing state with lots of auto- …aww fuck it. Sick of hearing about this place.

Ohio:

PPP: Obama 52/Romney 47

Rasmussen: Obama 49/Romney 49

Nate Silver: Obama 51.3/Romney 47.7

Actual Result: Obama 50.1/Romney 48.2

PPP Verdict: CORRECT

Rasmussen Verdict: INCONCLUSIVE

Notes: Romney outperformed a lot of polls in Ohio, including perhaps his own internals. However it was irrelevant to the final numbers, Karl Rove’s protests notwithstanding. At this point one could argue a very slight bias in PPP’s polling towards Obama, but not nearly enough to make an incorrect prediction. As for Rass, well I’ve decided that when a pollster predicts a tie, unless they were off by > 2%, it’s an inconclusive prediction. But in this case Rass’s Ohio poll was very misleading for the entire national picture, as it showed arguably the most important swing state in a dead-heat, when basically all other polls showed a small but steady Obama lead. Let’s see if it got better for Rass. (spoiler: it didn’t) 

Virginia: Governor forced ultra-sound? check. Heavy handed-state response to women protesting? check. Chance of Romney winning VA? Oh…

Virginia:

PPP: Obama 51/Romney 47

Rasmussen: Romney 50/Obama 48

Nate Silver: Obama 50.7/Romney 48.7

Actual Results: Obama 50.8/Romney 47.8

PPP Verdict: CORRECT

Rasmussen verdict: INCORRECT

Notes: If you stopped paying attention to the polls in early October, this result might have surprised you. If you were paying attention the last week or so, it didn’t. PPP obviously wasn’t surprised. Once again, highly accurate with the slightest underrating of Romney support. As for Rass, this was an epic fail. One that should really make people question his credibility as a pollster. You’re talking about a critical swing state that he got wrong by 5 points, all in the direction of Romney. He’d get a pass if it was his only bad miss of the night, or if there isn’t a history of this with his swing state polls. (spoiler: it wasn’t, and there is)

Michigan: According to friends who have lived there, this more or less sums up Michigan..

Michigan:

PPP: Obama 52/Romney 46

Rasmussen: Obama 52/Romney 47

Nate Silver: Obama 53/Romney 45.9

Actual Results: Obama 53.8/Romney 45.3

PPP Verdict: CORRECT

Rasmussen Verdict: CORRECT

Notes: Well, so much for the PPP-Obama bias. As you may recall, a few astro-turf polls came out in the waning days showing MI tied or Romney in the lead. This concerned the Obama camp so much David Axelrod put his mustache on the line to shore up support. And chalk up another good call for Rass. 2 so far.

Wisconsin: As opposed to commonly held perceptions, the people pictured do not represent the state as a whole

Wisconsin

PPP: Obama 51/Romney 48

Rasmussen: Obama 49/Romney 49

Nate Silver: Obama 52.4/Romney 46.9

Actual Results: Obama 52.8/Romney 46.1

PPP Verdict: CORRECT

Rasmussen Verdict: LOL

Notes: et tu, Paul Ryan? And with that, the PPP-Obama bias theory is dead. You know, categorizing PA, MI, WI, and MN (actually, I’m not even gonna bother with MN) as “swing states” is kind of misleading. They were only deemed so because of Rasmussen and the Romney Campaign and media memes, who desperately wanted this to be a neck-and-neck horse race. In these states it really wasn’t. No idea what Rass was looking at here, but it’s another huge, honking miss… this time by nearly 7 points!! Ouch.

Iowa: Where Obama kicks off and ends his presidential campaigning, and Howard Fineman is reminded of his childhood in post-Civil War America

Iowa

PPP: Obama 50/Romney 48

Rasmussen: Romney 49/Obama 48

Nate Silver: Obama 51.1/Romney 47.9

Actual Results: Obama 52.1/Romney 46.5

PPP Verdict: CORRECT

Rasmussen Verdict: FACEPALM

Notes: And now the bias would seem to have flipped, with PPP overestimating Romney’s support in some of these states. Which of course probably means there’s really no bias at all, and fluctuations are merely statistical noise. The most interesting thing to me in the closing days of the race in terms of Iowa, was when the great Howard Fineman flipped his shit over the Des Moines register endorsing Romney, which inexplicably led Fineman to muse that it was a sign Obama could very well lose the whole damn election. Good old Howard, hearkening back to a day when newspaper endorsements could really influence an election…. like maybe Grover Cleveland in 1888..

And.. oh Rasmussen.. Rasmussen, Rasmussen, Rasmussen. Shall I compare thee to a Summer’s Eve?

New Hampshire: Where clearly, not many fucks are given

New Hampshire

PPP: Obama 50/Romney 48

Rasmussen: Obama 50/Romney 48

Nate Silver: Obama 51.4/Romney 47.9

Actual Results: Obama 52.1/Romney 46.5

PPP Verdict: CORRECT

Rasmussen Verdict: CORRECT

Notes: Not much to say here, as both polls got the winner right, but by 4 points worse for Obama. Romney spent a lot of time campaigning here in the final week, as part of a strategy that would have also required picking up a whole bunch more states he had no real shot in. Or as his campaign called it: “Expanding the Map.”

North Carolina: Yea, we’re still the Deep South

North Carolina

PPP: Obama 49/Romney 49

Rasmussen: Romney 52/Obama 46

Nate Silver: Romney 50.6/Obama 48.9 

Actual Results: Romney 50.6/Obama 48.4

PPP Verdict: INCONCLUSIVE

Rasmussen Verdict: CORRECT

Notes: North Carolina has rapidly changing demographics, and is no longer safe GOP territory. Romney had to fight here, and it was a critical state for him to win, not for Obama. Rass got it right, but overestimated Romney by 4 points. Definite pattern emerging.

Oh, and look at Silver’s #s. Holy fuck.

Florida: Trust me, it’s much better to visit than to live here

Florida

PPP: Obama 50/Romney 49

Rasmussen: Romney 50/Obama 48

Nate Silver: Obama 49.8/Romney 49.8

Actual Results*: Obama 49.9/Romney 49.3

PPP Verdict: CORRECT

Rasmussen Verdict: INCORRECT

*as of 11/8/2012 at 2:14 pm CST, Florida has not officially called the state for Obama. And I mean.. really..why should they? It’s only November 8th

Notes: Governor Rick Scott shortened the state’s 14 early voting days down to 8, because… yea well, you guess. South Florida voters responded by showing up, waiting in line for upwards of 6,7,8 hours, and extending a solid middle finger in Governor Scott’s face in the form of a vote for Obama.

And once again.. Nate Silver…..                  sorcery

Colorado: Rocky Mountain High

Colorado

PPP: Obama 52/Romney 46

Rasmussen: Romney 50/Obama 47

Nate Silver: Obama 50.8/Romney 48.3

Actual Results: Obama 51.2/Romney 46.5

PPP Verdict: CORRECT

Rasmussen Verdict: GOOD GOD MAN

Notes: Colorado is one of those states that was very conservative years ago, and some pollsters just kinda refuse to accept it’s now much more young, much more liberal, and much more Hispanic. Pollsters like Rass, who really did another bang-up job here, missing by a whopping 7 points.. again..  apparently saw fit to find a “likely voter” model analogous to the 1980’s. Keep in mind Obama won this state in 2008 by 9 points. At a certain point there’s ignorance and being behind on trends, and then there’s willful ignorance.

Nevada: We get polled a lot now

Nevada

PPP: Obama 51/Romney 47

Rasmussen: Obama 50/Romney 48

Nate Silver: Obama 51.8/Romney 47.3

Actual Results: Obama 52.3/Romney 45.7

PPP Verdict: CORRECT

Rasmussen Verdict: EVEN A BLIND SQUIRREL..

Notes: Let’s leave this on a semi-positive not for our buddy Scott Rass.. Sure, he was off by nearly 5 points in another key swing state, but at least he got the winner right. If you are interested in the changing political landscape in Nevada, and why pollsters like Rass and Gallup have no clue how to poll there, check out Ralston Reports.

Bottom line: the anti-Obama crowd stayed confident, arrogant, sometimes downright smug over all the good news that Scott Rasmussen delivered to their Twitter feeds and inboxes with his polls. This same crowd inevitably, and this went right up through professional “pundits” had themselves good laughs dismissing PPP and Nate Silver. It seemed odd to me, because it was very easy to research each’s respective track records for recent elections. And if you did, then these results aren’t terribly surprising, although Rass was even more off than 2008 and 2010, and Nate Silver even more accurate.

Now once the shoe is on the other foot, and PPP and Nate Silver start showing us liberals #s in an election that doesn’t look so good for our candidate, will we substitute reality and inject our own, like our friends of the other political persuasion, or will we accept this wicked partisan sorcery (solid math and fundamentals of statistics) as truth? We shall see..

Advertisements

Gallup doesn’t think I’m going to vote, because I’m a flaky young whippersnapper


Oh yea.. you’re gonna vote Mr. “young urban professional?” Fat chance, kid.

 

 

 

 

 

 

According to Gallup, I am not a “Likely Voter.” 

 

 

 

 

 

 

 

 

 

Let’s take a look as to how, and why (my comments in red, Gallup’s stuff in white):

 

Gallup scores respondents one point for each question they answer in a way consistent with voting (the scoring scheme is detailed in subsequent paragraphs), resulting in overall likelihood of voting scores ranging from zero to seven. Gallup has then used various procedures to set a threshold for the pool of likely voters. The validity of setting a threshold based on a specific estimated turnout among the voting age population (VAP) or voter eligible population (VEP)is less clear than it was in the past, particularly given real-world changes in voting such as early voting and decreases in survey participation rates. For the 2012 election, Gallup currently considers respondents with the highest scores (six or seven) to be likely voters.

Questions Gallup Uses in Its Presidential Election Likely Voter Model

For the seven questions that make up the likely voter scale, respondents get one point on the likely voter scale for each question to which they give the response listed in parentheses (with a maximum of seven points possible). See the full question wording for each question in the “Question Wording” section.

  1. Thought given to election (quite a lot, some)

Obviously a point for me here

  1. Know where people in neighborhood go to vote (yes)

This is an asinine question for the millions of young, (like under 50) urban voters. Me, like most working Americans, has no need to know where my polling place is before election day. Add to that I lived somewhere else in 2008. Why would I need to know and have mapped out a route to where “people in my neighborhood go to vote?” Like pretty much everyone under 60, on election day I will plug the address into my GPS, and go. I don’t need to case the joint before-hand. – no point for me

  1. Voted in election precinct before (yes)

No. Newsflash Gallup. Young people are mobile. They move. Wayyyy more people vote in presidential elections than locals or primaries. Another ridiculous way of screening out shitloads of voters and not calling them “likely” – no point for me here either

  1. How often vote (always, nearly always)

Again, what are they referring to exactly? Municipal races? Local dog catcher? American Idol? Without context I’d have to assume they mean in general, and the answer for me wouldn’t be always or nearly always. Maybe.. kinda often? When I don’t know much about local candidates, I typically don’t vote. And guess what? A shitload of Americans, particularly younger Americans, do the same thing.  – no point for you, jerk-ass

  1. Plan to vote in 2012 election (yes)

Interestingly, my yes here ends up meaning no, according to Gallup.. keep reading. But – at least I get one more point for now

  1. Likelihood of voting on a 10-point scale (7-10)

Yes. 9.999 (barring catastrophic injury/illness/natural disaster – one point

  1. Voted in last presidential election (yes)

Yes – one point

That’s a total of 4, for those of you playing at home.. here are some qualifiers for the very young kiddos listening to Animal Collective and Barry Manilow (ironically): 

Gallup also measures the increasing trend of people voting before Election Day by asking people when they plan to vote and then considers this information when determining if people are likely to vote.

For the raw scores, Gallup makes the following adjustments:

  • Respondents who are not registered to vote receive a score of zero.

At this stage of the game, understandable. Most states are past the deadline now.

  • Respondents who do not say they plan to vote (see item No. 5) receive a score of zero.

Fair enough as well.

  • Respondents who report they already voted receive a score of seven.

Well duh. But I’m betting most people who already voted probably won’t bother with a survey at this point. Just a hunch.

  • Given the rise in voting by mail, respondents who say they do not know where people in their local district go to vote receive credit for that question if they say they plan to vote before Election Day and say they have voted in their precinct in person or by mail in past elections.

In other words, old people who likely answer would yes to both questions anyway. Again, I’m punished because I lived in another district in 2008, and I haven’t cased out my polling place yet. Does that really make me an anomaly? Gallup thinks so.

  • Gallup adjusts younger respondents’ scores to account for their ineligibility to vote in some or all past elections. In other words, even though the model identifies voters based on past voting history, Gallup does not penalize younger voters for not being of voting age in past election years.

Well, that’s good to know. But it also tells me Gallup is again grossly undercounting youth votes. I think plenty of 22-26  year olds who didn’t bother to vote last time around will be more inclined to now.

  • If aged 18 to 19, Gallup converts their scores as follows: 1=2, 2=4, 3=5, 4=7

That seems like a good approximation.

  • If aged 20 to 21, Gallup converts their scores as follows: 1=1, 2=3, 3=4, 4=6, 5=7

Yea ok neat-o.. but that doesn’t affect me. The 30-something who wasn’t in this district 4 years ago and will find my polling place via Google Maps or GPS. (or more likely I’ll know exactly where it is, I just haven’t looked at the address yet, because I’m mobile and am sure I won’t have difficulty getting to it)

 

So in short, I am not counted as a “Likely Voter” by Gallup. Not even close (I have 4 points, needed 6 or 7) And why?

1)      I’m new to the area and haven’t voted in this precinct before. That’s a point off for just about every single undergrad (and grad) student in the country, actually. Not to mention the millions who have moved in the past 4 years because of the economy and real estate crisis. And forget about me who moved to a new state, this potentially counts you out if you just moved a few miles away in your same town.

2)      The morning I go to vote, I’ll get the address off my voter registration card and head to my precinct to vote. I haven’t mapped out how to get there, or planned my day or week around finding it. I live in a big city, am mobile, and have a GPS on my phone, and an internet connection to use any number of map services. Is this unusual? Are young working people really penalized or considered “unlikely” if they haven’t written out their Election Day plans beforehand?

 

Bottom line: I vote in presidential elections, and I’m sure as hell not sitting this one out, even though there’s no way my state is going blue (yet.. give us a few more cycles) Yet Gallup doesn’t think so, for the reasons stated above. I bring this up because I just don’t think my situation (new to district, doesn’t always vote in local races, hasn’t yet mapped out where polling place is and only will on or night before Election Day) is all that unique for young professional adults or college students. And that group, as you know, overwhelmingly leans to Obama. This is my theory on why Gallup’s LV model may be fucked.