If we had a different electoral system, the question wouldn't be whether marginal candidates could effect the outcome. The question would be, would they still BE marginal?

The first past the post system combined with major party entrenchment is a large part of why they're marginal, after all.
 

Ah yes, the Andrew Cuomo swing vote.
 

So not only is the even-smaller-than-the-libertarians Constitution Party is going to be a spoiler throwing the election to Obama, but the reason they'll be a spoiler is that they were endorsed by Ron Paul?

Come on. Can we have some serious discussion of the issues, please?
 

Sandy:

I hate to wreck a fun illusion, but according to the latest AP polling:

The poll shows that while Obama has gained ground among Clinton's supporters — 69 percent view him favorably now, up 9 percentage points from June — this has yet to translate into more of their support.

In part, this is because their positive views of Republican presidential nominee John McCain have also improved during this period. Those supporting McCain have also edged up from 21 percent to 28 percent, with the number of undecided staying constant, the survey showed.

[In other words, the number of Hillary supporters committed to Obama is shrinking.]

Clinton backers' reluctance to support Obama helps explain why he is having a tougher time solidifying partisan supporters than McCain. Overall, 74 percent of Democrats say they will vote for Obama, compared to 87 percent of Republicans behind the Arizona senator. About nine in 10 Clinton supporters are Democrats.

McCain's pick of Palin has secured the GOP base. The Paul supporters fell in line during the convention. You did not hear a squeak from Paul there.

The actual story of party division is between the left and center of the Dem Party. AP has 28% of Hillary supporters, 90% of whom are Dems, going for McCain, with another 14% undecided.
 

The question, of course, is how many Paul-ites were going to vote for McCain, weren't going to vote for Barr, but now would vote for the Constitution Party guy. Hard to figure how to compute that. But we all remember what a seemingly small number of Nader voters did in 2000.

Bart: As I'm sure you kow, polls in the last week-10 days have shown a lot of movement to Obama, who is now in the lead in at least most polls and projections. It's still a close race and there's a long way to go, but even by your standards that was a toothless troll.
 

Overall, 74 percent of Democrats say they will vote for Obama, compared to 87 percent of Republicans behind the Arizona senator.

The Borg is always easier to keep in line; the RWA mindset is as good as those funny headpieces.

That's the price we lib'ruls pay for actually letting people think....

But I'm not worried; the Republican brand name itself is severely tarnished, and even McSame refuses to use the words "Republican" (or worse yet, "Bush"). But not to worry, the public can figure it out whose side he's on.

Cheers,
 

Prof. Levinson,

Baldwin is on the ballot in 37 states, Bob Barr in 45 states, and Ralph Nader in 46 states. Baldwin's total is a little misleading because it does not include California (10 per cent of the electoral votes). Watch Ballot Access News for updates to this information.
 

Re Elliot's comment: I for one regard as a "serious issue" our seriously defective system of electing presidents (and all other public officials inasmuch as we rely almost exclusively, save for a few outliers, on single-member first-past-the-post geographical districting). If he does not, I'm curious why he comes to that conclusions. (What message does he draw from the fact that almost no political system organized in the past 25 years has emulated our particular electoral system?)
 

Susan,

While the National Popular Vote plan might, in the long run, prove to be the right incremental step, it does not address Prof. Levinson's main point here, which is that vote splitting again (as in 1992 and 2000) threatens to elect the wrong candidate. Whether we get there indirectly via NPV or directly by constitutional amendment, we need a majority method rather than plurality.
 

Oh, come on DePalma! People can use statistics to prove anything. 14% of people know that!

Come now, you've tried your best, and failed miserably. The lesson is, never try.
 

jslater:

Bart: As I'm sure you kow, polls in the last week-10 days have shown a lot of movement to Obama, who is now in the lead in at least most polls and projections. It's still a close race and there's a long way to go, but even by your standards that was a toothless troll.

Actually, not.

Go to RealClearPolitics.com's assembly of horse race polling. Ignore all polling of registered voters because 40% plus of them will not vote.

You will then get two sets of "likely voter" polling. Rasmussen (tie) and Battleground (McCain +2) on one side and the Dem media polls CNN, CBS and Quinnipiac with outsize 4-5 point Obama leads.

Rasmussen and Battlegound ended the last cycle within a point of the final results. Indeed, Battleground is a joint poll by professional GOP and Dem pollsters. This is the first poll I consult when I want a realistic take. They both have realistic weights of Dems (around 39%) and GOP (around 33-34 percent). They are both far more stable over time. This race has essentially been tied or with a slight McCain lead since the GOP convention.

The Dem media polling is a joke that arrives at their totals by overweighing Dems and underweighting GOP and polling at times when conservatives are less likely to be at home. (Google "Shy Tory Factor.") Check their internals, if they both publishing them.

Indeed, the Gallup polling of registered voters (a category which has historically overstated Dem support) has a lower 3% lead than the Dem media polling's alleged "likely voters."

Something else that should be flashing red lights to even partisan observers is how polling like AP can show McCain splitting the independents which Bush lost and gaining more Dems than Bush could ever dream of and still somehow be trailing overall when Bush won by 3%. I know these people were probably taught the New Math, but anyone half awake might be excused for thinking that something is not adding up right.

We see this game being played every cycle since the 70s. If as a Dem news consumer it makes you feel better to believe these bogus Dem media polls (which BTW is their purpose), go right ahead.

This GOP voter is more than pleased with the results to date. I am waiting to see if the debates move the numbers and then I will offer my prediction with vote percentage projections.

Normally, these things are pretty predictable. Unlike congressional outcomes which flow back and forth between the parties, you can bank on the Dems falling between 46-49% starting with Bush I (absent the big Perot showing in 1992), tightening to an even more predictable 48-49% over the past three cycles. Thus, you simply need to peg the Dem vote at 48 or 49%, predict the small amount a third party might draw and then give the rest of the votes to the GOP candidate.

However, we are in far more uncertain territory this time around. The Dems are running the first African American candidate ever, the most liberal candidate since the 80s GOP blowouts and are the most split since 1968. On the other hand, the GOP brand is not exactly at its height of popularity and the GOP is running their first woman on the ticket.

The effect of running a very liberal African American is the biggest variable. Frequently, when liberal Dem (but not conservative GOP) African Americans run statewide races, the polling substantially overestimates their actual votes in what is called the Bradley or Wilder Effect.

[I am not going to speculate whether this effect is caused by (1) Dems who are not racist, but falsely tell pollsters that they prefer the African American candidate to avoid being thought of a racist, or (2) these Dems are racist and simply do not want to admit it. Once again, there is no evidence of this among GOP voters, who vote party line regardless of race.]

This effect has worked both ways for Obama during the primaries depending upon the percentage of African American voters in the state. In Southern states with a high percentage of African American voters, Obama gained more votes than the polling indicated. In the vast majority of other states with a lower percentage of African American voters, Obama's votes fell 3-7% below the polling.

The Dems have been privately assuming that the polls in overall and especially in the battleground states are overestimating Obama's lead by 3-6%. If this is the case, McCain is going to win in a walk and I cannot rationally determine the final percentages of the vote.
 

Jeez, DePalma, do I gotta tell you again what with the deal on the statistics??? People, learn a lesson from this man. He's teaching something very valuable; how to weasel out of things. It's what separates us from the animals... except the weasel.
 

Anyway, Prof. Levinson, if you'd taken a second to think this out, you would've realized that this endorsement is good news for McCain. The danger was that Paul might endorse Barr, a halfway legitimate candidate. This Baldwin character is (a) someone who no one has ever heard of, (b) running for a party no one's ever heard of, and (c) believes the government may be responsible for 9/11. Now granted, plenty of kooks think the same thing, so you might say, why that should stop Paul people from voting for Baldwin, but if you plotted Paulites and 9/11 conspiracy nuts on a Venn diagram, you'd probably only get 50/50 overlap. Also, this damages Barr.

And bart, the Bradley Effect is dead. I'm not going to give you the links, but look up Obama's performance in the primaries vis-a-vis his polls. He overperformed more often than not.
 

Bart: This isn't a polling site --I recommend fivethirtyeight.com for that -- but let's review the most recent national polls: CNN, Obama +4; Big Ten, O +1; CBS/NYT, O+5; Quinn. O+5; Gallup O+4; Rassmussen tied. In *every* case, there has been movement toward Obama in the past week-10 days. Cherry pick what you want, but calling several pollsters "Dem biased" but not noting the biases of RCP, for example, is silly. Recent state polls -- especially from the COL, NM, IOWA trio Obama nees to win -- look good for Obama too.

But why rely on me? Look at the move toward Obama on Intrade.com in the last week or so, from under 50 at the height of McCain's convention bounce back over 52 today. Even today, Obama is going up, McCain is going down.

It's still a close election, and either side could win, but Obama is ahead at the moment, and McCain and Palin have both, in their own ways, hurt themselves a lot recently. Note Palin's plumetting approval numbers.

Again, this isn't the time or place for a long discussion of polls, so I'll let the numbers and markets speak for me, and return you to your regularly scheduled topic.
 

Tray, I suspect you're right about the potentially greater importance of Paul's endorsing Barr instead of the hitherto unknown (to me at least) Chuck Baldwin. I do wonder, though, if Paul's public rejection of McCain (however unsurprising) might not also be at least a little boost to Barr, who could plausibly claim that Paul picked Baldwin in part because Baldwin had been a Paul supporter in the primaries, which I don't think was the case the Barr.
 

Oh, you're entirely right about that - but he rejected McCain a couple weeks ago at the convention. They called him, he told the press he turned them down. So that's old news. That being the case, this endorsement is preferable (from McCain's perspective) to about any endorsement he could've given. Unless he'd endorsed Cynthia McKinney, of course, but that would be hoping for too much.
 

"Bart":

How'd your "predictions" and "analysis" go in 2006?

Cheers,
 

jslater said...

I recommend fivethirtyeight.com...

By far the best of the lefty sites. His methodology is sound, but he uses too many garbage inputs from registered voter polls. Garbage in, garbage out.
 

Garbage in, garbage out.

# posted by Bart DePalma : 5:13 PM


Physician, heal thyself!
 

This comment has been removed by the author.
 

How'd your "predictions" and "analysis" go in 2006?

Since you asked, Arne, I thought I would look it up.

[I]f you want to save yourself yet another cycle of heartbreak, I would not assume that the Donkeys are even taking the house. The media polls consistently oversample Donkey respondents who never show up at the polls. I see no evidence in actual voting in special elections and primaries of increased Donkey turnout. In contrast, the Elephant GOTV effort seems to be at or above 2004 levels. Early and absentee voting, which Elephants disproportionately use, is up from previous elections. If the media polls keep tightening, the Elephant GOTV effort will most likely add 3-5 points onto what the Elephant candidate is getting in the final media polls make this a status quo election.

If I had to guess, the Elephants will lose 8-12 seats in the House and 2-3 in the Senate. If either Guliani or McCain run as the GOP candidate in 2008, the GOP should recover all of those losses.

-Bart DePalma, 10/30/06
 

Mr. DePalma,

By far the best of the lefty sites. His methodology is sound, but he uses too many garbage inputs from registered voter polls. Garbage in, garbage out.

likely voter metrics are not more reliable than a registered voter method. see http://www.fivethirtyeight.com/search/label/likely voters for nate and sean's reasoning.

its debatable, and utterly dependent on the quality of the pollster's likely voter metric, but your blanket assertion is incorrect. "likely voter" polls, depending on their methodology, are susceptible to much greater inaccuracies depending on the pollster's assumptions. in the ideal world, the likely voter model should add nuance to the predictions, but the adjustment is subjective and somewhat arbitrary, and not necessarily reflective of reality. Though they might be.
 

Professor Levinson- did you see the Washington Times article on the possibility of a tie in the electoral college. I imagine that such a scenario, should it occur, will win you some converts to the "Constitution is broken" movement.
 

nerpzillicus said...

BD: By far the best of the lefty sites. His methodology is sound, but he uses too many garbage inputs from registered voter polls. Garbage in, garbage out.

likely voter metrics are not more reliable than a registered voter method...its debatable, and utterly dependent on the quality of the pollster's likely voter metric, but your blanket assertion is incorrect.

My blanket assertion was that polling of registered voters historically overstates actual voter support for Dems and is per se inaccurate.

I have no disagreement with the statement that "likely voter" polling can be even more inaccurate than registered voter polling. See my lengthy 3:27 PM post explaining how the Dem media polls of "likely voters" are cooked by over weighting Dems and under weighting the GOP to come up with a result that overstates Dem voting even more than the Gallup tracking poll of registered voters.

However, the fact that "likely voter" polling can be cooked does not mean that polling of registered voters is accurate. Even Gallup admits this and switches over to "likely voters" when the election is imminent. Their excuse is that there isn't much difference between registered and likely voters until the election is imminent, which is humorous given that their monthly likely voter polling consistently shows lower support for Dems than does their tracking polling of registered voters covering the same dates. In reality, Gallup is catering to their Dem media clients with the registered voter polling running up to the election, but Gallup maintains their reputation by pointing to the final "likely voter" polling on election eve, which is generally pretty accurate.
 

nerpzillicus:

For a perfect example of a cooked Dem media poll, take a gander at the really egregious Wash Post/ABC News poll released today, which has a weighted sample with party members and leaning independents 54% Dem and 38 GOP to - surprise, surprise - end up with a 52% to 43% Obama advantage.
 

If I had to guess, the Elephants will lose 8-12 seats in the House and 2-3 in the Senate. If either Guliani or McCain run as the GOP candidate in 2008, the GOP should recover all of those losses.

-Bart DePalma, 10/30/06

Baghdad, what cooked poll did you use to make that prediction?
 

Bart:

So let's review. The new ARG poll today shows Obama +2 -- that's a 5 point movement from the last ARG poll, which showed McCain up 3. As you noted, the newest ABC/Post poll has 0+9; today's Rasmussen has O+2 (up from tied yesterday); and as noted earlier, the most recent CNN has 0+4, CBS/NYT 0+5, and Quinn O+4. See also fivethirtyeight.com for good news for Obama in yesterday's state polls as well.

Your position is that ALL this good news for Obama shows Dem bias, but the one poll you look at is THE ONE?

Also, whoever you think is "winning" at this time, my point was that there had been real movement away from McCain and toward Obama in the last week-10 days, and I believe every single poll bears that out.

Also, polls show Palin's approval numbers sinking.

It's still a close election and could go either way, but to deny that Obama has risen into a small but real lead lately is to deny reality.

Here's another hint that Repubs understand they are losing. They are going off about "biased media." That's what parties -- both -- always do when they think they are behind.
 

My thanks to Bob Richard for the information about ballot access. An electoral vote tie presumably produces an Obama victory, given that Democrats will almost certainly control the delegations of 26 states (including, of course, Vermont, whose one representative has equal voting power with the entirety of Texas's majority-Republican (presumably) delegation). Query, though, if one elector would choose to vote for the popularl vote winner in order not to produce a tie (and, therefore, not make the US a further laughingstock in the entire world for the idiocy of its electoral college system).
 

Mr. DePalma-

For a perfect example of a cooked Dem media poll, take a gander at the really egregious Wash Post/ABC News poll released today, which has a weighted sample with party members and leaning independents 54% Dem and 38 GOP to - surprise, surprise - end up with a 52% to 43% Obama advantage.

Consistent with your history here, you are changing the subject. Assuming arguendo that you are correct that the weighting in the WAPO/ABC poll is egregiously tilted, that is not evidence of the relative reliability of "likely voter" models vs. registered voter models. I personally think weighting by party ID is extraordinarily subjective and difficult. That said, denying the structural advantage for dems in party ID this year is tantamount to an annulment from reality. gut says it should be about 5-7 points, but really, i have no idea.

My comment to you was regarding your statement that fivethirtyeight was "garbage in, garbage out." my retort is that the registered voter model is not necessarily worse than the likely voter model. indeed, for many of the reasons from nate and sean, this year, the registered voter model may be more accurate, and most certainly more conservative due to extrinsic factors that are, quite honestly, anyone's guess. Most importantly, nate and sean have said they will switch to a likely voter model as the election gets closer, because, at that point, those models become more accurate. however, at this point in an election, the registered voter polls are historically more accurate.

The main problem is that all (i think) likely voter models skew older, thus skewing mccain. historically, with low young voter turnout, this is a reasonable assumption. however, this year, the evidence demonstrates that the youth vote may not be disproportionately low. If this is the case, most likely voter models will inaccurately give mccain 1-2 points. if history holds up, then perhaps they will be accurate.

another factor in the likely voter model is past history. once again, historically, a voter who has not voted before is a likely reasonably safe candidate to forego voting at an election. this year, however, there is once again evidence to show that enthusiasm is up among first-timers, who tend to skew democrat. this assumption in the likely voter model also gives mccain additional credit.

add in the cell phone factor, and well, there is simply a lot of noise in the system.

therefore, you are incorrect in stating that registered voter models are per se inaccurate. if one understands the inherent limitations of the model, one can adjust for them. most likely voter models are done in house, and the methodology is unknown to outside analysts. that being the case, there is no good way of accounting for any of the bias inherent in them. nate and sean can take the registered voter results, and apply their own regressions on those numbers, therefore yielding results they think most closely resemble reality. performing the process on likely voter results from pollsters is actually a lot closer to your "garbage in, garbage out" comment than with registered voter results, due to the "known unknowns" in the numbers.

there are a lot of factors that weaken either model. the biggest is that, while history does show certain trends, each election is its own function. the danger of the registered voter model is that it does not accurately reflect who will actually go to the polls. the danger of the likely voter model is that it bases its assumptions on historical data that may not be applicable to this election. more significantly, most of the time, a third party analysts cannot see the assumptions made in the likely voter model, and test to see if those assumptions are grounded in the current reality. there is doubt in both approaches, but I'd be really careful about putting all your eggs in the likely voter basket. the strength of the registered voter numbers is that those numbers reflect the reality, and third parties can apply their own assumptions to them. in short, they are more accurate "raw" numbers. likely voter models have inherent assumptions built in, making them less usable for independent analysis. sean and nate know stats, and I have read their reasoning. I'm betting on them.



p.s. assuming no october surprises or a blow out performance at the debate, obama wins with about 322 EV.

from the best madden commercial of all time
"Man, that don't show my breakaway speed"
"Our numbers say you lost a step."
[Brooks shrugs, walks away disgusted]
"It's the numbers."
 

Bart wrote:

take a gander at the really egregious Wash Post/ABC News poll released today, which has a weighted sample with party members and leaning independents 54% Dem and 38 GOP...

Amusingly, if you take a closer look you see that the internals of this poll look much the same as every other iteration of the same poll. Going back to the first of the year, this poll has NEVER shown less than 50% Dems + Dem-leaning Independents. And it has NEVER shown more than 42% Reps + Rep-leaners, with the latter number under 40% more often than not.

In fact, the spread of 54-38, which Bart decries as "really egregious" and an obvious sign of a "cooked poll," is the exact same spread which was found in the June version of the same poll. That poll - with the ridiculous, egregious R/D split that could only indicate a cooked poll - gave McCain a 1-point lead.
 

Mr. DePalma,



For a perfect example of a cooked Dem media poll, take a gander at the really egregious Wash Post/ABC News poll released today, which has a weighted sample with party members and leaning independents 54% Dem and 38 GOP to - surprise, surprise - end up with a 52% to 43% Obama advantage.


with all due respect, you read polls like you read law - poorly. this is not a weighted poll, but a question presented to those claiming to be independent about which party they lean towards. I thought you were actually claiming the WAPO/ABC poll weighted the results this ways, which would be egregious. instead, as it should come as no surprise based on the political status of the country, independents are currently breaking democratic, which is the antithesis of egregious. the actual respondents are 38-28 dem/repub. this seems a little high, but the liberal/moderate/conservative numbers are in line with the average (maybe even a little low on the liberal side). probably there are some disaffected repubs, and some shy tories. nevertheless, calling a poll cooked, and it actually being cooked are two different things. I see nothing wrong on the face of this poll. seems pretty pedestrian. you're gonna need hella better evidence than the fact that independents are not breaking in your favor before a poll is cooked.

what's actually most stunning is the dissipation of McCain's lead in the "better handling of terrorism/unexpected event" question. that presents huge problems for him, though I think it may just be icing on the cake at this point. If people are starting to believe Obama reacts better to crises (probably a perception currently reinforced by the wild swinging of McCain on the financial markets vs Obama's rather cool and collected approach, as expressed by George Will), McCain has lost his best angle.
 

My blanket assertion was that polling of registered voters historically overstates actual voter support for Dems and is per se inaccurate....

"... because it overstates actual voter support for Dems...." ;-)

I'd note for the brain-dead that while it's intuitively more "accurate" to use 'likely voters' -- because everyone knows that not every registered voter votes, and you want to get closer to the sentiments of those that are actually going to cast ballots -- this doesn't mean that raw "likely voter" data is in fact more accurate. Both "likely voter" data and "registered voter" data is massaged and adjusted for various types of sampling bias, and this sampling methodology and statistical adjustment is validated against actual results in previous elections. Such adjustment will account for any supposed tendency towards one party or the other in terms of actual results. The best thing that can be said for "likely voter" data is that it may provide less variability and be less prone to sampling or other methodological biases or irrelevant factors that influence the less likely voters more.

One thing to note is that pretty much all polling does this adjustment and validation; essentially nothing you ever see is "raw data". And the validation is based on the assumption that this year is going to play out pretty much the same as previous years. Whether this is true in a year of enormous changes and challenges, a time of multiple wars and a seven year stagnant economy, the country on the brink in many ways, and a young, charismatic candidate, the first ever major party minority candidate for president, remains to be seen. One thing to note is the immense variability between polls this years, with different polls disagreeing by over three SEMs from each other.

Cheers,
 

jslater:

[to "bart"]: Here's another hint that Repubs understand they are losing. They are going off about "biased media." That's what parties -- both -- always do when they think they are behind.

"Bart" whistles like a champ!:

["Bart"]: For a perfect example of a cooked Dem media poll, take a gander at the really egregious Wash Post/ABC News poll released today, which has a weighted sample with party members and leaning independents 54% Dem and 38 GOP to - surprise, surprise - end up with a 52% to 43% Obama advantage.

Obviously biased. People can't possibly be thinking the Republicans are a bunch of crooks, thugs, and incompetents (and war-mongerers cum religious nutcases) to boot.... The Republican brand name is a popular as it's ever been....

Cheers,
 

For people interested in election integrity, leaving the politics aside momentarily, there is an October 20 moot court planned, with its own website, principally keyed to address new ways to adjudicate disputed polling day rules when election outcome hinges on determinations from an 'impartial' bench. See Prof Foley's announcement at Univ of OH Moritz site.
 

Sandy Levinson: ... if one elector would choose to vote for the popular vote winner in order not to produce a tie (and, therefore, not make the US a further laughingstock in the entire world for the idiocy of its electoral college system).

Certainly possible. It would only take one out of 538. On the other hand, these 538 people are selected for the strength of their partisan loyalty.

While still slim, the odds that we might find out appear to be increasing somewhat. Today Pollster.com notes, "Most interesting about the current estimates is that if we ignore the classifications [lean versus toss up] and just examine which candidate has a numeric lead, the electoral votes as of today would divide in a perfect 269-269 tie."
 

Bob Richard:

[Prof. Levinson]: ... if one elector would choose to vote for the popular vote winner in order not to produce a tie (and, therefore, not make the US a further laughingstock in the entire world for the idiocy of its electoral college system).

Certainly possible. It would only take one out of 538. On the other hand, these 538 people are selected for the strength of their partisan loyalty.

This assumes they know that there will be a tie, and that they know who's supposed to do this. But this is all supposed to take place simultaneously in separate locations. What is someone else defects? Then the person who switches might make it a "toss-up" again. And then there's the more significant possibility of fraud, bribery, blackmail, etc. to sway the vote either way for reasons less "legitimate" than a popular vote victory. If we assume that the votes can be swayed, we have to be open to those possible reasons for such as well.

Cheers,