Your Daily Poll Warnings, Take Three

by: Chris Bowers

Wed Oct 08, 2008 at 12:26


I've been posting daily poll warning threads recently, and looking at the comments in Quick Hits on the tracking polls, clearly I need to keep doing so. There is nothing wrong with worrying about daily tracking polls--I do so, too. The key is to know what to worry about, and what to ignore.

Helpful tips for watching daily polls in the extended entry.

Chris Bowers :: Your Daily Poll Warnings, Take Three
Keep in mind the following:

  • 8.5% is the maximum victory: First, as I warned on Monday, please keep in mind that an 8,5% national victory is the maximum. In the last sixteen national elections (U.S. House and Presidential), the largest victory was Bill Clinton's 8.51% victory in 1996. The simple fact is that no one wins by double digits anymore. This goes for the large Republican victories in 1988 (President--7.72%), 1994 (5.9%), and 2002 (4.6%). It also goes for the large Democratic victories of 1992 (President--5.56%), 1996 (President--8.51%) and 2006 (7.9%). "Landslides" are now 5-8% national victories, not anything larger. Given that a very real percentage of Democrats and Independents won't vote for him because he is black, it was always absurd to think that Obama was going to break this mark. When Obama reached an 8% national lead, the only place for him to go was down.

  • Polls aren't static:  Second, as I warned yesterday, even if you accept that it is impossible for Obama's lead to continue rising, don't expect it to remain static, either. Polls are using non-static means (statistical ranges with margins of error) to measure a non-static subject matter (public opinion). So, when there is nowhere to go but down, when the polls inevitably start moving again, it isn't a shock to see them move down.

  • Peaks are called that for a reason: Within the last three days, Obama had surpassed or equaled his all-time high in every single tracking poll: Research 2000 (+12% two days ago), Gallup (+9% yesterday), Rasmussen (+8% yesterday), Battleground (+7% yesterday), Hotline (+7% three days ago), and Zogby (+3% yesterday). With Obama hitting his all-time high in four tracking polls yesterday, one two days ago and one three days ago, for both of the reasons mentioned above, the idea that his standing in these polls was gong anywhere but down is wishful thinking. When dealing with polls that have a track record of several weeks and, in some cases, seven months, it is highly unlikely that a record will be immediately followed with another record. This is especially given the two reasons explained above.

  • Less grandiose, and more specific to the Hotline poll, keep in mind that it shifted 4% toward McCain in a single day (yesterday). That is a shift of 11-13% in favor of McCain for Monday's individual day of polling from Friday's individual day of polling. When Monday's polling drops out of the averages (Friday morning), then let's see where the poll stands. Given the absurd one-day shift, I have a feeling Friday will snap back to Obama.

There are good reasons to be worried about polling in this election. For one thing, my research shows that even polling averages miss the final margin by an average of 2.0%, meaning that a lead of less than 2.0% isn't a lead at all. Second, until about three weeks ago, the campaign was still in that 2.0% range, meaning that Obama's lead is very recent and thus not very stable. New supporters are the shakiest supporters of all. Third, if 8.5% is an unattainable maximum, and 2.0% isn't a lead at all, then we are only dealing with a narrow 5-6% range where Obama can lead. And, as I said, Obama's movement into that range is new and shaky.

For the reasons I listed in the previous paragraph, and with the financial crisis now being forced to share headlines with McCain campaign attacks, there are good reasons to be worried. However, make sure you are worried for the right reasons, and not just wringing your hands because it feels comfortable to wring hands. Fretting over the Diageo-Hotline tracking poll, or even internals in that poll, is really foolish. Fretting over a 5% Obama lead national lead across either tracking polls or all national polls is also foolish. That is about as big as one should ever expect Obama's lead to be, and it still is as much of a landslide as any election has been since 1984. If you want to watch daily polls in our current age of politics and elections, then these are the margins you have to learn to deal with.


Tags: , , (All Tags)
Print Friendly View Send As Email
Thank you, Chris for your cool head... (0.00 / 0)
I needed that! :-)  

REID: Voting against us was never part of our arrangement!
SPECTER: I am altering the deal! Pray I don't alter it any further!
REID: This deal keeps getting worse all the time!


Weren't you the one who was complaining about Obama's lead being too big yesterday? (0.00 / 0)
And didn't I call you out for it? :)

[ Parent ]
The poll anxiety is *insane* (0.00 / 0)
In most of the polls lately, Obama has been near his likely ceiling, while McCain was near his likely floor.

It was almost statistically impossible for there not to have been a coordinated drop off in several of them in one day.

Gallup will close too.  You can almost book it.  Why?  Same reason.  It will close or at least stagnate.


Well It Didn't Happen Today! (0.00 / 0)
Obama's lead in the Gallup poll is now 11%, the highest lead by either candidate in the entire election!

Rasmussen is down slightly from 8% to 6%. DailyKOS/Research 2000 is now down slightly from 12% to 10%. Big deal.

The polling shows Obama won the debate last night and should actually move UP in the polls starting Friday if there's any effect.

Meanwhile despite your point that "Obama's lead is very recent and thus not very stable. New supporters are the shakiest supporters of all" in fact both Gallup and Rasmussen have found the election "very stable."

Galllup: Obama has now held a statistically significant lead since Sept. 24-26 polling and has not trailed McCain since Sept. 13-15, roughly coinciding with the intensification of the financial crisis.

For the past thirteen days, Obama's support has ranged from 50% to 52% while McCain has been at 44% of 45% every day

Voters have made up their minds. This economic crisis isn't going to go away, and no attempt by McCain to "get past the economy" and talk about trivial crap like Ayers and Democratic "electoral fraud" and personalities, is going to work.

Nobody gives a damn! I expect Obama's lead in the Gallup poll to reach 14-15% before next Monday.

It will certainly decline as the election nears, but a 5% lead translates to a landslide in the electoral college anyway.


[ Parent ]
what exactly (0.00 / 0)
...is a landslide victory in electoral college? just curious.

[ Parent ]
People need to get a grip (4.00 / 1)
We've been spoiled by these insanely good numbers for two weeks now; four weeks ago people would have been overjoyed with a three-point lead.  The fact is that the MSM is becoming convinced that McCain is headed for a huge defeat, and that meme is spreading through the public as well.  Morale can't be high for the McCain campaign.

I want to shout this from the rooftops: (4.00 / 2)

For one thing, my research shows that even polling averages miss the final margin by an average of 2.0%, meaning that a lead of less than 2.0% isn't a lead at all.

Corollary: shifts in the race of less than 2.0% aren't actually shifts at all.  Shifts in an individual poll of less than 5.0% aren't really shifts at all.  

This is my number one pet peeve of poll watching--people fretting over small shits in an individual poll that are clearly within that poll's margin of error.  

Also, Obama will win the post-debate coverage in addition to the debate.  What's the McCain soundbyte from that debate?  The overhead projector?  Obama had that very good comment about the scalplel and the hacksaw.  


Ugh. I wish I could edit. That isnt' the best typo ever. (4.00 / 1)


[ Parent ]
Scalpel/hatchet (4.00 / 1)
That might not get much attention, because Obama already used it in the first debate.  I agree that it is a good metaphor.

But it's probably not that big a deal either way, as I don't think very many people are taking McCain's "spending freeze" proposal seriously at all.  Apparently he isn't either, since he started talking about buying up mortgages in the debate last night.


[ Parent ]
Shifts Within The Margin Of Error DO Matter!` (0.00 / 0)
That is just flat WRONG!

While ONE poll showing an Obama lead within the margin of error isn't significant, when you have ten or twenty successive polls ALL showing the same narrow lead, all within the margin of error, the probability that McCain is actually leading drops towards zero!

As n (the number of trials, in this case the number of polls) increases, P, the probability depicted converges rapidly towards the real probability.

If you have 100 polls all showing a similar Obama lead (2-3% say, within the margin of error), the probability of a McCain lead is nil.

That's what we have here. We have multiple polls converging on an Obama lead of around 6% (Pollster.com). This tells us that Obama's lead is stable right now at over 5%.

Three of the 4 debates have come and gone and McCain/Palin have lost all three. The economy continues to tank and McCain has no solution, but to talk about Obama's "dangerous ties to Ayers."

As for voters not voting for Obama because he's black, that's already expressed in the polls. THERE IS NO Bradley Effect!:

No More Wilder Effect: No More Wilder E ffect, Never a Whitman E ffect: When and Why Polls Mislead about Black and Female Candidates
Daniel J. Hopkins
Post-Doctoral Fellow
Department of Government
Harvard University: http://people.iq.harvard.edu/~...

In the early 1990s, there was a pronounced gap between polling and performance for black candidates of about 2.3 percentage points. But in the mid-1990s, that upward bias in telephone surveys disappeared.

Racists today can find lots of reason to vote for McCain if they want. They don't have to lie to pollsters!  


[ Parent ]
When twenty polls show the same thing (0.00 / 0)
you are aggregating the twenty polls and averaging them, and then the standard deviation of the aggregate is less than the standard deviation in any of the individual polls.  

So shifts within the margin of error still don't matter, it's just that the margin of error of the aggregate is much smaller than the margin of error of the individual polls.  And this happens in a calculable way, so long as you neglect systematic errors.


[ Parent ]
100 polls? (0.00 / 0)
"If you have 100 polls all showing a similar Obama lead (2-3% say, within the margin of error), the probability of a McCain lead is nil."

In reality, are we really talking about a hundred polls? OR Are you just speaking hypothetically?  


[ Parent ]
Awesome analysis Chris (0.00 / 0)
Everything you write makes perfect sense.

I wasn't alarmed by the polls, but I'm definitely reassured by your analysis.


what do you make of these guys: (0.00 / 0)
http://www.colleyrankings.com/...

it's not a poll. they use the data from RCP and use medians to calculate things. i'm sure you've given it a look before (perhaps you've commented on it)......but what are your thoughts?


Gallup (0.00 / 0)
Obama 52
McCain 41

So calm the fuck down, guys.  We're doing fine.  Plus, the polls don't include the debate.


Agreed (0.00 / 0)
There simply is no doubt that Barack is solidly ahead by any historical measure at this point, and the national political climate is so toxic for Republicans that even the most cynical attacks will ultimately fail.  

Furthermore, there are 2 massive advantages Obama has that weren't mentioned above. Firstly, the enthusiasm gap is so large that it cannot be ignored. Enthusiastic supporters not only vote, but help get others to vote too.  Secondly, Obama's organization and volunteer base is so much stronger than McCain's, that it will almost surely provide a 1-2 point boost on election day (and perhaps even more in states where the Obama team has proven its chops like Iowa, North Carolina, Virginia, and Indiana).

Lastly, you can always pick and chose tracking polls to prove a point. Today's Morning Call Pennsylvania poll has Obama up 12 (a 2 point increase from yesterday), but that doesn't lead me to believe that Obama is surging anymore than he has been, just like McCain gaining a point in Rasmussen doesn't make me believe that his nasty attakcs earlier this week have captured the hearts of the American people.


electoral college (0.00 / 0)
All good points, Chris. And here's another: the President is selected by the electoral college not the national popular vote. Obama has a virtually insurmountable lead in the EC what with the Kerry states all moving into single digits plus Iowa, NM, and VA trending strongly his way and looking good in Ohio, Florida, North Carolina, Colorado, and Nevada too. Until Obama's base state electoral count drops below 270 there is nothing much to worry about.

edit wru (0.00 / 0)
That should be "the Kerry states moving into double digits."

[ Parent ]
I am no expert but i think you could be wrong (0.00 / 0)
As I said, i am no expert, and certainly your predictions make a lot of sense, but of all of the past elections you cited, none of them come even close to the importance of this present election.  So really there is nothing to compare with a situation where USA is in such a perilous condition, and a president who has had dismal ratings over the last 3 to 4  years and a wrong track record year, i really don't think even you experts can accurately predict where this is going to land.

Having said that, i really respect your ongoing analysis.


I don't know (0.00 / 0)
The 2000 election turned out to be pretty damn important.

And I heard almost as much hype about how critically important the 1992 election was, in the wake of the fall of the USSR and the instability in Eastern Europe.  

I think it can be really hard to determine how important an election is before the fact--if obama get elected, and we merely go back to the clinton years, isn't that just leaving the confrontation for some point in the future?


[ Parent ]
I don't know (4.00 / 1)
I'd argue that 2000 and 2004 turned out to be pretty freakin important. In fact, as much as I like Obama, I'd gladly trade an Obama win in 2008 for a Kerry win in 2004 or a Gore win in 2000.

Besides, I don't see where Chris says the present election isn't important.


[ Parent ]
they were important after the fact (0.00 / 0)
In 2000, I can remember even myself being kind of not all that interested in Gore, and even at one point thought i could see voting for Bush because of all of the bipartisan talk he had. (I know, i can't believe i even considered it). 2000 was pre 911, and the close election of 2000 wasn't known until after voting.  Prior to voting in 2000, we had no idea of the historic importance it had.

2004 was an important election for many of us, but many others it wasn't a tipping point yet and it was status quo.  Only in 2008 presidential elections is there a tipping point of historic proportions, and below, Rosenberg seems to be echoing similar thoughts.  So not being an expert, I'm not completely off base, :-)  .


[ Parent ]
What Calculator are you (0.00 / 0)
using,Chris Bowers?

My 2008 "super heavy duty" 60000 GB calculator tells me Obama will beat McNasty by 14 percentage points come Nov. 4th.  The highest in history for the winner, and it won't me Gumpy Old Fart McShame.


Standing ovation, Chris (0.00 / 0)
Absolutely spot on.

No & Yes & And (4.00 / 1)
The whole point of all my writing about realignment is that comparisons to "recent" elections may go out the window.  Religning elections aren't necessarily landslides--and, indeed, 1896, 1932 and 1968 were all followed by much more decisive elections.  But they do signal a change in the political climate, so it's not unreasonable to hope for a bigger margin of victory.

So that's my relatively rare "no" to what you wrote, Chris.

But the "yes" is to your larger point about not panicking, and not deepending on high expectations.  Based on enthusiasm levels, and the general tenor of the campaign, Election Day could well see turnout spike for the Dems and tank for the Reps, giving us an extra 2-3 points or more.  That could turn an 8-point margin into 10 or 11.  But you don't count on that.  You just keep working.

The "and" is that the election is held via the electoral college, and that seems pretty much out of reach, AND there are a lot of other elections out there, too.  So think big picture folks, it's not just about POTUS, it's about everything.

"You know what they say -- those of us who fail history... doomed to repeat it in summer school." -- Buffy The Vampire Slayer, Season 6, Episode 3


margin of 2%? (0.00 / 0)
There are good reasons to be worried about polling in this election. For one thing, my research shows that even polling averages miss the final margin by an average of 2.0%, meaning that a lead of less than 2.0% isn't a lead at all.

Unless you've already corrected for this, a margin of 2% discrepancy in no way defaults to a +2% bias as Chris suggests. It could also be a -2% bias, meaning the margin is actually 2 points greater as opposed to 2 points smaller.

I also think the "effects" rationale could be flawed, based on the noted research recently that dismisses a major Bradley effect--and also because it doesn't factor in either the "Reagan effect" or the "Winner effect." In 1980 most folks had made up their minds about Carter, but just didn't have enough info on Reagan to decide. After the first debate, people had their information--he looked presidential and safe to them, so they bit and bit in a pretty huge wave. The "Winner effect" can come out when one candidate blows open a big lead close to the election. I think we probably need another week or so, but if Obama maintains a wide enough, consensus-based enough lead, people who were on the fence may likely decide they want to back a winner, or will stay at home rather than take the time to back a sure loser. Although somewhat more longitudinal, it's the same effect that led networks to not call the Presidency when polls are still open--why go vote if it's over? (The nets calling it early is a more decisive signal, but when all polling shows a sizeable lead going in, I think that's a similarly determinant situation.

I do however agree with the counsel to RELAX. Obama's going to win, and win fairly easily. Maintaining the margin is important downticket, but the chances of it swinging back to McCain at this point are almost nil.

Help us Optimize McCain! Use these widgets to make it crazy-easy...


Do NOT relax - be energized! (0.00 / 0)
This isn't about winning elections, its about social change.  

There are decades of pent up needs for decent legislation, from quality schools and health care as human rights, to reversing economic inequality, to a non-imperial foreign policy, saving the biosphere, on and on.

The Repugs have a proven record of obstruction by any means necessary to prevent progressive Dems from winning legislation that benefits most voters, and locks in the majority.

To break the gridlock, we need a realignment and a supermajority.  The multiple train wrecks of the Cheney/Bush junta have given us the best opportunity since the mid 1960s.

The last thing we need to do is relax, except in the sense that analysts of sprinting use the term to mean running with utter fluidity, as fast as one is capable.

BTW, what you call the Winner Effect I have always heard called the Bandwagon Effect, which seems to be starting up. We can help it along.



There is no such thing as a free market.


[ Parent ]
Orwell might have something to say about the last 32 years (0.00 / 0)
It's good commentary and your overall point isn't damaged, but I think you made an error.

In the last sixteen national elections (U.S. House and Presidential), the largest victory was Bill Clinton's 8.51% victory in 1996.

Reagan in 1984 won by 18%, 59-41. Clinton's was the second-largest over that period, though.


So...irritating (0.00 / 0)
First, as I warned on Monday, please keep in mind that an 8.5% national victory is the maximum. In the last sixteen national elections (U.S. House and Presidential), the largest victory was Bill Clinton's 8.51% victory in 1996. The simple fact is that no one wins by double digits anymore.

Hmm, I'm going to ignore, for the moment, congressional elections, which are a) very different, and b) harder to find information on, but this strikes me as basically a non-argument...

In 1872, Grant beat Greeley by about 12 points (considerably less than the margin by which Reagan beat Mondale.

What was the next time a candidate won by more than 10 points?

1904, 32 years later, when TR beat Alton B. Parker by 19 points.  1904's version of Chris would have been even more justified than Chris is in saying that double digit victories were impossible.  And he would have been completely wrong, because Teddy stomped Parker.

Continuing onwards, in 1936, FDR beat Alf Landon by 24 points.  Although there were some sizeable victories in the years that followed (FDR over Willkie, Eisenhower over Stevenson both times, but especially the second, when he won by about 15 points), by 1964 nothing remotely comparable to FDR's stomping had occurred in 28 years.  So 1964 Chris would have been quite justified in arguing that, given the experience of the last 25 years, there was absolutely no chance of Johnson beating Goldwater by more than 20 points.  1964 Chris, like 1904 Chris, would have been completely wrong.

Effectively, Chris has no idea what Obama's ceiling is.  I doubt he's going to manage a landslide on the order of 1984, 1972, 1964, or even 1956.  And, indeed, at this point a win of less than 10 points seems more likely than anything larger than that.

But is a 10 point win (like Reagan's in 1980) really totally out of the realm of possibility?  1980 is more recent than Grant's big win in 1872 was in 1904, and exactly as far away as FDR's big 1936 win was in 1964.

Basically, Chris's argument here amounts to "It hasn't happened in a while, so it's impossible that it'll happen this year."  That's totally ridiculous.

Note: I am not arguing that Obama is likely to win by double digit margins.  I am merely arguing that Chris has not presented a minimally competent argument that such a victory is impossible.







Donate to Open Left




blog advertising is good for you
blog advertising is good for you
USER MENU

SEARCH

   

Advanced Search