Can the Polls Be Trusted?
It never fails that, as Election Day draws close, the supporters of the candidate trailing in the polls start telling us that the polls aren’t accurate. Part of this is Pauline Kaelism, with most people associating with likeminded people and therefore floored that more than half the country could possibly prefer the other guy when everyone they know supports their guy. Part of it is selective memory of past polling.
RealClearPolitics, a Republican-leaning site which popularized the aggregation of polls to minimize variation among individual surveys, shows Obama with a wide national lead and leading each and every battleground state. Further, every single national poll shows Obama ahead; the only question is by how much.
The new NBC/Wall Street Journal poll, the public survey I most trust (both because I know Neil Newhouse personally but also because both he and his Democratic cohort, Peter Hart, make their living doing survey research for high level political candidates and therefore have to get it right) has Obama up 10 among registered voters and up in every category that decides elections: “Obama’s current lead is also fueled by his strength among independent voters (topping McCain 49 to 37 percent), suburban voters (53 to 41), Catholics (50 to 44) and white women (49 to 45).” A new Pew poll also has Obama’s lead widening and, more importantly, showing increasing doubt about McCain’s age, judgment, and campaign style.
Ann Coulter recently argued that national polls historically undercout Republicans, an argument I’ve seen echoed in my comments section of late.
Reviewing the polls printed in The New York Times and The Washington Post in the last month of every presidential election since 1976, I found the polls were never wrong in a friendly way to Republicans. When the polls were wrong, which was often, they overestimated support for the Democrat, usually by about 6 to 10 points.
The trouble with that is she then demonstrates her conclusion by cherry picking individual polls that understated the final Republican vote count, choosing a Gallup poll here, an AP poll there, and a NYT poll elsewhere. When, though, was the last time that the aggregated polls were significantly off in a presidential election?
Michael Barone, a Republican-leaning analyst who’s nonetheless much respected for his insights, argues that yes, we can trust the polls but that we have to understand what we’re looking at. He’s right that polling is more art than science in close elections. There are always judgment calls to be made and this contest brings up several known unknowns.
- Are the likely voter screens any good? This is perhaps the hardest job professional pollsters have. It’s very difficult to distinguish attitude from action. Everyone who says they’ll vote for a given candidate won’t. This election, with it’s “enthusiasm gap” makes it especially difficult. Usually, we figure that if people have voted in previous elections, they’ll vote this time and, if they haven’t voted before, they won’t do it this time, either.
- Will Obama’s young supporters actually turn out? They usually don’t but they seem unusually enthusiastic this go-round.
- Will people actually vote for the black guy when it comes down to it? All indications are that they will and serious analysis of the “Bradley effect” seems to undercut the idea that it ever existed.
- Will the polls themselves affect the outcome? Will would-be McCain voters sit out the election, depressed by the conventional wisdom that their guy has lost? Will there be a bandwagon effect with undecidededs swinging toward Obama so they can be on the winning side? Will Obama supporters get overconfident and not show up? Will these factors cancel each other out? We honestly don’t know.
- Will early voting impact the race? I voted last Friday in Virginia, a battleground state. There was a short line. Many states now allow early balloting, either in person or by mail. We really don’t know what this means for the dynamics of the race. The conventional wisdom, though, is that it helps the Democrats. First, Democrat-leaners are less likely to show up on Election Day and wait in line. Anything that makes it more convenient to vote, then, should help the Democrats. Second, Obama has been ahead for a while. People who vote for him now can’t be affected by McCain’s advertising or his next “Hail Mary” play.
These factors and others are hard to calculate. The bottom line, though, is that Obama has a significant lead both nationally and in the state-by-states. For McCain to win, the pollsters are going to have to be wrong in quite a few states — or voter attitudes are going to have to change significantly in less than two weeks.
UPDATE: Rick Moran argues, in a post titled “THE GOP AND THE ‘DEAD PARROT’ SCENARIO,” that the best bet is that the polls understate Obama’s numbers. For those looking for a more upbeat assessment, D.J. Drummond provides it with “Gallup and New Coke.” I share Andrew Sullivan‘s assessment of the latter, unfortunately.
Can the Polls Be Trusted?: It never fails that, as Election Day draws close, the supporters of the cand.. http://tinyurl.com/5sfvty
Here is a good primer on tracking polls.
The trouble with that is she then demonstrates her conclusion by cherry picking individual polls that understated the final Republican vote count, choosing a Gallup poll here, an AP poll there, and a NYT poll elsewhere. When, though, was the last time that the aggregated polls were significantly off in a presidential election?
I guess the answer to your question, depends totally on your definition of “significantly”, James. And I think you tend to miss the point she was making, which I took to be that about this far out from the election is usually vastly different than the reality of it. As I said in my Nightly Ramble last night…
That’s really what this is about, not the polling numbers showing on election day. The suspicion is that there’s a lot of folks using ‘polling data’ to create the impression of moementum… and as Coulter points out the image usually spins toward the candidate farthest to the left.
Then again, it always seems to ‘tighten up’ just before election day, with the more conservative candidate gaining huge numbers in a very short time, for no action that anyone can logically attribute.
I remember predictng Bush 41’s victory in his second race, as an example, suggesting the numbers would change back to reality less than 48 hours before the actual election, lest some bright person get the idea of comparing the polling numbers to the actual vote, and a lot of pollsters being out of work, or at least exposed, thereby.
That’s a trend that I first noted in Reagan’s first term race, and seems to have been the pattern ever since. Isn’t it logical to ask why?
FWIW, and this goes to the “will the youngsters vote”, my wife works at the local university where she can vote early. When she went to vote yesterday, there was a 2-3 hour wait in the line. She’s going to wait a few days. This is in a battleground state, too.
What do you mean by that/Why do you say that, are democrats historically impatient or something? I missed that study/observation/whatever your source may be. I do hate waiting in line though…
Democrat-leaners come disproportionately from the ranks of the poor, less educated, and young. They’re less likely to vote, period. Any additional obstacles — it’s raining, it’s cold, the lines are long — make them even less likely to vote.
lol, no free smokes.
thanks for the clarification, that makes sense.
Could this then actually spell the end of the great experiment?
1]There were no primary elections
2]subsequently no conventions
3]And now with corruption in registration, early voting,mail-in ballots etc., no election day.
In Iraq the people held up a purple index finger as a symbol of self determination.
In America, the government and it’s soon to be absolute propaganda arm,the media,hold up a different finger to that very proposition.
As a result,one profane party with fasces in hand may soon control all branches of the federal government, leaving the American people out on a limb.
Statistically speaking an aggregation like the RCP poll has much better statistical properties than any single given poll. Even if a single national poll is biased then its impact is lessened by the rest of the polls that are not. For Coulter, et. al.’s argument to work a significant number of polls have to be biased which I find unlikely. Which I see Bithead is already working on.
Also, look at that In-Trade spread. McCain is toast at this point. Barring some huge unforseen scandal erupting around Obama, this thing is over.
Early or mail-in voting will probably help to increase young voter turn-out, simply because instead of having to go someplace to vote, they can do it in peer groups with peer pressure getting them to vote.
The scandals, gaffe’s, etc. are there, but no one cares.
Gaffes too.
I read the 538 polling synopsis (linked by DC) last night. It gives a good breakdown on each of the polls and their inherent biases and is well worth reading.
In the battleground states listed above only Colorado and Virginia are outside the MOE, though I do like the lean in the other states.
Obama’s battleground numbers have been deteriorating in recent days — a Mason/Dixon poll in VA shows him up by only 2, and Florida is quite close. Colorado has also been dwindling.
I dunno if these reflect the recent decline in national numbers (which seem to be rebounding), or perhaps the effect of relentless negative TV ads in those states.
Anyway, much as I support Obama, I’m not counting the Cabinet members just yet.
Did anyone notice, as if on cue, AP producing a poll saying the race is tied again?
It’s happened so often over the years, calling it this time was no great feat.
When InTrade and the RCP national average start saying the race is a tie, then I’ll get excited.
I just checked RCP and they have that poll, and it is the only one that is that close. The next closest is the GWU/Battleground poll that has Obama at +2 and then IBD/TIPP which is at +4.
and Fox has Obama up by 9.
It isn’t wise to hang your hat on any one poll or even poll aggregators like RCP or 538, though I think the latter are far more likely to be accurate than the former.
Grewgills; it is not productive to depend on any poll, short of the one on Nov.4.
Whether corrupted or not, it is the only one that counts.
As for your hat….”Hang on to your hat!”
Not for long.
Sit back, and watch. Assuming the pattern holds, the rest will start following suit in the next five days, usually less.
bitsy how did that big post debate palin bounce work out for you?
Better, I suppose, than was reported.
Got the 401k blues? Buy tin foil futures. A slam dunk.
Re: the AP poll, if you include 44% evangelicals in your sample when 23% voted in 2004, then yeah, you’re going to get a tight race.
Garbage in, garbage out.