Poll trolls

There is every reason to expect a landslide in next week’s US elections. I expect Obama to win with more than 300 EVs, and the Dems to pick up maybe 8 Senate seats (VA, NM, AK, NC, CO, NH, OR, maybe MN and GA at a pinch) and 30-40 House seats.

But, Faux News is at it again, touting the race as a “toss up”, narrower than the margin of error. So, one more time, here’s why they’re way out.

First, they’re reporting just one poll. Guess what? The others don’t paint the same picture. [Update: Fox’s poll has already been shown to be a dud, with party identification at 41% Democratic, 39% Republican.]

Here’s the picture you get aggregating the daily tracking polls:

Obama opened up a large gap in mid-September, and has maintained a large lead ever since, with a slight narrowing in recent days. This narrowing was expected, and is considered quite normal. Underpinning Obama’s continuing lead has been a realignment of perceptions of the candidates, reflected in the candidates’ favourability and trust ratings.

Not only has the “horse race” been a picture of remarkable stability, but things have been getting worse for McCain in recent days. The race is about winning electoral votes state-by-state, and not the popular vote. Ask Al Gore.

As of today, aggregating state-level polls shows Obama ahead by at least 8% in states with 272 EVs — enough to win — and by 6%–8% in states with a further 29 EVs and ahead in states with 56 EVs. And Obama is above 50% in the states where he is 6% or more ahead, except Ohio (49.5%). Apart from Missouri and Indiana (which remain toss-ups), the news in the last few days has been good for Obama: his lead has widened slightly in Florida, and remained steady in Ohio, Virginia, Colorado, New Mexico and Nevada.

Here’s RealClearPolitics’ changes in the status of swing states in recent days:

10/30 Pennsylvania Solid Obama »»» Leaning Obama
10/29 Nevada Toss Up »»» Leaning Obama
10/29 Georgia Leaning McCain »»» Toss Up
10/28 New Hampshire Leaning Obama »»» Solid Obama
10/27 New Hampshire Solid Obama »»» Leaning Obama

But the national polls are showing a great deal of variation, you say (correctly). Suppose there’s a lot of hidden support for McCain? Well, you’re right to have doubts about the polls, but if there’s any hidden support, it’s likely Obama’s…

Polls: Caveat Emptor

First, unlike NZ, turnout in the US has historically been a problem, so the data used are “likely voters” rather than registered voters. How do you know who’s “likely” to vote? Traditionally, pollsters put more weight on past voting behaviour as opposed to expressed intentions. Fine, for your average election. But this is not going to be an average election.

Obama’s margin appears to be over 30% with young voters and over 80% of African American voters. Young voters and African Americans have tended to have low turnouts in the past. However, voting in the primaries, early voting so far in the 31 states that allow it, and unprecedented volunteer activity all point to this being an exceptional year for young and African American turnout.

How much difference will this make to the result? Well, Gallup started producing two figures a couple of weeks ago, using the traditional likely voter model and an expanded one that gave greater weight to expressed intentions. The latter has been tracking at around 3–4% above the former, so that’s how much you could add to Obama’s figures in most national polls if you think that next week will see a massive turn-out.

The so-called Bradley effect? As Mark Blumenthal says, “Much of the recent debate centers on whether the effect ever really existed (see the skeptical take by ABC’s Gary Langer) or whether it existed and then disappeared 10 or 15 years ago (see the exhaustive report [PDF] by Harvard political scientist Daniel Hopkins).” There wasn’t any evidence of such an effect in the primaries either, please note.

Then there’s the cellphone-only voters. If anything, this adds another couple of percent to Obama’s total in the polls (here for earlier post).

But what about those polls showing a narrow gap? Like that IBD/TIPP poll that Faux News was salivating over last week? Well, it showed McCain leading by 52 percentage points among 18- to 24-year-olds. Incredible. As in, not credible.

Zogby, another Faux News favourite in recent times? Turns out that Zogby weighting his data by party ID (a common practice stateside) based on 2004 partisan identification. So he’s been weighting so that Republicans are about even with Democrats. Rasmussen, using current partisan identification weights his data 40.0% Democratic, 32.8% Republican, and 27.2% unaffiliated. (Even so, Zogby has Obama 7% ahead today.)

The late deciders you say? (Well, McCain’s claiming they’ll vote his way when they get into the voting booths.) The best analysis suggests that they will split fairly evenly, maybe a little towards McCain. Anyway, even if he got every last one of them, he’d still be behind in the swing states, as Obama is over 50% already.

This is a truly exciting contest, and it may be close, but it ain’t close yet.

Update: Friday evening US-time: McCain claims, “We’re coming back strong.” Reality: RealClearPolitics poll average had Obama’s lead slipping from 8.0% on 25 October to 5.8% on 30 October. Now it’s back up to 6.5%. McCain’s own state is now categorized as a “toss up” and Obama’s campaign has started running ads there. Comeback?! Strong!?

Advertisement

Radio NZ gets US election news wrong

Yesterday Radio NZ News reported that:

“Democrat Barack Obama’s lead over Republican John McCain in the United States presidential race has dropped to 3 points. That’s according to a Reuters/C-SPAN/Zogby poll which says Senator Obama leads his rival by 48 to 45% among likely American voters, down 1 percentage point. The four-day tracking poll has a margin of error of 2.9 points.”

What’s wrong with that?

Well, for a start, reporting any single poll as though it is fact — “has dropped to 3 points” — is arrant nonsense. It’s an insult to the listeners’ intelligence. Report the averages of the polls, if you must.

Worse, not only is Zogby just one tracking poll out of seven currently publishing results in the US (not to mention all the other regular polls being reported at this point in the race), it has consistently shown Obama’s support as lower than the others:

See those black dots below all the others, Radio NZ? That’s the poll whose results you have chosen to highlight. Dumb.

Why cherry-pick the most unfavourable poll? This is exactly what Fox News and Drudge (here for details of Drudge’s bad spin) have been doing all week, in order to try to launch their “Comeback” narrative for McCain. But Radio NZ isn’t part of the Republican propaganda machine, is it?

No, it did this because it used a news service that, like the others, misuses polls to try to make elections more interesting. It was lazy and/or ignorant.

Reporting their daily tracking poll results (Obama 51% to McCain 46%) this morning, highly respected pollster Rasmussen notes:

“Obama’s support has ranged from 50% to 52% every day for twenty-four straight days while McCain’s total has been between 44% and 46% during those days. Prior to today, Obama had led by four or five points every day for the past week. The week before, the Democrat was up by five to eight points each day (see trends).”

Not nearly as interesting is it?

We taxpayers pay good money for a news service that is a cut above the commercial crowd, Radio New Zealand. You must do better than this.

The full picture, courtesy pollster.com

Update: Since posting yesterday, the RealClearPolitics poll average has increased Obama’s lead from 5.0% to 5.8%. And the Zogby poll? It now shows a 6% gap. And no mention of polls by Drudge for the first time in a week or more. Funny that.

Update2: Oops! Monday night’s Zogby poll has Obama 50.3%, McCain 42.4%. Nary a mention on Drudge or Faux News. Or Radio New Zealand.

Two loveable rogues say “Game is on”!

Well that didn’t take long. Hot on the heels of the Morgan poll comes the TV3 poll (a.k.a. the accurate one) showing the gap between Labour and National halving, to six per cent (45% to 39%, with the Greens on 6.8%).

The key may be in the responses to the question, “who do you trust to best manage New Zealand for the next three years?”

“Labour and Michael Cullen just won that battle with 42.7 percent. National and Bill English were a close second, with 41.2 percent.”

When we last had an economic “meltdown” of this magnitude, the voters went initially with the conservatives, a cautious response to the economic turmoil. As in the US, the election looks to be turning on the question of who it is that voters feel safest with.

If so, after the bungled tax package, expect the next poll (done after today) to be worse for National, not better.

[Hat-tip to ak for “loveable rogues.]

Game on!? Nats ahead only 3% in poll shock

The latest Roy Morgan poll is a stunner! It puts National’s vote down 7% at 40.5% and Labour’s up 1 point to 37.5%, since the last Morgan poll two weeks ago. (Hat-tip Adam.)

The big winners appear to be the Greens, up to 2.5% to 9%, and ACT, up 2% to 3.5%. NZ First lose a point to 4%, but would still be competitive if this poll is accurate.

And that’s the problem. It’s just one poll. And one poll does not a summer of left-wing joy make. It may very well turn out to be a “rogue.” Then again…

Latest poll (N = 923) taken between September 22–October 5, 2008.

More on cell-phone skew

I posted in July about the Pew Research Center’s findings that non-inclusion of cell phone respondents “modestly affects” poll estimates of voting support in the US presidential race

They’ve done another poll with cell respondents included, and updated their findings. Their key finding is unchanged. Including cell respondents makes a little difference–some 3% in the September poll (see table below).

But they are now able to combine the data from the July and September polls to look more closely at the differences between young people responding on land-lines and cells. This is important because one way that the pollsters overcome the problem of excluding cellphone only respondents–as is the case in NZ–is to weight the data. But if they use the young respondents on landlines to weight the data, then they are skewing the estimates to the extent that there is a difference between young land-line and cell phone respondents.

The tuning fork election

Well, jafapete took off for Waiheke and it seems he’s not coming back. AndrewE has suggested that he drowned himself in a pail of ale after realising that Helen has no principles. Not quite. More on that later. This post’s for Andrew.

Last night’s Robert Chapman Lecture “Polls versus Expectations – Howard’s End in Australia and are there lessons for New Zealand?” was interesting, mostly because it is always good to stand back and contemplate recent history for patterns you might have missed. In this we were assisted by Australian election commentator Antony Green.

The crux of his analysis is to be found in the tuning fork graph that he used (courtesy wikipedia) which shows polling results since the last election:

Green noted that in the 2004 and 2005 elections in Australia and NZ, the polls had been very volatile, and that when that is the case it is easier for an incumbent government with waning popularity to turn things around during the election campaign. He didn’t state the converse so explicitly, but I shall.

Like the PM and many others, I can count Bob Chapman amongst my mentors. Many have commented on Bob’s grasp of the finer detail, but it was his capacity to distill this vast knowledge into easily understood wisdom that I valued most. (“The best ideas are the big, simple ideas”, he said once.) Bob always maintained that if the opinion polls clearly had one party ahead of the other for a year or more before an election campaign, then the party that had enjoyed the lead would win. This has always proved to be the case.

So the message from the graph above is clear, even if it’s not in tune with my sentiments.

There were some valiant attempts to avoid the message last night. “What about all the people who didn’t have landline phones, worked evening shifts, etc?” Green patiently explained that pollsters weighted their data to compensate for these known biases as best they could. After all, they have a lot to lose if they get it wrong. “What about the extra 100,000 unenrolled voters ferried to the polls in South Auckland?” asked another, seemingly oblivious to the tuning fork graph in front of her, or maybe not able to convert the percentages into numbers.

However, as Green noted, in an MMP context National’s higher vote need not necessarily lead to their taking government. It is also quite possible, for example, that National holds office as a true minority government, with the Maori Party sitting on the cross-benches vetoing anti-Maori or anti-low paid legislation, but not bringing the government down.

As I look at the tuning fork, that second-best scenario is some consolation.

6.5% and closing

Another poll shows the gap between National and Labour closing, to 6.5%. It confirms the trend revealed by other recent poll results.

The latest Roy Morgan poll — of 841 electors over 18–31 August — shows National at 44.5% (down 3.5%) and Labour at 38% (up 4%). Bad news for NZ First, down 4% to 2.5%, with the others largely unchanged: Greens at 8% (up 0.5%), Maori Party at 3.5% (up 1.5%), ACT at 1.5% (no change) and United Future at 1% (up 1%).

The previous Morgan poll was conducted between 28 July and 10 August, so would not have included much of the impact of the loose lips debacle. The current one does, along with much of the Peters saga.

Another factor may be increasing confidence in the Government:

“The Roy Morgan Government Confidence Rating has risen strongly for the third New Zealand Morgan Poll in a row, rising 11.5pts to 115. It is now at its highest level since being at 118.5 in late February. For the first time since February, a majority of New Zealanders 50.5% (up 6%) say the country is “heading in the right direction” compared to 35.5% (down 5.5%) that say the country is “heading in the wrong direction.””

The interest rate cuts may have helped, and there are more of those on the way.

It is only one poll, but should it prove accurate then the Peters saga has affected NZ First but not Labour or ACT. This would not be a surprise, as most NZers probably understand only too well that had Labour done as National suggests, and shut out Peters, then we would simply have had an early election. This doesn’t seem to be wished for outside the right-wing blogs and John Key.

According to Tane’s calculations the latest result puts a Labour-led coalition in by a nose. Morgan on the other hand say that, “If an election were held now New Zealand would have a Coalition Government likely to be led by the National Party.” Even though Tane doesn’t factor in the overhang from the Maori Party seats, it looks as though the Maori Party would hold the balance of power.

But this is really beside the point. Political events in NZ are moving at a breathtaking pace, and anything could happen between now and the election, whenever that is.

Latest poll goodish news for Labour

Labour’s “on the move.” So says TV3 News, reporting the latest 3News poll.

The poll confirms the trend revealed by other recent poll results. A slight bounce for Labour, National maintains a decisive advantage, with the loose lips debacle not having made much impact… yet.

National remains on 48%, Labour is up 2 points at 37%, the Greens are down one point on 6%, NZ First’s on 3%, Act 2%, and United Future 1%.

A little under 3 months out from the election this is good news for Labour, who need to start closing the gap. If the Greens and NZ First make it back into Parliament, then it doesn’t have too much more ground to make up before the election starts to look very exciting.

The minor party results probably don’t mean too much, although it doesn’t look like the donations saga has done Peters any good.

The poll of 1000 voters was taken between August 14 and August 20, and has a margin of error of 3.1 percent.

Poll confirms gap narrowing, a little

The One News Colmar Brunton poll just out confirms my reading of the two other polls released late last week.

Taken more recently (Saturday week ago to last Thursday), the Colmar Brunton poll confirms a slight bounce for Labour, up 2 points to 37%. This appears to be at the expense of the minor parties, however –with none of the smaller parties polling above the 5% threshold in this poll.

Further, National’s support has not been affected by the secret tapes and loose lips affair. At 51% National is still polling well enough to govern alone.

I don’t expect either situation come election time. NZers have proven themselves extremely adept at fine-tuning the final result, and we can confidently expect sufficient green-leaning Labour voters to peel off to save the Greens, if this looks necessary on the day.

Likewise, I’d be surprised if the voters give National a total mandate to govern alone. Why would they? They are still too much of an unknown quantity to be allowed to govern without check (and ACT doesn’t qualify as a check for more than 90% of the electorate, probably).

[Update: Pierson at The Standard posts on the polls.]

No bounce for Labour yet

The two mid-August polls released to date show little evidence of a bounce for Labour out of the “loose lips” debacle for National.

Roy Morgan’s poll shows National steady at around 48% and Labour returning to its May level of around 35% after a few months in the 30-32.5% range. Perhaps a bounce, but pretty small at best.

More interestingly, confidence in the Government has bounced back “strongly” to 103.5, after falling from over 120.0 last September to bottom out at 87-88% in late June-early July. And more NZers (44.5%, up 6%) say the country is “heading in the right direction” than say it is “heading in the wrong direction” (41%, down 6%). So some good news for Labour.

Not so the Fairfax Nielsen poll. This shows Labour steady on 35% and National up 3 points to 54%.

Like the Morgan poll, the FN poll showed confidence in the economy improving: “the number of voters who are more optimistic increased by three points to 45% over April, while pessimists declined from 39% to 30%.”

The Morgan polling period was 28 July–10 August (N = 834), so the effects of the loose lips stories that broke on the evening of 3 August would be diluted. The FN poll was conducted over 6 August–12 August (N = 1102) and includes the effects of the loose lips — although some respondents would have been polled before the story had fully unravelled. It is also more likely to be accurate given the larger sample size.

The next TV3 poll should therefore help shed light on a confused situation. As Inventory2 points out, Duncan Garner’s blog yesterday hints that the news will not be good for Labour (Garner is the 3 News political editor, after all). Garnber opines that:

“National was quick to nail Labour and Helen Clark to the secret recorder. Key has no proof of course – but he didn’t have to. It was a bit of a dog whistle – but it may have worked. I reckon National might just hold up in the polls due to be released over the next 10 days.”

Still, even if there is no immediate pay-off for Labour from the loose lips debacle, the episode may prove valuable in the long-run. Garner imagines Labour putting together advertisements that feature English and his comments as part of a campaign based on trust, leadership and privatisation. I’d say that last week’s events might prove very useful to Labour as it seeks to build a narrative around “hidden agendas”.

[Note: Neither poll includes the “refused to says” or “don’t knows”, the latter of which stand at 11% in the FN poll.]