Wednesday, January 09, 2008

The New Hampshire Polls Weren't Wrong, the Media Was

ABC's Gary Langer writes in The New Hampshire's Polling Fiasco : "There will be a serious, critical look at the final preelection polls in the Democratic presidential primary in New Hampshire; that is essential. It is simply unprecedented for so many polls to have been so wrong. We need to know why."

Sadly Mr. Langer's "look" doesn't seem very serious. I'd like to know why the Mainstream Media (MSM) as represented by ABC should be believed? I suggest that any allegation made by the MSM that the polls were wrong is suspect, because it was the MSM that reported on those polls wrongly. The MSM is now trying to cover its misreporting by claiming the polls were wrong. As I see it, the New Hampshire polls weren't wrong. It was just that the MSM didn't know how to read them.

Let's analyze rather than blindly follow the MSM's fictions.

The results of the election were:

Clinton 39%
Obama 37%
Edwards 17%
Richardson 5%
Kucinich 1%
(unassigned 1%)

The Real Clear Politics (RCP) pre-election polling averages had for 01/05/08 to 01/07/08 was:

Obama: 38.3
Clinton: 30.0
Edwards: 18.3
Richardson 5.7
(unassigned 7.7)

First, in the RCP polling average there is only one percentage number (Clinton's) out of four that is outside the expected margin of error. Obama's, Edwards' and Richardson's numbers were within a reasonable margin of error. So how is that, in Langer's words, either "simply unprecedented" or "so wrong"? It's not.

Next, how do the numbers in the RCP poll add up? 38.3+30.0+18.3+5.7= 92.3 leaving 7.7% unassigned going to either "undecided" or Kucinich or other candidates like write-ins.

Compared to the RCP poll averages, the results showed that Kucinich got 1% leaving 6.7% unaccounted for. Those 6.7 added to the poll's 30.0 for Hillary gives 36.7% for Clinton. Then add the 1.3 from Obama and 1.3 from Edwards and the 0.7 from Richardson's (all three amounts were within the margin of error but they went to Clinton) for 3.3 and then subtract the 1% not assigned and you have 30.0+6.7+3.3-1.0=39% which is the actual result for Clinton.

So, according to the averages, the polls weren't wrong at all for Obama and Edwards, it was only that of the 7.7% that was unaccounted for, most of it went to Clinton and none to Obama, Edwards, or Richardson. That is not a big mix up in the polls. It was the MSM reporters who jumped to a conclusion because they didn't know how to read the polls.

If the media had analyzed the polls correctly instead of trying to create flashy fireworks to report, they would not have focused on the number that gave Obama an apparent lead, but on the relatively large number of 6.7%undecideds. Had that large number of 6.7% been correctly reported there would have been speculation about which candidate would get it. Even if people speculated that it would all go to Obama, or be evenly distributed, the only surprise compared to the polling is that this number went all to Clinton.

A poll provides an abstract snapshot using numbers. Each number has a margin of error. Add to each number's margin of error the number of undecided and you have the reasonable range that the numbers of the poll must be considered to have. The MSM seems to have mistaken logic for probability by thinking it was not logical that an amount equal to the undecided would all go to Clinton. But just as a toss of a coin may come up heads seven times in a row, logic actually dictates that all the undecided voters may go to only one candidate.

So the number of undecideds must be added to each candidate's margin of error number to know what the possible swing is that is portrayed by the poll's snapshot. That is the "logic". Now, the probability is more unlikely than likely that the undecideds will all go to one candidate, but before an election that is not a question of logic. Before an election, logic dictates that even improbable but not impossible outcomes must be considered to be within the range of numbers that the poll is reporting.

But under no circumstances if the polling was read correctly, should the media have reported that this 6.7% was a lock for any of the candidates and it should have been reported that the 6.7% was the wild card that meant the election could go any direction. The media should not have reported "the polls predict Obama will win," but only, "this poll shows a likelihood of Obama winning but also it is possible within the margin of error and if the undecideds break for Clinton that Clinton could win." Then it is up to the audience to decide how probable or not the outcome based on the length of time between the poll and the election. By the same token, the polls did show that even if all the undecideds went to Edwards and his margin of error was considered, that he would not win. So on that score too, the polls were not wrong.

But because the MSM wanted a story, it created its own narrative of a big swing to Obama in the polls and totally ignored the 6.7% that was unaccounted for. By ignoring that 6.7% the media created the story instead of reporting it. The MSM should have reported that the polls showed Obama and Clinton in a close race to the finish and that due to the number of undecideds and the margin of error it was still too close to call.

Now, the media and pundits of the MSM are once again creating a story, this time that "the polls were wrong", when in fact it was not the polls that were wrong but only their mistaken interpretation of the polls.

As we have seen, it looks like most of the undecided voters and a few of the voters for each of the other candidates shifted to Clinton. By the way, I am not saying that in fact all the undecided voters went to Clinton. I am saying that any of the undecided voters who went to the other candidates can be accounted for by a shift from those candidates to Clinton that was well within both the general margin of error and the inherent uncertainty of polls due to considerations of the polled support being softer than harder.

Also, we must remember that there is no longer the length of time between Iowa and NH that there used to be. Thus the Iowa "bounce" for Obama was not able to be polled in a stable state. The MSM should have known the bounce was unstable and reported that any bounce toward Obama would, like a pendulum, necessarily have some amount of back swing. The question that the MSM didn't report is whether the polls that had swung so much in Obama's favor from Clinton would swing back any before the election.

The election showed that a completely reasonable number of people who had polled in the bounce for Obama, either came back to their pre-Iowa pro-Clinton feelings or had their soft support switch to Clinton. All this should have been considered in reading and reporting on the polls, but this is too complicated for the MSM which hates complications and must find a single simple story line that is no more complex than a slogan to sell to the masses as the view of reality.

Alternatively some pundits like Tim Russert are creating a story that the polls weren't wrong and instead the Clinton "victory" was stunning. This is just as mistaken and results in saying such absurdities as this.

"NBC's Tim Russert, subdued for most of the night, resumed some of his post-Iowa-caucus exuberance shortly after Clinton's victory speech. 'One of the greatest upsets in American political history. Underscored,' he said on MSNBC. 'This is the political equivalent of Ali-Frazier.'"

No, Mr. Russert, this was not at all an upset, much less "one of the greatest upsets in American political history."

The polls said the race was close because there was an average 8.3% difference with 6.7% unaccounted. Clearly if the 6.7% went to Clinton then the result was statistically too close to call. That is actually what appears to have happened, so it was not a "greatest upset." Again, only by buying the crazy idea that the 6.7% should be ignored can someone like Russert believe that there was an upset or like Langer that the polling was unprecedented.

Lastly, let's look at the MSM fairytale that Clinton won New Hampshire.

In fact, New Hampshire was a tie. The MSM theme that Clinton was way behind was an illusory fraud, so the new theme that she is now a "comeback girl" is also a fraud.

Clinton was virtually tied in the Iowa results since she came in second place and only one delegate (out of 4,049) behind Obama (16 to 15), while Edwards was in third place in Iowa with 14 delegates (not counting superdelegates for any of them which Clinton has many more of). So with 15 delegates meaning she was only one delegate behind Obama's 16 and with a candidate needing 2025 delegates to win the nomination, the Iowa results did not put Clinton at all far behind Obama. One out of 2025 is not the "huge" win or the huge lead that the MSM and the Obama campaign made it out to be. Obama's campaign should never have allowed themselves to be sucked into that fairytale whirlwind.

In primaries, the actual difference in the number of votes doesn't count if the percentage is less than the percentile needed to win a delegate. Thus in New Hampshire, where the vote difference was not enough to award and extra delegate, the result between Clinton and Obama was a practical tie with 9 delegates each. The so-called 2% "stunning victory" in New Hampshire was an illusion since Clinton really needed about a 4.25% lead just to get one more delegate than Obama for a real win.

Since New Hampshire was an actual dead-heat tie for delegates between Obama and Clinton coming off a virtual tie in Iowa, there was no great "fall behind" and no "great upset" win for Clinton. Both angles on the Iowa-New Hampshire story are made up by the MSM and pundits like Russert and analysts like Langer create illusory sizzle out of the plain facts.

No comments: