**"Iowa Poll, the Steve Harvey of polls."**

Congratulations to Ted Cruz for organizing a winning ticket in the Iowa caucus. That's leadership, and that is what presidents do. I thought Donald Trump would steamroll the competition. I was wrong.

Now, by a show of hands, how many of you think the Des Moines Register is the evening's biggest loser?

Everyone?

That's because not only did it once again back the wrong pony with its editorial endorsement (Rubio who finished third) but because its vaunted "Iowa Poll" was off. Not only did it miss the winner, but its numbers fell well outside the "margin of error" that it claims is only 4 points.

It predicted Donald Trump would win with 28 percent.

WRONG! He got 24 and the poll was off by four.

It predicted Cruz would finish second with 23 percent.

WRONG! He won with 28 and the poll was off by five. That is outside its stated margin of error.

It predicted Marco Rubio would finish third with 15 percent.

WRONG! He finished third with 23 and the poll was off by eight.

Guess what? That is the average mistake this poll makes. Counting the Democratic race, where it missed Sanders by seven points, the poll has averaged a margin of error of 7.5 points in the last 11 contested races from 1988 to 2016, inclusive.

Everyone makes mistakes, but this newspaper's failure to learn from its mistakes -- the same woman has been running this slop for 28 years -- is journalistic malpractice. The poll is useless politically, but a great marketing tool for a newspaper whose reputation exceeds its ability.

I wrote about this on Sunday, and here it is again:

Why do we bother with the Iowa Poll? The final Iowa Poll before the Iowa Caucus has been off by an average of 7.5 points in each of the nine contested contests since 1988, according to numbers at 538.com.

So much for its claim that "The margin of error is plus or minus 4 percentage points."

The poll happy 538 site called it the best poll in Iowa. That is like being the tallest midget.

In the 2012, this vaunted poll assured everyone that Rick Santorum would finish third with 16 points.

He won with 24.5.

Only off by 8.5 points, oh vaunted poll.

In 2008, this vaunted poll assured everyone that John McCain would finish second with 26 points.

He was third with 13.

Only off by 13 points, oh vaunted poll.

In that same race, this vaunted poll assured everyone that Mitt Romney would finish third with 13 points.

He was second with 25.2 points.

Only off by 12.2 points, oh vaunted poll.

Well, it is hard to poll in Iowa, right. Yet the Real Clear Politics Average of Polls had McCain at 11.8 (off by 1.2) and Romney at 26.5 (off by 1.5).

In fact, the Real Clear Politics Average of Polls was a pretty good indicator. Its numbers produced a margin of error of 5 points, plus or minus in 2008, when the Iowa Poll was off by 13. Go check it out.

To be fair, everyone misunderestimated Santorum in 2012. But still, what is the purpose of saying the Iowa Poll is the most accurate when it averages a miss of 7.5 points every time?

I think polls have jumped the clichÃ©.

My wife and I are watching (for about the 4th time) our DVD of James Clavell's Shogun. We have reached the point in the story where the samurai Kasigi Yabu commits seppuku. I recommend this, figuratively of course, to the editors of the Register and its pollsters, who have been so wrong for so long and have so dishonored themselves and their profession that they cannot live any longer with the shame.

ReplyDelete"I was wrong." Nothing new there Don.

ReplyDeleteAs for polls, from Bloomberg, "The Bloomberg Politics/Des Moines Register Iowa Poll was in line with the average of eight polls compiled by RealClearPolitics, which showed Trump with a 4.7 point advantage heading into caucus day, and Clinton with a 4 point lead. The polls included in the average, though, showed the volatility and uncertainty of measuring voter intentions. The results ranged from an 11-point lead for Clinton in a Gravis Marketing survey to a 3-point edge for Sanders in a Quinnipiac University poll."

We know you are sour about being wrong (again) but why continually beat up the Des Moines Register? Obvious there are a lot of polls and a lot are wrong. Like bloggers.

Same reason you come to troll me. Difference: I man up and use my name, and admit my mistakes.

DeleteNo surprise that a crappy paper has crappy polls. The Register once published an op-ed piece by Donald Kaul which called for the torture and murder of gun owners and Republicans.

ReplyDeleteMark Twain had the right idea: "There are three kinds of lies: lies, damned lies, and statistics." (Polls can be thought of as a kind of statistic, no?) - Elric

ReplyDeleteBoth results on Monday night were within the margin of error. Nothing wrong with the polling, only analysts who don't understand statistics.

ReplyDelete8 is not 4. It is twice what 4 is. Difference between Trump and Rubio was 1, not the 13 predicted -- a 12-point failing. But thank for playing. Buy my books http://www.amazon.com/Exceptional-Americans-People-Need-Know/dp/1511533056/ref=sr_1_1?s=books&ie=UTF8&qid=1431232532&sr=1-1&keywords=donald+surber

DeleteCompared with the claimed 4 percentage point error of the poll, an 8 point difference in the election result is a lot more erroneous than a mere factor of 2. Assuming 4% is the 1-sigma error in a Gaussian distribution, there is a 68% chance the final result of the caucus would have agreed with the prediction by the poll within anywhere from 0% to 4% difference. An 8% difference between the predicted and actual outcome of the caucus, on the other hand, corresponds to a 2-sigma error, which has only a 5% probability of coming true. Now 68%/5% = 13.6, which is a lot bigger than 8/4 = 2. In terms of probabilities, the poll was off not by a factor of 2, but by a much greater factor of almost 14. Anyone who is that bad in polling should find another line of work.

DeletePeople advertise (lie) about their stuff all the time, what's new? Amazon review, Red bull's "Energy drink", McDonald's "healthy" choices, Homeopathy and diet-free weight loss programs. But Statistics as a tool is still the best tool we have in searching for truth and is prone to abuse. We can moralizing abuse and make everyone make a statement to condemn it but the best defence is to learn and educate people so they know how to conduct experiments/survey and interprete result themselves. Exposing methods to lie with statistic also helps preventing people from doing so. Darrell Huff wrote a book 50 years ago called How to Lie with Statistics you should check it out.

ReplyDeletePerhaps the Des Moines Register should announce a "margin of error" of 12% to make sure it comes close. What a joke!!!

ReplyDeleteTheir margin could be 100% and they would still choke.

DeleteQuit blaming us Poles! GOD, I hate this time of the political seasons. We get blamed for every thing that is wrong with American politics.

ReplyDelete