EXPOSED: Reuters/IPSOS Poll Over-Samples Dems by 13% – and LIES about Percentage!

UPDATE: Shortly after this was posted, Real Clear Politics removed this poll from their list of averaged polls. Unfortunately, they replaced it with a Democratic Poll by Democracy Corps that sampled 50% Democrats and 43% Republicans (with leaning) that also under-sampled independents. At least it is honest about it’s skew.

I have been watching, analyzing and digesting polls for a long time. And one of the saddest trends that has been happening more and more often are polls that use sample sizes to distort the results. Why would any poll want to distort results?

Mostly to attempt to influence opinion rather than simply measure it.

Groups like PolitiFact pretend to “check facts” while most honest observers know they are spinning for the Democrats. Once you move away from true or false and go into degrees of accuracy, it is no longer fact checking. It involves opinion. And when your lean is decidedly left, there simply is no value in the facts you pretend to check.

The same is true of polls, though it is a lot harder to lie. You must actually manipulate the sample size to get the desired outcome. And most of these polls skew to the left – some badly. And when you look at the numbers behind the poll results they publish, you can easily see this manipulation. Usually these pollsters simply call a lot more Democrats than Republicans to make the Democrats look good.

And an even newer trend I have seen is for the pollsters to actually poll a pretty close number of Republicans and Democrats, but they poll a tiny number of Independents. Now that wouldn’t matter if the Indy’s were splitting 50-50 for Romney and Obama – but they are not! The Independents are going for Romney by double digits! So if you sample fewer Independents, you essentially skew the poll towards Democrats.

And normally, these shady polls count on you only looking at the results, not the mix. And the devil is always in the details.

But I have never seen one actually fudge the percentages to make it look like their sample was not out of touch with the political makeup of the country.

Until now!

A brand new Reuters/IPSOS poll goes to a whole new level. Not only did they badly over-sample Democrats and horrifically under-sample Independents, they lied about those numbers. If this were not a polling firm that makes their living (and reputation) from calculating and crunching numbers I would give them the benefit of the doubt. Accuracy is their life-blood. And if they didn’t intentionally miscalculate the numbers they published, then they didn’t lie, they are simply incompetent.

Either way – this is not good.

Let’s start with the bogus percentages of Democrats vs. Republicans vs. Independents. Did they think people wouldn’t check their math?

At the top of page 1 of the Reuters/IPSOS Poll we find the following Demographic Information:

Interview dates: Sept 7-10, 2012
Interviews: 1,089 adults; 942 registered voters, 873 likely voters
572 Democrats; 439 Republicans; 78 Independents

If we total the last line of Dems, Repubs and Indy’s we get 572 + 439 + 78 and that totals 1,089, which is the number of people they called. So far so good.

Now we need to figure out the percentages of each political leaning group they called. For the mathematically challenged, have no fear. It is REALLY easy.

To calculate the number of Democrats called out of total number of people called, you simply divide the number of Democrats – 572 –  by the total number of people called – 1,089. So, in your calculator enter the number of Dems – 572, press the “/” key (which means “divided by”) and enter the total number of people called (1,089). And if you did everything right, you should get a long number like this: 0.52525252525252525252525252525253 

And since we want the percentage we need to multiply by 100, but we can do that in our heads by moving the decimal TWO PLACES TO THE RIGHT. So instead of .525 (point five-two-five) we now have 52.5 (fifty-two point five). (Note, just ignore the first zero and we really only need the first 3 numbers. No need to make this really hard.)

And let’s round 52.5 to 53 and add the percent sign. We get 53% Democrats polled in this survey.

(My apologies to the math majors, I want to make sure others can understand these calculations.)

And we figure the other numbers the same way. 439 Republicans divided by 1,089 total called times 100 (for percentage) and round to nearest whole percentage we get 40% Republicans and 78 Independents divided by 1,089 total called times 100 = 7%.

So, we see they polled:

53% Democrats

40% Republicans

 7% Independents

(And if you add the percentages up you get 100% – just checking!)

So, at this point we will not address the glaring fact that way more Democrats were called than Republicans – and almost no Independents were called.

We will first look at the number of Republicans, Democrats and Independents IPSOS polled in this survey. This is where the lie (or incompetence) comes in.

On the very last page of the survey, we find completely different numbers from what we just calculated. (The document is here if you already closed it.)

(Click Table to Enlarge)

We are concerned with the first column of numbers labeled “ALL” (since that is what we calculated). And we will look at their numbers for Total Democrats, Total Republicans and Independents.

The table below compares the figures IPSOS reported here with the raw number percentages we calculated from the first page of the poll.

So, IPSOS and Reuters would have you believe that they sampled 47% Democrats and 38% Republicans, “only” a 9 point advantage for the Democrats. But their breakdown on the first page tells us that they polled 53% Democrats and 40% Republicans – a whopping 13 point advantage for Democrats!

Did IPSOS intentionally fudge the number of Democrats and Republicans or is this a simple case of polling professionals being sloppy? Human error? Incompetence?

Does it really matter?

The fact is, this poll is rubbish and needs to be discarded. Real Clear Politics should reconsider the polls they include in their average.

Garbage in – garbage out.

What this poll should tell you is that Obama is in serious trouble. In a poll that gives Democrats a 13 point head start, he only manages to lead by 3 points. Even of you consider the 9 point lead of Democrats IPSOS figured, this is bad for Obama.

And to take this one step further, the tiny sample of Independents also works to Democrats advantage. While most polls show a very high percentage of self proclaimed Republicans and Democrats are planning to vote for the candidate of their party (most have this in the high 90% range – 97% or so) Independents are consistently going for Romney by double digits – 15% or 20% more in poll after poll.

So, if you under-sample Independents where Romney has a pretty good advantage and over-sample Democrats by as much as 13% and only come up ahead 3 points, then you are going to lose. No creative and deceptive polling results are going to change that.

And while we are on the topic, exactly what should the percentages be to have the polls accurately measure public opinion?

In 2008 Obama enjoyed an advantage in the number of people who identified themselves as Democrats and he won the election by 7 points, 53% to 46. But in 2010 the elections showed a strong turnaround. A Gallup poll showed (in 2010) that the US was made up of 31% Democrats, 29% Republicans and 38% Independents.

But in August 2012, Rasmussen found that Republicans actually outnumbered Democrats 37.6% to 33.3%, a 4.3 point Republican advantage. The same poll showed Independents at 29.2%.

So in order to be accurate, a poll’s Demographics should closely reflect the makeup of America. And while there may be more Republicans than Democrats for the 2012 election, rarely does a poll gives an advantage in numbers to Republicans. And if there is, it it only one point.

This IPSOS poll is so far off from being anywhere close to the Demographics of the country that it is worthless unless you consider the huge advantage Democrats have and the 3 point margin for Obama in a poll skewed so badly.

Reuters should fire IPSOS and find a polling firm that can at least do percentage calculations.

Related Posts

• Poll Analysis: Adjusted Polls from CNN – ABC/WaPo and IBD/CMS/TIPP Show a 4 Point Lead for Romney
• New PPP Poll Shows Manipulation And Bias
• New Poll – George Allen up in Va Sen by 5 Points and Romney Up by 5
• Allen Increases Lead Over Kaine: Now Up 8%
• Which Polls Can You Believe?