« BACK  |  PRINT

RS

MEMBER DIARY

Are the Polls Skewed?

I don’t know if the polls are skewed or not but they certainly are in disagreement with each other.

Below are some facts and figures with lots of links so you can make up your own mind.

Thanks to polls, Harry Truman will always be the patron saint of political underdogs, yet by and large most polls come reasonably close to being accurate.  However, sometimes they are wrong, sometimes wildly so, and sometimes suspiciously so.

There are several questions that must be asked anytime the result of a particular poll are at issue.

Sample size. Margin of error. Demographic breakdown of those polled etc. Party affiliation is a huge issue. It doesn’t take a high functioning brain to understand that a poll made up of 500 Republicans and 500 Democrats will almost certainly give different results than a poll made up of 400 Republicans and 600 Democrats.

It also is reasonably clear that to be accurate a poll should closely mirror the entire population in a broad range of demographics.  However party affiliation is a uniquely difficult nut to crack.  Except in rare circumstances party affiliation is alone in being a major demographic that sees changes on almost a daily basis. A person’s mood doesn’t determine what sex they are or how old they are, or where they live, or whether they are married or not.

But even if it was known with certainty that 45% of the population are Democrats and 43% are Republicans, that still would not tell us in what proportions they will vote in a particular election.

So what’s a pollster to do? If I understand it correctly, most pollsters have a secret decoder ring that tells them what the approximate partisan mix is and use those numbers to forecast turnout based on various likely voter screens.

All of the foregoing is relatively easy to understand, comprehend and accept as a valid practice in what is obviously an inexact exercise.

Here is where it gets murky and inscrutable. We have to go back to that secret decoder ring. How is party ID determined? Does the pollster call up 1000 people and ask whether they are an R a D or an I? Or does the pollster call up and ask  “Do you agree with Mitt Romney that 47% of Americans suck?” and then ask about their party ID?

Also how does the pollster draw the line between either party and Independent. Does the pollster leave that strictly up to the person being polled or do they ask a series of questions designed to screen for people who say they are independent but almost always vote for one party?

The Pew Research Center has a nifty slide thingy that enables you to see the changes in party ID from year to year. Whether you think Pew’s numbers (before 1990 they are actually Gallup’s numbers) are accurate or not doesn’t matter. What is interesting is how their numbers move.

What is immediately apparent is that the number of Independents increase when one party or the other has an unusually bad situation on their hands. Currently the number of independents (according to Pew) is at an all time high of 38%.

YEAR    D      R       I
2004    33    30    30
2006    33    28    30
2008    36    25    32
2010    33    25    36
2012    32    24    38

What I find intriguing is that over the past eight years BOTH parties have hemorrhaged supporters. Even more interesting is that over the past four years the Democrats have lost 4 points while Republicans have only lost 1 point. And finally I’d point out that the current D-R spread is exactly the same 8 points as it was in 2010 which was a pretty good year for the GOP.

Since the polls are all over the place which one or ones should we rely on? Let’s take a look at how Scott Rasmussen did over the last six weeks of 2008. By that time the race had settled down, all the convention bounces were over, Lehman Brothers had gone bust, and it was universally understood in the media that Sarah Palin was a hopeless dunce.

Below are the seven polls listed here done by Rasmussen in that time frame:

Date                        Obama    McCain
September 26    50              45
October 13          51              45
October 28          51              46
October 29          50             47
November 1        51             47
November 2        51             46
November 3        52             46

Notice that of those seven different reference points, only one was what might be called an outlier and even it showed Obama at 50% and three points ahead.  Conclusion: In 2008 Scott Rasmussen consistently hit the mark.

In summation,  this race is closer than most polls indicate.  Based on the over-all polling data Obama is probably slightly ahead. Rasmussen is the best firm to watch. If Obama starts to open a large lead in Rasmussen that would be intensely bad news. However at this point Obama and Romney are tied at 46 and that indicates its anyone’s game.

Finally, here three links I’d recommend to get a a cross-section of thought on this subject.

Jim Geraghty at NRO yesterday.

David Lauter at the L.A. times a couple months ago.

Jay Cost at the Weekly Standard today.

(I’ve been working on this off and on all day as my pesky business permits and I see now that while Rasmussen moved the race to even today 46-46, Gallup has gone hog wild this afternoon and now shows Obama up 50-44. My observation of Gallup over the years has taught me that they occasionally have hiccups like this. I suspect that margin will contract over the next several days. No way should a seven day tracking poll go from 48-45 to 50-44 in one day. )

Get Alerts