Re: Gallup organization on polling methods

[dennis roberts]

Ladies and Gentlemen of the Jury ... I come to praise Cesear ... not to stone him ... hmmm, that's a different talk.

I think people really need to read (if interested) the link that Jill originally sent ...

http://www.gallup.com/poll/FromtheEd/ed0211.asp

It was written by Frank Newport, EDITOR IN CHIEF OF THE GALLUP POLL ... he has a strong and vested interest in putting pollsters in general and the Gallup poll in particular, in the most positive light. He is going to highlight the positives and downplay (or not mention at all) the negatives. That is his job. I will assume that he is a good and honorable person ... but, that STILL is his job.

We all know of the many problems with polling ... NONE of which he mentions occurs in his poll ... such as "slamming", the phenomenon of people who answer the phone "slamming" down the phone as soon as they discover what you want. And, that is not the only problem. Let's look specifically at what Newport said:

--------------
On the nights of Oct. 31 through Nov. 3, The Gallup Poll talked with exactly 1,221 Americans on the telephone and asked them to indicate for whom they planned to vote in their local congressional race in the Nov. 5 election. But these weren't just any 1,221 Americans. They were selected using a random sample of all adults living in the country, using a process called Random Digit Dialing, or RDD.
--------------

(We know of course that the statement " ... of all adults living in the country ... " is not true ... telephone polling does NOT have the potential of reaching all adults living in this country)

My first question is: where on earth did the Gallup poll come up with a number like 1221? It certainly does NOT square root nicely as, some sampling error formulas for proportions would like.

He first says that the Gallup poll talked with exactly 1221 Americans ... (and later) ... were selected using a random sample of all adults living in the country ...

Now, I can believe that they talked to 1221 people, presumably Americans (if you call and talk to enough, you can GET 1221 eventually) ... but, IS THAT THE NUMBER AND SAMPLE THEY ORIGINALLY SELECTED OR ... IS THAT HOW MANY THEY ENDED UP WITH? This distinction is very important.

If we assume that they indeed did PRE select these 1221 ... there is no way on earth that they could have reached all 1221 and EVEN if they did, that all 1221 would have talked to Gallup. So, we have to dispel this notion that Gallup preselected 1221 Americans ... from the telephone lists ... and talked to those 1221.

Since they have no way of knowing BEFORE hand whether there will be a person in the house who is an eligible voter ... you have to wait and see AFTER someone is willing to talk to you IF they indeed are a registered voter ... to then follow up with your real questions.

So I ask: HOW MANY TELEPHONE CALLS DID IT TAKE ... REACHING PEOPLE WHO WOULD TALK AND WHO WERE REGISTERED VOTERS ... TO ALLOW YOU TO >>>>>> end up with <<<< 1221?

Mr. Newport makes no mention of this ... in fact, makes it sound like they selected 1221 and got all 1221. So, we know that can't be true. How many people did they call and reach who refused to answer some of the questions or, simply hung up the phone? Mr. Newport makes no mention of this ... makes it look like no one did this.

Mr. Newport has not really told us how they got their sample of 1221 ... let's face it.

What I think happened was: they selected out a large batch of phone #s (maybe even at random) ... and, made some rough ahead of time decision to find at least a 1000 ... maybe somewhat more ... and then, kept dialing and dialing from this list, and even over and over again in some cases, UNTIL THEY FINALLY GOT WHAT THEY THOUGHT WAS ENOUGH TO QUIT. It just happened to end up at 1221.

So, we really have no idea of whether they are "typical" of the "population of likely voters" or not. It was NOT a SRS of Americans who were "likely voters".

Of course, you might look at his article at the bottom and see their results compared to the final REAL results and since they are so close, say ... they must have done something right. But, please notice that in the paper, they whittled down the 1221 to 715 ... and, they just say that they used "their methods" to decide who looked like a likely voter or not. ONE CAN ALWAYS AFTER THE FACT ... MAKE YOUR SAMPLE CONSISTENT WITH ACTUAL RESULTS ... and we have no way of knowing based on this article, how the final group of 715 was determined.

I say, they either helped themselves OR ... just got lucky.

If you want to really examine the accuracy of polls, you need to take a large cross section of polls done in local, state, and national elections ... at various times ... and compare the PUBLISHED poll figures say a week before the vote and, what actually happens and, I think you will find that polls are not nearly as "accurate" as we tend to "think" they are.

Certainly, if we are interested in accuracy, we cannot put much weight on ONE result that the Editor of the Gallup poll spread out before us from THEIR poll ... as THE evidence we need.
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================