Monday, August 25, 2008

All winners!!

It's back to politics today at the start of the Democratic Convention. I may be paranoid, but I see alot of 'partisan' politics here.

The name of the game lately has been "gain or loss." Obama extends lead; McCain cuts Obama lead". The actual figures get little attention. So let's say I am a Republican leaning pollster. Today I want Obama a few points ahead so after the convention I can report a near tie, so the Convention did not help Obama. And the reverse. If I am a Democratic leaning polster I want it almost a tie today. Then after the convention if Obama is 4 points ahead I can say "Convention gives Obama big boost.

As you know I am not afraid to go to the limb. After the convention, Obama will be 2-3 points ahead on average. So both sides win.

Republican leaning poll: Rasmussen.
Democratic leaning poll. Zogby.

And don't forget margin of error (MOE). Everyone is right using MOE.

5 comments:

  1. These days it's all I can do not to shoot the TV. Feel my pain please Jack, I am suffering. Thanks for watching, you are the first person I have heard from that actually is keeping tabs on the convention. Check out my rant.

    ReplyDelete
  2. John

    Not that I should complain, but perhaps you are being a little more cynical than necessary.

    The main polling services are commercial organizations that sell polling services. The no cost to viewer presidential polls are advertising. People see that they do a good job and want to hire them for their own polling. They can’t play very many games with the numbers or it will defeat the adverting intent. As you noted the margin of error is important, they all seem to be in the margin of error of each other. They seem to move with eeach other.

    Differences can usually be figured out by reading the methodology. Rasmussen is showing definite and leaners. If you read their definition and then read the methodology of the other polls they are following one approach or the other with similar results to what Rasmussen gets for that group. Also the polls that are reporting “likely voters” (rassmussen) vs “registered voters” (Zogby)shows more support for McCain, which more likely means Obama needs to strengthen his "get out the vote" operation rather than any bias. This is a historical problem for Democratic candidates.

    The polls that are usually flaky are the political party sponsored or the one time news service sponsored polls that often seem designed to produce a result to go with a news story. They are usually obvious because the results are out of line with main polling services.

    ReplyDelete
  3. Hank, thank you so much. I post questions on "538" and never get an answer amid all the 'trash' talk. I can't get hold of you because I don't now how to register. Okay, I'm out of it, but URL is a mystery to me.

    My point is that presidential polls have no way of being checked EXCEPT by final results, and that the MOE protects almost all of them. Would it not be very easy to manipulate any poll with just a 'flick of the pencil.'

    My problem with Rasmussen is all his official 'commenters' seem very pro-Republican. And did not Zogby release a poll on election day in 2004 based partly on exit polls. Also on Rasmussen, their Ohio poll of a few weeks ago showing McCain up by ten seemed a bit strange. I may be wrong but he never seems to mention that result. I was going to post a SOS for you but you read my mind.

    Also, naked prejudice but Rasmussen describes himself as "evangelical." On a poll I am suspicion of having to factor in RELIGIOUS belief. Thanks. Jack

    ReplyDelete
  4. Joe

    You can always leave a message in the top post or send me an email message the address is in the How to Comment section at the top of the side bar. I usually check that mailbox once a week or so. I do not have anything going now except a password requirement, which is the password you login to for you own Blog.

    At the bottom of my side bar there is a HTML tutorial, which as saved me from disaster many times.

    In 2004 when Rasmussen’s undecided numbers were extrapolated to the candidates based upon their relative strength in the last poll, it missed the vote results by much less than 1%. Whatever his editorial opinion, the firm has a good reputation for having the math correct.




    Margin of error is pretty simple

    If you take a perfectly random sample it should match the true value of larger population. Of course that is impossible, and each time you take a sample it will very a little different. The MOE is the amount it will very. This is straight math it is hard to repeatedly fudge without it becoming obvious. Any one who took a college stats course, and remembers the material, can check it.


    A standard sample of 1000 items will have a MOE of three percent. The way it is usually done this means that 19 of 20 samples will be with in 3 percent of the true value and most within 2 %.

    If there is a daily tracking poll from two polling operations there results for any day will not match because they are different samples but they should both be with in three percent of the true number (Max 6% i.e 3+3) and usually with in three percent.


    Pollster and Real Clear Politics are poll averages, there averages stay pretty close to the daily tracking numbers. If one polling company were playing a game it would swing out of line from the poll average.

    The commercial firms that would hire them to do polls have people to do the math and check the numbers. We mortals may miss it but they would see a game and it would cost the polling firm cash paying business.

    I think we can be sure that the numbers are pretty solid, but of course don’t follow just one poll, you get a better picture from several

    ReplyDelete
  5. Hank, thanks for responding. (BTW, I'm Jack--no big deal).I was making two points in my querry to you.

    One. The ONLY way to evaluate a poll is by the final result of the election. So if I were a pollster, I could say Obama is ahead by, say, 15 points. But then the day before election I say Obama ahead by 2. He actually wins by 3. Good poll; very close and within the margin of error. Pretty good poll, people would say. I shall do this Nov.4.

    I understand the MOE. But what does it tell me. Rasmussen says 47-44 Obama today. But of course this could be Obama 49, McCain, 42; or Obama 45. McCain, 46. Is this not right? So what have I learned? I know something of statistics, my point was simply there is no way to evaluate a pollster till AFTER the election. BTW, I still have seen no explanation of Ras's 52-42 lead in Ohio a few weeks ago. Also, I love the term "noise." Is this just not another 'super' MOE.

    Stll love your site and thanks for taking the time to respond. I am going to comment on your blog today. Don't faint!!!Jack

    ReplyDelete