I am pretty sure this is by far the largest Texas lead that I have seen so far. I decided that I am not going to trust this poll after reading the article about it, but the major margin is worth this post.
Arlington, Texas (Feb 21, 2008)- Barack Obama has a double digit lead over Hillary Clinton in Texas, according to a statewide survey of 678 registered voters planning to vote in the March 4 Democratic Primary. The survey was conducted February 20 and 21 by Decision Analyst, a major national survey research firm. Decision Analyst projections indicate that if the election were held today Obama would win 57% to 43% over Clinton. The survey's margin of error is 3 percentage points, plus or minus, at a 90% level of confidence.
Already here I saw a problem (although it is a minor one). Almost every [political] public opinion poll uses a 95% confidence level. Usually the level of confidence is not even reported for this reason. I do not know enough about public opinion research to say how much that would affect the Margin of Error, but to read this poll accurately -- the margin of error would clearly find itself above 3% normally.
This was just a minor issue for me, though. On a normal poll, I would imagine that a 678 sized sample for the state of Texas wouldn't get any obscene margin of error. What bothered me was the paragraph I read about the poll and the firm.
Results based on scientific sampling of registered voters from American Consumer Opinion® (www.acop.com), one of the largest online research panels in the world. The sample was carefully balanced by gender, ethnicity, age, and geography; the data were weighted as necessary to fully represent the different demographic groups.
About Decision Analyst (www.decisionanalyst.com), based in Dallas-Fort Worth, is a leading international marketing research and marketing consulting firm specializing in advertising testing, strategy research, new product development, and advanced modeling for decision optimization. In addition, Decision Analyst owns and operates American Consumer Opinion® Online, one of the largest research panels in the world with over 7 million members.
A "scientific sampling?" I realize that American Consumer Opinion has over 7 million members worldwide, but I'd be willing to bet that there are over 7 million registered voters in the state of Texas (I could be way off, but I'm sure the number is in the millions). I'm sure the number of American Consumer Opinion's Texas panelists is a minute one compared to the actual population of registered voters.
Scientific samples are supposed to be a random sample of the population, and this was not random -- instead, it was a sample of Texas registered voters who decided on their own accord to register with www.acop.com. Some pollsters believe that public opinion research will eventually get to become electronic based, but we are not yet there.
--
Ok, so what does this survey tell us? Well, it probably tells us that Barack Obama is winning the internet users of Texas Democratic voters by a wide margin.
As if we already did not know that.
(I'm only an amateur on analyzing public opinion. I take a class on it at UT from Daron Shaw. It's possible that he has taught me wrong thus far, but I doubt it. I trust the man on this - if I remember correctly, he worked on polling for President Bush's presidential campaigns, and you have to be good if you can help get that man elected.) |
