August 7, 2008
Never Send to Know Whom the Poll Sells...
...It sells Barack H. Obama -- by hook or by crook.
The aptly named CBS News -- when you watch the network, you CBS -- released a poll today showing Obama leading John S. McCain by a strong six points, 45 to 39. And mirabile dictu, it actually includes a bunch of internals!
Perhaps too many... for they even included a figure that, well, tends to cast a bit of doubt upon their objectivity -- and their polling results, past, present, and future.
The line that caught my eye (ouch) was right at the end. Here is what I read:
Huh. So when they conducted the poll, they ended up with 381 Democrats and 317 Republicans, an advantage of 64 to the former. But CBS decided that wasn't heavily enough tilted towards the left... so they tossed in a multiplier that nearly doubled that advantage to 122 -- 406 Democrats and 284 Republicans.
Here's the same table, substituting percentages (out of 1,034 total respondents):
|Total Republican %||30.7||27.5|
|Total Democratic %||36.8||39.3|
|Total Independent %||32.5||33.3|
In other words, CBS considers it a reasonable prediction that in November's general presidential and congressional elections, 39.3% of voters will be Democrats, while only 27.5% of them will be Republicans... Democrats will outnumber Republicans by half again as many.
The only possible explanation is that CBS predicts one of the greatest Democratic turnouts in post-WWII American history... coupled with one of the most dismal Republican turnouts ever.
Say, might this weighting equation might affect their conclusion that Obama led the race by six points?
Here is where the poll internals came in amazingly handy. Page 3 of the pdf breaks down the candidates' support by different groups -- sex, race, age, education -- but also by political party. It was the work of a couple of minutes to make an algebraic formula to see how changing the party percentage would affect the final results.
First, I used the weighted percentages; I got 45.4% for Obama and 38.7% for McCain... which by rounding turns into 45% to 39%, just as CBS reported. This confirmed that my formula worked.
Next I plugged in the original, unweighted party percentages; this changed the results to:
- Barack H. Obama 43.6%
- John S. McCain 40.6
Using the unweighted sample -- which already has a substantial advantage for Democrats -- reduces Obama's lead from six to three points... which by an amazing coincidence is exactly the margin of error.
So the CBS poll, using the raw sample, found Obama and McCain in a statistical tie... but after some quick manipulation of the party percentages, increasing the Democratic advantage markedly, they ended instead with a significant and fairly substantial lead for Barack Obama.
Just for giggles and grins, I also tried pluggin in the "null-hypothesis" assumption -- which I don't actually think very likely -- that just as many Republicans as Democrats will vote; I changed the Democratic and Republican numbers to 349 each, leaving the Independent figure at the raw level of 336. Under this scenario, the results would be:
- Barack H. Obama 41.7%
- John S. McCain 42.7 (McCain ahead by 1%)
Consider this the outside right edge of the CBS poll; I consider the raw percentage they actually found as the real outside left edge. So as far as I'm concerned, this poll's actual result is a range from Obama up by 3 to McCain up by 1, all within the margin of error.
Until, of course, one adds the appropriate weighting -- to get the "correct" result of Obama up by 6.
Hatched by Dafydd on this day, August 7, 2008, at the time of 4:48 PM
TrackBack URL for this hissing: http://biglizards.net/mt3.36/earendiltrack.cgi/3187
The following hissed in response by: kimsch
Dafydd, of course there's more Democrats. New Democrat registrations rose in droves during the primaries!
The pundits are always talking about how many new Democrats there are. About how droves of Republicans were switching their registrations!
And they believe that all those new registrations are real.
Of course Rush Limbaugh and Operation Chaos had nothing to do with it. Neither did the fact that registration means nothing in the general election.
I think the Democrats will all be *not* pleasantly surprised come November.
The above hissed in response by: kimsch at August 7, 2008 5:33 PM
The following hissed in response by: TerryeL
I saw this poll and told myself that since it was CBS give McCain about another 6 points. I have wondered why the polsters jack with the numbers like this, and I think part of it is fundraising. They want people to think Obama is a sure winner so that the money will keep coming in. And they want to make it easier to claim they were robbed if they do in fact lose.
I have no idea who is ahead at this point, but it is a looong campaign. Anything can happen.
The following hissed in response by: GW
Nice catch. I remember reading one of Dick Morris's books where he dissected several NYT polls on precisely these grounds. At any rate, toss in the Bradley effect, and the One ought to be a bit concerned.
The above hissed in response by: GW at August 8, 2008 12:11 AM
The following hissed in response by: Rovin
Newest chapters in the evolution of journalism----Stay with in the margin of error, AND, "reputable" sources need not be revealed when publishing fabrications if it's in the best interest of the general public---two full chapters on Eason Jordan's school of thought.
The above hissed in response by: Rovin at August 8, 2008 4:50 AM
The following hissed in response by: TerryeL
I heard or read somewhere that according to Rasmussems Democrats lost 2% on the overall party ID last month. Some say it is the alienated youth thing. It seems Obama is a politician and they are just now catching onto that.
The following hissed in response by: kimsch
Yep, they are finding out that he's just an empty suit. And he's supposed to be my junior senator. He's really good at taking no stand at all. He voted present more than yeas or nays while in the state senate. He got the state senate job by disqualifying everyone else (including the incumbent) in the primary. He got the US Senate seat because his main primary rival's divorce papers were made public, then his Republican rival's divorce papers were made public. Then the IL GOP idiotically brought in Alan Keyes to run against Obama.
The above hissed in response by: kimsch at August 9, 2008 9:04 AM
The following hissed in response by: RunningRoach
I have 4 kids...aged 27 to 37....along with 3 grandkids. They are spread up and down the east coast...RI,CT,NY,NJ and we were all toghether for one of the little grandkid's birthday parties yesterday. So between kids and spouses I counted 8 plus my ex wife, her two sisters and one brother. None of the "adults" have less than a masters degree in whatever. Two of my kids are VERY liberal along with their spouses. (RI and NY.) I brought up Obamanaa and said, "He might be good for the country. You know....major change and all that"...and that CBS has him up in the polls. My most liberal daughter said' "Dad, have you flipped out? My oldest son said "Say what!!!" " what have you become???" My ex wife just laughed under her breath and wispered " bananna head" to her sister, another lib. So my small sample of mixed liberals and conservatives, not counting the kiddies has O-man down 12 zip. So much for the CBS poll. Just another dune of donkey dust bought and paid for by those who wish to lead but don't have a clue how.
The above hissed in response by: RunningRoach at August 10, 2008 6:32 PM
The following hissed in response by: scrapiron
So Can't Broadcast Sh** (CBS) can't read or do simple math. What a shock, and they wonder why they are losing viewers by the millions. I'll bet they learn to count when the paychecks start to get smaller or don't show up. No news is better than lying news.
The following hissed in response by: Jon S.
Kimsch above has a very good point about the newly registered voters, some of whom (maybe 15%, maybe more?) are Rs who registered via Operation Chaos. Many more, I'd guess, are simply low-income folks who were approached by ACORN types in housing projects, parks, etc, and who have no intention of voting (or may have some intention but will not follow through). And then there are those who are not low income per se -- especially students -- who will by partying on Election Day instead of voting. So my guess is of all the many new registrants, no more than 1/3 or so will actually vote.
Next, there has always been a lead in Dem party ID; while it was quite small in 2004, for most of the past 30 years it has been quite a bit bigger. That many Ds ultimately end up staying home or voting for the Republican is clear (see elections of Reagan, Bush I, and Bush II). I suspect that this time we'll see the same phenomenon; it's a referendum on Obama and the guy, for many reasons, cannot close the deal. At the very least McCain will win by a similar margin than did Bush in '04; at best it'll be a 40-42 state blowout with some very surprising blue states turning red.
Post a comment
Thanks for hissing in, . Now you can slither in with a comment, o wise. (sign out)(If you haven't hissed a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Hang loose; don't shed your skin!)
© 2005-2009 by Dafydd ab Hugh - All Rights Reserved