Captain Ed fisks CBS and the NYT latest effort in creative interpretation:
CBS and the New York Times have a new poll out that looks at the Democratic primary race and at the general election. In the former, it uses a rather small sample, but in the latter the sample gets weighted — as usual — in favor of Democratic voters. Barack Obama has taken a lead in the national numbers for the primary, not exactly breaking news:
“A new CBS News/New York Times poll finds Barack Obama with a 16-point lead over rival Hillary Clinton among Democratic primary voters nationwide.
Obama, coming off 11 straight primary and caucus victories, had the support of 54 percent of Democratic primary voters nationally. Clinton had 38 percent support.
In a CBS News poll taken three weeks ago, shortly before Super Tuesday, Obama and Clinton were tied at 41 percent. Clinton led by 15 points nationally in January.
The former first lady has lost her advantage among women, according to the poll: The two leading Democrats now have even levels of support among female primary voters.”
How did CBS reach this conclusion? They polled 427 Democratic voters. That isn’t an exceptionally strong sample, and it produces a conclusion that is a likely outlier. Gallup, AP, and Rasmussen all show Obama leading but in a much closer race.
The problems increase when the poll includes Republicans. They show Obama beating John McCain by twelve points, 50-38. However, the sampling and weighting explains the strange notion that John McCain would only get 38% of a general election vote. The sample of 1152 respondents comprises 358 Republican voters, 420 Democrats, and 337 independents. Here’s how they weight the sample:
Democrats – 419
Republicans – 318
Independents – 325Get the math? They deducted 40 Republicans, 12 independents, and a grand total of 1 Democrat for their weighted sample. The original configuration would have made Democrats 37.6% of the sample, and Republicans 32% – almost exactly how Rasmussen
breaks out party affiliation. Instead, the weighting makes Democrats 39.5% of the sample, and Republicans just a shade under 30%.If your butcher did this, you’d demand that he take his thumb off the scale. These results are completely useless, and once again CBS and the New York Times report more on their own credibility than on the mood of the electorate.
Even better news is to be found from the crosstabs of the latest LA Times poll:
In this case, the sample also has problems. It uses 1246 registered voters, not as reliable for predictive models as likely voters, which usually puts Republicans at a disadvantage. It also only includes 290 Republican primary voters against 436 Democratic primary voters, a definite advantage for Democrats.
And yet, the results show McCain ahead of both Obama and Hillary, and not just on the war and foreign policy. He leads Obama on the economy by eight points, 42%-34%. He also beats Obama on illegal immigration. (Clinton edges McCain on both issues.) And on leadership, McCain beats both Hillary and Obama, with the survey showing him as the “strongest leader” for the country.
These results come from a sample and survey methodology that should have helped the Democrats. Imagine what a clean and balanced sample of likely voters will find.
For the record, what was the actual partisan breakdown of the electorate on election day for the past three cycles?
February 27th, 2008 at 12:36 pm
Yeah what the hell is this poll? The New York Times is such a laughingstock….even
funnier when they get together with CBS….
February 27th, 2008 at 12:42 pm
Kavon,
Pollster.com does a great service with its weighted average for Bush’s job approval. You can look at each polling company and see if their methodology has tended to produce a result that favors Bush relative to the trendline or is below it.
http://www.pollster.com/presbushapproval.php
CBS/NYT has consistently been below the trendline. In fact, for the past three years, in the past forty-nine iterations, it appears to have produced a result above the trendline three times (and barely so at that)
February 27th, 2008 at 12:53 pm
I looked at today’s Rasmussen poll, and the numbers remain pretty good. It is only a matter of time before McCain is considered the frontrunner and Obama the underdog in public perception for the general election. McCain does not always do that well as the frontrunner, but general election polls are far less fluid than nomination polls.
February 27th, 2008 at 1:28 pm
I don’t know if I’d agree that the LA Times poll is particularly biased. The registered voters sample certainly skews things, and in the past polling only registered voters has tended to favor Democrats. But, I think the Republican vs. Democrat primary samples are probably pretty fair. I mean, they polled half and again more Democratic primary voters. Then again, that comports pretty reasonably with the actual primary breakdown. It doesn’t strike me as unreasonable to assume that Democratic primary voters are as likely to show up in the general election (relatively speaking) as their Republican counterparts. The CBS/NYT poll is idiocy though.
February 27th, 2008 at 1:35 pm
The only good information you can glean from CBS/NYT polls or AP-Ipsos polls because they consistently favor more Democrat partisans in their sample is to check the trendlines from each iteration of their polls.
Looking at one poll and trying to figure out an actual snapshot in support is like trying to find up or down in outer space.