Bad Polls

Poll Underdog?

The Drudge Report is linking to this Reuters article with a headline that reads “Poll Underdog”:

Democrat Barack Obama has a narrow 5-point lead on Republican John McCain in the U.S. presidential race, but holds a big early edge with the crucial swing voting blocs of independents and women, according to a Reuters/Zogby poll released on Wednesday.

Two weeks after clinching the Democratic nomination and kicking off the general election campaign, Obama leads McCain by 47 percent to 42 percent. That is down slightly from Obama’s 8-point advantage on McCain in May, before Sen. Hillary Clinton of New York left the Democratic race.

On the last line of the three-page story sits an incredibly important bit of information:

The national survey of 1,113 likely voters, taken Thursday through Saturday, had a margin of error of 3 percentage points.

That means Obama could be as low as 44 and McCain as high as 45. Translation: A statistical dead heat. McCain’s only a “poll underdog” if you pretend that the statistical laws surrounding margins of error are meaningless.

But hey, let’s not let math get in the way of a sensational headline or a compelling story.

By | June 18th, 2008|Uncategorized|0 Comments

The lede in the Des Moines Register:

Barack Obama has pulled ahead in the race for Iowa’s Democratic presidential caucuses, while the party’s national frontrunner Hillary Clinton has slipped to second in the leadoff nominating state, according to The Des Moines Register’s new Iowa Poll.

Here’s the actual numbers:

Obama, an Illinois senator, leads for the first time in the Register’s poll as the choice of 28 percent of likely caucusgoers, up from 22 percent in October. Clinton, a New York senator, was the preferred candidate of 25 percent, down from 29 percent in the previous poll.

Former North Carolina Sen. John Edwards, who led in the Register’s May poll, held steady with 23 percent, in third place, but part of the three-way battle.

Wow, pretty close race really. Better look at the margin or error:

The telephone survey of 500 likely Democratic caucusgoers was conducted Nov. 25 to 28 and has a margin of error of plus or minus 4.4 percentage points.

That’s a huge margin of error. It means that every candidate might really be in a nearly 9-point (8.8) range. Edwards could be as high as 27.4 and Obama could be as low as 23.6. So, the new poll actually shows all three candidates in a statistical dead heat.

Why didn’t the Des Moines Register point this out? Because being honest about polls doesn’t make for good headlines.

All the major news outlets are guilty of similar shenanigans. No wonder media credibility numbers are in the toilet.

By | December 2nd, 2007|Uncategorized|0 Comments

Bad poll

Incredibly poor news judgment shown by the editors of the Wall Street Journal Online. This story‘s headline reads: “Poll Shows Strong Public Support For Range of Health Practices.” One of these health practices is a 75 percent support for universal health care, the message touted in the email alerting me to the poll.

But, the poll was conducted online — presumbably by computer users who chose to log on and answer the questions. By no means does this represent a statistically representative sample of the “public.” (Very little detail was given regarding the collection of the poll, but online surveys as a rule can’t be called representative because no attempt at getting a cross-section of the public is made.)

The fact that the editors chose to make this into a news story either shows (a) their bias in favor of universal health care or (b) their illiteracy in reading polling data. At the very least, the story should point out that the poll isn’t representative. And the headline shouldn’t have the word “public” in it.

By | October 20th, 2005|Uncategorized|0 Comments