AP Poll Biased: Anti-Bush, Anti-Republican

Today (Saturday) there are stories in numerous papers based on an AP-Ipsos poll just released. Typical of the lot is an article in the New York Post, whose lede and third paragraphs are here:

Dissatisfied with the nation's direction, Americans are leaning toward wanting a change in which political party leads Congress - preferring that Democrats take control, an AP-Ipsos poll found. Democrats are favored over Republicans 49 percent to 36 percent.

President Bush's job approval remains low - 40 percent in the AP-Ipsos poll, with only one-third saying the country is headed in the right direction. Bush also remains low on his handling of Iraq, where violence against Iraqis and U.S. troops has been surging.

This and other results in this poll would be bad to very bad news for the Bush Administration, if the poll were statistically accurate. It is not. The poll is stacked in most of its demographics to favor pro-Democrat results.

Source: http://breakingnews.nypost.com/dynamic/stories/P/POLITICS_AP_POLL?SITE=NYNYP&SECTION=HOME&TEMPLATE=DEFAULT&CTIME=2006-01-07-02-06-45

Another AP story drawn from the same poll and published on the Internet, had this lede and third paragraph:

A majority of Americans want the Bush administration to get court approval before eavesdropping on people inside the United States, even if those calls might involve suspected terrorists, an AP-Ipsos poll shows.

Yet 56 percent of respondents in an AP-Ipsos poll said the government should be required to first get a court warrant to eavesdrop on the overseas calls and e-mails of U.S. citizens when those communications are believed to be tied to terrorism.

Source: http://apnews.excite.com/article/20060107/D8F02GJO6.html

AnkleBitingPundits did the heavy lifting to examine the inputs for this AP-Ipsos poll. Only 81% of respondents were eligible to vote, with no indication whether even those were registered and did vote in the last election. Respondents were 52-40 percent Democrats over Republicans, even though in the last election the split was 37 each (the rest being unaffiliated or third party).

Every single one of the following crosstabs from the poll are wrong, and in every instance the error favors the Democrat responses. Each is compared to actual voters in 2004. Religion: 19% said they had “none,” versus 10% in 2004. Age: 31% were 18-34, versus 17% for 18-29 (closest available data). Income: 15% under $15,000, versus 8%.

Marriage: 56% married, versus 63%. Geography: 17% rural, versus 25%. Race: 71% white and 12% Hispanic, versus 77% and 8%. (Remainder were black, and other.)

Source: http://www.anklebitingpundits.com/index.php?name=News&file=article&sid=2916&mode=nested&order=1&thold=0

A first-year student in statistics would be flunked for turning in poll results as obviously bad as this AP-Ipsos poll. Yet thousands of articles will appear in all media, across the nation and the world in the next few days, based on this poll.

At the very least, the AP should withdraw the poll and apologize for it. At worst, those who purchase the “news” services of AP should demand their money back.

John_Armor@aya.yale.edu