Saturday, April 19, 2008

Statistics Makes Liars, Too!

If you don’t think method matters, look at two polls conducted this week by SurveyUSA. A Monday poll shows Hillary Clinton up 16 percent in Indiana. By Friday, Obama was up by five percent. Did Clinton get a DUI outside Bronko's?

Nope. Slate.com explains that the Clinton-leading poll dialed numbers at random, meaning it got unregistered voters in the mix. In contrast, the Obama-leading poll used numbers provided from voters themselves when they registered.

Yeah, I’ll take the latter set of numbers every day of the week. I predict those who provided phone numbers WILL be voting, if for no other reason than to stop the barrage of phone calls from operatives and recorded elected officials telling them to "remember to vote."

Here's your story:

http://www.slate.com/id/2175496/


Share/Save/Bookmark

5 comments:

Anonymous said...

Your post points out that people reading about political polls must take care to understand the methodology of how those polls are put together. But, you misunderstood the impact of including unlisted phone numbers in a research project, like a political poll. The survey company, if they decide to include listed and unlisted phone nuimbers, creates a potential sample of numbers to call. Those numbers are picked randomly. In a political poll, when answered, the interviewer would ask IF the voter is registered and likely to vote in the election. Random dialing of listed and unlisted numbers is a normal practice of researchers and pollsters. I was surprised that the poll from IUFW did NOT include unlisted numbers. Minorities (Blacks, lations) and younger people are more likely to have unlisted (by choice or got their number after a directory has been published). The latest poll (IUFW) would underrepresent those groups. Which means Obama could be doing better in Indiana.

Anonymous said...

the poll also indicates that only 78% of African Americans plan to vote for Obama...this is very very low from the results in other states......The number will be closer to 90%.

Chris Worden said...

To Anon and Anon:

Good comments both. 2nd Anon, I agree. When I read the cross-tabs, I said there is NO WAY Obama is only 78% African-American support. That would absolutely defy the national numbers, so we know this survey is off.

Anonymous said...

The Obama Team volunteers in one particular township have and continue to be overwhelmingly African-American. And no, I am not referring to Center.

varangianguard said...

There are three kinds of lies: lies, damned lies, and statistics.


Benjamin Disraeli