About Me

A blog wherein a literary agent will sometimes discuss his business, sometimes discuss the movies he sees, the tennis he watches, or the world around him. In which he will often wish he could say more, but will be obliged by business necessity and basic politeness and simple civility to hold his tongue. Rankings are done on a scale of one to five Slithy Toads, where a 0 is a complete waste of time, a 2 is a completely innocuous way to spend your time, and a 4 is intended as a geas compelling you to make the time.

Saturday, May 12, 2012

Margin of Error

I thought I'd share with the world an e-mail I sent to the Public Advocate at the NY Times regarding margins of error in polling...

Dear Mr. Brisbane:

I am getting really tired of articles, like ones earlier this week with a paragraphs pasted below my signature block, that consistently mis-represent the meaning of the margin of error in polling. Every single time a poll shows candidates apart, even in the mid to high single digits, the articles imply that the race is tied, within the margin of error. But that's only half true. The margin of error can just as easily go the other way. Scott Walker could be five points either ahead or behind Tom Barrett in Wisconsin. He could have been tied with Kathleen Falk or could actually have had a landslide twelve point lead over Kathleen Falk. In Virginia, President Obama could be in a very close dead heat with Governor Romney because he has a seven point lead in a poll with a four percent margin of error or he could be ahead of Governor Romney by almost fifteen points.

I'm not taking sides to say in any of these cases which way the margin of error wind is actually blowing, what I am saying is that the NY Times is distorting the meaning of the term when it implies in every single instance when it uses a margin of error in polling that the margin will always and only serve only to narrow the gap between two candidates or two sides in an opinion poll and will never extend it. I realize that the approach the Times is taking is the best "cover your ass" approach, there will be a lot more complaints about the inaccuracy of polling if you say someone is up by six points and ends up winning or losing by one point than if they ended up winning by twelve, but it's not the right way to give the public a clear and proper understanding of how to evaluate polling for themselves. And more to the point, it's just not true.

Joshua Bilmes

That advantage, however, was less apparent in a poll conducted last month by Marquette University Law School that showed Mr. Walker and Mr. Barrett essentially tied in a general election matchup. Mr. Walker led Ms. Falk 49 percent to 43 percent among likely voters, a six-point advantage that is within the poll’s margin of sampling error of plus or minus four percentage points on each candidate.


Barack Obama won Virginia four years ago, the first time a Democrat had won the state in more than 40 years. This year, it looks like Mr. Obama and his Republican challenger, Mitt Romney, may have a competitive contest for the state’s 13 electoral votes.

Mr. Obama is backed by 51 percent of voters surveyed by The Washington Post from Saturday to Wednesday, and Mr. Romney, the former governor of Massachusetts, is the choice of 44 percent. Mr. Obama’s seven percentage-point advantage is within the poll’s four percentage-point margin of sampling error for each candidate.

No comments: