Monday, September 03, 2007

Insignificant other

It's perfectly all right for editors to demand that reporters know what they're saying and say what they mean -- not in the Orwellian sense of "if you don't know what a word means, find a dysphemism for it" but in the sense of "be sure it doesn't mean something you don't want it to mean." Polling season, which is looking more and more like a year-round effort, reminds us to be particularly careful:

U.S. Sen. Hillary Clinton can claim stronger support for president in the Charlotte region than any other leading candidate in either party, according to an Observer/WCNC-TV poll conducted last month.

OK so far. Next?

The two other Democrats named in the poll, U.S. Sen. Barack Obama of Illinois and former U.S. Sen. John Edwards of North Carolina, trail Clinton significantly. Overall, 14 percent said they'd vote for Obama and 13 percent picked Edwards.

OK. Clinton came in at 28% on this question (asked of all respondents): "Of the following Democratic candidates for president, which one -- if any -- would you be most likely to vote for?" So "significantly" is a fair cop, assuming that the writer means that given the margin of sampling error for the entire sample, the difference has only a 5% chance or less of having come about by accident. The variance in the sample is an accurate reflection of variance in the population, given the confidence level.

Of the Republican candidates, the poll shows none with significant support.

Whoa now. Given that Giuliani and McCain are both around 13% in the total sample and 19-20% among Republicans, what do we mean by "significant support"? Looks as if the writer is changing meanings. He wants the one that says "doesn't look like a big deal to me." Bad idea, especially if you have a lot of foamy-mouthed readers around who still think the paychecks arrive in a secret envelope from Moscow and want nothing more than a clear-case of Librul Bias to cap off their Wheaties.

It's worth noting that -- what, half a dozen writers have now gotten bylines on various tales stemming from one set of polling data? All the more reason for the desk to insist that all of them know the basic terms and concepts they're dealing with. Can't pass a 10-question quiz with 80% or better? Then don't write about survey data.

(The Home Office is happy to field inquiries about those workshops, by the way).


Post a Comment

<< Home