Friday, January 12, 2007

It's always poll season

Public opinion surveys don't go away when the polls close; for good or ill, they're with us all year round. So let's call this one the first assignment of the semester, and let's give CNN a B+, with the reminder that the midterm won't be graded so gently.

Poll: Two-thirds of Americans oppose more troops in Iraq
WASHINGTON (CNN) -- Two out of three Americans oppose President Bush's plan to send more troops to Iraq, a CNN/Opinion Research Corp. poll released Friday indicates.

Nearly two-thirds of those polled also say Bush has no clear plan for Iraq.

While his numbers have inched up slightly on that question since the previous poll last week, Bush's address to the nation Wednesday night seems to have made little difference.

Put a marker on this one for revisiting because of the inference. "Inched up slightly" is irritating (as opposed to "inched up dramatically"?), but that's bad writing, not bad data.

Nearly half of those who saw the speech say their minds were not changed, while the rest are evenly split over whether they'd be more or less likely to support his policies.

... This is the first poll gauging Americans' positions on the strategy following Bush's address. The telephone survey of 1,093 adult Americans was conducted Thursday. The sampling error on all the questions in the poll is plus or minus 3 percentage points. (Read the complete poll results -- PDF)

Full credit for mentioning sample size, population, time in the field and margin of sampling error. Extra credit for the link to the PDF. Point off for omitting confidence level. Three points off for misinterpreting the poll's statement of sampling error. Opinion Research said it was 3 points for "results based on the total sample," not for "all the questions in the poll." As the preceding graf makes clear, at least one of the questions involves a subset ("those who saw the speech"), for which the confidence interval is always higher -- in this case, about 4.5 percentage points.

The president argued that the increase in troop strength would be the best chance to succeed in a war the U.S. cannot afford to lose.

But Americans, the poll indicates, do not see it that way. Asked their positions on sending more troops to Iraq, 66 percent of respondents said they oppose the move, while 32 percent said they favor it.

Two points off for bad juxtaposition. "Do you favor or oppose the plan?" isn't the same question as "Is the increase the best chance to succeed in a war the U.S. cannot afford to lose?"

... Asked whether Bush has a clear plan for Iraq, 63 percent said no, while 35 percent said yes.
A week earlier, 72 percent said no and 25 percent said yes.

But that slight rise is apparently not attributable to having watched Bush's speech Wednesday night. Among those who watched the speech -- which was a little less than half the people surveyed -- 45 percent said it made no difference. Meanwhile 27 percent said they were more likely to support his policies -- and 27 percent said they were less likely.

Three points off for deck-stacking. "Slight rise" is the sort of editorial judgment that leaves you wide-open to anybody who wonders why 10 points is a "slight rise" for Republicans and a "leap" for Democrats. Assuming that the week-ago poll was of comparable size, there are no non-chance cases of overlap, and there's about a 5-point difference (the confidence interval is smaller for a 75/25 split than for a 50/50) between the biggest non-chance value of the previous week and the smallest of this week. With the "yes" camp being about 40 percent larger in this poll, that sounds like a pretty striking change.

Five points off for drawing a conclusion from a false comparison. The question asked of the speech-watching subset was "Did the speech make you or less likely to support Bush's policies in Iraq?" That is emphatically not the same thing as "Do you think Bush does or does not have a clear plan for handling the situation?"

Let's call it an 88 and give CNN the aforementioned B+. Reminders for the rest of the term: Don't throw in adjectives you can't justify; they can make you look either dumb or biased (or both). Don't shift your pivot foot when you're trying to draw inferences. The person who writes the follow-up is likely to stop at the third graf and say that a January poll found that Bush's speech didn't affect opinions about whether he has a "clear plan." We have no idea whether there was any such effect. When we say that we do -- in other words, when we make stuff up about the data -- we're poisoning the well for anybody who uses this story for reference in the future.

And whatever it is that you put first on the list of Things that Ail Journalism, "making stuff up" doesn't fix it.


Blogger Guy Barry said...

The thing is its not made up,rather its reported from different angles and slants?

8:40 AM, January 13, 2007  
Blogger fev said...

Some things are "different angles" and some aren't. Whether a 10-point change is "slight" or "big" is a framing question -- like whether a 10-point win on the sports page gets a "trounces" or a "slips past." Drawing conclusions from data that aren't there, on the other hand, isn't a matter of slant. It's fiction.

10:29 AM, January 13, 2007  
Anonymous Anonymous said...

runescape money runescape gold runescape money buy runescape gold buy runescape money runescape money runescape gold wow power leveling wow powerleveling Warcraft Power Leveling Warcraft PowerLeveling buy runescape gold buy runescape money runescape itemsrunescape accounts runescape gp dofus kamas buy dofus kamas Guild Wars Gold buy Guild Wars Gold lotro gold buy lotro gold lotro gold buy lotro gold lotro gold buy lotro gold runescape money runescape power leveling runescape money runescape gold dofus kamas cheap runescape money cheap runescape gold Hellgate Palladium Hellgate London Palladium Hellgate money Tabula Rasa gold tabula rasa money Tabula Rasa Credit Tabula Rasa Credits Hellgate gold Hellgate London gold wow power leveling wow powerleveling Warcraft PowerLeveling Warcraft Power Leveling World of Warcraft PowerLeveling World of Warcraft Power Leveling runescape power leveling runescape powerleveling eve isk eve online isk eve isk eve online isk tibia gold Fiesta Silver Fiesta Gold
Age of Conan Gold
buy Age of Conan Gold
aoc gold

china tour
beijing tour
beijing travel
china tour
tibet tour
tibet travel
computer monitoring software
employee monitoring

11:25 PM, December 30, 2008  

Post a Comment

Links to this post:

Create a Link

<< Home