Guardian vs Daily Mail: who is the one who made the mistake?
Written by Mark Pack on 20th August 2008 – 10:08 amEarlier this week, Guido poked fun at The Guardian for publishing an opinion poll where the voting intention figures added up to 101. Of course, there isn’t really any mystery or error in all of this, because figures are nearly always rounded off to the nearest whole number when these sorts of polls are published. It’s therefore actually quite common to have polls were the rounded-off numbers add up to 101 or 99 or even sometimes 98 or 102.
So far, so silly. However, today’s Daily Mail carries a piece in their Ephraim Hardcastle column repeating the story (though without crediting Guido) and concluding, “Perhaps the unworldy types at The Grauniad are above such mundane things as simple addition.”
So I feel it is only fair to point out that the Daily Mail too has a record of publishing polls where the numbers don’t add up to 100, for similar rounding reasons. Take this story from their November 2007 poll for example:
Have you or do you intend to change your bank account as a result of the HM Treasury lost data fiasco?
I will not change them 79%
I intend to 11%
I have changed them 3%
Don’t know 10%
That’d be 103%. Oops.
Personally, I’d score that as no mistakes from The Guardian and one mistake (for criticising the Guardian figures) from the Daily Mail.
Posted in News
20th August 2008 at 10:30 am
The figures in the lost data question from the Mail ARE impossible. Even with rounding, the maximum a four option question could round to is 102%.
20th August 2008 at 11:47 am
George C, it’s simple maths - one rounding down, three roundings up means a four-way poll can round to 103. Stop over-playing your hand.
20th August 2008 at 11:54 am
Guido is getting increasingly ridiculous isnt he?
Not as bad as Iain Dale’s petty rants about Ed Davey though. They’d be hilarious, if he wasnt trying to bring troops in to them.
20th August 2008 at 1:00 pm
OranjepanL
“George C, it’s simple maths - one rounding down, three roundings up means a four-way poll can round to 103. Stop over-playing your hand.”
George C. is quite right. Rounding up a number can increase it by at most 0.5. Rounding up four numbers can increase their total by at most 4 x 0.5 = 2.
20th August 2008 at 1:18 pm
CCF, well that depends upon whether you’re rounding halves and whether you’re rounding consistently.
This is the DM we’re talking about, you’re assuming they’re rigorous in their logic!
20th August 2008 at 1:39 pm
So the Mail prints untruths. I’m looking forward to the article on bears’ defecatory habits in woodland and the religion of the Pope.
20th August 2008 at 4:57 pm
I feel the bottom of a barrel is being scraped. If Guido wants to look like a tool, then let him, I find it funny.
20th August 2008 at 5:08 pm
BTW Iain Dale is become a bit tetchy isn’t he.
20th August 2008 at 5:10 pm
“CCF, well that depends upon whether you’re rounding halves and whether you’re rounding consistently.”
There’s nothing complicated about this. Rounding can’t increase a number by more than 0.5, however you’re doing it.
20th August 2008 at 5:29 pm
Can you hear a loud “WHOOSH!” sound John & Jay?
20th August 2008 at 5:35 pm
@CCF
It can if you’re fully rounding up or down (i.e. to the integer that is the next up/down respectively, rather than the one that is nearest).
I.e 3.2 would be rounded to 3 using halves rounding, and also to 3 using down rounding, but to 4 using up rounding. Can anyone remind me of the technical terms, dammit?
20th August 2008 at 5:36 pm
CCF: http://www.diycalculator.com/popup-m-round.shtml would appear to disagree with you. The Round-Ceiling algorithm rounds any number over a whole integer up to the next whole integer. The whole thing is probably the fault of an idiot techie who used the ceil() function in place of the round() function (or whatever their platform-specific equivalents), but I suspect that we may never know!
20th August 2008 at 5:41 pm
That’s it. The Round-Ceiling algorithm. On the tip of my tongue.
20th August 2008 at 6:05 pm
No doubt if you use a peculiar enough rounding algorithm, you can get whatever results you like. I don’t believe anyone would consider applying such a strange algorithm to opinion poll results.
And, after all, George C was responding to a specific suggestion that the total of 103 could be caused by rounding “to the nearest whole number”.
20th August 2008 at 6:14 pm
Guido, how do you feel a sound ?
20th August 2008 at 6:46 pm
Ah, politics and mathematical programming together - does life get any better than this?
20th August 2008 at 10:22 pm
No!