By BOB PENNERN
In the movie “A Fish Called Wanda,” the main character Otto (played by Kevin Kline) is bragging about how America always wins. Another character says, “You mean, like Vietnam?”
“It was a tie!” Otto blurts.
The truth, it seems, is hard for some to admit.
In Canada, a recent poll from Léger Marketing showed the NDP in first place federally with 33%, the Conservatives second with 32% and the Liberals third with 19%.
Great news for the NDP. For the first time since 1988 – 24 years ago – the NDP placed number one in a federal poll. To me that calls for a screaming headline:
“NDP in first place in Federal poll for first time in almost quarter century!”
But no, the headline instead read:
“NDP in dead heat with Conservatives.” Isn’t a dead heat an exact tie? And this wasn’t an exact tie.
Then, in case you missed it, the sub headline (which usually adds new information to grab reader interest) repeats again: “Nationwide poll confirms statistical tie.” Shades of Otto, if you ask me. Or maybe that’s just the way close polls are reported these days. So the next poll out, this time by Ipsos Reid, had the Conservatives ahead with 34%, and the NDP close behind at 33%. And the headline?
“Conservative popularity sinks but Stephen Harper approval holds steady: poll.”
Not a great headline for the Conservatives, but no headline mention of the rising NDP, and certainly no mention of a “statistical tie.”
And now even worse, the most recent Harris Decima poll. The NDP are now ahead of the Tories by 3% (33% to 30%) indicating what well may be a trend for a rising NDP: And the headline? Yep, once again:
“Conservatives, NDP statistically tied in new poll.”
For some, the idea that the NDP could actually be in front, a contender for power in the next federal election, clearly might take some getting used to. It took decades for the Americans to get over Vietnam too.
And here is the thing about “Statistical Ties.” They aren’t. It’s a misleading term. It implies that if two parties in a poll have results within the posted margin of error of each other that it is equally likely that either party could be winning. But that is not true, and is a misunderstanding, or misreporting, of what the Margin of Error (MOE) actually is. Statistically, if one party is ahead of another party in a poll, it’s more probable that they are ahead, than behind (or in a tie), even if both values fall within the MOE.
Margin of Error doesn’t mean that all points within the margin of error have the same probability of being the true value – especially those at the outer edges of the MOE range. It is true that it much more likely that the true value is inside the MOE than outside it, but that is very different from saying all points within the MOE are of equal probability of being the actual result had everyone, rather than only a sample of them, been polled. That is not the case.
Here is a tie: “Party A 40%, Party B 40%”. The headline could read “Party A and Party B are in an actual tie!”
So what does what “Harris Decima 33% to 30%” really mean? It means the NDP is almost certainly ahead. And the recent polling trend is clear too. The NDP is rising significantly.
The bottom line? “Statistical tie” is just not a useful term, except to downplay the real meaning of poll results.
The Vietnam war wasn’t a tie. And Mulcair’s NDP is in first place federally. It’s time for the media and pollsters to say so.
Bob Penner is the CEO of Strategic Communications Inc., and a campaign strategist and pollster.
Comments are closed.