Donegal On Sunday

There are lies, damned lies, and statistics. Benjamin Disraeli came up with that memorable quote, but the irony is, no one remembers what he meant by it.

The Victorian prime minister wasn’t saying that the numbers collected by statisticians were lies, but that if you didn’t understand what the numbers meant, you’d get an inaccurate picture of what was going on.

Take the headline news in last Thursday’s Donegal Democrat for example. The numbers seem clear enough. Mary Coughlan (FF) will top the poll in Donegal South West with 33% of first preferences, Dinny McGinley (FG) and Pearse Doherty (SF) are in second and third place, and Pat the Cope Gallagher (FF) looks like he’s in serious trouble with only 15% of the vote. Niall Blaney (FF) is predicted to be in similar trouble in Donegal North East.

But hold on, what’s this on the internet? A news report saying that an opinion poll shows the Cope will get 31% of first preferences, with Mary Coughlan not far behind and Dinny McGinley lying in third?

How can both polls be right? The answer is simple. The second poll was taken five years ago by TG4, just before the 2002 election was called.

But that’s not the whole story either. As the election junkies out there will remember, Pat the Cope didn’t get 33% of the vote in 2002. The Pat and Mary Show got 42% of the first preferences in 2002, quite a bit short of the 55% the TG4 poll predicted.

So why did the pollsters get it wrong last time? Well, they didn’t. For one thing, Tom Gildea didn’t run in 2002, and he was included in the TG4 sample ballot paper. Still, his 5% hardly explains the difference.

But, the 2002 poll also underestimated the final support for Sinn Féin. They were supposed to get 9%, they got nearly 11%. The results were off for the other candidates too.

The poll, and what happened a few weeks later, are a handy illustration of the limits of the science. Pollsters will always warn about the dangers of putting too much faith in the numbers. The poll, you will be reminded, is a snapshot. It tells you what people thought a week ago.

But a week is famously a long time in politics. In a week, people who haven’t yet made their minds up will choose a candidate, and those choices will affect the final result. Not only that, but some people will change their minds. How many people might have read the 2002 poll for instance, decided Pat the Cope was safe, and decided to vote for another candidate instead?
On the other hand, there’s the ‘bandwagon’ effect. One candidate appears to get a bit of momentum, and undecided voters opt for that candidate, simply because it seems to be the popular choice.

Its hard to tell who is more worried by opinion polls. Politicians will tell you they take no vote for granted until it’s safely counted. If you’re topping the opinion polls, your supporters may stay home or vote for someone else, because you’re safe. If you’re a couple of points behind, voters may decide a vote for you is a wasted vote, and look elsewhere.

Meanwhile, the bandwagon effect can gather even more support if you get out in front, and the rearguard can pick up votes if the poll motivates campaign workers to canvas harder. And let’s not forget the margin of error.

The truth is, there’s a limit to how accurate an opinion poll can be. A national poll with a sample size of 1000 for example is accurate to within 3%, give or take. So next time you read a party is up or down 2%, remember there may have been no change at all. Or even that the change was in the other direction.

At local level, the result is even hazier. With a sample size of only 400 or 500, the margin of error can be over 5%. In real terms, that means that if two candidates are 10% apart in the poll, then its possible they’re locked in a dead heat.

So next time you read an opinion poll, take it with a pinch of salt. Not because its a lie or a damned lie, but because there’s a limit to how much the numbers can tell you.

Its a hackneyed old line, but there really is only poll that counts. Maybe the pollsters have it right, maybe they’ve got it wrong. It all depends on what you decide next Thursday. So whatever you decide, don’t forget to vote. It matters.