The media are jumping up and down. “It’s a real fight”, they were saying with delight as they published the latest poll figures on Friday.

“Howard closes the gap on Rudd”, The Age roared. “Howard gets back in the fight”, trumpeted The Australian. Is this correct? The answer, strangely, is yes and no.

As The Age said in very small print under these latest poll results, with a random sample of 1,126 electors, the margin for error is plus or minus 2.9%. The two-party-preferred vote this time, compared with the previous AC Nielsen poll in The Age conducted on October 4 to 6, has moved 2 percentage points Labor’s vote has gone up 1 point and the Coalition’s 2 points.

However, Kevin Rudd’s approval rating has gone down 5 percentage points and John Howard’s has gone up two points. Some of these figures, particularly Rudd’s approval rating, are significant; the others are not.

Let us stand back a bit. Since July this year the Labor two-party-preferred vote as measured by Nielsen, has been: July, 58%; August, 55%; September, 57%; October 8, 56%; October 19, 54%.

Taken together, these figures do suggest a slight softening of the Labor vote, but, to date, no more than this. And the specific two-party-preferred figure this time, compared with last time, is well within sampling error.

Our experience suggests that there is a tendency among many journalists and most media to strongly, if not sensationally, report poll results: the results are theirs exclusively, they have paid good money for them, so why not make the most of them?

Some are more cavalier than others. Some do not seem to understand that a one or two-point movement may mean little or nothing. Some seem more interested in urging on the fight than on reporting it, or sometimes in the case of The Australian — putting the best spin on it for the Government.

Let us suggest a new caveat emptor: let the reader beware.

  • Don’t accept the headline until you have studied the figures.
  • Look to see whether the movement is within or outside sampling error.
  • Look to the trend over time as well as to the latest individual figures. (If the trend alters, wait for another poll before jumping to conclusions.)
  • Look at the sample size (anything short of 1,000 is probably not too reliable).
  • Look carefully at the question asked.

Having done all that, put your own bias away if you can, take a deep breath, and think carefully about what the poll is really saying.

You might then find that the headlines are not as helpful or indeed as accurate as they could be.