Sometimes we get it right and occasionally we get it wrong. Yesterday in our article “Where do polls come from? Media must keep readers in the loop” we got it wrong, and we apologise to the National Foundation for Australian Women and to Farah Farouque for doing so.

Our article called for poll data to be published in conformity with the Press Council guidelines for the publication of poll material. We implied that these guidelines had not been met in this case. This was wrong. In fact they had been met entirely.

We also suggested that the questions could have been improved. There is always room for argument among professional pollsters about question formulation. This was one such example. Certainly we do not claim infallibility on this subject and we are sure the professionals who designed the questions do not do so either.

Having said all this, we stand by the point we made at the beginning of the article:

Today’s Newspoll shows a rise of one percentage point in the Coalition primary vote from 39% to 40% and an equivalent fall in the Labor vote from 48% to 47%.

This means nothing. The sample — like most of those used in polls like this — was 1,152, meaning that the real result of this poll lies somewhere between plus or minus 3% of this figure. This is called sampling variance or error.

Any movement of less that three percentage points with a sample of this size falls within sampling variance and cannot be said to represent real change. Editors and journalists ought to know this, and if their words suggest otherwise it is perhaps their political views which are speaking or their wish to create a story.