Lies, damned lies and tabloid statistics …

For months, SABC’s Interface programme called on viewers to vote in an SMS poll. The results were then delivered by the presenter during the show, in percentages — like “40% agree with the statement, and 60% disagree”.

Whether a particular percentage was of a universe of 10 SMS messages — or 10 000, we weren’t told. Of course, it makes quite a difference as to whether the tallies represented four “yes” SMSs, and six “no’s”, versus a total of 4 000 and 6 000 votes.

Worse still, the results were also presented as representing those viewers who did not send in SMSs. So you would squirm on your sofa as you heard the announcer blithely advise that “40% of our viewers agree … ”. That’s despite the fact that even 10 000 SMS votes would still not necessarily be representative of the show’s total audience.

Finally, we never heard whether the SABC disqualified more than one SMS coming from the same number — in other words, the extent to which possible vote-rigging was rooted out.

That particular abuse of statistics has now been corrected. But Business Day delivered equally badly on reporting statistics last week. That paper’s readers were met with a story-brief headlined: “Whites, coloureds have little confidence in Zuma — survey”.

The text that followed this cue was taken directly from the top paragraphs of a Sapa story and sourced this political finding to “the latest Ipsos-Markinor survey”. The study, it continued, was released on February 26.

Now, read on, to see what’s wrong with that. It’s a problem that popped up in two other papers as well:

On Saturday, The Citizen used the same Ipsos-Markinor survey to underpin its editorial “Most ANC voters believe JZ guilty”. That piece said that the statistics on Zuma’s guilt emerge from a “new survey”. It added that the result “flies in the face of the prevailing perception” that the party is 100% for JZ.
If you turn to Friday’s Sowetan, you get the same study being reported under the headline “Confidence in Zuma low — survey”. It’s the same Sapa story as Business Day, but in more length. Five paragraphs into it, you discover — finally — that the survey was actually conducted way back in October last year.

At least the same piece also includes the information that the research sample was 3 500 people, which helps put additional perspective on the findings. But you wouldn’t know either the date of the survey or its sample size from skimming the headlines of the three papers.

If a day can be a lifetime in politics, then old survey findings — predating so many recent developments — can be expected to contrast with what The Citizen calls the “prevailing perception”.

The basic problem is presenting the Ipsos-Markinor research findings as if they have current resonance. Interesting, if you examine Friday’s Daily Sun, you find the survey’s status reported accurately. This tabloid paper has the headline: “Mbeki’s shadow over the ANC. In this poll, he was still the man voters liked!”

Note the phrase “was still”, which points to the fact that the report is about history, not the present. And right up front, in the second paragraph, the Sun’s writer tells us that the information dates to October. His fifth paragraph explicitly cautions: “ … the nationwide research was carried out five months ago when the situation was a lot different to today”. When the country’s biggest tabloid reports statistics properly, it makes you wonder what makes for misleading reporting in the “serious” press.

My guess is that it’s incompetence. But many people would say the mainstream media is quick to make a meal, even when quarter-baked, of statistics that suggest problems for Zuma’s leadership. For them, journalists have a political agenda against the new ANC.
Perhaps wish-fulfilment does come into the picture — which is probably why the mainstream missed predicting the extent of the Zuma victory at Polokwane.

Ultimately, the misuse of statistics is probably caused by a combination of reasons, yet even politically partisan journalism needs to do better if it wants to carry some credibility.

A case making the point by negative example was the Sunday Times this weekend. Seven young people’s mug shots and political views were presented on page four. That paltry sample was deemed sufficient for the headline to declare: “Gauteng youth give the ANC a wide berth”.

Readers are now supposed to believe that seven people reflect “Gauteng youth”?

Business Day, The Citizen, Sowetan and the Sunday Times — four newspapers that need to pull up their statistical socks. And it’s the leading tabloid paper that avoids tabloid treatment of statistics.