Statistician-general Pali Lehohla. Picture: SAPA
STAYING PUT: Statistician-general Pali Lehohla has had his contract extended by three years. Pictures: PEGGY NKOMO

STATISTICIAN-GENERAL Pali Lehohla's contract has just been extended by three years - despite growing doubts about his competence and in particular his handling of Statistics SA's most important project, Census 2011.

"I'm on top of things," Mr Lehohla insisted this week. But two demographic experts from the University of Cape Town, actuarial scientists Tom Moultrie and Rob Dorrington, are not so sure.

They believe Mr Lehohla has made a hash of things by releasing the results of the R3.4bin survey without allowing time for the findings to be properly interrogated.

As a result they contain so many apparent flaws and inconsistencies that their usefulness to government planners is highly questionable.

The primary objective of the census is to tell planners how many people live in the country. But, four months after the results were released, we still do not really know.

According to Stats SA's figures, the estimate is 51.7-million people living in the country, plus or minus 1.95-million. This means there is a massive range of uncertainty of 3.9-million concerning the national population.

"I would say this is fairly major in terms of government allocations and government policies," says Mr Moultrie.

According to Census 2011, we are, remarkably, five times less certain about the population than we were in 2001, which, says Mr Moultrie, "simply does not make sense".

Census 2001 gave us a population of 44.8-million plus or minus 390,000 — in other words with a range of uncertainty of 780,000.

According to Census 2011, there are 20% more black African children and 10% more white children in the nought to four age group than in the five to nine age group.

This means there has been a sharp increase in fertility, which bucks a long downward trend and, says Mr Moultrie, is "demographically improbable and not supported by any evidence".

Lesotho-born Lehohla, 55, has been statistician-general since November 2000. Doubts about his competence were raised in 2003 when Stats SA miscalculated the inflation rate.

This was a hugely expensive error, which meant South Africa was paying too much on its overseas debt.

Amiable as ever, Mr Lehohla defends his corner with a trademark combination of laughter, bluster, insult and confusion.

Mr Lehohla acknowledges that the outside experts contracted by the Stats SA council to vet the final data received them long after the promised date  — but says this is immaterial.

Fourteen experts received the data at the same time the two UCT professors did, he says, and 12 of them managed to check it - "and I'm satisfied with their work" - in the time available.

"Only the two UCT professors said they needed more time. The other 12 said, 'We've done our work'."

Allowing more time "was never going to change anything".

He says he did not need the UCT experts to give the census results their stamp of approval.

"I and my internal experts supervised the work and I'm satisfied with the work. The internal expertise in this organisation is amazing."

But in that case, why were two senior officials dismissed for supposedly bungling the all-important post enumeration survey?

Because, says Mr Lehohla, the results they gave him were "bunkum".

In particular, what infuriated him was that these results implied an undercount of more than 18%, which would have been hugely embarrassing for Mr Lehohla who had predicted an undercount of less than 2%.

He brought in a new team which reduced the undercount to 14.6%.

But an undercount of even this magnitude is unheard of, even in other developing countries. It suggests an extraordinary level of incompetence with regard to the execution of Census 2011 and reinforces suspicions about the quality of the data.

"The quality of the census is directly connected to the magnitude of the undercount," says Mr Moultrie.

Mr Moultrie says it is nonsense for Mr Lehohla to claim that 12 out of 14 experts being satisfied with the data makes it good enough.

The fact is that only two of the 12 were demographers. One of these, Nigerian Professor Eric Udjo, worked at Stats SA until 2003.

"And the demography is crucial. If you get that wrong then the economic data will be wrong too."

Mr Lehohla says he rates his two sharpest critics highly. "They're good actuarial scientists, make no mistake."

But he blithely dismisses their criticism as "stupid" and with "no scientific basis whatsoever".

At the same time he exudes what, in the circumstances, seems a questionable confidence in his own mastery of the data combined with sometimes dubious references to that data to prove that while others are wrong he is always right.

The "good actuarial scientists" from UCT may have decided that the range of uncertainty about the national population is five times greater than it was in 2001, but Lehohla insists that it is actually five times lower.

"Unfortunately the statistician-general is simply wrong," says Mr Moultrie. "He can't say we've made a mistake. We're using the numbers published in Stats SA's own report."

While Mr Moultrie and Mr Dorrington find census results suggesting a sharp increase in the number of children in the nought to four age group relative to children in the five to nine age group "demographically improbable" and "distinctly odd", Mr Lehohla appears serenely unfazed.

"They don't have a basis for that. It doesn't fit their model, that's all.

"Fertility can have a long trend of falling. When circumstances change it doesn't take long for fertility to pick up."

But what circumstances exactly are not clear.

Mr Lehohla supports the view of Udjo, who ascribes the increase to "better data capturing" in Census 2011 than in earlier censuses.

"I think I would also relate it to the distribution of anti-retrovirals" for HIV-positive mothers, he says.

Mr Moultrie responds: "We've looked at that and there is absolutely no possibility of that being the cause ... How do you explain a 10% increase in the number of white children aged nought to four relative to five to nine? Child mortality from HIV was not a big factor in the white population of South Africa, so how do you explain that increase?"

"The professors at UCT are reading the world differently," laughs Mr Lehohla.

Mr Lehohla's census data suggests that "the world has been a whole lot more interesting in South Africa in the last decade than anyone thought possible", responds Mr Moultrie.

* This article was first published in Sunday Times: Business Times