Stats+SA+credibility+under+fire,+Ethel+Hazelhurst,+Business+Report



=Stats SA credibility under fire=


 * Ethel Hazelhurst, Business Report, Johannesburg, 9 January 2008**

Johannesburg - Statistics SA's 2007 community survey, which has been described by demographic researchers as "unreliable" and ridden with "errors", is only the latest in a series of bungles by the agency.

The Star yesterday reported criticisms of the R600 million survey, which used 6 000 field workers and visited 280 000 households.

Gerhard Bijker, a demographic researcher at Global Insight Southern Africa, said the survey showed a population of 48.5 million - a figure which he regards as "way too high".

Stats SA itself produced a figure of 47.9 million in its mid-year population estimate for last year and did not provide any explanation for the discrepancy in the figures - "which is unusual", he said.

The agency has made a number of serious errors in the past, including providing incorrect inflation data that influenced the direction of monetary policy in 2002 and early 2003. Interest rates were kept unnecessarily high for many months, until the mistake was discovered by John Stopford, a portfolio manager at Investec Asset Management.

In May 2003, Stats SA admitted CPIX (the consumer price index minus mortgage costs) was overstated by 1.9 percent. The revision reduced March 2003 inflation from 11.2 percent to 9.3 percent.

Carel van Aardt, a research professor at the Bureau of Market Research at the University of SA, said the community survey and Stats SA's labour force survey, as well as the survey of income and expenditure, often produced three different figures for critical variables.

He said: "Any statistical unit that publishes an array of conflicting findings creates a credibility gap."

Stats SA's first high-profile problem was the preliminary 37.9 million population estimate, published in 1997, based on the 1996 population census - a figure far lower than was generally expected. The data were reworked and, in 1998, Stats SA published a figure of 40.6 million which was still regarded as unrealistically low.

The exercise was repeated in 2001.

Mike Schussler, an economist at T-Sec, said: "Stats SA admitted to a 17 percent undercount in the population census estimate. It adjusted the figure to 44.8 million. A lot of people believe the population was even higher and that the census missed a lot of people living in the richer areas."

A research project was launched in November 2006 at the University of Cape Town to clean up historic errors.

Martin Wittenberg, an associate professor of economics, is examining the October household surveys and has found major errors.

Wittenberg said: "In 1994 researchers conflated data on daily, weekly and monthly income to come up with misleading estimates of monthly income." And he found "an inconsistency between the levels of agricultural employment in 1995 and 1996".

Reliable figures are vital for private and public policy makers. President Thabo Mbeki recently used the community survey figures to show improvements in delivery and refute suggestions made by John Kane-Berman, the chief executive of the SA Institute of Race Relations, that poverty has increased since 1996. But weaknesses in some areas of the report cast doubt on the rest of the data.

Despite the many criticisms of the agency, Schussler said there have been recent improvements in the availability of information.

Wittenberg pointed out conflicting survey findings were not confined to South Africa but were also a source of controversy in other counties.


 * From: http://www.busrep.co.za/index.php?fArticleId=4198256**

563 words