Statistics South Africa on the Income and Expenditure Survey

Meeting report

Presentation by Statistics South Africa
Mr Pali Lehohla (Statistician-General of Statistics South Africa) dealt with the institutional arrangements concerning Statistics South Africa (SSA). He said that in the work they did, although the statistics were improving, the institutional arrangements were becoming increasingly important. He listed the key role players in the environment in which they operated. These were:
- Top Political Authority - who determined policy
- Planning Authority
- Mass media - who put information in the public domain
- Specialist groups - who examined the information produced and determined whether they were delivering
- Resource Authority
- Statistical Authority

He referred to recent comments by economist, Mr Russell Lamberti, who alleged that the Minister must have had presight into the results of the figures feeding into the Producer Price Index (PPI). He said that these comments were absolutely untrue. He said that knowledge had to be made freely accessible to everyone at the same time so that it did not affect the markets. The creators of ideas had to be separate from those generating ideology. Theory had to be kept separate from practice. Knowledge creators had to be kept separate from Government.

The environment in which SSA operated included the following elements:
- Intellectual
-Technological
- Logistical - the samples they used were representative of the entire population of South Africa
- Political - The journalist Karyn Maughan had placed SSA in the middle of the political arena. While they were not a political body they had to remain aware of the political environment in which they operated
- Administrative

He referred to the differences in methodologies used in the survey:
The recall method relied on people remembering what they had spent in a specific period. The diary method entailed people accounting for their expenses in writing (in the diaries). The differences in these methods were partly responsible for the differences in the surveys of 2000 as opposed to 2005/06. This was because the 2000 survey relied on the recall method, while a combination of the two methods was used in the 2005/06 survey.

Mr Lehohla described the impact of media articles (such as the one written by Karyn Maughan) on public perception. Staff enumerators had, since the news article, found that people became unwilling to supply information because the image of the organization had been tarnished.

Analysis of results
Mr Rashaad Cassim (DDG for Economic Statistics: SSA) explained that the survey had been sophisticated and very cumbersome. The focus of the survey had been on (i) income (ii) expenditure and (iii) inequalities.

The bulk of the work had been done in the area of expenditure because this would input into the Consumer Price Index (CPI). The survey was done every five years because the CPI was re-weighted every five years.

The survey showed that income from work done constituted 74% of total income. Other forms of income included social grants. There had been an overall increase in per capita income since the 2000 Income and Expenditure Survey (IES).

Expenditure on Housing, Transport and Food and Non-Alcoholic Beverages constituted 60% of total expenditure. Transport increased considerably as a proportion of total spending. Food on the other hand decreased as a proportion of total spending. As far as inequality was concerned, Mr Cassim said that South Africa had one of the highest levels of inequality in the world. The income of the top 10% of earners was 94 times higher than the bottom 10% of households. In addition 10% of the population earned over 50% of the country’s total income. Social grants however served to reduce this inequality to some extent.

As with all surveys, the IES had its limitations and had to be interpreted carefully. The team worked very hard to ensure that the survey was of a good quality and the problems arose simply from the nature of data collection. One had to bear in mind that they were encroaching on the private lives of people and people often did not reflect accurate values. The most inaccuracies were found with the very poor (who had irregular employment and could therefore not report on this matter very accurately) and the very rich (who were reluctant to provide information on income from dividends and second properties, etcetera). It was found that when reporting on income they encountered bias, and when reporting expenditure they encountered error (as people often forgot what they spent). Thus when the survey had to be analysed, caution had to be exercised when interpreting the information.

Mr Cassim explained that decile one referred to ten percent of the poorest in South Africa, while decile ten referred to ten percent of the richest. There was above average growth in per capita incomes in deciles one to three. This was positive on the one hand since it indicated that Government was looking after the poor by the payment of social grants. On the other hand it presented the problem of dependency on social grants.

In high growth periods there was an increase in levels of inequality. This was because although absolute poverty decreased, these periods also presented great opportunities for highly skilled people.

It is not always possible to compare the results of the 2000 IES with those of the 2005/06 IES. This was because SSA was constantly innovating new techniques in their surveys, which would sometimes mean that results would not be comparable. It was evident that food as a proportion of total expenditure decreased, although it is difficult to describe the extent of this. One contributory factor could be that when income increased, people did not necessarily increase their spending on food proportionally. Instead they spent the difference on cellular phones or cars, etcetera. This was evident from the fact that there had been a staggering increase in the ownership of cars.

Mr Cassim went through the graphs dealing with expenditure patterns in the country. He pointed out that Western Cape and Gauteng spent less on food as a proportion of total expenditure, despite the fact that these were the two richest provinces. The expenditure patterns were looked at in relation to income groups, race and province.

For transport expenditure, the majority of purchasing power was with the ‘whites’, although the spending by ‘blacks’ was fast catching up. The purchase of motor vehicles accounted for 57, 6% of spending on transport. This was largely a middle class phenomenon.

Expenditure on food and non-alcoholic beverages indicated an unusual phenomenon. If one looked at the inverse relationship between the spending patterns on food as opposed to income, and compared it to the 2000 IES, it would seem that either the figures in the 2000 survey were exaggerated, or that 2006 figures were underreported. It could also be due to the differences resulting from use of recall and diary methods. The recall method relied on memory and was therefore not very reliable. The diary method was reliable, but there were problems associated with this method, for example: (1) fatigue caused people to forget to include items and (2) in richer households people did not want to take time to complete diaries, so they gave receipts to enumerators to capture the figures for them. Very often they did not provide all the receipts which resulted in the exclusion of certain items. Thus the over-reporting in the 2000 IES could be as a result of the recall method while the underreporting in 2006 was as a result of the diary method. However despite the lack of comparability between the surveys, they remained useful. The SSA could have said that the food expenditure figure should be higher and could have included the estimated figure in the survey results, but decided to report the figures as they were in the interests of transparency. However, when feeding the figures into the CPI the figure would have to be adjusted scientifically.

Inequality between the population groups was determined by looking at what percentage of the population each racial group constituted, and then comparing it to the percentage of total income for each group. The Gini Coefficient in South Africa was 0.72, which was indicative of a very unequal society.

Mr Lehohla said that both the Community Survey and the IES would be used to determine if the goals of the Reconstruction and Development Programme had been achieved. There was evidence that this was indeed the case although there were still inequalities in society.

Mr Howard Gabriels (Chairperson of the Statistics Council) thanked SSA for the professional way in which the survey had been conducted. Much time was spent on this survey, as it was the basis for determining the inflation indexes. It would be used in both the public and private sectors. The new method used in this survey was stable and could therefore serve as a basis for future surveys. One major innovation had been the communication campaign around the survey. The analysis team had done a very good job in data confrontation and had compared the results of this survey to almost every other survey and source of information available in order to ensure internal consistency.

Discussion
Mr K Moloto (ANC) asked how they explained the reduction in personal care expenditure, given the fact that it appeared that more people (especially males) were spending more on personal care.

Mr Cassim agreed that people were spending more on personal care, but said that people were reluctant to provide information on this and there was therefore underreporting in this area.

Mr Moloto asked for more information on expenditure in the area of financial services (especially banking).

Mr Cassim replied that SSA captured the amounts that were paid by the customer.

Mr S Marais (DA) asked how the Gini coefficient in South Africa compared to those of other African countries, as well as globally.

Mr Cassim said that in African countries the inequality was driven by the fact that society was constituted of large amounts of very poor people on the one hand and very rich on the other. These societies were characterized by the absence of a middle income group. The Gini coefficient in African countries was 0, 6. In Latin America the Gini coefficient was 0, 7 while it was 0, 4- 0, 5 in developed countries. South Africa has one of the highest levels of inequality if compared to other middle income countries.

The Chair asked what quality control measures SSA had in place to ensure that enumerators obtained accurate information.

Mr Marais referred to issues raised previously about skills shortages in the SSA which members had felt could affect the credibility of the results of surveys. He said that if inaccurate information were obtained in this survey, it would mean that this information would be relied upon in future.

Mr Lehohla responded that they had to express the limitations of the study in a way which did not harm the credibility of the study. The limitations had to be made clear in the interests of transparency, but this had to be done in a way that did not erode the perception of the coherence and reliability of the data. However the figures for food expenditure looked suspicious and would be further investigated.

Mr Gabriels said that the results were extremely credible. They had looked at internal consistencies, data consistencies, as well as benchmarks within and outside the SSA. They had surveyed 24 000 households which allowed them to make generalizations for 12 million households. Limitations of the survey related to the sample chosen or the methods used, but still met the objectives it sought to achieve. They could have used other methods and made available only the adjustments, but in the interests of transparency had decided to make available the figures without adjustments. The way in which they made data available was part of their process of building credibility. Mr Gabriels said that he would not hesitate to say that the data was based on sound methodology as required by the Act.

Mr Desmond Booysen (Executive Manager: IES Field Collection: SSA) said that people had generally been very willing to participate in the survey. SSA did not like to use the Statistics Act (No 6 of 1999) to coerce people to participate. Instead they relied on well-trained fieldworkers who were trained to build relationships from the very first contact. There was also a dedicated publicity team who ensured that people bought into the process. This would ensure that they remained committed by showing them how the survey would benefit their particular community. The design of the survey was very deliberate and much flexibility had been built into the instrument. Although the process required six visits, they made other arrangements where people displayed reluctance to meet all six times. Although the process entailed recording information in diaries, they were flexible where people were illiterate or displayed reluctance to do so. In these instances, the enumerators would use till slips to enter the data into the diaries. They built a rapport with households very early on in order to obtain sensitive information on income, which was done later in the process.

Mr Marais expressed concern that these were the results of a survey done in 2005/06. He said that much had changed since then. He asked how they could be sure that they were not dealing with outdated information.

Mr Cassim replied that major patterns of expenditure changed very little over the short term. Big changes would only result from the changes done to the selection of samples. The lag period would have very little impact on the re-weighting of the CPI.

Mr B Mnguni (ANC) asked why they had excluded mortgages in the survey.

Mr Cassim explained that information on mortgages was not very reliable. Mortgage payments could be divided into capital and interest portions. People very often could not provide specific information on the respective amounts for each. SSA then determined the imputed rent by trying to determine the amount of rental a household would be able to derive from their property if they chose to rent it out. In 2000 they had used the mortgage payments, which was unreliable. This survey relied on imputed rental.

Mr Mnguni asked if they had considered that they could also have been surveying illegal immigrants.

Mr Cassim said that this was irrelevant, as the focus of the survey was on income and spending in South Africa.

Mr M Johnson (ANC) asked why consumption of alcoholic beverages had not been included in the survey.

Mr Patrick Kelly (Executive Manager: CPI: SSA) said that spending on alcohol and tobacco was generally under-reported as people seemed embarrassed to provide this information. For the purposes of collecting information for the CPI, they therefore looked at excise taxes, suppliers, manufacturers and retail associations to obtain more accurate figures.

Mr Johnson referred to the fact that rich people often hid information on additional income and assets. He asked why the SSA could not link with other state agencies the way the South African Revenue Service (SARS) worked with the banks.

Mr Lehlohla said that some countries used employment information as a frame for household based survey. Visits would then still be used to provide additional information. They were looking at ways to link data sources but were not quite there yet.

Mr Cassim said that in Canada income reported was compared to tax records to which the surveying body had access. These negotiations with SARS were important, as obtaining this information from them was vital.

Mr Lehohla said that use of administrative records in these surveys was not done in all countries. In Scandanavian countries the rate of use was 90%, while in Canada it was 50%. South Africa fell in the bottom 10%. The move in this direction should be encouraged as information was more dependable and cheaper to acquire. However, these methods were not data rich in terms of variables as the household surveys were.

In reply to Mr Johnson asking if SSA was on track for the achievement of targets set out in the Millennium Development Goals, Mr Lehohla said that this was on track. These would be presented and information would be tabled within the following week.

Ms N Mokoto (ANC) explained that despite interventions by Government, savings had not increased. She asked what Government could do to encourage savings.

Mr Cassim replied that responses on savings had been unreliable and that this information was best obtained from the Reserve Bank. It was not possible to provide this information based on this survey.

Mr N Singh (IFP) asked at which point the references to race would be abolished. One could not make these generalizations based on race given the fact that there were an increasing number of poor whites and rich blacks.

Mr Gabriels said that it was not SSA’s role to determine if race should be included. Once the relevant bodies decided upon this, this would be implemented by SSA. The Council had looked at the SSA method of sampling and was satisfied that its methodology had been scientific.

Mr Singh asked if the decrease in food spending could be attributed to overall healthier lifestyles or the fact that people could be producing their own food. It could also be attributed to the fact that people were poor and hungry.

Mr Cassim responded that these factors did not contribute to the decrease in food expenditure in any significant way. This decrease had not been due to demographic issues. SSA believed that food consumption had in fact increased.

Mr Singh asked if there was a legal compulsion on people to complete surveys.

Mr Gabriels replied that such compulsion existed in terms of the Statistics Act. Since the response rate had exceeded 90%, it had been unnecessary to invoke this legislation.

Dr D George (DA) asked how SSA chose samples. He asked if they selected a certain number of households per race.

Mr Lehohla said that 24 000 households were surveyed. Although participants had not been selected on race, it had (due to the sample size) emerged that racial groups were proportionally represented.

Dr George asked how they tested for the validity of their results.

Mr Lehohla replied that this was done by examining internal consistency of data. Validity was also ensured in the design of the instrument. There were random checks on households previously visited.

Dr George asked if SSA collected qualitative data.

Mr Gabriels felt that this was not the role of SSA. Collection of qualitative data involved normative elements and required normative judgments, which were not in the scope of this research institution.

Dr George asked if SSA did a separate technical report on the data analysis.

Mr Lehohla said that there was no analytical report at present, but that they would still produce this report. The question which arose was whether SSA should be involved in analysis of data and to what extent this should be done. It had been suggested that they should be involved in this process, but SSA would have to employ analysts to do this work.

Mr Kelly provided clarity on the issue of medical aid contributions in the survey. These had not been dealt with under health care, as it had been considered to be a type of insurance. Medical expenditure was recorded by taking into account the payment made for the service and not the usage of the service.

In reply to Mr Johnson asking how many cars there were on South African roads, Mr Lehohla said that there were ten million cars.

Mr Johnson said that since undertaking the survey in 2006, there had been many economic developments for example the introduction of the National Credit Act, which had resulted in repossessions of houses and cars. He asked what the net impact of this Act would be on the information provided.

Mr Cassim replied that the data was still current and that the impact of this Act would only be seen during the next survey.

Comments on the Producer Price Index
Mr Lehohla said that the two-month limit SSA had estimated for the completion and release of the PPI had been unrealistic. When they had done the re-weighting from the data they had in the system, it had shown the emergence of an unlikely pattern. They had asked to delay the PPI in order to prevent a situation where inaccurate data would be released. They had then requested further delays and had only received the new PPI the previous day. It was difficult to say what the impact of this delay had had on the markets, but he assured the Committee that the delays had been requested for good reason and not due to sinister motives as suggested by the media.

Mr Cassim added that SSA had decided to err on the side of caution, as they had not been confident about the results. Although the PPI was important, its significance had been elevated by the media’s suggestions that it had affected the market to such a huge extent. The PPI excluded the service sector, which was a major part of the economy. It was strange that the media had elevated the PPI to the level where it was seen as being able to move markets.

The Chair adjourned the meeting.

Appendix:
Validity of R600m survey called into question    

Karyn Maughan January 08 2008 Independent Newspapers

The R600-million state-funded survey - which President Thabo Mbeki has used to show that the government was winning the fight against poverty - has been exposed as "unreliable" and ridden with "errors".

Now Statistics SA, the government department responsible for the "Community Survey 2007" report, has itself warned potential users of the data to "be more cautious" when using the study.

The warning and release of a "revised" report comes two months after Finance Minister Trevor Manuel handed the survey to Mbeki, who later described it as an "important document about the progress we are making to meet the basic needs of the masses of our people".

Quoting from the report in an article on his website, Mbeki said the survey showed that South Africans "say in these areas of delivery of meeting basic needs that 2007 is better than 2001 and indeed 2001 was better than 1996. Today is better than yesterday."

Mbeki also urged people to read the survey, which he suggested was an effective assessment that "would give us the scientific basis to determine whether we should change any of our policies and the direction that such possible change should take".

Concerns raised by the Statistics Council of South Africa have, however, raised doubts about whether it can provide a "scientific basis" for future government policy.

According to a statement by Statistics Council of South Africa chairperson Howard Gabriels:

·  The distribution of households by province in the Community Survey has very little similarity with data previously recorded by the General Household Survey or the 2001 census, and there was a "maldistribution of the population by province".

·  The number of grants recorded by Stats SA "do not match the South African Social Security Agency data".

·  There were concerns over the levels of income recorded by the Community Survey, particularly the "unreasonably high levels of income for children".

·  "Unreliable" and "higher" unemployment figures were recorded in the Community Survey, apparently because Stats SA staff did not ask survey subjects the same question about their employment status.

·  Stats SA did not establish South Africa's "institutional population" - the number of people living in prison, army barracks and other institutions - and instead estimated it.

The revised survey report recorded half-a-dozen corrections made from the earlier survey.

Stats SA spokesperson Trevor Oosterwyk yesterday insisted, however, that the survey's overall finding of improved social conditions had not been destabilised by the "health warnings".

University of Cape Town economics professor Martin Wittenberg said the biggest concern about the survey was that the number of households it recorded for each province "apparently does not square".

"It's potentially quite disturbing... because if that is true, attempting to provide many services aimed at households will be like attempting to hit a moving target."

Professor Paul Mostert, from the South African Statistical Association, pointed out that it was not the first time Stats SA had come under fire. In 2005, Statistician-General Pali Lehohla reportedly faced dismissal over a series of costly mistakes involving official statistics, which included its release of figures showing that manufacturing production was down by 5,1 percent, whereas it had declined by 1,2 percent.