How current measures underestimate the level of poverty in South Africa

© Matt Ragen – 123RF.com

A study we recently undertook suggests that Statistics South Africa's upper-bound poverty line underestimates poverty. The Stats SA line indicates that approximately 53% of South Africans are poor, but ours suggests that this is closer to 63%. In 2011 Rands, the Stats SA line is R779 per person per month while ours is R1042 per person per month.

Our upper-bound line is equivalent to R43 per person per day in March 2015 prices. Our food poverty (or extreme poverty) line is equivalent to about R15 per day.

A poverty line generally refers to an amount of money per month or day that sets a boundary between being "poor" and "non-poor". These lines are often correctly criticised as a crude way of understanding poverty, but they are frequently a necessary evil when one does empirical work on the topic.

Our study is a methodological review of South Africa's existing poverty lines. The approach we followed is the internationally recognised "Cost of Basic Needs" method. The method is used to produce three lines:

• a food poverty line;
  
• a lower-bound poverty line; and
  
• an upper-bound poverty line.

Our view is that the lower bound is not an appropriate threshold for measuring poverty, and thus we omitted it from our proposed new poverty lines.

The food poverty line

The method is based on the cost of some minimum sufficient caloric requirement, which we set at 2100 kilocalories per person per day. This is then used to calculate the cost of necessary non-food expenditure.

With the caloric requirement set at this level, the food poverty line is then calculated by determining how much 2100 kilocalories worth of food would cost, while respecting the prevailing food consumption habits of the poor. This is quite an involved process, but we ultimately calculated a figure similar to Stats SA's.

While Stats SA calculates the food poverty line to be R335 per person per month in March 2011 prices, we calculate it to be R337 per person per month. The food poverty line should be understood as the minimum expenditure needed for sufficient calories if people spend their money only on food. Stats SA calls this an "extreme poverty" line.

The upper-bound poverty line

This is calculated by looking at the non-food expenditure of households which have per person food expenditure close to the food poverty line. In our paper, this meant looking at households which had food expenditure close to R337 per person per month in 2011. Their non-food expenditure is then added to the food poverty line to give the upper bound. In our case, this meant adding R705 of non-food expenditure to the food poverty line of R337 to get an upper-bound poverty line of R1042 per person per month.

The assumption is that households which just achieve their essential food needs, and therefore spend close to R337 per person on food, will also just achieve their essential non-food needs. While this assumption is certainly contestable, this method is probably the best way for researchers to approximate and include the cost of essential non-food needs, out of the few alternatives that exist.

The line can alternatively be interpreted as the level of total expenditure at which households tend to buy sufficient food. We propose that the upper-bound be understood as a measure of what might be called general poverty.

How we differ with Stats SA

The main difference between our lines and Stats SA's is due to one methodological decision. Stats SA originally calculated its upper-bound poverty line to be R959 per person per month, which is close to our estimate. It then decided that this figure was implausibly high, and revised it downwards by disregarding the non-food expenditure of households in expenditure deciles 1 and 8-10. This led to the R779 figure.

The Stats SA concern about the upper bound being implausibly high is based on legitimate data concerns. Using the same data as Stats SA, we found that some households which reported food expenditure close to the food poverty line also reported non-food expenditure as high as R11 000 per person per month. This is clearly a data reliability issue, and it has the potential to skew the upper-bound poverty line. However, we argue that Stats SA's fix for this issue seems to make the poverty line more unrealistic, rather than improving it.

With a food poverty line of R335 a month, the unadjusted Stats SA line of R959 suggests that poor people allocate 35% of their expenditure to food. The reduced line of R779 per month, together with the food poverty line of R335, suggests that poor people allocate 43% of their expenditure to food. We show in our paper that the 35% food share is much more representative of poor people's actual expenditure habits than 43%.

Implications of the new poverty lines

Table 1 below shows the existing South African "Cost of Basic Needs" lines, and compares them to our new set of lines, listed under 'SALDRU'. The table also shows each line's "headcount ratio", which is the percentage of South Africa's population classified as poor. All of the rand values are in 2011 prices.

Table 1: Different poverty lines and associated poverty headcount ratios.

Comparing our and the 2015 Stats SA upper-bounds, it's clear that each line implies a substantially different picture when it comes to the proportion of South Africans that are poor. Our estimate is closer to the headcount ratio suggested by the Hoogeveen and Ozler (HO) upper-bound poverty line, which is an older measure based on data from 2000.

HO's figures suggest that around 35% of the country lives in extreme poverty, while our lines, similarly to the Stats SA lines, suggest that it is closer to the 21%, which is obviously still very high.

Three things to note

While we argue that the Stats SA's upper-bound line is too low, this is not by itself sufficient basis to assert, as some have done, that the Stats SA lines were affected by political considerations. Though we disagree with the methodological decision which lowered their upper bound, the decision was motivated by an understandable technical concern.

Secondly, the conceptualisation of poverty discussed in this paper is just one way of thinking about who is poor and who is not. It may not even be the best way to think about the issue. But this method has been particularly influential in South Africa, and we think it valuable to get as accurate a measure of this kind of poverty as possible.

Lastly, the poverty lines developed here are not estimates of a "decent living level". These are just subsistence measures, and living just above the poverty line can hardly be seen as desirable.