The South African MPI - Statistics South Africa [PDF]

The South African MPI Creating a multidimensional poverty index using census data

The South Africa I know, the home I understand

Statistics South Africa Private Bag X44 Pretoria 0001 South Africa 170 Thabo Sehume Street Pretoria 0002 User information service: 012 310 8600 Fax: 012 310 8500 Main switchboard: 012 310 8911 Fax: 012 321 7381 Website: www.statssa.gov.za Email: [email protected]

The South African MPI Creating a multidimensional poverty index using census data

Statistics South Africa, 2014 Pali Lehohla, Statistician-General Report No. 03-10-08

The South African MPI: Creating a multidimensional poverty index using census data / Statistics South Africa Published by Statistics South Africa, Private Bag X44, Pretoria 0001 © Statistics South Africa, 2014 Users may apply or process this data, provided Statistics South Africa (Stats SA) is acknowledged as the original source of the data; that it is specified that the application and/or analysis is the result of the user's independent processing of the data; and that neither the basic data nor any reprocessed version or application thereof may be sold or offered for sale in any form whatsoever without prior permission from Stats SA. Stats SA Library Cataloguing-in-Publication (CIP) Data The South African MPI: Creating a multidimensional poverty index using Census data / Statistics South Africa. Pretoria: Statistics South Africa, 2014 Report no.: 03-10-08 58pp ISBN 978-0-621-42616-8 A complete set of Stats SA publications is available at Stats SA Library and the following libraries: National Library of South Africa, Pretoria Division National Library of South Africa, Cape Town Division Library of Parliament, Cape Town Bloemfontein Public Library Natal Society Library, Pietermaritzburg Johannesburg Public Library Eastern Cape Library Services, King William’s Town Central Regional Library, Polokwane Central Reference Library, Nelspruit Central Reference Collection, Kimberley Central Reference Library, Mmabatho This report is available on the Stats SA website: www.statssa.gov.za For technical enquiries please contact: Ross Jennings Tel: (012) 337 6342 Email: [email protected] Copies are obtainable from: Printing and Distribution, Statistics South Africa Tel: (012) 310 8093 (012) 310 8619 Email: [email protected] [email protected]

ii

Contents List of tables .................................................................................................................................... iv List of figures ....................................................................................................................................v Preface ........................................................................................................................................... 1 Introduction .................................................................................................................................... 2 What is multidimensional poverty? ................................................................................................ 2 The history of multidimensional poverty measurement by Stats SA .................................................... 2 What is the MPI? ......................................................................................................................... 3 Constructing the South African MPI ................................................................................................... 4 Data sources ............................................................................................................................... 4 Exploring the data ........................................................................................................................ 5 The dimensions, the indicators and their deprivation cut-offs............................................................ 5 The poverty cut-off ....................................................................................................................... 7 Limitations ................................................................................................................................... 8 Using the SAMPI .......................................................................................................................... 8 Profiling multidimensional poverty in South Africa ............................................................................... 9 The poverty measures................................................................................................................... 9 What is driving the poverty situation? ........................................................................................... 10 Poverty measures at a provincial level .......................................................................................... 11 Poverty profile: Western Cape ..................................................................................................... 16 Poverty profile: Eastern Cape ...................................................................................................... 20 Poverty profile: Northern Cape.................................................................................................... 24 Poverty profile: Free State ........................................................................................................... 28 Poverty profile: KwaZulu-Natal .................................................................................................... 32 Poverty profile: North West ......................................................................................................... 36 Poverty profile: Gauteng............................................................................................................. 40 Poverty profile: Mpumalanga ...................................................................................................... 44 Poverty profile: Limpopo ............................................................................................................. 48 Conclusion ................................................................................................................................... 52 The way forward ............................................................................................................................ 52

iii

List of tables Table 1: The dimensions, indicators and deprivation cut-offs for SAMPI.................................................... 6 Table 2: Deprivation headcounts on each of the indicators ..................................................................... 7 Table 3: The dimensions, indicators and their weights............................................................................. 8 Table 4: Poverty measures for Census 2001 and Census 2011 ............................................................... 9 Table 5: Intensity of deprivation amongst the poor ................................................................................. 9 Table 6: Poverty measures for Census 2001 and Census 2011 at provincial level ................................... 11 Table 7: Poverty measures for Census 2001 and Census 2011 in Western Cape .................................... 16 Table 8: Poverty measures for Census 2001 and Census 2011 at municipal level in Western Cape ......... 17 Table 9: Poverty measures for Census 2001 and Census 2011 in Eastern Cape ..................................... 20 Table 10: Poverty measures for Census 2001 and Census 2011 at municipal level in Eastern Cape......... 21 Table 11: Poverty measures for Census 2001 and Census 2011 in Northern Cape................................. 24 Table 12: Poverty measures for Census 2001 and Census 2011 at municipal level in Northern Cape ...... 25 Table 13: Poverty measures for Census 2001 and Census 2011 in Free State ........................................ 28 Table 14: Poverty measures for Census 2001 and Census 2011 at municipal level in Free State .............. 29 Table 15: Poverty measures for Census 2001 and Census 2011 in KwaZulu-Natal ................................. 32 Table 16: Poverty measures for Census 2001 and Census 2011 at municipal level in KwaZulu-Natal ...... 33 Table 17: Poverty measures for Census 2001 and Census 2011 in North West ...................................... 36 Table 18: Poverty measures for Census 2001 and Census 2011 at municipal level in North West ............ 37 Table 19: Poverty measures for Census 2001 and Census 2011 in Gauteng .......................................... 40 Table 20: Poverty measures for Census 2001 and Census 2011 at municipal level in Gauteng ............... 41 Table 21: Poverty measures for Census 2001 and Census 2011 in Mpumalanga ................................... 44 Table 22: Poverty measures for Census 2001 and Census 2011 at municipal level in Mpumalanga......... 45 Table 23: Poverty measures for Census 2001 and Census 2011 in Limpopo .......................................... 48 Table 24: Poverty measures for Census 2001 and Census 2011 at municipal level in Limpopo ............... 49

iv

List of figures Figure 1: Dimensions and indicators of the MPI ..................................................................................... 3 Figure 2: Contribution of weighted indicators to SAMPI 2001 at national level ....................................... 10 Figure 3: Contribution of weighted indicators to SAMPI 2011 at national level ....................................... 10 Figure 4: Poverty headcount (H) at provincial level in 2001................................................................... 12 Figure 5: Poverty headcount (H) at provincial level in 2011................................................................... 13 Figure 6: Poverty headcount (H) at municipal level in 2001................................................................... 14 Figure 7: Poverty headcount (H) at municipal level in 2011................................................................... 15 Figure 8: Contribution of weighted indicators to SAMPI 2011 in Western Cape ...................................... 16 Figure 9: Poverty headcount (H) in Western Cape at municipal level in 2001 ......................................... 18 Figure 10: Poverty headcount (H) in Western Cape at municipal level in 2011 ....................................... 19 Figure 11: Contribution of weighted indicators to SAMPI 2011 in Eastern Cape ..................................... 20 Figure 12: Poverty headcount (H) in Eastern Cape at municipal level in 2001 ........................................ 22 Figure 13: Poverty headcount (H) in Eastern Cape at municipal level in 2011 ........................................ 23 Figure 14: Contribution of weighted indicators to SAMPI 2011 in Northern Cape ................................... 24 Figure 15: Poverty headcount (H) in Northern Cape at municipal level in 2001 ...................................... 26 Figure 16: Poverty headcount (H) in Northern Cape at municipal level in 2011 ...................................... 27 Figure 17: Contribution of weighted indicators to SAMPI 2011 in Free State .......................................... 28 Figure 18: Poverty headcount (H) in Free State at municipal level in 2001.............................................. 30 Figure 19: Poverty headcount (H) in Free State at municipal level in 2011.............................................. 31 Figure 20: Contribution of weighted indicators to SAMPI 2011 in KwaZulu-Natal ................................... 32 Figure 21: Poverty headcount (H) in KwaZulu-Natal at municipal level in 2001 ...................................... 34 Figure 22: Poverty headcount (H) in KwaZulu-Natal at municipal level in 2011 ...................................... 35 Figure 23: Contribution of weighted indicators to SAMPI 2011 in North West......................................... 36 Figure 24: Poverty headcount (H) in North West at municipal level in 2001............................................ 38 Figure 25: Poverty headcount (H) in North West at municipal level in 2011............................................ 39 Figure 26: Contribution of weighted indicators to SAMPI 2011 in Gauteng ............................................ 40 Figure 27: Poverty headcount (H) in Gauteng at municipal level in 2001 ............................................... 42 Figure 28: Poverty headcount (H) in Gauteng at municipal level in 2011 ............................................... 43 Figure 29: Contribution of weighted indicators to SAMPI 2011 in Mpumalanga ..................................... 44 Figure 30: Poverty headcount (H) in Mpumalanga at municipal level in 2001 ........................................ 46 Figure 31: Poverty headcount (H) in Mpumalanga at municipal level in 2011 ........................................ 47 Figure 32: Contribution of weighted indicators to SAMPI 2011 in Limpopo ............................................ 48 Figure 33: Poverty headcount (H) in Limpopo at municipal level in 2001 ............................................... 50 Figure 34: Poverty headcount (H) in Limpopo at municipal level in 2011 ............................................... 51

v

please scroll down

Preface The main purpose of this report is the provision of poverty maps and poverty data at provincial and municipal levels. In so doing, this report details how the South African Multidimensional Poverty Index (SAMPI) was conceptualised and constructed using data collected by Statistics South Africa (Stats SA) through the censuses of 2001 and 2011. The SAMPI, based on the Alkire-Foster method, provides another powerful tool in Stats SA's ongoing efforts to measure poverty and deprivation in the country. The strength of this index using census data rests in the ability to reliably map poverty down to whatever geographical level required and thus, will greatly assist municipalities in understanding the unique challenges facing their areas and how the situation on the ground has changed since 2001. The SAMPI was born out of the desire to develop a new product that could build onto the work started with the development of the Provincial Indices of Multiple Deprivation (PIMD) after Census 2001. Unlike the PIMD, the SAMPI has the advantage of being fully decomposable by demographic and geographic variables; this shows not only in what aspects the poor are deprived, but also reveals the interconnections among those deprivations. This will enable policymakers to better allocate resources and more effectively target policies. This report has benefitted from our on-going partnership with UNICEF. In addition to technical support on the analysis of child poverty, UNICEF has also generously supported efforts to build capacity for Stats SA staff to construct this SAMPI.

1

Introduction What is multidimensional poverty? Poverty is often defined by unidimensional measures, such as income or expenditure. While this provides us with a very useful way of measuring absolute poverty, it does not capture the multiple aspects that constitute poverty. Multidimensional poverty is made up of several factors that amount to a poor person’s experience of deprivation – these can include poor health, lack of education, inadequate living standards, lack of income, disempowerment, lack of decent work and threat from violence. A multidimensional measure seeks to incorporate a range of indicators to capture the complexity of poverty, and thus provides a more robust tool to better inform programmes and policies designed to fight it. The indicators used for this measure can vary from country to country; ultimately, they should be chosen to fit the society and situation. The motivations for using a multidimensional approach include: • Poverty is a complex issue that manifests itself in economic, social and political ways • Income denotes the means to achieving a better life, but is not actually the better life itself • Poor people themselves describe their experience of poverty as multidimensional • The more policy-relevant information there is available on poverty, the better equipped policy-makers will be to reduce it • Some methods of multidimensional measurement can be used for additional purposes such as targeting or conditional transfers

The history of multidimensional poverty measurement by Stats SA Since 2000, Stats SA has undertaken or been involved in a number of initiatives that have sought to profile poverty or deprivation from a multidimensional perspective. These include: • Measuring poverty in SA (2000) – This saw the development of a household infrastructure index and a household circumstances index using Census 1996 data • Provincial Indices of Multiple Deprivation (2006) – A series of provincial indices of deprivation were constructed and ward-level analysis of deprivation within each province was undertaken across five domains: income and material; employment; education; health; and living environment using Census 2001 data • Men, Women & Children (2013) – Using the Bristol method for comparative purposes, a poverty index was created across eight dimensions using data from the Living Conditions Survey (LCS) 2008/2009

2

What is the MPI? The global Multidimensional Poverty Index (MPI) is an international measure of acute poverty. The model was developed by Alkire & Foster from Oxford University for the United Nations (UN) and has been used in over 100 developing countries. The MPI “complements traditional income/expenditure-based poverty measures by capturing the severe deprivations that each person or household faces with respect to education, health and living standards”.1

Figure 1: Dimensions and indicators of the MPI

Source: http://www.ophi.org.uk/multidimensional-poverty-index/

The MPI assesses poverty at individual/household level. If someone is deprived in a third or more of ten (weighted) indicators, the global index identifies them as ‘MPI poor’, and the extent – or intensity – of their poverty is measured by the proportion of deprivations they are experiencing. According to the architects of the index, the “MPI can be used to create a comprehensive picture of who and where people are that are living in poverty… [and it also]permits comparisons across countries, regions and the world and within countries by population group, settlement type, as well as other key household and community characteristics. This makes it invaluable as an analytical tool to identify the most vulnerable people – the poorest among the poor – revealing poverty patterns within countries and over time, enabling policy makers to target resources and design policies more effectively.” There are a number of reasons why the MPI is gaining traction around the world as the preferred multidimensional poverty measure, namely: 1. It provides both a headcount and a measure of the intensity/acuteness of poverty; 2. It can decompose the overall poverty index by geographic area and demographic characteristics;

1

http://www.ophi.org.uk/multidimensional-poverty-index/

3

3. It can look at the contribution of each dimension to the overall poverty situation and decompose this contribution at whatever geographic level; 4. The methodology shows aspects in which the poor are deprived and helps to reveal the interconnections among those deprivations. This enables policymakers to allocate resources and implement policies more effectively. This can assist with targeting the correct services to the correct areas and not assume that a blanket approach is suitable to poverty alleviation/eradication per se; and 5. It is relatively simple to change the assumptions around which the multidimensionally poor are identified and therefore, making it an intuitive and transparent measure of multiple deprivations. Finally, in a post-2015 Millennium Development Goal (MDG) context, the meeting of the UN General Assembly in 2013 to report on progress towards the current goals saw various countries and institutions call upon the UN to adopt a new multidimensional poverty measure, the MPI2015+, to track progress alongside the traditional income/expenditure measures toward the new goals adopted after 2015. At a national level, governments were urged to embrace these new multidimensional poverty measures due to their effectiveness for policy change, with those countries that have already done so stressing the need to complement any international MPI with nationally adapted indices for poverty reduction.

Constructing the South African MPI Data sources Against this background, Stats SA embarked on the creation of a South African Multidimensional Poverty Index (SAMPI) to improve poverty measurement for the country and to align ourselves with the growing international trend towards measuring poverty beyond the traditional money-metric method. We have embraced the Alkire-Foster methodology and have also adjusted the indicators, and included an additional dimension – that of economic activity – in line with the call to develop national indices that reflect countryspecific conditions and needs. We chose to use census data given its ability to disaggregate information to the lowest possible geographic area. Census 2011 was the third census to be conducted in South Africa since the post-democratic elections in 1994. The first census was held in 1996 and the second in 2001. These censuses were all conducted as a de facto census, which means that people were counted where they were found or stayed on Census night (the midnight of the 9th/10th October), or, if they were not at the dwelling on Census night and were not enumerated elsewhere, where they returned to the next day. Data collection took place from 9 to 31 October of the respective year, with a Post-enumeration Survey (PES) happening in November to assess the undercount. For 2001, over 83 000 enumerators and over 17 000 supervisors and fieldwork coordinators were employed to collect information from about 80 000 enumeration areas (EAs). In 2011, a fieldwork force of 120 000 enumerators and about 31 000 supervisors and fieldwork coordinators was required to do the count in over 103 000 EAs. Using common variables found in Census 2001 and Census 2011,2 the SAMPI seeks to profile poverty at household level as most of the indicators used are generated at the household level. It also allows us to measure change between the two censuses. In order to compare Census 2011 with Census 2001,we were required to align the 2001 data to 2011 municipal boundaries. This is because the country’s demarcations underwent a number of changes at provincial and municipal levels between 2001 and 2011.

2

Unfortunately using Census 1996 would have limited us to a smaller set of common variables.

4

Of the nine provinces, only two provinces (Western Cape and Free State) were unaffected by provincial boundary changes. The provincial boundary changes were mostly as a result of eight cross-boundary municipalities which were absorbed in full into respective provinces. In 2001, the geographical frame consisted of 262 local municipalities. This total has been reduced to 234 local municipalities in the 2011 geographical frame. The difference of 28 municipalities is explained as follows: • In total, 25 District Management Areas (DMAs) were absorbed into the existing provinces. • The City of Tshwane absorbed a further two municipalities (Nokeng tsa Taemane and Kungwini). • A new municipality (Kagisano Molopo – NW379) was established by merging NW391 (Kagisano) and NW395 (Molopo). For municipalities, 107 municipalities decreased in geographical area while 155 municipalities had an increase in geographical area.

Exploring the data Once the list of common variables in the two censuses had been identified, a series of principal component analyses were conducted to identify the most suitable data items to use in the SAMPI. The statistical method of principal components analysis (PCA) is mainly a variable-reduction procedure that serves to reduce redundancy among observed variables by combining correlated variables – which by implication measure the same construct – into fewer uncorrelated derived variables (called principal components). The principal components are linear combinations of optimally weighted observed variables. Because PCA uses covariation among the variables involved in generating the weights, the resultant principal components account for a progressively decreasing amount of variance among the observed variables, with the first component accounting for most of the variance. In the context of the SAMPI, PCA was run to determine variables that best capture deprivation in predetermined dimensions (health, education, living standards and economic activity). The approach taken was to first produce a series of poverty indices using variables contained in census data and relate them to the above-mentioned dimensions.

The dimensions, the indicators and their deprivation cut-offs Creating an index is obviously guided by a number of factors. In the case of the SAMPI, these included: • the global MPI and its dimensions and indicators; • the South African context and issues affecting poverty; • the availability of data items in both censuses; and • the suitability and robustness of these data items after data exploration, confrontation and consultation. As a result of these factors, the following dimensions, their indicators and the deprivation cut-offs were established as the most suitable:

5

Table 1: The dimensions, indicators and deprivation cut-offs for SAMPI Dimension

Indicator

Deprivation cut-off

Health

Child mortality

If any child under the age of 5 has died in the past 12 months

Education

Years of schooling School attendance

If no household member aged 15 or older has completed 5 years of schooling If any school-aged child (aged 7 to 15) is out of school

Standard of living

Fuel for lighting Fuel for heating Fuel for cooking Water access Sanitation type Dwelling type Asset ownership

If household is using paraffin/candles/nothing/other If household is using paraffin/wood/coal/dung/other/none If household is using paraffin/wood/coal/dung/other/none If no piped water in dwelling or on stand If not a flush toilet If an informal shack/traditional dwelling/caravan/tent/other If household does not own more than one of radio, television, telephone or refrigerator and does not own a car

Economic activity

Unemployment

If all adults (aged 15 to 64) in the household are unemployed

In comparison with the global MPI, the SAMPI is identical in the following aspects: • The health dimension has the same mortality indicator. • The education dimension is identical with the same two indicators.

Regarding the standard of living dimension, the SAMPI is similar to the global MPI in the following aspects: • Whereas the global MPI has general access to electricity and cooking fuel as two indicators, the SAMPI looks at fuel for lighting, heating and cooking as three separate indicators. • The SAMPI includes the use of paraffin in the deprivation cut-off for lighting, heating and cooking purposes due to the health and safety implications. • Like the global MPI, the SAMPI also includes a water indicator but our deprivation cut-off is narrower, focusing on those households with no access to piped water in the dwelling or on their stand. There are those that may argue that this indicator should be in line with the minimum standards as articulated by the Reconstruction and Development Programme (RDP) – piped water within 200metres. However, given that this was the short-term aim of the RDP and its longer term vision “is to provide every South African with accessible water and sanitation” (The Reconstruction and Development Programme, paragraph 2.6.8), we believe our cut-off is appropriate and better suited for long-term measurement. • The SAMPI also has a narrower definition of deprivation when it comes to the sanitation indicator – as in the case of water above, we believe this is in line with the longer term goals of the RDP. • With respect to the assets indicator, the SAMPI does not include a bicycle in the deprivation cut-off as neither census collected information on this.

6

The SAMPI differs from the global MPI as follows: • In the health dimension, the censuses do not provide any information on nutrition, hunger or food security. As a result, the SAMPI only has one indicator in this dimension. We have explored other options, such as regarding water and sanitation as health issues or the use of disability data, but upon review opted not to use these. On the use of water and sanitation, we wanted to keep these in the standard of living dimension to be in line with the global MPI. With regard to the disability data, the questions differed significantly from the Census 2001 to the Census 2011, and so cannot be compared. Therefore, we have used just the mortality indicator. • The global MPI includes an indicator on the type of flooring in the standard of living dimension. In the absence of such data being collected by either of the censuses, we have included an indicator on dwelling type instead. • The SAMPI includes an economic activity dimension, as this dimension is critical to the current priorities of the country. The indicator identifies a deprived household where all adults in the economically active age cohort (15 to 64) are defined as unemployed using the expanded definition of unemployment (which includes those defined as unemployed as well as discouraged work-seekers). If there are any adults who are not economically active, such as still in education, retired or looking after the home, they would not be defined as unemployed and therefore, the household would not be classified as deprived on this indicator.

The poverty cut-off Given the indicators and their respective deprivation cut-offs, the following headcounts for households were generated for each indicator at a national level:

Table 2: Deprivation headcounts on each of the indicators Dimension

Indicator

Headcount 2001

Headcount 2011

Health

Child mortality