Getting ahead or falling behind: Findings from the second wave of the National Income Dynamics Study

Abstract

The National Income Dynamics Study is the first national panel study of South African individuals. Its objective is to track these individuals over time to study social mobility. This paper documents the survey design and a successful recontact record in Wave 2 before providing an overview of the key findings from the other papers in this issue of the journal. Those from the top of the income distribution were hardest to recontact. The papers show that average real incomes grew slightly between 2008 and 2010. However, life satisfaction and expectations of future upward mobility declined. Being unemployed and moving into unemployment is associated with the lowest level of life satisfaction. Aggregate employment did not decline much but there was significant labour-market churn. The National Income Dynamics Study data reveal high levels of grade repetition and a slow transition from school to work. Relocating is shown to be an important part of schooling and employment decisions.

Keywords:

• mobility
• panel survey
• South Africa
• National Income Dynamics Study

1. Introduction

The National Income Dynamics Study (NIDS) is the first national panel study of individuals of all ages in South Africa. Its main objective is to measure and understand why some people are making progress and others are not; whether younger generations are better off than their parents; which livelihood strategies are associated with escaping poverty and which social policies are having the greatest impact. As a panel survey, NIDS seeks to track changes over time in the livelihoods of a fixed group of respondents and their households. NIDS goes to field every two years, with each round of the survey constituting a ‘wave’ in an ongoing statistical investigation. A key feature of the study is its ability to follow people as they move out of their original households; as the panel unfolds, it reveals the dynamic structure of households in South Africa and the changes in the living conditions and well-being of household members.

NIDS is a national research endeavour established by The Presidency of South Africa. The implementation agency for the first three waves is the Southern Africa Labour and Development Research Unit (SALDRU) at the University of Cape Town. The project is overseen by a Steering Committee comprised of representatives of several government departments and Statistics South Africa (Stats SA), as well as local and international academics. A description of the data – along with access to questionnaires, technical and discussion papers, and the NIDS data itself – is available online.⁴

Wave 1 of the NIDS survey took place in 2008 and provides the baseline data on the well-being of 28 247 individuals (‘permanent sample members’) in 7301 households against which to measure all future changes. As it happened, the NIDS Wave 1 survey took place in 2008, the year in which countries worldwide entered a devastating recession. South Africa was not long to follow, and the NIDS data from 2010 thus offer a window into how South Africans fared during this challenging period.

The preamble to the National Planning Commission's (2011) diagnosis of human conditions in South Africa observes that while constitutional democracy has brought many positive gains, ‘the legacy of racial, economic, gender and spatial exclusion continues to shape human development among South Africa's poor majority’. It goes on to raise questions that define the core rationale for establishing NIDS and which set the stage for this special edition of findings from the Wave 2 survey conducted in 2010:

Social and economic exclusions are both outcomes and causes of poverty and inequality. Race, class, gender and spatial inequalities combine with new risks and vulnerabilities to reduce the freedoms and opportunities available to the vast majority of South Africa's people. … In this context, is it possible for the majority of the population to achieve an acceptable standard of living? To what extent can this be achieved through a household's own initiatives, especially private earnings from work or business, or through other means? To what degree must social protection provide a safety net to prevent people from falling into deeper poverty? (National Planning Commission, 2011)

NIDS helps to lay a foundation for engaging with these and related questions so that the answers that emerge can contribute to evidence-based policy-making. This issue of the journal includes nine papers that explore a wide range of thematic areas, namely income mobility, wealth, subjective well-being, migration, labour-market outcomes, education and health. These findings combine to tell a broader story about change in South Africa.

2. Survey design and response rates

In Wave 1, 10 367 dwellings in 400 areas were selected to be approached to take part in NIDS. Of those dwelling units, 491 (4.5%) were found to be multi-household dwellings. Of the 10 858 eligible households, 7301 agreed to participate (see Table 1; data provided by NIDS staff).

Table 1: Wave 1 household responses

	Number	Percent
Total dwelling units sampled	10 367
Plus multi-households added to sample	491
Total potential sample	10 858	100
Less vacant and out-of-scope dwelling units	523	5
Less no-access areas	119	1
Less non-contacts	1 214	11
Less refusals	1 698	16
Participating households (as % potential sample)	7 301	67
Participating households (as % potential sample less vacant and out of scope)	7 301	71

Within the participating households, 31 163 individuals were identified as household members. However, 2916 people were not resident members and were excluded from the study. This was to avoid double counting, as they had a chance of being selected for the study at their ‘usual’ place of residence. A resident member was defined as a person who usually resides at the dwelling four nights a week. In addition, non-residents who were currently residing in institutions that are regarded as ‘out of scope’ (such as a hospital, prison or student hostel) were included in the sample. All these sample members, including children, are termed ‘continuing sample members’ and form the sample with whom interviews are attempted at each subsequent wave (see Table 2; own calculations on NIDS data).

Table 2: Wave 1 individual responses

	Number	Percent
Household members	31 163
Less non-resident members	2 916
Continuing sample members	28 247	100
Adult questionnaire	15 633	55
Proxy questionnaire	1 753	6
Child questionnaire	9 408	33
Individuals refusal or unavailability	1 453	5

Of the possible 28 247 continuing sample members from Wave 1, 22 050 were re-interviewed in Wave 2. When excluding those that moved out of scope or died between waves, the attrition rate is 19%. Table 3 summarises individual outcomes between Wave 1 and Wave 2.

Table 3: Wave 1 and Wave 2 individual outcomes

	Wave 2
Wave 1	Success	Refused/not available	Household non-response	Moved outside South Africa	Dead	Total
Success	21 098	520	4 279	51	846	26 794
Refused/not available	952	91	365	2	43	1 453
Not in Wave 1	6 591	209	23	0	140	6 963

It is important to note that non-respondents in Wave 2 (the sum of refusals, household non-response and those that moved outside South Africa) are not lost to the panel in perpetuity. Their names and contact details remain on record and they will be attempted again in future waves. The reasons for household non-response are shown in Table 4.

Table 4: Reasons for household non-response at the individual level

	Number	Percent
Refused/not available	1 814	38.9
Not located	2 224	47.7
Not tracked	629	13.5
Total	4 667	100.0

The biggest reason for individual non-response was household-level non-response, and the major reason for household-level non-response (48%) was that the household could not be located. Some of this was due to significant redevelopment or relocation of informal settlements. The reasons for attrition between Waves 1 and 2 are shown in Table 5.

Table 5: Reasons for attrition

Reason	Number	Percent
Refusal	2 136	37.5
Non-contact	2 714	47.7
Deceased	846	14.9
Total	5 696	100.0

Table 5 shows three categories of attrition. ‘Refusals’ are individuals who were not interviewed in Wave 2 because of an individual or household refusal. ‘Not contacted’ individuals consist of respondents who were not tracked, not located or moved outside South Africa. Finally, there are respondents who died between waves.

Of the nine provinces, the province with the highest attrition rate was the Western Cape at 28%. Gauteng was next at 25%, followed by the Eastern Cape at 23%. Limpopo, with 11%, had the lowest rate of attrition. Further insight into the incidence of non-response is presented in Table 6, where we disaggregate attrition by income decile (where decile 1 consists of the poorest 10% of households and decile 10 consists of the richest 10% of households).

Table 6: Attrition by Wave 1 income decile

Decile	Number of refusals	Number of non-contacts	Number of deceased	Total number of individuals	Attrition rate (%)
1	125	322	67	514	18.6
2	124	266	78	468	17.2
3	79	228	81	388	14.4
4	158	230	70	458	17.0
5	122	212	93	427	16.1
6	148	251	116	515	19.1
7	174	260	91	525	19.8
8	191	271	122	584	22.1
9	310	327	77	714	27.3
10	705	347	51	1103	41.6
Total	2 136	2714	846	5696	21.3

Table 6 shows that non-contact, rather than refusal, is the main reason for attrition for deciles 1 to 9. Interestingly, however, this is reversed for the top decile, where the ratio of refusal to non-contact is about 2:1.

Analysing the attrition rate by Wave 1 income deciles shows that the richest 10% were far less likely to participate in the second wave than all other households. A total 41.6% of the top decile were not successfully re-interviewed in Wave 2, whereas attrition rates in deciles 1 to 7 were in the 14 to 20% range.

The racial distribution of attrition is presented in Table 7. Here we see that ‘non-contact’ is the dominant reason for attrition among African respondents, while ‘refusal’ dominates for all other race groups. The population groups with the highest attrition rates are Whites and Asians/Indians.

Table 7: Attrition by racial group

Population group	Refusal	Non-contact	Deceased	Total	Attrition rate (%)
African	1 054	2 067	710	3 831	18.1
Coloured	480	427	100	1 007	26.1
Asian/Indian	112	30	8	150	39.5
White	490	190	28	708	53.1
Total	2 136	2 714	846	5 696	21.3

Between Waves 1 and 2, the overall attrition rate in the NIDS was 19%. This is reasonably high compared with attrition rates elsewhere in the world for household interview surveys (Lepkowski & Couper, 2002). For example, in the most recent survey comparable with NIDS – the Household, Income and Labour Dynamics in Australia survey – the attrition rate between Wave 1 and 2 was 13% (Watson & Wooden, 2004). Furthermore, the profile of attrition in NIDS is fairly unique: it is most prevalent among high-income households, and is disproportionately concentrated among White and Indian households (see Brown et al., 2012).

Attrition is a form of missing data, and consequently has the potential to bias point estimates and statistical inference from observed data alone (Hirano et al., 2001). In NIDS, the attrition weight is developed to compensate for the loss of respondents through attrition (for a discussion of the panel weight, see Brown et al., 2012). Researchers utilising the panel dimension of NIDS need to utilise this weight in order to avoid the potential biases that attrition can cause.

3. Data quality

As with any survey, the NIDS data are vulnerable to measurement error. Ongoing technical checks are undertaken within the SALDRU office to ensure the reliability of the data. The panel dimension of NIDS provides it with an in-built capacity for correction: variables that are poorly answered in earlier waves can be revisited in later ones. The quality of the data thus increases over time.

Two notes of caution must be sounded, however, as they are directly relevant to certain of the analyses in this issue. The first is that the data show a dramatic decline in unemployment, in particular a large – and unexplained – reduction in the number of individuals in the ‘searching unemployed’ category. Although there is not an exact comparison available in published Stats SA documents, their statistics do not indicate a similar trend and there is no reason to expect this decline in the searching unemployed to be accurate. Instead, the distribution of results suggests that some of the searching unemployed were not correctly classified by fieldworkers in the Wave 2 fieldwork. Given these concerns, the discussion of the labour market (Cichello et al., 2013) focuses on the rate of employment rather than the rate of unemployment. The second note of caution is that the high rates of attrition in the top decile (elaborated on above) may bias the analysis of income changes between 2008 and 2010.

4. Key findings

In this section, we briefly introduce some of the interesting findings from the analysis of the first two waves of NIDS. These are elaborated on in the papers contained in this issue.

4.1 Income mobility

The paper on income mobility (Finn et al., 2013) takes care to consider the impact of the selective attrition discussed above and to demonstrate that their findings are robust. They show that, in spite of the global recession, average real per-capita monthly income rose between 2008 and 2010. There were winners and losers in all groups, and the distribution of income changes was wide. More than two-thirds of respondents who were living below the poverty line in Wave 1 were still poor in Wave 2. Positive mean and median changes in average income were evident for all racial groups other than whites and across all settlement types. The highest percentage growth in median income was experienced by the lowest two income quintiles, with the top quintile seeing the smallest percentage increase. Looking only at mean income-change, all quintiles made gains except for the richest 20% in the sample. As a group, whites experienced the greatest fluctuations in income between 2008 and 2010, although their incomes are typically higher than those of other racial groups and these larger changes might not have the same impact on livelihoods that the smaller positive and negative changes have for the other groups.

On the face of it, these patterns of distribution seem tightly aligned with stock features of South African society, notwithstanding the fact that the NIDS sample saw an overall and across-the-board growth in mean income during a period of heightened economic adversity. More interestingly, the data show that it was not inevitable that the rich would stay rich and the poor would remain poor. Across all groups there were winners and losers, along with those who saw only relatively small positive or negative income changes and the many for whom things stayed as they were. A crucial implication which follows from this is that people's economic situation is not necessarily immutable.

A key benefit of the NIDS panel study is that by tracking individuals over time, as opposed to periodically sampling a different group of respondents, it is able to contribute to the identification of specifically vulnerable and specifically successful people through the profiles generated by the data. It can do so because, principally, the study registers change at the level of the individual. NIDS can identify who is experiencing change, in what ways, through what probable causes, and with what observable consequences. NIDS panel data are therefore also able to pinpoint those whose plight displays no significant change at all.

Of those who fell below the lower poverty line of R515 per capita per month in 2008, 70% had not been able to escape this poverty by 2010; of the 30% who did move above this lower poverty line, two-thirds remained below the upper poverty line of R948 per capita per month. By contrast, there was considerable movement among those originally lying between the lower and upper poverty lines (i.e. with incomes of between R515 and R948 per capita per month), with only 31% found there again in 2010. Among this group, 28% climbed out of poverty while 41% fell further into poverty and were now observed to be below the lower poverty line.

In summary, between Waves 1 and 2 the sample experienced positive real income changes on average. However, the distribution of change was very wide, and there were winners and losers across all groups during a period of notable adversity – a trend which suggests that people's economic outcomes are neither as inevitable nor as intractable as are sometimes assumed.

4.2 Wealth

The second wave of NIDS included a module on household wealth that had not been included in the first wave. This module provided much-needed information in an area noted for the paucity of data in South Africa. The fruits of this endeavour represent a major opportunity for expanding knowledge about wealth in South Africa, particularly about the composition and distribution of personal and household wealth.

Accruing wealth depends on the acquisition of assets, and while income and expenditure are important in determining an individual and household's current well-being, assets are key indicators of sustainable consumption in the future. Acquiring assets is itself dependent on the ability to access credit and the capacity to save, which in turn is a prerequisite for investing. Expected returns on investments vary over time and asset classes. In other words, investments carry risks and investors generally hedge their bets by diversifying their investment portfolios. Analysing people's composition of assets, especially investment assets, therefore sheds light on how, and to what extent, they insulate themselves from risk by planning for an uncertain future and ensuring later well-being.

Wealth accrual is linked to individuals' asset-acquisition capacities, access to credit, ability to save, diversification options and time and risk preferences with regard to investments. It is also closely related to age, in as much as these factors vary over the course of an individual's lifecycle to produce a typical picture that shows a rise in wealth between early adulthood and early middle age, a peak shortly before retirement, and a decline thereafter. Unlike income or expenditure, which is always non-negative, net worth can be positive, zero or negative depending on whether assets exceed liabilities or not. Hence, people typically have negative net worth in early adulthood as they set about accruing assets such as vehicles and repaying liabilities such as student loans. By middle age they typically transition into positive net worth; before retirement they tend to reach the peak of their net worth; and after retirement they generally begin to dissave.

This universally observed pattern is the basis for the lifecycle hypothesis, which predicts that individuals' wealth profiles follow an inverted U-shape as they get older. However, the pattern is disturbed when people lose the means to earn an income, such as when they are or become unemployed. During times of unemployment, individuals continue to consume and thus dissave. Where unemployment is widespread, national aggregate wealth accrual will be crucially affected.

There are six major asset types in the NIDS instrument: real estate, business, retirement annuities, financial, vehicular and livestock. A key finding in the paper by Daniels et al. (2013) is that, for poorer households, asset portfolios are concentrated in financial assets. For richer households these portfolios are dominated by real-estate assets, except in the case of the very wealthiest households, where real estate drops to just over one-half of all assets and retirement annuities begin to feature with significance in aggregate asset profiles. This suggests that it is only at the top end of the asset distribution that households diversify their assets.

The major liabilities in NIDS are fourfold: real-estate debts, business debts, financial debts and vehicle debts. Self-evidently, debts are incurred in order to make purchases. Less evidently, they are useful indicators of the types of goods for which households run up their largest debts. The finding of Daniels et al. (2013) is that the distribution of total debt is highly skewed. Financial debts play an important role for the mainstream of households, and it is only among the wealthiest that real-estate debts become dominant, a correlation to be expected in view of the qualifying conditions attached to home loans: individuals who can securitise these debts are more likely to be employed and thus to be wealthier.

The paper on wealth shows that wealth accrual is a complex process that generally describes a pattern of rise and decline in the course of individual lifecycles. Wealth over the age distribution shows a non-linear trend, suggesting a bequest motive in the financial behaviour of the aged. Accrual is adversely affected by unemployment. The distribution of total debt is skewed with household asset portfolios largely defined by the presence or absence of housing. Limited access to credit and high unemployment serve as likely barriers to home ownership and concomitantly housing finance.

4.3 Subjective well-being

The NIDS dataset presents an opportunity to assess economic well-being in South Africa in new ways. Over and above the fact that it is a panel study, NIDS augments income, asset and wealth measures with information about individual panel members' subjective, self-assessed well-being. Moreover, data of this kind were obtained in both Wave 1 and Wave 2, making it possible to explore changes in subjective measures over time by correlating them with changes observed in other categories of data. The result is that NIDS is capable of providing a more holistic reading of well-being and quality of life than has been previously possible.

Although NIDS is not the only national household survey in South Africa that collects information on subjective well-being, other surveys have posed their questions at the household level, thereby inviting the respondent to answer on behalf of the entire household. These questions rest on two problematic assumptions – first, that the respondent is willing and able to report accurately on the household's level of satisfaction; and, second, that the household enjoys a unified and identical state of well-being among all its members. What makes NIDS distinctive is that it frames its question at the level of the individual, a strategy that enables it to circumvent the difficulties associated with questions posed at household level.

Posel's (2013) paper in this issue engages with the dataset along three lines of enquiry: it looks at life satisfaction, perceived economic status and expectations of upward mobility. She finds that more than one-half of the resident adults were less satisfied with their lives in 2010 than in 2008, with Africans being much more likely than whites to report lower satisfaction. About two-thirds of adults did not perceive an improvement in their economic ranking in the period, although evidence suggests that individuals underestimate their relative positions. Expectations of future upward mobility declined considerably among adults, and especially among Africans. These subjective assessments appear to be interlinked, in as much as people who think they are richer than others, and who anticipate ranking higher in the future, will be more satisfied with their lives.

Regarding the second group of findings about self-perceived economic status, whites were more likely than Africans to perceive their economic status as having stayed unchanged between Waves 1 and 2; Africans were more likely to view it as having declined. In both waves, more than one-half of the adults in the sample thought they ranked among the poorest third of South Africans, indicating that a sizeable group of people are better off than they think are.

In a comparison of objective income rank with subjectively perceived economic rank, large divergences between the two were evident. For example, only 6% of adults among the richest third believed themselves to be on the uppermost rungs of a hypothetical economic ‘ladder’. The largest correspondence in income rank and perceived economic rank was among adults in the bottom third of the income distribution.

A striking inter-wave change in the responses to questions about perceived status concerns expectations of future mobility. In Wave 1, almost three-quarters of adults anticipated being on a higher rung in two years' time; by 2010, this had fallen to 50%. While South Africa may have weathered the economic storm of the financial crisis, it seems clear that this took a toll in terms of subjective perceptions about this 2008–2010 period as well as about the future.

4.4 Labour market

The paper by Cichello et al. (2013) uses the first two waves of NIDS to study the dynamics of the labour force in South Africa between 2008 and 2010. The focus is on transitions in the labour force, so individuals who had a change in labour-market status and those who were employed over both periods but had a change in type of employment are of particular interest.

Individuals between the ages of 20 and 55 in 2008 who remained in the sample in 2010 are considered part of the balanced panel. Their labour market statuses take on one of four options: ‘not economically active’ (NEA), such as students or homemakers; ‘discouraged unemployed’, who are able and willing to work but are not actively seeking employment; ‘searching unemployed’, those actively searching for work; or ‘employed’. Of those who were employed in Wave 1, 71.6% remained employed in Wave 2 and 28.4% fell out of the employment category. Of those searching for work in 2008, 32.3% reported being employed in 2010. Twenty-eight per cent of the discouraged job-seekers became employed between waves, while the majority of NEA did not move out of their category and only 22% moved from NEA to employed. Aggregate employment dropped by only 1.7% from 2008 to 2010, which highlights that broad measurements such as this one do not indicate the underlying churn of the labour market.

The story can be further unpacked by considering the race and gender breakdown of the various employment statuses. Noteworthy findings include the employment rate, which is far lower for women than it is for men (41.9% for women versus 63.3% for men in 2010). Women are also more mobile than men, with one-half changing their status from the first time period to the second, while only 38% of men made a transition.

The paper goes on to investigate not only the labour-market status of prime-aged individuals between 2008 and 2010 but also the types of employment that workers find themselves in. The three types are regular employment, self-employment and casual employment. Only 6.5% of those in casual employment in 2008 remained in that category in 2010, indicating this is a transitory state. Of all workers, only 7.2% were classified as casual workers and this decreased to 3.8% in 2010. Similarly self-employment dropped by three percentage points from Wave 1 to Wave 2, to 8.7%. Only regular employment increased to 87.5% of the employed, with 76.2% of those in regular employment in 2008 remaining there through to 2010. A transition matrix between the two waves shows that most of the shifts taking place were individuals moving into regular employment from the other categories.

The attraction of regular employment is apparent when earnings in the different work categories are taken into account. The mean earnings of those who moved out of regular employment decreased, while those who remained in regular employment, or moved into that category, experienced an increase in mean earnings.

The movement between the different sectors is also analysed in this paper. The primary sector consists of agriculture and mining/quarrying. The secondary sector comprises manufacturing utilities and construction while the tertiary sector covers wholesale and retail trade and various services. A final category of private households is included to encompass exterritorial organisations and all other activities. The majority of workers remained in the same sector between 2008 and 2010, except for those in the secondary sector in the first wave. Only 48.9% remained, while 38% moved into jobs in the tertiary sector. The decline of employment in this sector is probably driven by a decline in total employment in manufacturing.

The evidence gained from the NIDS panel survey shows that women display greater mobility into and out of the workforce and employment, while men who remain employed in both periods demonstrate more mobility across employment types.

4.5. Subjective well-being and unemployment

The paper by Lloyd & Leibbrandt (2013) looks at a topic that sits at the interface of the papers by Posel (2013) and Cichello et al. (2013) in that it uses the questions on subjective well-being and labour-market status from both waves of the survey to investigate the relationship between being unemployed and subjective well-being. The fact that this issue can be explored in NIDS serves as an important illustration of the many possibilities that there are in NIDs for doing important and innovative work on the non-economic burdens of poverty or on the role of subjective well-being and mental health in holding back or empowering progress through school, effective job search or career advancement.

The relationship between unemployment and subjective well-being seems to be some distance away from hard labour economics. This is not the case at all. In line with international practice, Stats SA measures unemployment in terms of those who searched for a job in the last two weeks. This is the so-called narrow definition of unemployment. Stats SA augments this definition with a broader definition that includes the narrow unemployed along with those who are not searching for employment but say that they want to work and would take a job if it is on offer. These non-searching unemployed that are captured in the broader definition are sometimes called the ‘discouraged unemployed’, in that they want work but the high costs of job search and the low probability of success have discouraged their search behaviour. This resonates with an understanding of the social and economic pressure that being unemployed puts on an individual as well as their household. It is stressful and unpleasant to be unemployed, especially discouraged unemployed. This is testable and it is this key insight that led Kingdon & Knight (2006) to use questions on subjective well-being from a 1993 South African survey to test whether the non-searching unemployed are as dissatisfied with their lives as the searching unemployed.

Lloyd & Leibbrandt (2013) repeat this exercise and find that the non-searching unemployed are indeed ‘discouraged’ rather than choosing to be idle. Indeed, they are the more dissatisfied than the searching unemployed and are the most dissatisfied group amongst the working-age population. Moreover, those who move into this category between the waves experience the greatest decline in subjective well-being than other transitions. This confirms and strengthens the finding of Kingdon & Knight (2006) that the non-searching unemployed are better thought of as unemployed than non-economically active.

4.6. Education

The NIDS Wave 2 dataset provides the first longitudinal information ever collected on education in a national household survey in South Africa. NIDS is also the first study of this kind to contain detailed questions about post-schooling shifts into the labour market. Using Wave 2 data, Branson et al. (2013) assess the changes in rates of grade progression, repetition and dropout between 2008 and 2010 and examine what respondents do when they leave school. They are able to study transitions across grades and transitions in and out of school in ways that have not before been possible. Their analysis of the data focuses on progress through school and into work, and also examines access to schools and the targeting of school funding.

In overview, the initial findings are that progress through school is slow. There are high rates of grade repetition throughout the grades, dropout increases systematically from Grade 7 onwards, and few youth successfully complete matric, with even fewer pursuing alternative vocational study. On the other hand, exiting the school system does not offer better options. Most of the respondents who were in Grade 12 in 2008 were so-called ‘NEETs’ (not in employment, education or training) in 2010.

The paper by Branson et al. (2013) has a particularly interesting gender dimension. They find that females have higher pass rates than males in every grade; females progress through school faster than males; and females are also more likely to further their post-schooling education. However, despite this relatively better educational performance, they are less likely than males to find employment when they exit the education system.

Post-apartheid education policies aim to furnish all learners with high-quality education and seek to achieve this by, inter alia, improving access to education and redressing inequalities in school funding. Schools are assigned to quintiles based on the income, employment rate and education level of their neighbourhoods. They are then allocated budgets according to quintile ranking, with lower quintile schools receiving larger allocations per learner than higher ones; schools in quintiles 1 and 2 – ‘no-fee schools’ – may not charge fees, but are compensated by the state. Against this backdrop, Branson et al. (2013) investigate access to schools and assess school outcomes. They are able to do this because the NIDS data have been integrated with certain data from the Department of Basic Education to provide information from the administrative records about schools, such as their locations and their quintile rankings. The overall finding is that education policies are successfully targeting the poor in terms of spatial access and funding, but that the school choices of poor learners are restricted and the basis of the quintile system is open to question.

Branson et al. (2013) find that the wealthier have a wider choice of schooling and travel further to attend schools of their preference. Most children do not attend their closest school, choosing to rather attend higher quintile schools which are less likely to be no-fee schools. What this suggests, especially in the case of poor learners, is that despite the relative ease of spatial access to schools, households are often prepared to have learners travel greater distances in order attend schools of perceived higher quality, but are willing to forfeit the school-fee exemption to which they might otherwise have been entitled had the learners gone to the closest school.

In relation to school funding, they compare respondents' socio-economic characteristics with the school quintile of their closest school and find that schools with the lowest quintile status have learner populations from the poorest households. School funding is therefore accurately targeted to poor neighbourhoods. However, the neighbourhood characteristics of quintile 1, 2 and 3 schools are similar in terms of income, employment and education; in other words, it is not clear that those in quintile 2, even quintile 3, schools are significantly less disadvantaged than those in quintile 1 schools. This finding raises some questions about the basis of school funding, because it suggests that in these cases the quintile system assumes a distinction between the socio-economic status of learners that does not seem to be an accurate reflection of the reality on the ground.

4.7. Health

The NIDS panel data hold great potential for deepening knowledge about the relationship between health status and socio-economic status in South Africa. Ardington & Gasealahwe's (2013) paper explores the socio-economic correlates of mortality and investigates the relationship between mortality and self-reported health.

Regarding the socio-economic correlates of mortality, the authors find that deaths are disproportionately concentrated among very young children and older people. The mortality rate decreases after early childhood and shows a roughly linear increase from age 15 and after; however, consistent with other South African datasets, it exhibits a hump between ages 20 and 40 most probably due to deaths associated with the AIDS pandemic. Africans have the highest mortality rate, followed successively by coloureds, whites and then Indian/Asians. Reporting any chronic conditions in Wave 1 is associated with a 3.2% increased risk of dying by Wave 2 (i.e. about double the odds of dying). Smokers in Wave 1 were 2.1% more likely than non-smokers to have died by Wave 2.

Regarding the correlation between self-reported health and mortality, the paper finds that self-rated health is a significant predictor of two-year mortality. This is consistent with evidence from many industrialised countries and a few developing countries. This association remains even after controlling for socio-economic status and several other subjective and objective measures of health. This finding is important in validating the usefulness of the self-reported health measures that are collected in NIDS.

4.8. Migration

This issue includes two papers on migration. The paper by Grieger et al. (2013) shows that 10.5% of the NIDS sample moved between Wave 1 and Wave 2 and three-fifths of the sample experienced a change in household composition. Among non-movers, compositional change is more likely to be experienced by blacks and coloureds, young adults and children, females, urban individuals, and individuals with lower incomes. They find that moves among white households appear to be most strongly correlated with income, while the correlates of moving among other racial groups are much more diverse. Most notably, age appears to be a major correlate of moving among African respondents, with young adults and young children moving more frequently.

To take the analysis of moving further, movers are divided into three age groups based on age in Wave 2 (13 to 17 years, 18 to 25 years, and 26 to 59 years) and four statuses are defined for these panel members (in school, employed, unemployed, and other). A series of transition matrices by age groups is used to compare the Wave 1 and Wave 2 status of these movers. As expected, 82% of movers 13 to 17 years old who were in education in Wave 1 remain in education in Wave 2. In addition, for 67% of those who were not in education in Wave 1 their move was associated with them being in education in Wave 2. It seems that educational decisions are key drivers of moving between waves.

Another key driver is employment and job search. The 18 to 25 year age group is the group most likely to move between the waves. In this group of movers, 64% of those who were employed in Wave 1 remained employed in Wave 2. Then, 20% of this group who were in education in Wave 1 and 21% of those who were unemployed in Wave 1 were employed in Wave 2. This is a key age group for transitions into employment and these transitions into employment are positive. Less positive, in the short run at least, is that 22% of movers in this age group were unemployed in Wave 2. This includes 22% of those who unemployed in Wave 1 and 26% of those who were in education. Under the assumption that moving is correlated with a higher probability of employment than not moving, this is still a potentially positive transition. However, data from Wave 3 and beyond will allow this issue to be resolved.

The paper by Clarke & Eyal (2013) explores further the relationship between migration and employment among prime-aged African adults. A lack of employment in Wave 1 is shown to be positively correlated with the respondent having migrated by the time of the Wave 2 interview. This relationship is stronger for men than women. Women are more likely to migrate for reasons unrelated to job search or employment. They also find that receiving a government housing subsidy or (Reconstruction and Development Programme, ‘RDP’) house substantially reduces the likelihood of migration, particularly among women. This is an intuitive finding because owning a house increases the opportunity cost of leaving the area. They also find that social assistance grants to the household are associated with a lower probability of migration.

5. Conclusion

This paper has sought to provide glimpses into the richness and power of the NIDS panel data. While NIDS is still in its infancy, there are interesting lessons to be learnt from the first two waves. In this paper we have shown that NIDS reveals a society that is characterised by substantial income mobility, with winners and losers right across the income distribution. Despite the fact that average real incomes did not fall between the two waves, South Africans were significantly less satisfied with their lives in 2010 than in 2008, and two-thirds of them did not think their economic situation had improved in this time. Expectations of future upward mobility were notably lower in 2010 than in 2008, especially among Africans. The aggregate level of employment showed little decline. However, about one-third of those individuals who were employed in 2008 were out of work in 2010, suggesting significant labour-market churn. Importantly, people who moved residence fared better than stayers in finding work and remaining employed.

In schools, there were high rates of grade repetition, an increase in dropout from Grade 7 onwards, and low completion rates. The vast majority of those in Grade 12 in 2008 were neither employed nor studying in 2010. Female learners perform better than males, but are less likely to find work after leaving school. Education policies are successfully targeting the poor in terms of spatial access and funding, but the poor have less choice in schooling than the non-poor.

The NIDS data show that there is a significant amount of migration and fluidity in household formation. More than three-fifths of South Africans experienced a change in their household composition over the two-year period. Change can be positive; for example, when it involves migrating in order to take up a new job or to attend an educational institution. However, it can also signal disruption, particularly for children.

In conclusion, the NIDS data present a dynamic picture of a society undergoing significant economic and social change. The data present evidence of a society in which individuals and the households in which they live are aspiring to do as well as they can within their socio-economic constraints through the choices that they are making about schools for children, place of residence and employment. However, for the poorer sections of our society, these socio-economic constraints are very limiting. Further waves of NIDS data will continue to track these individuals and will reveal which of these decisions and actions were successful in positive trajectories going forward. The papers in this issue only scratch the surface in terms of what is possible with the NIDS data, but they provide an overview of some of the key themes and strong encouragement to the broader research community to make use of these publicly available data to explore these key issues of socio-economic mobility in contemporary South Africa.

Notes

⁴See http://www.nids.uct.ac.za.