Abstract
We use longitudinal data from the National Income Dynamics Study to document the extent of recent short-term residential and household compositional change in South Africa. We analyse the demographic correlates of these transitions, including population group, age, urban/rural status, and income. We examine educational and labour-market transitions among movers and the prevalence of the four major types of compositional change – births, addition of joiners, deaths, and loss of leavers. We find that short-term household change is prevalent in South Africa. During a two-year period from 2008 to 2010, 10.5% of South Africans moved residence and 61.3% experienced change in household composition. We find that moving is more common among blacks and whites, very young children, young adults, urban individuals, and those with higher incomes. Among non-movers, compositional change is more likely for blacks and coloureds, young adults and children, females, urban individuals, and individuals with lower incomes.
1. Introduction
Under both apartheid and the preceding policy of separate development, long-range migrant labour played a central role in the process that trapped the majority of South Africa's population in remote, overcrowded pockets of the country. Thus, for a long time South African social scientists have given detailed attention to the consequences of residential movement (Crush, Citation1999; Wilson, Citation2001). The international literature, too, from Lewis's seminal work on migration, has been integral in capturing the processes of population transition that are part of longer-run economic development (Lewis, Citation1954; Harris & Todaro, Citation1970; Stark, Citation1991).
In the development literature, an explicit focus on changes in household composition has elicited less attention. However, this is not the case in developed countries as these compositional arrangements and changes are viewed as key to understanding the perpetuation of processes of disadvantage and the strategies employed by individuals to cope with disadvantage. The South Africa literature, too, has been sensitive to the importance of compositional changes because of the clear links between oscillating migrant labour and split and stretched household structures, with consequent distortions in the functioning of households of many South Africans at both the urban and the rural ends of this migration (Posel, Citation2001a, Citation2001b; Posel & Casale, Citation2006; Posel et al., Citation2006; Budlender & Lund, Citation2011).
Given the history of migrant labour, there have been a number of puzzling aspects to post-apartheid migration trends. First, it has been hard to understand why the aggregate levels of migration have not been larger given over a century of policies that dammed so many South Africans into the rural peripheries. Also, much migration has retained its character of stretching and splitting households (Kok et al., 2003, 2006; Posel, Citation2003). Of course, it is not only migration that drives changes to household composition; and the HIV/AIDS epidemic has focused attention on the consequences of deaths of economically active members of the household and to understanding household adjustments and support structures around debilitating illness. This is true both internationally and in South Africa (Bachman & Booysen, 2003; Hosegood et al., Citation2007; Chapoto & Jayne, Citation2008).
To gain an understanding of household behaviour, including migration strategies and coping strategies around deaths, births, and other ‘shocks’ to the household, recent social science has understood that panel data are enormously useful. There is no underestimating the value to be gained from following people over time as they migrate, as they respond to a death, or as the composition of the households in which they live alter in response to external events and internal decisions.
In this regard, social analysis in South Africa has lagged. Anthropologists have made a major contribution to our understanding of post-apartheid household change dynamics. Using careful observation of South African households in specific communities, they have detailed households under pressure with rapid dissolution and re-constitution (Spiegel, Citation1995) and multi-stage migration strategies (Cross, Citation2001). However, larger-scale, quantitative empirical work has been limited to cross-sectional description of falling household sizes and migration flows, usually based on census data (Simkins, 2001; Ziehl, Citation2004; Naidoo et al., Citation2008).
There are a few examples of studies that employ panel data to examine migration. Ardington et al. (Citation2009) use panel data from a demographic surveillance site in rural KwaZulu-Natal to show how the arrival of a state old age pension releases a credit constraint and facilitates the migration of working-age adults – especially females – in search of work. Madhavan et al. (Citation2012) use panel data from rural Mpumalanga to show that children of non-resident or temporary migrant mothers are more likely to migrate than children whose mothers reside with them full-time.
This paper seeks to add to this small corpus of panel data work. We use the first two waves of the National Income Dynamics Study (NIDS) (SALDRU, 2012a, 2012b) to detail residential dynamics and changes in household composition at the national level in South Africa. The nationally representative sample of 28 247 individuals who were interviewed in the first wave in 2008 and for a second time in 2010/11 provides the evidence for these changes.
We first explore the extent of both residential and compositional change among NIDS respondents overall and separately by population group. Next, we examine the demographic correlates (such as age structure, gender, income, and urban/rural residence) to residential and compositional change. In order to gain a deeper understanding of why South Africans might be moving, we describe the educational and labour-market transitions among movers. Finally, in order to more deeply understand household compositional change, we examine the prevalence of four major types of compositional changes: births, the addition of other household ‘joiners’, deaths, and the loss of other household ‘leavers’.
2. Data and measures
2.1 About the NIDS data
NIDS is a nationally representative study of South Africans, comprised of 28 247 respondents in about 7300 households. The study consists of several components: a household roster, an adult questionnaire for respondents age 15 and older, a child questionnaire for respondents under age 15 (reported by a caregiver), and a proxy questionnaire for absent household members. Questionnaire modules examine income dynamics, education, health, consumption, and more. A longitudinal study, NIDS seeks to interview respondents at two-year intervals. Currently, information has been collected over two survey waves – the first in 2008 (SALDRU, 2012a) and the second in 2010 and early 2011 (SALDRU, 2012b). As described above, NIDS is uniquely situated to examine household residential and compositional change, because it is currently the sole nationally representative panel data source in South Africa.
NIDS employs a stratified sampling approach, and survey weights are provided in order to obtain estimates generalisable to the South African population. We use these sample weights in all our analyses. Further information about the NIDS sampling procedures and other metadata are available online.Footnote5
2.2 Analytic sample
In this analysis we use individuals, as opposed to households, as our primary focal point. It would be impossible to categorise a wave 1 household as a ‘mover’ unless all members of the household moved to a new location, as in a ‘whole house’ relocation to one or more new residences. While this type of residential change does occur, it is not the only possible scenario (in fact, our results indicate that it is one of the least common scenarios). Therefore, it makes more sense to think about household changes from the perspective of the individual. Compositional change is not as complex because all members of a household should fall into the same category of composition changer or non-changer. Still, we choose to think about these changes as pertaining to individuals: an individual either did or did not experience a change in the makeup of his or her household between waves 1 and 2, just as an individual either did or did not move residence during this same time period.
Our analytic sample consists of individuals who were successfully interviewed in waves 1 and 2 of the NIDS study. Out of the 31 163 individuals originally selected for a wave 1 interview, 2916 were non-resident household members and were excluded from the panel because these individuals were not considered by NIDS staff to be ‘continuing sample members’. These non-residents were, however, eligible to be re-sampled in wave 2 if they returned to their wave 1 household. After this exclusion, there were 28 247 wave 1 sample members. We further excluded 6227 of these individuals because they were not successfully interviewed in wave 2, yielding a final analytic sample of 22 020.
Sample loss to follow-up is a big concern for any panel study, and NIDS staff made many efforts to ensure high response rates and to adjust for sample biases that might be induced by selective attrition. In both waves, multiple contact listings were collected for each respondent, including geo-codes for residential location. Between waves, NIDS maintains and updates sample members’ contact information via mail (change of address forms) and occasional telephone follow-ups. In the field, interviewers were required to make at least three attempts to reach someone at home. In cases of a household split, interviewers attempted to obtain new contact information for the absent household member. Following initial fieldwork, NIDS staff re-contacted all non-response households and individuals via telephone. Following fieldwork, statistical methods were employed to develop weights to correct the sample for attrition.
We employ two strategies to alleviate potential concerns stemming from panel attrition. First, we use the panel weights provided by the NIDS staff, which are designed to adjust the sample for non-response. Second, we compare the descriptive statistics of our analytic sample of 22 020 individuals to the full wave 1 sample of 28 247 to examine the ways in which selective attrition may induce bias into our results (discussed below).
2.3 Measures
The main dependent variables for this analysis measure whether an individual experienced residential or compositional changes in their household between the first and second waves of the study, a period that represents about two years. We constructed two binary variables to capture residential and compositional changes. In the case of residential change, respondents were categorised as either ‘movers’ or ‘non-movers’ based on whether or not their physical residence was in the same dwelling in waves 1 and 2. In the case of compositional change, respondents were categorised either as ‘changers’ or ‘non-changers’, based on whether or not their household membership was the same in waves 1 and 2. Non-resident household members, excluded from this study, were considered to be living outside the household.
In addition to our main dependent variables, we also obtain information about respondents’ demographic and socio-economic characteristics. Population group (black, coloured, Asian/Indian, white), gender (male or female), age, household location (urban or rural), and household income are all obtained from the wave 1 household roster.Footnote6 For age, we divide the sample into one of six different groups, each representing a distinct part of the life course: 0 to 5 (young child), 6 to 12 (child), 13 to 17 (adolescent), 18 to 25 (young adult), 26 to 59 (adult), and 60+ (older adult). We obtain information about respondents’ schooling and employment status at waves 1 and 2 from the adult or child questionnaires that were administered for each panel member in each of the waves.
3. Results
3.1 Sample characteristics
documents the descriptive statistics of our analytic sample of 22 020 respondents. At wave 1, blacks represented the largest proportion of our sample (80.7%), followed by coloureds (8.6%), whites (8.1%), and Asian/Indians (2.6%). Females represent a slightly higher proportion of our sample than males (52.4 versus 47.6%, respectively). About 39.5% of our sample was under the age of 18 at wave 1, while 53.1% were between the ages of 18 and 59 and 7.4% were 60 or older. In addition, more of our sample members lived in urban areas (57.6%) than rural ones (42.4%) at wave 1. Finally, those in lowest income quartile had a household per-capita income of R276 or less per month, those in the second income quartile had an upper bound on household per capita income of R510 per month, and those in the third income quartile had an upper bound of R1292.
Table 1: Characteristics of sample members
To ensure that our analytic sample is not seriously affected by panel attrition, we also show the descriptive statistics for the entire wave 1 sample in the second column of . According to the table, our analytic sample (column 1) is very similar to the full wave 1 sample (column 2) on the demographic characteristics shown. The small differences indicate that our analytical sample includes slightly fewer blacks and coloureds, women, young children, young adults, older adults, and urban dwellers because of attrition between waves.
3.2 Overview of short-term residential and compositional change: Movers, non-movers, changers, and non-changers
displays the extent of short-term household residential and compositional changes, for the full analytic sample and separately by population group. Our results indicate that short-term residential change is not negligible in South Africa: 10.5% changed residence during the two years between 2008 and 2010. This suggests that South Africans are mobile, with one in 10 moving to a new household in a two-year period. Household compositional change in the short term is extremely common: 61.3% of our sample members changed household composition between waves. Population group differences in residential change were relatively small compared with the very large differences in compositional change. In terms of residential change, the proportion of sample members who moved between waves is lowest among coloureds (9.1%) and highest among whites (13.4%). On the other hand, 65% of blacks and 58% of coloureds experienced short-term compositional change, while only 27% of whites changed household composition between waves.
Table 2: Proportion of individuals who experienced a residential or compositional change between 2009 and 2011
According to , there is a positive relationship between residential change (moving) and compositional change. This is not surprising given that a move will always be accompanied by a compositional change except in the case of a ‘whole-house’ move. However, compositional change is prevalent even among non-movers. Overall, 78% of movers experienced a compositional change, while 59% of non-movers experienced a compositional change.
The majority of respondents experienced at least one type of change (residential or compositional) in the two-year span. Most common were non-movers who experienced compositional change (53.1%), followed by non-movers who did not experience compositional change (36.4%). However, there were significant population group differences. While the majority of blacks and coloureds experienced some sort of residential and/or compositional change, nearly two-thirds (66.2%) of white respondents did not move or experience a compositional change during the two-year period.
3.3 Correlates of residential change
examines the correlates of residential change in South Africa at the aggregate level. The first five columns of are an efficient way of presenting bivariate associations between residential change and demographic characteristics. The coefficients represent the differences in the proportion of respondents who moved as compared with some reference group (which is omitted and represented by the constant). Essentially, each these columns represents a bivariate linear probability model (or ordinary least-squares regression with a binary outcome). The sixth column is comprised of estimates from a multivariate linear probability model with all demographic controls included. As a robustness check, and to address issues related to functional form inherent in linear probability models (namely, the linearity restriction), we present estimates from a logistic regression in the seventh column. We also estimated the same models, but stratified them by population group (shown in Appendices A to C).
Table 3: Residential change between waves 1 and 2
The coefficients in the first column of illustrate the unconditional proportion of respondents who moved, which we also presented in . The estimates indicate that the proportion of white respondents who moved is almost three percentage points (20%) higher than blacks (column 1); however, this difference is entirely explained by demographic differences between blacks and whites (column 6). Even after controlling for demographic characteristics, coloured respondents were still the least likely to have moved – that is, coloured respondents’ probability of moving is 3.1 percentage points less than black respondents of similar age, gender, household location, and income (column 6). These findings are robust to model specification, as indicated in column 7.
The coefficients in the second column of illustrate the unconditional proportion of respondents who moved depending on age. These estimates indicate that young adults (age 18 to 25) were the most likely to have moved (15.9%) followed by very young children (age 0 to 5, 11.8%) – and that older adults were the least likely to have moved (age 60+, 3.2%). This was true even after controlling for demographic factors and using different model specifications (columns 6 and 7). This is not surprising given that young adulthood is traditionally the period in the life course when many individuals transition out of their parental home. It is also unsurprising that older people moved much less, as that is the period in the life course when many individuals are no longer as mobile as they might have been during their younger years. Further examination indicates that these age differences are driven entirely by coloured and black respondents; white respondents of all age groups were statistically equivalent in terms of their likelihood of moving, except for older adults who were much less likely to have moved (see Appendices A to C). This might be explained by differential fertility rates and timing across population groups. For example, black and coloured young adults are more likely than white young adults to have young children, and this may explain why respondents of these two age groups are the most likely to have experienced a move among blacks and coloureds, but not whites.
The coefficients in the third column of illustrate the unconditional proportion of respondents who moved depending on gender. Overall, there does not appear to be a gender difference in residential moves, even after controlling for other demographic characteristics. The coefficients in the fourth column of indicate that, overall, respondents living in urban areas were more likely to have moved than similar rural respondents – and this is true after controlling for demographic characteristics and regardless of model specification (columns 6 and 7). However, further examination shows that this is true only for black and white respondents. Rural coloured respondents were more likely than urban coloured respondents to have moved (see Appendices A to C).
The coefficients in the fifth column of illustrate the unconditional proportion of respondents who moved depending on their household per-capita income (logged) in wave 1. Overall, respondents living in households with higher incomes were more likely to have moved, even after controlling for other demographic characteristics and using alternative model specification. Because of the vast differences in income by population group, it is particularly important to examine the association between moving and income separately by race group (see Appendices A to C). Further examination shows that the positive relationship between income and moving is true for black and coloured respondents, but for white respondents we observe a negative relationship between moving and income. This is probably explained by the vastly different housing realities faced by white and non-whites; that is, high-income whites may already live in the residence of their choice, whereas high-income non-whites (who have substantially lower income than high-income whites) may be moving to better or worse homes depending on the stability of their income.
To gain a better understanding of why these moves might have occurred, we examine changes in the education and employment status of movers between the two waves. To do this, we develop and examine education and employment transition matrices for respondents who moved between waves 1 and 2. Because educational and employment status is different depending on a respondent's place in the life course, we stratify the transition matrices by age group, with one matrix for each of the following groups: age 13 to 17 (), 18 to 25 (), and 26 to 59 () at wave 2.
Table 4: Educational and labour-market transitions for movers aged 13 to 17 at wave 2
Table 5: Educational and labour-market transitions for movers aged 18 to 25 at wave 2
Table 6: Educational and labour-market transitions for movers aged 26 to 59 at wave 2
According to , the vast majority of movers aged 13 to 17 (in wave 2) who were in education in wave 1 were still in education in wave 2 (82%). This is not surprising given the very high levels of school enrolment in the country. A smaller number (14.1%) reported being out of education and the labour force altogether (‘other’), while an even smaller number transitioned to the labour market (2.5% unemployed and 1.1% employed at wave 2).
According to , for those aged 18 to 25 (in wave 2) moves were more often accompanied by educational and labour-market transitions. According to the first row of the table, of those who were in education at wave 1, just 33% were still in education in wave 2, while 26% were unemployed, 20% were employed, and 25% were neither in education nor in the labour market.
The second row of shows that the majority (57.2%) of movers who were unemployed in wave 1 did not have employment in wave 2 (20% remained unemployed and 37.2% left the labour market and education altogether). About equal proportions of movers who were unemployed in wave 1 returned to education (22%) or were employed (20.8%).
The third row of shows that those who were employed in wave 1 were somewhat more stable. The majority of movers who were employed in wave 1 (64%) were still employed in wave 2; however, 35% of this employed group were either unemployed or left the labour market for reasons other than education by wave 2.
Finally, the fourth row of shows that the majority of movers (62.2%) who were neither in the labour market nor in education at wave 1 re-entered the labour force, although only about two-thirds (43.5%) were employed. About one-quarter (26.7%) of these movers remained out of the labour force and education in wave 2.
According to , movers aged 26 to 59 (in wave 2) also experienced fairly frequent educational and labour-market transitions between waves. The first row of shows that only a very small proportion (2.6%) of movers in this age group were in education at wave 1. About one-quarter (25.2%) of those who were in school at wave 1 were still in school at wave 2, while 35.3% were employed, 14.9% were unemployed, and 24.6% were neither in education nor the labour market.
The second row of shows that, for this same age cohort, a plurality (42.9%) of those who were unemployed in wave 1 found work by wave 2. However, about one-third (34.5%) of these respondents left the labour market altogether, and 22.5% remained unemployed.
The third row of shows that the vast majority (76.5%) of movers who were employed in wave 1 retained employment in wave 2. Of the remaining quarter, about one-half became unemployed (11.2%) or left the labour market altogether (12%).
Finally, the fourth row of shows that the majority (67.2%) of movers who were out of the labour force and education in wave 1 rejoined the labour market, but only about two-thirds of those (45.1%) were actually employed in wave 2. About one-third (32.8%) of movers in this group remained outside the labour force and out of education in wave 2.
Although we do not produce a transition matrix for movers who were over age 60 (in wave 2), the vast majority (over 80%) were out of the labour market and out of education at waves 1 and 2; and most of the remainder were employed.
3.4 Correlates of compositional change
Having examined the correlates of residential change in South Africa, we now consider changes in household composition among those who did not move between waves. We do not examine changes in the household composition of movers because, by definition, a mover will experience a household compositional change except for in the rare case of a whole-house move. According to our results in , of the 89.5% of respondents who did not move, 59.3% (53.1/89.5) experienced a change in household composition. examines the correlates of compositional change in the households of non-movers. As with , the first five columns of are an efficient way of presenting bivariate associations between compositional change and demographic characteristics. We also estimated the same models, but stratified by population group (shown in Appendices D to F).
Table 7: Compositional change between waves 1 and 2 among stayers
The coefficients in the first column of illustrate the unconditional proportion of non-movers who experienced a compositional change, which we also presented in . The estimates in column 1 of indicate that non-whites are nearly three times as likely as whites to experience changes in household composition. Blacks were most likely to experience changes in household composition (63.3%), followed by coloureds (55.5%), and whites (23.6%). After controlling for other observed demographic characteristics, about two-thirds of the black/coloured difference and one-half of the black/white difference is explained (column 6). Thus, even after accounting for demographic differences between the population groups, white non-movers were much less likely to experience compositional change than non-whites with similar characteristics.
The coefficients in the second column of illustrate the unconditional proportion of non-movers who experienced a household composition change depending on age. Overall, these estimates indicate that young adults (age 18 to 25, 64.7%) were the most likely to have experienced a change in composition, while older adults (age 60+, 49.8%) were the least likely to have experienced a change in composition between waves. These differences persist but are reduced after controlling for other observed characteristics (column 6). Further examination indicates that these age group differences depend on population group (see Appendices D to F). For example, the higher occurrence of household compositional change for young adults (age 18 to 25) is driven entirely by blacks, and the lower occurrence for older adults (age 60+) is driven entirely by non-blacks. This could be due to relatively high fertility rates among black young adults and/or the high prevalence of joining older individuals’ households as a poverty-coping strategy. Statistically, the proportion of non-mover whites that experienced a household compositional change between waves was flat across age groups.
The coefficients in the third column of illustrate the unconditional proportion of non-movers who experienced a household compositional change depending on gender. Overall, female non-movers are more likely than male non-movers to experience a household compositional change, and this is true even after controlling for other observed characteristics (column 6). However, this gender difference is completely driven by blacks (see Appendices D to F), which is probably a result of the high prevalence of female-headed households in this subpopulation.
The coefficients in the fourth column of illustrate the unconditional proportion of non-movers who experienced a change in household composition depending on urban/rural household location. While the estimate indicates that rural non-movers were more likely than urban non-movers to experience a change in household composition, this difference was entirely explained after controlling for other demographic characteristics. A further examination by population group (see Appendices D to F) reveals that this urban/rural difference is only present among blacks, with no significant difference between urban and rural coloureds and whites.
Finally, the coefficients in the sixth column of illustrate the association between income (logged) and household compositional change for non-movers. The estimate indicates that non-movers with higher incomes were less likely to experience a compositional change, and this was true after controlling for other observed factors (column 6) and was driven mainly by black and white respondents (see Appendices D to F).
To gain a deeper understanding of compositional change, we decompose compositional change into four types: the addition of one or more household members due to birth, the addition of one or more household members for reasons other than birth (other addition), the loss of one or more household members due to death, and the loss of one or more household members for reasons other than death (other loss). We examine the extent of these types of changes among those who did not move residence between waves but did experience a change in household composition. presents these results for blacks, coloureds, whites and the full sample.Footnote7
Table 8: Types of compositional change
According to the first row of , among black non-movers who experienced a compositional change, 31% added at least one new household member by birth, 59% added at least one new member by non-birth, 38% lost at least one member by death, and 23% lost at least one member by non-death. These results suggest that in the two years between waves, many blacks experienced multiple types of compositional changes, with the most prevalent type being the addition of new members by non-birth (other joiners).
According to the second row of , coloureds experienced the different types of compositional change in similar proportions to blacks, with a slightly higher percentage of coloureds experiencing birth(s) within the household and slightly lower percentages experiencing death(s) and the loss of one or more members due to reasons other than death (other leavers). The largest difference between blacks and coloureds is in their experience with ‘other joiners’ entering the household (a seven percentage point difference in proportion). Still, as with blacks, the most common type of compositional change among coloureds is the addition of one or more new members for reasons other than birth (other joiners).
It is also important to re-emphasise the fact that the sample for this analysis is limited to non-movers who changed household composition between waves. So while only a small proportion (20%) of whites did not move residence but did experience compositional change, the third row of provides insight into the predominant types of compositional changes experienced by this subset of the population. According to the table, and in contrast to blacks and coloureds, compositional change among white non-movers was most often due to death. We know from our examination of residential change that a large number of white non-movers were in the older adult age category (age 60+), and this probably explains why most compositional change in these households is due to death (first column of Appendix C).
4. Discussion
4.1 Summary of findings
This analysis provides insight into short-term household change in South Africa by examining the extent and correlates of residential change and compositional change, exploring educational and labour-market transitions among movers, and deconstructing the types of compositional changes experienced by non-movers. Four key conclusions arise.
First, residential change across a two-year period is non-negligible among individuals in South Africa and compositional change is very common, especially among non-whites. About one in 10 South Africans moved residences in the period between waves. Among the non-movers, about six in 10 experienced a change to their household composition in the period between waves. White non-movers are only about one-third as likely as non-whites to have experienced some changes in their household membership.
Second, although a greater proportion of whites than non-whites moved residences between waves, white moves appear to be most strongly correlated with income, while the correlates of moving among non-whites are much more diverse. Most notably, age appears to be a major correlate of moving for non-whites (particularly black respondents), with young adults and young children moving more frequently.
Third, economic transitions are very common among movers and the results are mixed. On the one hand, about four in 10 unemployed movers transitioned to employment or education (mainly employment); however, about three in 10 (greater for younger respondents) employed movers transitioned out of employment.
Finally, the types of compositional changes varied substantially by population group. Among non-movers, compositional changes in the households of non-whites were very common and came in all varieties (births, deaths, other additions, other losses). Compositional change in white households is not common and is mostly due to the loss of household members, mainly by death (because most white non-movers were older adults).
4.2 Implications, limitations, and future directions
In terms of moving, black and white individuals were much more likely than others to relocate to a new residence in the two-year period between 2008 and 2010. However, our results uncover that these moves are probably very different in nature depending on population group. Whites were more likely to move if they had low incomes, while blacks were more likely to move if they had higher incomes or were younger. Among blacks, moving is also strongly driven by age (younger people move more). As we noted before, high young adult fertility rates among this group are a likely contributor.
Unfortunately, our descriptive study does not identify whether individuals are moving to better or worse housing or employment opportunities, and future work should examine this complicated question further. There is some analysis of this question by Finn et al. in this issue, suggesting that moving is associated with real income gains on average. Additionally, we do not differentiate in our analysis whether the moves were across the neighbourhood, or whether moves were long distance. Given what we know about long-distance migration in South Africa, it is likely that long-distance moves are much more common among black than white movers (Naidoo et al., Citation2008).
As for compositional change, white households appear to be much more stable than non-white (especially black) households in terms of their membership. Black (and to a lesser extent coloured) households are quite turbulent, with multiple additions and subtractions occurring. Black households are unique because young and old individuals experienced compositional change, as compared with coloured individuals, for whom most compositional change was located within the younger end of the age distribution. This probably reflects the creation of multi-generational households, which are common among blacks (Budlender & Lund, Citation2011). These mutigenerational households are known to be an effective poverty-alleviation strategy, as the South African old age pension is a reliable and relatively substantial source of income (Case & Deaton, Citation1998). Our results paint a limited picture of compositional change in South Africa, and further studies are needed to examine whether or not ‘joiners’ and ‘leavers’ are related, romantic partners, and so forth, because the ‘roles’ of these gained/lost household members could be differentially consequential.
It is clear from our findings that the younger years are a very turbulent time for South African children in terms of household stability. Because a move or compositional change could disrupt or improve a child's development depending on the situation, an important line of future research will produce studies with more nuanced information about the residential and compositional transitions faced by children.
Acknowledgements
Murray Leibbrandt acknowledges the Research Chairs Initiative of the Department of Science and Technology and National Research Foundation for funding his work as the Research Chair in Poverty and Inequality. April Williamson acknowledges the Yale Global Health Initiative and the Lindsay Fellowship for Research in Africa for funding her work at the Southern Africa Labour and Development Research Unit.
Notes
5See http://www.nids.uct.ac.za/home/. Last accessed: 25 October 2013.
6Because of the small sample size for Asians/Indians, we do not discuss results for this population group.
7Because these categories are not mutually exclusive, these rows need not sum to 100%.
Related Research Data
References
- Ardington, C, Case, A & Hosegood, V, 2009. Labor supply responses to large social transfers: Longitudinal evidence from South Africa. American Economic Journal: Applied Economics 1(1), 22–48. doi: 10.1257/app.1.1.22
- Bachmann, M & Booysen, F, 2003. Health and economic impact of HIV/AIDS on South African households: A cohort study. BMC Public Health 3(1), 14. doi: 10.1186/1471-2458-3-14
- Budlender, D & Lund, F, 2011. South Africa: A legacy of family disruption. Development and Change 42(4), 925–46. doi: 10.1111/j.1467-7660.2011.01715.x
- Case, A & Deaton, A, 1998. Large cash transfers to the elderly in South Africa. Economic Journal, Royal Economic Society 108(450), 1330–61.
- Chapoto, A, & Jayne, TS, 2008. Impact of AIDS-related mortality on farm household welfare in Zambia. Economic Development and Cultural Change 56(2), 327–74. doi: 10.1086/522894
- Cross, C, 2001. Why does South Africa need a spatial policy? Population migration, infrastructure and development. Journal of Contemporary African Studies 19(1), 111–27. doi: 10.1080/02589000123491
- Crush, J, 1999. The discourse and dimensions of irregularity in post-apartheid South Africa. International Migration 37(1), 125–51. doi: 10.1111/1468-2435.00068
- Harris, R & Todaro, M, 1970. Migration, unemployment and development: A two sector analysis. American Economic Review 60(1), 126–42.
- Hosegood, V, Preston-Whyte, E, Busza, J, Moitse, S, & Timaeus, IM, 2007. Revealing the full extent of households’ experiences of HIV and AIDS in rural South Africa. Social Science & Medicine 65(6), 1249–59. doi: 10.1016/j.socscimed.2007.05.002
- Kok, P, O'Donovan, M, Bouare, O, & van Zyl, J, 2003. Post-apartheid Patterns of Internal Migration in South Africa. HSRC Press, Cape Town.
- Kok, P, Gelderblom, D, Oucho, JO & van Zyl, J, 2006. Migration in South and Southern Africa: Dynamics and Determinants. HSRC Press, Cape Town.
- Lewis, W, 1954. Economic development with unlimited supplies of labour. The Manchester School of Economics and Social Studies 22(2), 139–91. doi: 10.1111/j.1467-9957.1954.tb00021.x
- Madhavan, S, Shatz, E, Clark, S & Collinson, M, 2012. Child mobility, maternal status, and household composition in rural South Africa. Demography, 49(2), 699–718. doi: 10.1007/s13524-011-0087-3
- Naidoo, N, Leibbrandt, M & Dorrington, R, 2008. Magnitudes, personal characteristics and activities of Eastern Cape migrants: A comparison with other migrants and with non-migrants using data from the 1996 and 2001 censuses. Southern African Journal of Demography, 11(1), 3–38.
- Posel, D, 2001a. How do households work? Migration, the household and remittance behaviour in South Africa. Social Dynamics 27(1), 165–89. doi: 10.1080/02533950108458709
- Posel, D, 2001b. Intra family transfers and income pooling: A study of remittances in Kwazulu-Natal. South African Journal of Economics 69(3), 501–28. doi: 10.1111/j.1813-6982.2001.tb00023.x
- Posel, D, 2003. Have migration patterns in post-apartheid South Africa changed? Paper prepared for Conference on African Migration in Comparative Perspective, 4–7 June, Johannesburg, South Africa. http://pum.princeton.edu/pumconference/papers/1-Posel.pdf Accessed 5 February 2004.
- Posel, D & Casale, D, 2006. Internal labour migration and household poverty in post-apartheid South Africa. In Bhorat, H & Kanbur, R (Eds.), Poverty and Policy in Post-apartheid South Africa. HSRC Press, Cape Town, pp. 351–65.
- Posel, D, Fairburn, J & Lund, F, 2006. Labour migration and households: A reconsideration of the effects of the social pension on labour supply in South Africa. Economic Modelling 23(5), 836–53. doi: 10.1016/j.econmod.2005.10.010
- SALDRU (Southern Africa Labour and Development Research Unit), 2012a. National Income Dynamics Study 2008, Wave 1, V4.1. SALDRU, Cape Town [producer]. DataFirst, Cape Town [distributor].
- SALDRU (Southern Africa Labour and Development Research Unit), 2012b. National Income Dynamics Study 2010–2011, Wave 2, V1. SALDRU, Cape Town [producer]. DataFirst, Cape Town [distributor].
- Simkins, C, 2001. Can South Africa avoid a Malthusian positive check? Daedalus 130(1), 123–50.
- Spiegel, M, 1995. Migration, urbanisation and domestic fluidity: Reviewing some South African examples. African Anthropology 2(2), 90–113.
- Stark, O, 1991. The Migration of Labour. Basil and Blackwell, Cambridge, MA.
- Wilson, F, 2001. Minerals and migrants: How the mining industry has shaped South Africa. Daedalus 130(1), 99–121.
- Ziehl, SC, 2004. Documenting changing family patterns in South Africa: A census Data of any Value. African Sociological Review/Revue Africaine de Sociologie 5(2), 36–62.