Rural-Urban Migration as a Means of Getting Ahead

Full text

1
Rural-Urban Migration as a Means of Getting Ahead
Justin Visagie and Ivan Turok
October 2017
Forthcoming HSRC Press
1. Introduction
One of the most striking legacies of Apartheid is the persistent mismatch between the
geography of population and the geography of jobs. Stark spatial divides have continued
since the 1990s, despite state efforts to undo the damage of the past by establishing common
constitutional rights, universal social protections, a unitary system of sub-national
government, sizeable fiscal transfers between leading and lagging regions, and many national
programmes intended to lift people out of poverty and promote development. Indeed the gulf
in economic conditions between and within regions appears to have widened since the 1990s.
These inequalities are matched by stubborn gaps in the availability and quality of social
infrastructure, amenities and public services, and hence in living conditions and well-being.
There is little doubt that these spatial divisions exert a profound influence on people’s life
chances and on their ability to achieve their true potential.
The government’s most substantial response has been to skew public resources towards rural
communities and towns through social grants and disproportionate spending on healthcare,
schools, housing and basic services. Although redistributive expenditure on social welfare
and human settlements has done much to alleviate extreme hardship and misery, it has not
tackled the underlying causes of poverty and social exclusion in the severe shortage of
economic opportunities in rural areas. The education system has also arguably failed to
prepare young people for the changing labour market, epitomised by the lack of regular, paid
employment.
In these circumstances, moving from the countryside may offer people the prospect of a
better future. The rate of rural-urban migration certainly accelerated after influx controls were
lifted from the cities in the 1980s (Turok 2014). Yet economic progress for migrants is by no
means assured, since there are multiple barriers to entry into urban labour markets. People
leaving rural areas face substantial transport costs and difficulties in gaining access to

2
affordable housing in the cities. Migrants with poor educational qualifications and vocational
skills are bound to struggle to compete for anything but the most precarious, low-paid, entry-
level jobs. Many may be confined to generating their own meagre livelihoods through self-
employment and other survivalist activities on the periphery of cities. Living in rudimentary
shacks in informal settlements without basic amenities means a hazardous existence exposed
to fire, flooding, disease, crime and other threats to personal safety.
This chapter examines whether rural-urban migration is associated with an improvement in
economic circumstances for rural-urban migrants. It uses an invaluable source of longitudinal
data from the National Income Dynamics Study (NIDS). This follows a large group of
individuals every two years starting in 2008. The paper asks whether migrants are more likely
to escape poverty and experience upward social mobility than those who remain in rural
areas. Does migration represent a useful way of promoting economic upliftment and social
inclusion in South Africa? Is this form of ‘adjustment’ a more efficient way of narrowing
spatial disparities than trying to rebalance economic activity between cities and rural areas?
The structure is as follows. The next section discusses the challenges and opportunities of
urbanisation, migration and development. Section three explains the method of analysis and
source of data. The following section examines patterns of social mobility associated with
migration. Section five explores gross flows backwards and forwards between sending and
receiving communities. Section six summarises the main findings and draws conclusions.
2. Urbanisation and development in South Africa
Around the world, urbanisation has been closely associated with economic and human
development for more than two centuries. Cities currently account for 54% of the world’s
population, yet over 80% of global GDP (UN-Habitat 2016). This is because the
concentration of population and production creates positive externalities or ‘agglomeration
economies’ that boost productivity, innovation and entrepreneurial dynamism, leading to
faster economic and employment growth (World Bank 2009, Glaeser 2011, Storper 2013).
In-migration fuels a larger labour supply, bigger consumer markets and economies of scale
and scope in production, economic infrastructure and service delivery. These advantages
attract investment, spur output growth, raise household incomes and generate tax revenues
for improved public infrastructure and transport connectivity. The positive feedback loops

3
between urbanisation and economic development can in turn enhance public health,
subjective well-being and the quality of life (UN-Habitat 2016).
Yet this ‘urban dividend’ does not emerge automatically through the spontaneous actions of
households and firms. Unstructured and haphazard forms of urban development can generate
serious congestion, conflict and contagion (Glaeser & Sims 2015). Overcrowded informal
settlements illustrate how unregulated density creates major environmental and health risks
and social insecurities. To reap the benefits of agglomeration requires the investments made
in housing, business premises and physical infrastructure to be coordinated so as to ensure
proximity and accessibility (Collier & Venables 2017). An efficient spatial arrangement of
the city curbs commuting distances, lowers the cost of trade between firms, reduces traffic
congestion and limits pollution (UN-Habitat 2016). In other words, compact and connected
cities counteract the negative externalities of unplanned and dispersed urban growth.
In South Africa, urbanisation was viciously restricted under Apartheid through influx
controls and forced removals. A punitive system of oscillating migrant labour was engineered
to provide cheap transient labour for the mining areas of the Transvaal and other industrial
cities and towns (Adepoju 1988, Posel 2004). This fostered a temporary form of African
urbanism whereby many migrant workers developed some attachment to their urban
locations, while retaining a strong social affiliation with their rural homelands. The rate of
rural-urban migration rebounded somewhat during the 1980s and 1990s after the physical
restrictions on urban movement were lifted (Turok 2014). It remains unclear whether South
Africa still has a lower level of urbanisation than one might expect, considering its level of
economic development. Some observers argue that urbanisation should be actively
encouraged by government as one way of accelerating national economic growth and job
creation (Centre for Development and Enterprise 2017).
Another distinctive feature of urbanisation in South Africa is the distorted physical form of
cities. The imposition of racial segregation created an unusually fragmented spatial structure,
with crowded black townships located on the periphery, far from jobs and amenities. Even
today, residential densities tend to rise with distance from the urban core. This is a source of
inefficiency in the use of land and in the provision of infrastructure and service delivery. It is
also a feature of social injustice. Low income households typically spend between 20-40% of
their monthly earnings on transport costs (Kerr 2017). The government’s main housing

4
programme has been unable to keep pace with the growth in the urban population, resulting
in the proliferation of informal settlements on marginal land and shacks in township
backyards. Since urban land is such a scarce resource, there is intense pressure on vacant land
and empty buildings, which generates many disputes and regular violent conflicts between
shack dwellers, property owners and the police.
For all these reasons, South African cities remain inhospitable, relatively high cost
environments for rural migrants. With relatively few rungs on the housing ladder or in the
labour market, they hamper people’s efforts to get ahead. Their segregated structure hinders
rather than helps social inclusion. The spatial mismatch between jobs and population within
cities makes it extremely difficult to operate viable public transport systems. The sprawling,
fragmented form of cities also undermines agglomeration economies, which stem from the
intense interactions that are only possible in dense and diverse urban districts.
Despite these limitations, the question arises as to whether South African cities are still more
conducive environments for economic development and human progress than towns and rural
areas? They may not be well-structured or particularly functional, but they may still be more
favourable locations for productive investment, business formation and the entry of
additional labour into employment or self-employment. The issue of labour market entry and
progression over time is vital for a country where the level of social inequality is extremely
high and the level of social mobility is generally very low (Keswell et al. 2013, Piraino
2015). We are particularly interested in whether people moving from rural to urban areas are
more likely to access employment and escape hardship than people remaining in the
countryside.
There has been surprising little systematic research on the impact of migration on household
welfare in cities and rural areas, perhaps because of the lack of longitudinal data until
recently (Roberts 2006, Kok & Collinson 2006, Rogan et al. 2009). Posel and Casale (2006)
and Pendleton et al. (2006) found higher levels of poverty among migrant-sending
households compared to non-migrant households in rural communities, possibly because of
the loss of an income generator. Rogan et al. (2009) made use of the longitudinal KwaZulu-
Natal Income Dynamics Survey, and also found that labour-sending households did not fare
well over time. Mixed messages emerge from research on the benefits for migrants settling in
urban areas. Cornwell et al. (2004) reached positive conclusions in that households who

5
moved to an urban area within the last year were just as successful at finding employment as
existing urban households. Mulcahy and Kollamparambil (2016) used the NIDS for 2008 and
2012 and found that migration led to an increase in income, but a decline in subjective well-
being. They suggested that the decline in life satisfaction might be attributable to migrants
having unrealistic expectations and experiencing an emotional cost from being away from
their family and home environment.
3. Data and approach
The National Income Dynamics Study (NIDS) follows a large group of individuals (rather
than households) over time. The first wave was carried out in 2008 and this has been
followed up every two years until the fourth wave in 2014. The NIDS tracks individuals who
reside within a household, defined as those staying at least four nights a week within that
household. A low proportion of residents (less than 3%) report being away from home for
more than one month in any of the waves and no panel member reports being absent from
their household for more than one month in all of the waves. One can be fairly confident that
people’s life-chances are being shaped by the geography of the households in which they
report residence.
The following analysis is based on a balanced panel constructed from the geographical
information made publicly available in the NIDS dataset.
1
This includes information on the
district (or metropolitan municipality) of residence for each panel member, and a
categorisation of the settlement’s geographic status by urban, traditional or farms according
to the Census 2011.
2
The sample is restricted to individuals of working age (19-64 years) in
all four waves.
3
Weights are applied to correct for sample attrition. The weights also allow us
1
The precise geographical coordinates of each household are not made publically available to protect the
individual’s anonymity.
2
Urban: defined as “A continuously built-up area that is established through township establishment such
as cities, towns, ‘townships’, small towns, and hamlets. The areas are identified by “erf/erven/cadastre”
from the Surveyor General or Municipal planning units.”
Traditional: defined as “Communally-owned land under the jurisdiction of traditional leaders. Settlements
within these areas are villages.”
Farms: defined as “Land allocated for and used for commercial farming including the structures and
infrastructure on it. The areas are identified by farm and farm portion cadastre from the Surveyor General.”
3
We do not include those aged 15 – 18 years in wave 1 to allow members of the sample to have completed
their secondary schooling. We also restrict the sample to those who were not older than age 56 in wave 1
due to exiting the labour force by the fourth wave. If we relax the age restrictions we reach the same
conclusions.

6
to provide total numbers of the overall scale of change for the national population. This helps
to capture the significance of the socio-economic trends underway.
Three cohorts have been created for comparative purposes:
 ‘remained urban’ – people who were in an urban area within the same district/metro
in all four waves.
4
 ‘remained rural’ – people who were in a traditional area in the same district in all four
waves (or otherwise in the same farm area).
5
 ‘rural-urban’ – people who were rural in the first wave and urban in any of the
subsequent waves (in a linear progression).
We have not included a category for linear ‘urban-rural’ migration due to the small size of
the sample (approx. 100 observations). However in Section 5 we tentatively consider gross
migration flows between rural and urban communities which have larger samples. A
limitation of the dataset is that we are unable to account for migrations that could have
occurred between any of the waves (i.e. any short-term migration of less than 2 years). For
example, we may miss people who moved to urban areas in an attempt to find gainful
employment, but who subsequently returned because they were unsuccessful. Omission of
this group may tend to inflate the apparent effects of migration on social outcomes.
Table 1 presents descriptive statistics for each cohort in wave 1. The sample size is quite
small for rural-urban migrants (450 observations), so 95% levels of confidence are reported
to demonstrate that the findings are robust. Attention is specifically drawn to instances where
the estimates are statistically insignificant.
The cohorts represent distinct groups in several respects. The average age of rural-urban
migrants was younger than of the other two cohorts (27 years compared with 36 years).
Younger people are likely to have fewer social ties and may have less domestic
responsibilities, which simplifies the migration decision. Black Africans comprised the
overwhelming majority of the rural and rural-urban cohorts. Consequently rural-urban
migration tended to reduce racial differences between regions. Migrants were just as likely to
be women as men.
4
In urban areas, the sample is split into roughly 50/50 between metros and other small towns or cities.
5
In rural areas, roughly 15% of the sample are in farmland, and 85% are in traditional areas.

7
Labour market status in wave 1 differed across the three cohorts. Migrants were less likely to
be employed and more likely to be searching or non-searching unemployed than individuals
who remained in rural areas. The unemployment rate for migrants was 50% in wave 1,
reflecting the chronic shortage of work for migrants before their decision to migrate. In
comparison, far more urban residents were in employment and far fewer were economically
inactive, reflecting the better economic conditions in urban areas. More than a third of
individuals who remained rural or migrated to an urban area said they were economically
inactive (not looking or available for work) in 2008. They may have been studying, had
domestic responsibilities or been discouraged from seeking work. Poverty was extremely
high among both of these groups, with four out of every five people living below the poverty
line.
6
Meanwhile, only half of urban residents were below the poverty line, a poignant
reminder of the rural-urban income gap which motivates people to move to the cities.
The last important distinction between the cohorts is their duration of schooling. Urban
residents and rural-urban migrants had on average completed grade 10 level education.
Approximately two-fifths of migrants had completed secondary school. Meanwhile rural
residents had on average completed grade 8 and roughly one-fifth had finished secondary
school. Therefore migrants were slightly better educated when looking for work in the city
than people who stayed behind. Yet higher educational attainment was not correlated with a
superior labour market position in wave 1.
The fact that rural-urban migrants had the same levels of poverty in 2008 as the remaining
rural residents may suggest that these groups faced similar socio-economic deprivations
before any migration occurred. It is useful to bear this in mind as their subsequent socio-
economic trajectories are described, notwithstanding that it is not possible to infer causal
relationships.
6
We employ an upper-bound poverty line of R1140 per capita in December 2012 prices based upon a
basic-cost-of-needs approach as developed by Budlender et al. (2015).

8
Table 1: Descriptive statistics between migration categories at the wave 1 baseline
Remained rural
Rural-urban
Remained Urban
Sample
Percentage share
29.8
4.8
65.4
[25.04, 32.17]
[3.91, 5.48]
[58.53, 66.3]
Total number (thousands)
4,322
703
9,487
[3,817; 4,827]
[584, 822]
[8,363; 10,611]
Observations
2881
450
3271
Age
Average age
36.2
27.0
36.1
[35.61, 36.89]
[26.03, 28.01]
35.21, 36.94]
Gender
Female
62.4
50.7
55.2
[60.08,64.57]
[44.66,56.76]
[52.49,57.81]
Male
37.7
49.3
44.8
[35.43,39.92]
[43.24,55.34]
[42.19,47.51]
Race
African
96.8
94.0
76.2
[94.23,98.21]
[89.09,96.81]
[68.41,82.51]
Coloured
2.4
4.4
11.4
[1.198,4.676]
[2.176,8.721]
[7.303,17.22]
White/Indian
0.9
1.6
12.5
[.3349,2.145]
[.4985,4.801]
[8.058,18.84]
Labour Market
Employed
42.1
31.3
56.8
[38.93,45.39]
[25.46,37.72]
[53.03,60.52]
Searching unemployed
15.7
25.7
18.9
[13.58,18.05]
[19.91,32.43]
[16.39,21.65]
Non-searching unemployed
7.6
7.2
6.3
[6.316,9.159]
[4.313,11.64]
[4.973,7.897]
Not Economically Active
34.6
35.9
18.0
[31.61,37.66]
[29.76,42.56]
[15.7,20.61]
Unemployment rate
35.86
50.37
30.28
[32.54,39.32]
[42.22,58.5]
[26.92,33.86]
Schooling
Average years
7.6
10.0
9.5
[7.25, 7.85]
[9.56, 10.40]
[9.25, 9.77]
Poverty Rate
81.78
82.81
53.61
[78.13,84.93]
[77.09,87.34]
[47.52,59.6]
Source: NIDS wave 1; own estimates
Notes: 95% confidence intervals in parentheses

9
Figure 1: Number and percentage of total rural-urban migrations in each wave
Source: NIDS waves 1 - 4; own estimates
Notes: See table A1 in the appendix
Figure 1 shows the levels of migration from waves 1 to 4. It is worth noting that far more
migration occurred between waves 2-3 and waves 3-4 than between waves 1-2, possibly
following signs of an economic recovery in the cities between 2010 and 2014. Cumulatively,
it is only by wave 4 that the whole rural-urban cohort would have migrated to urban areas.
Scaled-up to national totals using the sample weights, this equals roughly 700,000 migrations
over the period. Therefore, any effects of rural-urban migration are amplified across each
wave when making descriptive comparisons between the cohorts. Transition matrices
between waves 1 and 4 can assist in understanding patterns of mobility.
4. Descriptive evidence linking migration to social mobility
Figure 2 shows the proportion of people in each cohort who were living in poverty in 2008,
2010, 2012 and 2014 (defined in footnote 6). It reveals a sizeable gap between the incidence
of poverty in urban and rural areas. The trend over time has been positive for people living in
both types of area. However, the proportion of people living in poverty has declined faster in
urban areas. The extent of poverty among rural-urban migrants was similar to rural residents
in 2008 and 2010 (the slight decline among migrants was not statistically significant)
although the majority of migrations had still to occur. The fall in poverty among migrants
116
275
313
0
50
100
150
200
250
300
350
0
10
20
30
40
50
60
W1-W2 W2-W3 W3-W4
Thousands
Percent (%)
Number of migrants % of migrants

10
Figure 2: Percentage below the poverty line by migration cohort, wave 1 – wave 4
Source: NIDS waves 1 - 4; own estimates
Notes: See table A2 in the appendix
Figure 3: Poverty transitions for each migration cohort, wave 1 to wave 4
Source: NIDS waves 1 and 4; own estimates
Notes: See transition matrix A3 in the appendix
20
30
40
50
60
70
80
90
100
2008 2010/11 2012 2014/15
Percent (%)
Remained rural Rural-urban Remained urban
21.3
66.5
43.0
27.7 30.3
14.1
0.0
10.0
20.0
30.0
40.0
50.0
60.0
70.0
Remained rural Rural-urban Remained urban
Percent (%)
% poor W1 became non-poor W4 % non-poor W1 became poor W4

11
was very substantial in 2012 and 2014, corresponding to the majority of migrations over the
period. Eight out of every ten migrants were living in poverty in 2008, but this fell to just
three out of ten by 2014. By the time the full cohort had moved to urban areas, the incidence
of poverty was the same as those who had remained in urban areas throughout. The extent of
progression out of poverty experienced by migrants is very striking.
A similar message emerges from figure 3, which shows the transitions into and out of poverty
between waves 1 and 4 (the corresponding transition matrix is provided in table A3 in the
appendix). The first column in each set shows the percentage of poor people who moved out
of poverty by wave 4. The second column shows the proportion of people who were not poor
at the outset but who moved into poverty by wave 4. The key finding is that two-thirds of
migrants escaped poverty compared with one-fifth of rural residents and two-fifths of urban
residents. In absolute numbers, this amounts to 385,000 migrants escaping from poverty by
wave 4 out of 580,000 who were originally poor in wave 1.
7
Interestingly, urban-dwellers
were more resilient to falling back into poverty. When the sample of migrants is sub-divided
into those who moved to a metro and those who moved to a secondary city or town, the
former cohort experienced a bigger reduction in poverty.
8
This suggests that the size of a city
matters to migration outcomes.
These differential changes occurred during a period of overall expansion of social grants to
poor households. Indeed rural communities are bound to have benefited more from these
transfers because of the higher incidence of poverty there. The findings probably imply that
the better labour market conditions in urban areas are more influential than social welfare in
reducing poverty.
Of course, changes in income poverty may be driven by a wide range of factors besides the
labour market. A shift in household composition is one of them. Rural-urban migrants may
leave children behind with relatives in sending regions, which may improve per capita
incomes because they would now have fewer dependents, although income from child
support grants (which follow the child) may counter this effect. Alternatively, migrants might
attach themselves to new, better-off households, which would improve poverty levels,
7
The direction of trends over time are arguably more resilient than raw totals which may change
considerably depending on the precise definitions.
8
Although this finding is robust at a 95% of confidence, the sample size becomes very restrictive.

12
without any change in their employment status. The higher cost-of-living in urban areas (e.g.
for housing and transport) might also overstate the impact of any rise in incomes.
Some light can be shed on these factors by examining changes in labour market status over
the period. Most important, the unemployment rate among the cohort of rural-urban migrants
fell rapidly from 50% to 13% between 2008 and 2014. This was partly linked to a general
decline in unemployment experienced by all cohorts. Rural residents experienced a 9
percentage point decline and urban residents a 16 percentage point decline over the same
period. However, rural-urban migrants experienced a much larger 37 percentage point decline
in unemployment. This is quite dramatic.
Figure 4: Rate of unemployment by migration cohort, wave 1 – wave 4
Source: NIDS waves 1 - 4; own estimates
Notes: See table A4 in the appendix
The rate of unemployment for migrants was noticeably higher than for rural residents at the
start of the period. However, it was considerably lower by wave 4. This may be part of the
reason why poverty was slow to decline between waves 1 and 2 among migrants compared
with rural residents, as shown in figure 2. Extremely high unemployment at the outset may be
linked to the pressure to move to a city to find employment.
0.0
10.0
20.0
30.0
40.0
50.0
60.0
70.0
80.0
2008 2010/11 2012 2014/15
Percent (%)
Remained rural Rural-urban Remained urban

13
Figure 5 shows the change in employment status for each cohort between 2008 and 2014.
The horizontal axis shows their status in 2008, i.e. unemployed, employed or not
economically active. The breakdown of each column shows their status in 2014. The first set
of columns reveals that almost 80% of rural-urban migrants who were unemployed in wave 1
had a job by wave 4. This amounts to 175,000 jobs for the 225,000 migrants who were
looking for work in wave 1. The equivalent figure for unemployed rural residents was 40%
and for unemployed urban residents it was 50%. This is another very striking finding.
Figure 5: Employment transitions by migration cohort, wave 1 to wave 4
Source: NIDS waves 1 and 4; own estimates
Notes: See table A5 in the appendix
A similar pattern is apparent for people who were not economically active (the last set of
columns). Figure 5 shows that three-fifths (61%) of migrants who were inactive in 2008 had
39.5
77.5
49.5
63.3
85.8
76.8
30.8
60.8
40.3
24.6
9.4
21.2
10.8
4.9
6.1
20.6
17.6
12.2
35.9
13.1
29.4 25.9
9.4
17.1
48.7
21.6
47.4
0.0
10.0
20.0
30.0
40.0
50.0
60.0
70.0
80.0
90.0
100.0
Remained
rural
Rural-
urban
Remained
urban
Remained
rural
Rural-
urban
Remained
urban
Remained
rural
Rural-
urban
Remained
urban
W1: Unemployed W1: Employed W1: NEA
Percent (%)
W4: Employed W4: Unemployed W4: NEA

14
a job in 2014. The equivalent proportion for inactive rural residents was 31% and for inactive
urban residents it was 40%. The middle set of columns relates to those who had a job in
2008. It reveals that rural residents were less likely to remain in employment compared with
migrants and urban residents.
To summarise, rural-urban migrants were worse-off than rural residents in terms of their rate
of unemployment at the start of the period. However, they experienced much greater
improvement in their labour market status subsequently. Moving to a city seems to have
helped migrants to get well ahead of rural residents. By the end of the period their rate of
unemployment matched the lower rate of unemployment experienced among urban
residents.
9
An additional 300,000 people had moved into employment when migrating to an
urban area amongst the 700,000 migrants estimated from our panel.
Figure 6: Proportion of employees in precarious jobs, wave 1 – wave 4
Source: NIDS waves 1 - 4; own estimates
Notes: Precarious employees are those who are either self-employed (non-professional),
casually employed or have no written contract. See table A6 in the appendix.
9
Moving to a metro enabled better unemployment-to-employment transitions compared to moving to other
urban areas by roughly 10 percentage points. This finding was not robust at a 95% level of confidence due
to the small sample size.
0
10
20
30
40
50
60
70
2008 2010/11 2012 2014/15
Percent (%)
Remained rural Rural-urban Remained urban

15
The quality of people’s employment or livelihoods is just as important as the quantity. Figure
6 shows the percentage of the employed in each cohort who were in precarious livelihoods,
defined as self-employed (non-professional), casual employment or with no written contract.
Investigating job quality in the panel is rather tentative as the sample becomes limited to the
share of the employed (which is particularly constrained in wave 1 amongst migrants).
The percentage of employed in precarious jobs was consistently about 20% lower for urban
residents than for rural residents. This indicates that the rural-urban divide is apparent in the
quality of work and not merely in the amount of work. Rural-urban migrants straddled the
middle-ground between the extremes of rural and urban residents, with no consistent increase
or decrease in job quality over the period, although the margin of error is too wide to be
conclusive. A possible lack of improvement in the incidence of insecure employment among
migrants provides a qualification to the very positive message emerging from the other
findings. Nevertheless, the fact that there may have been no substantial increase in precarious
jobs suggests that migrants were being absorbed into a mixture of secure and insecure forms
of work. It is not as if their stepping stone to the urban labour market comprised of only
precarious work.
The living conditions of rural-urban migrants are also important in forming judgements about
the extent of progression associated with moving to cities. Put simply, were migrants forced
to sacrifice their quality of life in order to gain access to the economic opportunities available
in cities? In particular, were they obliged to occupy shacks rather than durable houses
because of affordability constraints? Decent housing is important for protection from the
elements, health, well-being, privacy and security (Turok, 2016). Figure 7 shows the
proportion of individuals living in informal settlements or in shacks elsewhere (either in
backyards or freestanding). The proportion of regular urban residents living in shacks was
broadly unchanged between 2008 and 2014, at nearly a quarter (23%). Rural-urban migrants
were more likely to occupy shacks - more than a third of them (a total of 255,000 people) had
migrated into informal settlements or shacks by the end of wave 4, indicating some apparent
deterioration in their living conditions over time as more of them moved to cities. Compared
with regular urban residents, they were most likely to settle in backyard or free-standing
shacks. Nevertheless, this should not obscure the fact that nearly two-thirds of migrants were
living in durable structures rather than shacks.

16
Figure 7: Percentage of people living in shacks, wave 1 – wave 4
Source: NIDS waves 1 - 4; own estimates
Notes: See table A7 in the appendix
5.2 7.5
13.2
17.6
5.6
8.4
5.9
18.5
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.0
2008 2010/11 2012 2014/15
Percent (%)
Rural-urban
Informal settlement Shack dwelling elsewhere
16.1 14.9 14.8 14.5
7.7 9.4 8.8 8.5
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.0
2008 2010/11 2012 2014/15
Percent (%)
Remained urban
Informal settlement Shack dwelling elsewhere

17
Figure 8: Access to basic services in the dwelling by cohort, wave 1 – wave 4
0
10
20
30
40
50
60
70
80
2008 2010/11 2012 2014/15
Percent (%)
Piped water
Remained rural Rural-urban Remained urban
0
10
20
30
40
50
60
70
80
2008 2010/11 2012 2014/15
Percent (%)
Flushing toilet
Remained rural Rural-urban Remained urban
40
50
60
70
80
90
100
2008 2010/11 2012 2014/15
Percent (%)
Electricity
Remained rural Rural-urban Remained urban

18
Access to basic services is another important dimension of the living environment. Having
piped water, a flush toilet and electricity within the dwelling saves people time and makes a
big difference to their welfare, health, safety and dignity. Figure 8 shows the percentage of
people in each cohort with access to these essential services. It reveals that regular urban
residents have much better access to services than rural residents, particularly piped water
and flush toilets. The position of rural-urban migrants is somewhere in between the other two
cohorts. All three cohorts experienced some improvement in access to services over the
period 2008 to 2014. Rural-urban migrants were distinctive in consistently experiencing the
biggest improvements. By 2014, nearly a third of the cohort had gained access to piped
water, a flush toilet and electricity that they did not have in 2008. This represents a major
improvement in living conditions for this cohort of people.
5. Gross flows of forwards and backwards migration
The weight of the descriptive evidence suggests a close relationship between migration and
socio-economic progress. This has been evaluated from the perspective of a one-directional
flow of migrants from rural areas to take up residence in towns or cities. It is very well
known in the migration literature that there are gross and net migration flows between many
places (including between cities) and in both directions, driven by all sorts of motivations and
individual circumstances (Maritz & Kok 2014). Most of these flows occur between the
metropolitan municipalities or larger secondary cities (Pieterse 2015, Ginsburg et al. 2016).
Circular migration remains a distinctive feature of South African urbanisation where migrants
maintain social and cultural attachments with their respective rural-sending communities
(Bekker 2001, Posel & Marx 2013). There is some evidence of sizable backwards flows of
migrants from urban to rural areas (Kok & Collinson 2006, Ginsburg et al. 2016).
The sample size of the NIDS panel limits the extent to which detailed patterns of migration
can be feasibly measured. However it is possible to make some tentative observations about
the gross flows between rural and urban regions by examining aggregated migration
transitions as they occur between adjacent waves. Table 2 (and figure 9) present these
patterns.

19
Table 2: Patterns of migration between waves
W1-W2
W2-W3
W3-W4
Total
Rural-urban
Number
183,000
[132,000; 234,000]
405,000
[314,000; 497;000]
359,000
[287,000; 432,000]
948,000
[808,000; 1087,000]
Percentage
19.3%
42.7%
37.9%
100%
obs
124
221
220
565
Urban-rural
Count
243,000
[142,000; 345,000]
138,000
[67,000; 209,000]
287,000
[209,000; 364,000]
668,000
[513,000, 823,000]
Percentage
36.4%
20.7%
43.0%
100%
obs
56
52
125
233
Source: NIDS waves 1 - 4; own estimates
Notes: 95% confidence intervals in parenthesis
Figure 9: Gross migration flows in each wave
Source: NIDS waves 1 - 4; own estimates
The data suggests that there was a significant level of backwards migration in each wave,
although the sample sizes become very small. This is particularly prevalent between waves
2008 and 2010, where the level of out-migration from urban areas matched the extent of in-
migration. This could be a result of the 2008 global financial crisis in reducing job
opportunities in the cities and restricting resources available to rural households to send or
sustain migrants in search of employment. Fluidity between rural and urban areas warrants
further exploration and may have implications for how urbanisation shapes social outcomes
over the life-cycle of migrants and their respective sending and receiving communities.
183
405 359
243
138 287
0
100
200
300
400
500
600
700
W1-W2 W2-W3 W3-W4
Thousands
Rural-urban Urban-rural

20
The degree of rural-urban migration was much larger in latter waves. In total, there were
roughly 1 million gross migrations amongst the working population (aged 19 – 64 years in all
waves) into towns or cities from the countryside and possibly more than half of this number
moving the other way (the margin of error for these estimates is wide). This does not include
instances of multiple migration that could have occurred between any two adjacent waves in
the panel, or other geographical movements such as rural-rural and urban-urban migration.
The objective is not mainly to identify total numbers but rather to get a sense of gross flows
between urban and rural areas.
What do these gross migration flows imply about changes in social outcomes? Aggregating
rural-urban and urban-rural transitions into ‘before migration’ and ‘after migration’
categories allows for an assessment of changes in the levels of poverty and unemployment
before and after the event of migration. Every rural-to-urban migration event is included in a
gross rural-urban cohort irrespective of movements of the individual in latter waves and vice-
versa.
10
This has the useful effect of bolstering the sample size.
Unfortunately, a straightforward comparison between forwards and backwards migration
cohorts is problematic as changes in social outcomes between waves are affected by time-
specific trends which become conflated when using this approach. Of concern is that the
overall rate of poverty or unemployment declined more substantially in waves 3 and 4
compared to between waves 1 and 2 (see figures 2 and 4). Considering that the urban-rural
cohort includes a proportionately greater number of migration events between waves 1 and 2
this may negatively bias a descriptive comparison of changes in poverty or the probability of
finding work between the two cohorts. We include a variation of the urban-rural cohort which
excludes wave 1 to 2 transitions to assess how the comparisons might change.
10
For instance, an individual who is urban in wave 1, rural in wave 2, and urban in waves 3 and 4 would
be classified as an urban-rural migrant between waves 1 and 2 and at the same time as a rural-urban
migrant between waves 2 and 3.

21
Figure 10: Gross migration flows and the poverty rate
Source: NIDS waves 1 - 4; own estimates
Notes: See table A9 in the appendix
Figure 11: Gross migration flows and the rate of unemployment
Source: NIDS waves 1 - 4; own estimates
Notes: See table A10 in the appendix
70.6
46.0 43.7
37.1
48.3 48.6
0.0
10.0
20.0
30.0
40.0
50.0
60.0
70.0
80.0
90.0
100.0
Gross
rural-urban
Gross
urban-rural
Urban-rural:
excl W1/W2
Percent (%)
Before migration After migration
38.4 35.4
28.8
22.7
31.3
25.9
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.0
45.0
50.0
Gross
rural-urban
Gross
urban-rural
Urban-rural:
excl W1/W2
Percent (%)
Before migration After migration

22
Figures 10 and 11 show the relationship between backwards and forwards migration and the
rate of poverty and unemployment. It is comforting to see that the rate of poverty and
unemployment fell significantly between rural to urban migrations in line with our previous
findings. The evidence is less conclusive for urban to rural migrations. Levels of poverty or
unemployment did not change significantly, in part because of the small size of the sample,
but possibly also because the potential effects of backwards migration are less dramatic. The
static level of poverty amongst the urban-rural cohort was considerably lower in comparison
to the four-wave average of those who remained rural (refer back to figure 2) which suggests
that those moving to rural areas were better-off than rural communities. The conclusions
remain unchanged if we exclude waves 1 and 2 transitions for the urban-rural cohort.
The margin for error is wide but in summary it would appear that gross rural-urban migration
flows are positively associated with falling levels of poverty and unemployment, whereas
gross migration flows back to rural areas are not related to a change in individual fortunes,
although poverty-levels are typically lower amongst urban-rural migrants, some of whom
would be returning to their sending communities.
6. Summary and conclusions
South Africa is a country of apparently entrenched divides. Social and spatial inequalities are
very wide and upward social mobility is low. This is a source of economic inefficiency as
well as considerable social frustration and instability. The last decade has also been a period
of anaemic national economic performance and rising unemployment. There is a sense that
the country is trapped in a socially-divisive, low investment, low growth equilibrium that
inhibits shared prosperity and inclusive development. The shortage of opportunities to lift
people out of poverty in a way that is sustainable is greatest in rural areas. In this context, any
signs of socio-economic dynamism and upward mobility must be regarded as extremely
important.
Many rural residents have responded to the lack of jobs in the countryside by migrating
towards the cities in the hope of securing a better future. The chapter has analysed the extent
of economic and social progress achieved by a sample of adult migrants over the period
2008-2014. Despite the multiple barriers migrants face in trying to enter urban labour
markets, the findings show that many succeeded in getting jobs and increasing their incomes.

23
Four-fifths of them were living below the poverty line in 2008 compared with only one-third
in 2014. A total of 385,000 migrants managed to escape from poverty over the period. Many
rural-urban migrants also improved their living conditions in terms of access to basic
services. The cohort of migrants improved their economic and social circumstances much
more quickly than the remaining rural residents or urban residents. There is also tentative
evidence of a fair amount of fluidity between rural and urban areas with a fair number of
relatively better-off urban residents moving to rural communities.
Despite the compelling descriptive evidence, the association between mobility and human
progress is the outcome of many different factors and forces that determine whether
individual decisions to resettle in urban areas result in success or failure. Some of the
progress achieved may be explained by individual factors, such as the superior education of
migrants, their greater drive and determination, or their innate ability to succeed in difficult
circumstances, rather than the broader differences in economic and social conditions between
urban and rural areas. The NIDS panel is limited by the size of the sample as well as missing
some amount of short-term migration between any of the waves. Notwithstanding these
concerns, the weight of the association between migration and social mobility is compelling,
irrespective of the causal mechanism through which it may operate.
The results have potent implications for government policy towards migration and towards
cities. For many years national policy has been ambivalent about rural-urban migration,
partly because of social dislocation in the sending regions and pressures on land, housing and
public services in the cities. The history of the forced migrant labour system and the legacy of
rural neglect also continue to influence ruling party thinking. The most substantial policy
response to rural poverty has been to redistribute public resources in the form of social grants
and free healthcare, schools, housing and basic services. However, this has not addressed the
underlying employment problem.
Government policy should recognise the achievements of people’s spontaneous efforts to get
out of poverty by moving towards the cities. It should endorse the Constitutional right to
freedom of movement and do more to support urbanisation by accommodating the growing
urban population in reasonable living conditions. This would alleviate many of the social
pressures and environmental problems in overcrowded informal settlements and backyard
shacks. By working together, planning ahead and investing in essential infrastructure the

24
different spheres of government could also start reshaping the fragmented urban form in
order to create more efficient, inclusive and liveable cities.
References:
Adepoju A (1988) International migration in Africa south of the Sahara. In R Appleyard
(Ed.) International migration today: Trends and prospects (Vol 1). Perth: University of
Western Australia
Bekker S (2001) Diminishing returns: Circulatory migration linking Cape Town to the
Eastern Cape. Southern African Journal of Demography 8(1): 1–8
Budlender J, Leibbrandt M, & Woolard I (2015) South African poverty lines: a review and
two new money-metric thresholds. Southern Africa Labour and Development Research Unit
Working Paper No. 2015/151, University of Cape Town
Centre for Development and Enterprise (CDE) (2017) Opportunities first. Johannesburg:
CDE
Collier P & Venables AJ (2017) Urbanization in developing economies: the assessment.
Oxford Review of Economic Policy 33(3): 355–372
Ginsburg C, Collinson MA, Iturralde D, van Tonder L, Gomez-Olive FX, Kahn K & Tollman
S (2016) Migration and settlement change in South Africa. Southern African Journal of
Demography 17: 133–198
Glaeser E (2011) Triumph of the city: How our greatest invention makes us richer, smarter,
greener, healthier, and happier. London: Penguin
Glaeser E & Sims H (2015) Contagion, crime and congestion: Overcoming the downsides of
density. Growth Brief Series May 2015, International Growth Commission
Kerr A (2017) Tax(i)ing the poor? Commuting costs in South African cities. South African
Journal of Economics 85(3): 321–340

25
Keswell M, Girdwood S & Leibbrandt M (2013) Educational inheritance and the distribution
of occupations: Evidence from South Africa. The Review of Income and Wealth, 59: 111–137
Mulcahy K & Kollamparambil U (2016) The impact of rural-urban migration on subjective
well-being in South Africa. The Journal of Development Studies 52(9): 1357-1371
Kok P & Collinson M (2006) Migration and urbanisation in South Africa. Statistics South
Africa Report 03-04-02, Statistics South Africa
Maritz J & Kok P (2014) Using election registration data to measure migration trends in
South Africa. Town and Regional Planning 64: 31–42
Pendleton W, Crush E, Campbell T, Green H, Simelane D, Tevera & de Vletter F (2006)
Migration, remittances and development in Southern Africa. SAMP Migration Policy Series
No. 44, Cape Town: Idasa
Piraino P (2015) Intergenerational earnings mobility and equality of opportunity in South
Africa. World Development 67: 396-405
Pieterse A (2015) Population flows between cities. Centre for Scientific and Industrial
Research. Accessed September 2017, http://www.stepsa.org
Posel D and Casale D (2006) Internal labour migration and household poverty in post-
Apartheid South Africa. In Bhorat H & Kanbur K (Eds) Poverty and policy in post-Apartheid
South Africa. Cape Town: Human Sciences Research Council Press
Rogan M, Lebani L, & Nzimande M (2009) Internal migration and poverty in KwaZulu-
Natal: Findings from censuses, labour force surveys and panel data. Southern Africa Labour
and Development Research Unit Working Paper No. 30, University of Cape Town
Storper M (2013) Keys to the city: how economics, institutions, social interaction, and
politics shape development. Princeton: Princeton University Press

26
Turok I (2014) South Africa’s tortured urbanisation and the complications of reconstruction.
In Martine G & McGranahan G (Eds) Urban growth in emerging economies: Lessons from
the BRICS. London: Routledge
Turok I (2016) Housing and the Urban Premium. Habitat International 54: 234–240
UN-Habitat (2016) Urbanisation and Development: Emerging Futures. World Cities Report
2016, Nairobi: UN-Habitat
World Bank (2009) World development report 2009: Reshaping economic geography.
Washington, DC: World Bank

27
Appendix
Table A1: Percentage and number of migrations in each wave
% of migrant
cohort total
Cumulative % of
migrant cohort total
No. of migrations
(000’s)
Cumulative no. of
migrations (000’s)
W1 - W2
16.45
16.45
116
116
[12.25, 21.74]
[12.25, 21.74]
[80, 152]
[80, 152]
W2 - W3
39.07
55.52
275
390
[32.13, 46.48]
[48.82, 62.01]
[197, 352]
[303, 478]
W3 - W4
44.48
100
313
703
[37.99, 51.18]
-
[248, 377]
[583, 823]
Source: NIDS waves 1 - 4; own estimates
Table A2: Percentage below the poverty line by migration category, w1 – w4
Remained rural
Rural-urban
Remained urban
wave 1
81.78
82.81
53.61
[78.13,84.93]
[77.09,87.34]
[47.52,59.6]
wave 2
79.44
79.09
52.61
[75.81,82.64]
[73.42,83.82]
[46.84,58.31]
wave 3
75.43
50.21
46.63
[71.84,78.69]
[43.73,56.68]
[41.21,52.13]
wave 4
69.41
32.98
37.09
[65.58,72.99]
[26.68,39.95]
[32.57,41.86]
Source: NIDS waves 1 - 4; own estimates
Table A3.1: Poverty transition matrix for remained rural, w1 to w4
wave 4
wave 1
Non-poor
Poor
Total
Non-poor
72.31
27.69
100
[65.36,78.33]
[21.67,34.64]
Poor
21.30
78.70
100
[18.47,24.43]
[75.57,81.53]
Total
30.59
69.41
100
[26.96,34.48]
[65.52,73.04]
Source: NIDS waves 1 & 4; own estimates

28
Table A3.2: Poverty transition matrix for rural-urban, w1 to w4
wave 4
wave 1
Non-poor
Poor
Total
Non-poor
69.75
30.25
100
[48.81,84.8]
[15.2,51.19]
Poor
66.45
33.55
100
[59.93,72.4]
[27.6,40.07]
Total
67.02
32.98
100
[60.05,73.31]
[26.69,39.95]
Source: NIDS waves 1 & 4; own estimates
Table A3.3: Poverty transition matrix for remained urban, w1 to w4
wave 4
wave 1
Non-poor
Poor
Total
Non-poor
85.9
14.1
100
[81.83,89.17]
[10.83,18.17]
Poor
43.01
56.99
100
[38.67,47.47]
[52.53,61.33]
Total
62.91
37.09
100
[58.14,67.43]
[32.57,41.86]
Source: NIDS waves 1 & 4; own estimates
Table A4: Rate of unemployment by migration category, w1 – w4
Remained rural
Rural-urban
Remained urban
wave 1
35.86
50.37
30.28
[32.54,39.32]
[42.22,58.5]
[26.92,33.86]
wave 2
36.18
42.36
21.73
[32.36,40.18]
[34.46,50.68]
[18.42,25.45]
wave 3
35.07
28.99
23.12
[31.62,38.68]
[23.01,35.8]
[20.08,26.47]
wave 4
27.10
13.18
14.63
[24.32,30.07]
[9.7,17.66]
[12.39,17.19]
Source: NIDS waves 1 - 4; own estimates

29
Table A5.1: Labour market transition matrix for remained rural, w1 to w4
Wave 4
Wave 1
NEA
Unemployed
Employed
Total
NEA
48.67
20.57
30.76
100
[43.69,53.68]
[17.63,23.86]
[26.27,35.64]
Unemployed
35.91
24.59
39.5
100
[31.13,40.97]
[20.96,28.63]
[35.24,43.93]
Employed
25.89
10.8
63.3
100
[22.56,29.53]
[8.624,13.46]
[58.85,67.54]
Total
36.01
17.39
46.6
100
[33.11,39.03]
[15.63,19.3]
[43.46,49.76]
Source: NIDS waves 1 & 4; own estimates
Table A5.2: Labour market transition matrix for rural-urban, w1 to w4
Wave 4
Wave 1
NEA
Unemployed
Employed
Total
NEA
21.61
17.62
60.78
100
[14.31,31.27]
[10.74,27.54]
[50.08,70.53]
Unemployed
13.1
9.403
77.5
100
[8.289,20.08]
[5.404,15.86]
[68.12,84.74]
Employed
9.392
4.855
85.75
100
[4.897,17.26]
[2.062,11.01]
[77.16,91.47]
Total
14.95
10.88
74.17
100
[11.04,19.93]
[7.933,14.75]
[68.54,79.1]
Source: NIDS waves 1 & 4; own estimates
Table A5.3: Labour market transition matrix for remained urban, w1 to w4
Wave 4
Wave 1
NEA
Unemployed
Employed
Total
NEA
47.43
12.23
40.33
100
[40.72,54.24]
[9.093,16.26]
[33.93,47.08]
Unemployed
29.36
21.17
49.47
100
[24.71,34.48]
[16.81,26.3]
[44.84,54.1]
Employed
17.05
6.122
76.83
100
[14.64,19.77]
[4.846,7.706]
[73.79,79.6]
Total
25.43
10.95
63.62
100
[22.95,28.09]
[9.276,12.88]
[60.67,66.46]
Source: NIDS waves 1 & 4; own estimates

30
Table A6: Percentage of employees who are either self-employed (non-professional), casually
employed or have no written contract by migration category, w1 – w4
Remained rural
Rural-urban
Remained urban
wave 1
61.03
43.62
42.93
[56.35,65.52]
[32.54,55.38]
[38.4,47.58]
wave 2
55.26
41.25
34.81
[50.53,59.9]
[31.57,51.66]
[30.49,39.4]
wave 3
49.03
47.47
30.27
[44.54,53.52]
[38.89,56.2]
[26.01,34.91]
wave 4
51.13
41.22
33.7
[46.52,55.72]
[33.94,48.92]
[29.86,37.76]
Source: NIDS waves 1 - 4; own estimates
Table A7: Percentage living in shacks and shack settlements by migration category, w1 – w4
wave 1
wave 2
wave 3
wave 4
Rural-urban
Rural-urban
Informal settlement
5.19
7.48
13.17
17.60
[1.769,14.26]
[3.42,15.59]
[9.00,18.87]
[13.17,23.12]
Shack dwelling elsewhere
5.56
8.40
5.85
18.53
[2.41,12.34]
[4.68,14.63]
[3.82,8.86]
[13.68,24.61]
Remained
urban
Remained
urban
Informal settlement
16.09
14.88
14.75
14.50
[9.56,25.81]
[8.87,23.89]
[8.90,23.45]
[8.72,23.12]
Shack dwelling elsewhere
7.67
9.43
8.84
8.47
[5.29,11.01]
[6.86,12.84]
[6.29,12.30]
[6.26,11.37]
Source: NIDS waves 1 - 4; own estimates
Table A8.1: Percentage access to piped water in the dwelling by migration category, w1 – w4
Remained rural
Rural-urban
Remained urban
wave 1
13.81
20.65
57.76
[10.24,18.37]
[14.59,28.39]
[51.53,63.76]
wave 2
24.43
32.16
69.32
[20.34,29.04]
[25.4,39.77]
[63.49,74.59]
wave 3
14.78
35.86
63.28
[11.34,19.03]
[30.37,41.75]
[57.18,68.99]
wave 4
13.71
46.06
67.19
[10.77,17.31]
[39.72,52.52]
[61.65,72.29]
Source: NIDS waves 1 - 4; own estimates

31
Table A8.2: Percentage access to flushing toilet in the dwelling by migration category, w1 – w4
Remained rural
Rural-urban
Remained urban
wave 1
4.51
9.84
40.41
[2.75,7.323]
[5.19,17.87]
[33.69,47.52]
wave 2
9.08
18.08
60.75
[6.57,12.43]
[12.82,24.89]
[55.02,66.2]
wave 3
8.31
31.55
47.20
[5.662,12.03]
[24.99,38.95]
[41.52,52.96]
wave 4
7.20
41.21
47.69
[5.153,9.984]
[34.86,47.87]
[42.37,53.06]
Source: NIDS waves 1 - 4; own estimates
Table A8.3: Percentage access to electricity by migration category, w1 – w4
Remained rural
Rural-urban
Remained urban
wave 1
64.58
62.82
84.44
[56.89,71.59]
[54.16,70.74]
[78.43,89.01]
wave 2
66.51
68.99
86.27
[59.9,72.54]
[61.29,75.76]
[80.88,90.33]
wave 3
77.31
84.65
90.58
[70.54,82.9]
[79.04,88.96]
[86.22,93.65]
wave 4
80.05
88.47
93.01
[74.49,84.65]
[84.2,91.71]
[90.19,95.06]
Source: NIDS waves 1 - 4; own estimates
Table A9: Gross migration flows and the rate of poverty
Before migration
After migration
Gross rural-urban
70.6
37.1
[64.27, 76.21]
[31.77, 42.72]
Gross urban-rural
46.0
48.3
[35.38, 57.08]
[37.75, 58.91]
Urban-rural: excl W1/W2
43.7
48.6
[32.45, 55.68]
[36.94, 60.39]
Source: NIDS waves 1 - 4; own estimates

32
Table A10: Gross migration flows and the rate of unemployment
Before migration
After migration
Gross rural-urban
38.4
22.7
[31.34, 45.98]
[18.57, 27.52]
Gross urban-rural
35.4
31.3
[24.74, 47.76]
[22.59, 41.51]
Urban-rural: excl W1/W2
28.8
25.9
[18.07, 42.63]
[16.51, 38.29]
Source: NIDS waves 1 - 4; own estimates