The information on which the analysis is based has been taken from the 2011 Census and the 2016 Community Survey. The data as published are not ideally suited to the analysis of migration. Adjustments have to be made to both sources in order to put them into a coherent demographic framework, anchored to United Nations estimates and projections of the South African population. Once these adjustments are made, and two different approaches to the estimation of migration are applied, it becomes clear that the recording of moves within South Africa in the Community Survey is incomplete. While the two methods reveal approximately the same patterns of migration flows, there is considerable uncertainty about their size.
The main conclusions from the studies are as follows:
The apartheid pattern of movement was away from commercial farming (rural areas) towards tribal areas and encouragement of oscillating migration between tribal areas and urban areas. Has this pattern changed two decades after the end of apartheid?
An analysis of internal migration requires geographical division of the country. The grid used is the set of metros and local municipalities as they existed in 2011.
As indicated in the first brief, data from 2011 Census and the 2016 Community Survey have been adjusted to a five year interval, from mid-2011 to mid-2016 and have been anchored to the United Nations population estimates and projections of the South African population (2015 revision).
It distinguishes between three categories of movement:
Movers - Residents at the time of moving who have moved within the boundaries of metros and local municipalities, as they were constituted at the time of the 2011 census.
Migrants - Residents at the time of moving who have moved across the boundaries of metros and local municipalities.
Immigrants/Emigrants - People who have entered South Africa from another country, or left South Africa for another country. Short term visitors are not counted.
The first brief dealt with immigrants and emigrants. The point of distinguishing between migrants and movers in this brief is to bracket out movers who have established themselves within a local municipality and who make short distance moves within it, from migrants who, on average, make longer distance moves. The operational definition captures the conceptual distinction imperfectly, since people living close to the edge of one municipality may move a short distance into a neighbouring municipality.
A couple of conventions, generally used in the literature on migration, need to be mentioned. People who move from one municipality to another within the country are referred to as outmigrants and inmigrants, whereas people who move across the country’s borders are described as emigrants and immigrants. Rates of migration are expressed as moves per thousand of the relevant population. Thus, if the outmigration rate from a municipality is reported as 50, this means that 50 of every 1 000 residents left the municipality in the five year period. An inmigration rate of 30 means that for every 1 000 residents in the municipality, 30 people arrived in the municipality in the five year period.
There are two approaches to measuring migration. The first is direct measurement where respondents are asked about where they have come from and where their usual place of residence is. Within this framework, there are two approaches to capturing movement. The first is to compare the place of residence five years ago with the current place of residence, and to record a move if the places are different. The second is to ask for the date and former place of the most recent move within the interval. The community survey used the second approach, which means that earlier moves among people who moved twice or more in the interval are not captured. Measured inmigration and outmigration are under-estimates of actual movement over five years, probably of the order of 10%. Direct measurement enables one to identify both inflows and outflows into municipalities. The difference between the two represent net inmigration, which can be positive or negative.
The second approach is indirect measurement. This uses the population equation:
Population in 2016 = Population in 2011 + Births between 2011 and 2016 – Deaths between 2011 and 2016 + Net inmigration between 2011 and 2016 + Net immigration between 2011 and 2016
If one knows all the other magnitudes in the equation, net inmigration between 2011 and 2016 can be calculated. The indirect method does not allow one to calculate inflows and outflows separately.
The results of the analysis will be presented as a series of maps in the atlas which is at the end of this document.
Map 1 indicates net migration as measured by the direct method. It indicates substantial net movement into the three Gauteng metros and Cape Town, but not into the other four metros. There was also movement of more than 5% of the 2011 population into a couple of Limpopo municipalities, one Northern Cape municipality and one Western Cape municipality.
Map 1 indicates substantial net movement out of parts of Limpopo and Mpumalanga and one Gauteng municipality and movement of more than 5% of the 2011 population out of a dozen municipalities scattered across Northern Cape, Eastern Cape, KwaZulu-Natal, Free State and Limpopo.
Map 2 indicates net migration as measured by the indirect method. Map 2 indicates much more substantial net inflows and outflows than Map 1. In Map 2, there are inflows into Nelson Mandela Bay and Buffalo City as well as the three Gauteng metros and Ethekwini. Cape Town now registers net outmigration, and there is an outflow from Mangaung as well. There is net inmigration to municipalities around Cape Town, movement within the Northern Cape and substantial outmigration from many municipalities in Eastern Cape, Mpumalanga, Limpopo, North West and some in the Free State and KwaZulu-Natal.
The contrast between Maps 1 and 2 indicates that direct movement in the 2016 Community Survey has been substantially under-recorded. On the other hand, the indirect measurement of net movement is not perfect. Births and deaths are recorded in the municipalities in which they occur, and not in the municipalities of usual residence, and the pattern of fertility measured in the year before the Community Survey is assumed to hold for the entire five year period. The result may be some spurious measured movement. The correlation coefficient of the two net migration estimates across municipalities is 0.58, significantly different from zero at the 1% level, suggesting that the two measures reveal, very approximately, the same underlying pattern across municipalities . The issue is the extent of net inmigration.
Map 3, based on reported moves across municipal boundaries divides municipalities into five categories:
Other municipalities, divided into:
Areas with high outmigration relative to inmigration are found mainly in the Eastern Cape, KwaZulu-Natal, Mpumalanga, Limpopo and, to a lesser extent in the Northern Cape and Free State. Areas with high inmigration relative to outmigration run in a belt from western Limpopo through the Gauteng metros and extending to the east, and in the western and northern Cape. Interspersed with the northern belt are municipalities with moderate and high churn. There is also a belt of mostly moderate churn extending from the Ciskei across the Northern Cape. Unlike Maps 1 and 2, Map 3 contains no indication of absolute levels of migration, but merely indicates patterns of movement.
Map 4 considers the change in population shares. Municipalities are placed in one of eight categories:
Map 4 shows two important things. The first is there are more municipalities containing tribal areas where the share of the population in tribal areas has increased than municipalities where the share of population in tribal areas has decreased. The second is that in other municipalities, there are quite a few, notably in the Western Cape, where the share of population in urban areas has increased by at least 10%. These points will be developed below.
Up to this point, no account of has been taken of movement within municipalities. Of an estimated number of moves of 3.75 million captured directly, 54.5% were moves within municipalities and 45.5% were moves between municipalities. Map 5 divides municipalities into categories by directly measured moves as a percentage of the 2011 population. Unweighted medians by category are presented in Table 1.
Table 1 - Movements within municipalities |
||
Median percentage of 2011 population |
||
Predominantly outmigration |
1,27% |
|
Predominantly inmigration |
4,48% |
|
Low churn |
2,75% |
|
Medium churn |
3,17% |
|
High churn |
3,83% |
The pattern in Table 1 makes sense, although moves probably have been under-recorded . Municipalities which have seen much more outmigration than inmigration are stagnant or declining with little incentive to move within them. Municipalities which have seen much more inmigration than outmigration are dynamic and can be expected to have lots of internal movement. And municipalities with a greater balance between inmigration and outmigration have movement which rises with churn.
Map 6 presents the distribution of the foreign born in South Africa in 2016, showing a marked concentration in the northern parts of the country and the metros.
Indirect estimation leads to the population balance by geotype shown in Table 2. It presents indirect estimates of net inmigration by geotype, using the4 population equation.
Table 2 |
|||||||
Analysis by geotype |
|||||||
2011 |
Births |
Deaths |
Net |
2016 |
Net |
||
Population |
immigration |
Expected |
Actual |
inmigration |
|||
Metro |
20 349 245 |
1 892 256 |
1 040 717 |
409 209 |
21 609 994 |
21 613 766 |
3 772 |
Urban |
13 402 604 |
1 477 374 |
900 293 |
-159 617 |
13 820 068 |
14 264 711 |
444 643 |
Tribal |
15 901 884 |
1 996 958 |
1 279 208 |
-226 796 |
16 392 839 |
16 662 532 |
269 693 |
Rural |
2 579 267 |
211 211 |
128 582 |
-22 797 |
2 639 099 |
1 920 991 |
-718 108 |
Total |
52 233 000 |
5 577 800 |
3 348 800 |
0 |
54 462 000 |
54 462 000 |
0 |
Table 2 leads to three surprising conclusions:
Are these results plausible? In the technical report accompanying the Community Survey results, Statistics South Africa states that the sample design weights were adjusted to fit a demographic model at the national, provincial and municipal levels. To the extent to which this unpublished model is inaccurate, some distortions may have had implications for geotype totals. And some of the difficulties with indirect estimation have been noted above. So the magnitudes in Table 2 are suspect, even though net movement of the commercial farms almost certainly fuels net immigration into both towns and tribal areas.
The implications of Table 2 are that 64.6% of the population lived in metropolitan and urban areas in 2011, rising to 65.9% in 2016.
International comparisons of internal migration rates are hampered by the fact that measured migration depends on the coarseness or fineness of the grid used to measure it, and on the period over which migration is measured. The finer the grid, the more moves will take place across it. Fortunately, there is a formula which enables one to standardise other results to the grid of the 234 municipalities in this study.
Bell and Charles-Edwards found that New Zealand, Australia, Canada and the United States of America, all developed new world countries, exhibit the highest values of migration intensity. Asian countries, with the exception of Malaysia, exhibit low values. The picture is more varied in Africa, Latin America and Europe. The range is great as Table 3 indicates.
What is the CMI for South African between 2011 and 2016? The CMI based on number of directly counted moves between municipalities was 3.3 between 2011 and 2016. Adding 10% for moves which may have taken place between 2011 and the date of last move, the CMI rises to 3.6. On the other hand, comparison of net migration measured by the indirect and direct method suggests that use of indirectly estimated moves would increase the number of moves by a factor of 2.9, bringing the CMI to 10.4. Even this figure is lower than the estimate Bell and Charles-Edwards derive from the 2011 Census (13.2 for a grid of 52 metros and district municipalities).
Table 3 - Crude migration intensity |
|||
Selected countries |
|||
Country |
CMI |
Number of |
CMI standardised |
regions |
to 234 regions |
||
South Africa (low) |
3,6 |
234 |
3,6 |
Indonesia |
3,9 |
280 |
3,8 |
Brazil |
3,4 |
27 |
5,6 |
China |
6,7 |
347 |
6,2 |
Argentina |
7,2 |
511 |
6,3 |
Portugal |
7,1 |
308 |
6,8 |
Ghana |
6,0 |
110 |
7,0 |
South Africa (high) |
10,4 |
234 |
10,4 |
Canada |
12,5 |
288 |
12,0 |
USA |
8,9 |
51 |
12,3 |
Chile |
16,7 |
178 |
17,6 |
Australia |
16,5 |
69 |
21,3 |
|
|
|
|
Source: Bell and Charles-Edwards, Table 5. All the data (except for South Africa) are taken from the 2000 round of censuses
It is unlikely that the standardised CMI for South Africa would be as low as directly observed moves would imply, so the international evidence suggests that directly recorded moves have been under-counted. On the other hand, we are less mobile than people in developed new world countries.
Over the last hundred years, South Africa has kept up fairly well with international developments in the development of statistical series. But the quality of enumeration has always been lower than in developed countries and this shows up once again in this study. Some broad patterns are apparent, but they are not properly in focus, especially when it comes to the magnitudes of migration flows.
In many ways, the massive removals under apartheid have been stabilised. The proportion of people living in tribal areas remains over 30%, with no discernible movement. And people are moving off the commercial farms in net terms, putting pressure on urban areas outside the metros. The proportion of people living in metros has hardly increased, and would not have increased at all, had it not been for immigration from other countries. On the other hand, oscillating migration affects not only municipalities containing tribal areas. There are two additional belts of moderate or high churn areas – one surrounding the Gauteng metros and another stretching from the Ciskei across the Northern Cape.
If internal migration is not rapidly changing the geographical distribution of the population as affected by apartheid policies, the pattern of migration to and from Zimbabwe, Mozambique and Lesotho –though they would hate to hear it - bears a remarkable resemblance between movements in and out of the Bantustans under apartheid. Movement in when young, movement out when older and less strong, and frequent imprisonment and deportation in between.
Charles Simkins
Head of Research
charles@hsf.org.za