Post-Apartheid Patterns of. Internal Migration. in South Africa. Free download from

Transcription

1 Post-Apartheid Patterns of Internal Migration in South Africa

2

3 Post-Apartheid Patterns of Internal Migration in South Africa PIETER KOK, MICHAEL O DONOVAN, OUMAR BOUARE AND JOHAN VAN ZYL

4 Compiled by the Integrated Rural and Regional Development Research Programme, Human Sciences Research Council Executive Director: Mike de Klerk Published by HSRC Publishers Private Bag X9182, Cape Town, 8000, South Africa Human Sciences Research Council First published 2003 All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. ISBN Cover photograph by David Lurie, first published in his book Life in the Liberated Zone, published in Production by compress Printed by Print24.com Distributed in South Africa by Blue Weaver Marketing and Distribution P.O. Box 30370, Tokai, Cape Town, South Africa, Tel/Fax: (021) ,

5 Contents List of tables... List of maps... List of figures... List of graphs... About the authors... Preface... Overview... vii ix xi xi xii xiv xvii 1. Introduction... 1 Migration data generated by Census Context and scope of the study... 2 Purpose of the book... 4 Outline of the book Literature review... 8 Current status of migration research... 8 Definitions... 8 Data adequacy, reliability and appropriateness Theories and models of the causes of migration Economic factors that cause migration Economic factors that perpetuate migration Non-economic factors that cause migration Non-economic factors that perpetuate migration Evaluation Modified gravity model of migration Problems with migration intervals Local/area-specific data: guidelines for research Conclusions Population redistribution Urbanisation trends Metropolisation and inter-metropolitan migration... 35

6 Multivariate statistical techniques used in this study Inter-provincial migration Patterns of inter-provincial migration Multivariate profiles of inter-provincial migrants Inter-district migration Labour migration Migration proper Summary of the effect of distance Summarising the interactions between districts Other examples of migration modelling Conclusions Migration differentials Proportion of migrants in the population Migration intervals compared: and Migration selectivity Age-gender selectivity Employment, education and gender differentials Comparing migrants and labour migrants Differential migration: a multivariate analysis Conclusions The way forward Dealing with migration in future censuses Migration questions in sample surveys Conclusions Appendices A Inter-provincial migration: detailed MNA results B Data and definition issues C Information on the logistic regression D Modelling migration: further attempts Index

7 List of Tables 2.1 A suggested (partial) typology of spatial mobility encompassing both circulation and more permanent moves, and incorporating the more flexible approaches to defining migration Migration to, between and from the four main metropoles and non-metropolitan areas in South Africa ( ) Inter-provincial migration in South Africa ( ): number of people involved in every migration direction Age-specific comparison by province of the proportion of the population that has ever migrated Former migrants by population group ( & ) Results of the logistic regression: Comparing labour migration and migration Overall statistical results of the multiple classification analysis (MCA) of the probability of having ever migrated An MCA-based profile of the probability of having ever migrated: provincial location of the respondent An MCA-based profile of the probability of having ever migrated: type of locality An MCA-based profile of the probability of having ever migrated: population group An MCA-based profile of the probability of having ever migrated: gender An MCA-based profile of the probability of having ever migrated: age An MCA-based profile of the probability of having ever migrated: marital status An MCA-based profile of the probability of having ever migrated: highest educational qualification An MCA-based profile of the probability of having ever migrated: work status vii

8 Page 4.13 An MCA-based profile of the probability of having ever migrated: annual household income An MCA-based profile of the probability of having ever migrated: household size An MCA-based profile of the probability of having ever migrated: dwelling type An MCA-based profile of the probability of having ever migrated: home-ownership by the household of the dwelling lived in Migration questions asked during Census Suggested migration questions for future Censuses A1 Destinations of inter-provincial migration in South Africa: Some statistical details of the explanatory variables used in the multivariate nominal-scale analysis (MNA) A2 MNA results on inter-provincial migration destinations in South Africa: province of origin A3 MNA results on inter-provincial migration destinations in South Africa: type of locality (at destination) A4 MNA results on inter-provincial migration destinations in South Africa: population group A5 MNA results on inter-provincial migration destinations in South Africa: migration period A6 MNA results on inter-provincial migration destinations in South Africa: age at (last) migration A7 MNA results on inter-provincial migration destinations in South Africa: migrant worker status A8 MNA results on inter-provincial migration destinations in South Africa: gender C1 Variables (and their categories) used in the analysis D1 Estimated coefficients of the model D2 Elasticities derived from the model D3 Overall distribution of reasons for leaving the previous place of residence D4 Statistical details of a multivariate nominal-scale analysis (MNA): reasons for leaving previous place of residence D5 Reason categories for leaving previous place of residence: results of the multivariate nominal-scale analysis (MNA) viii

9 List of maps 1.1 Location of provinces, districts and former homelands Distribution of the population (1996): a district-based perspective Usual districts of residence of labour migrants (1996) Ratio of migrant workers to economically active population District destinations of migrants ( ) Origin districts of migrants ( ) Origin districts of migrants to Gauteng ( ) Usual districts of residence of Gauteng labour migrants ( ) ix

10 List of figures 2.1 (Extended) Value-expectancy-based model of migration decision-making De Jong s general model of migration decision-making x

11 List of graphs 3.1 Urbanisation levels ( ) Net inter-provincial migration ( ) Proportions of the various migrant categories Contribution rate and migration distance Transformed rate of contribution and migration distance: areas that were not part of former homelands Transformed rate of contribution and migration distance: former homeland areas (a) Age-specific migration rates ( ) based on the official 5 per cent sample of the 1980 census (b) Age-specific migration rates ( ) based on the migration community profile of Census District-based higher education and unemployment rates District-based in-migration and unemployment rates District-based out-migration and unemployment rates District-based out-migration and unemployment rates (in terms of the ratio of females to males in the district) Relative probability of having migrated (by income) xi

12 About the authors Pieter Kok Pieter Kok is a chief research specialist in the Integrated Rural and Regional Development research programme. He holds a Ph.D. in urban and regional planning from the University of Pretoria. Prior to joining the HSRC, Dr Kok worked as a town planner at the Borough of Newcastle (KwaZulu-Natal), where he was responsible for research and master planning. Pieter s areas of research specialisation are internal migration and urbanisation, and he has been the project leader for two large multidisciplinary studies on migration (since 1999) and urbanisation ( ). He was the leader of two provincial studies, one on the socio-economic profile of Gauteng ( ) and the other on development indicators for KwaZulu-Natal ( ). Pieter has been the author or co-author of a number of research reports (including a two-volume book on urbanisation), journal articles, formal papers and book reviews. Dr Kok also taught part-time in demography at the University of Pretoria and held the position of Professor Extraordinary from 1991 to In 1996 he received the HSRC Award for Research Excellence. Michael O Donovan Michael O Donovan is a research specialist in the Surveys, Analyses, Modelling and Mapping research programme. He holds a Master s degree in sociology from the University of the Witwatersrand. Prior to joining the HSRC, Mr O Donovan established the Township Databank, which aimed at the creation of an up-to-date demographic profile of Gauteng townships and settlements. Michael s area of research specialisation is the analysis, interpretation and management of data, with a particular emphasis on large datasets, and he has recently been the project leader for a study to develop a minimum data set on ageing in South Africa. Mr O Donovan was the co-recipient of the United Nations Development Programme (UNDP) Best Practice Award for a best practice in empowering urban communities, and between 1998 and 2000 he was a board member of Agrèment South Africa. xii

13 Oumar Bouare Oumar Bouare is a chief research specialist in the Social Aspects of HIV/AIDS and Health research programme of the HSRC. He holds a Ph.D. in economics (New School for Social Research, New York), a Ph.D. in mathematics (University of Paris VI), and an MA in philosophy (University of Paris I, Sorbonne). Oumar s area of research specialisation is economic analysis. One of his outstanding achievements is the development of a new theory of international trade, Profit, and profit and externalities as a basis for international trade (Instituto di Economia Internazionale, vol. LI, no 3 in 1998). In this article, he developed a theory that profit, and profit and externalities are a basis for international trade. Dr Bouare has recently been selected for inclusion in the 2002 edition of Great Minds of the 21st Century, compiled by the Governing Board of Editors of the American Biographical Institute (ABI). According to the ABI his selection is due to significant accomplishments within, and mastery of economics. Earlier he was also included in Who s Who in Economics in the 21st Century, compiled by the International Biographical Centre, Cambridge, England. Johan van Zyl Johan van Zyl is a research specialist in the Integrated Rural and Regional Development research programme. He holds a BA (Hons) degree in geography from the University of Pretoria. Prior to joining the HSRC, Johan was employed by the Bureau for Market Research at the University of South Africa (UNISA), and served in an advisory capacity for a number of local and international organisations. These include the SAARF Demographic Research Division of the Bureau of Market Research at UNISA, Statistics South Africa (Census 1991 and 1996 and October Household Surveys from 1995), the 1998 South African Demographic Survey and the Demographic Surveillance System (Wellcome Trust and the Africa Centre for Population Studies and Reproductive Health based at Hlabisa, KwaZulu-Natal). Johan s areas of research expertise include analytical demography (adjustment of census data for smaller geographical areas, population projections, estimation of fertility and mortality rates from survey data, applied demography and field surveys), interpreting the needs of clients in order to design a survey (questionnaire design, sample specifications, and management of fieldwork) and a detailed knowledge of the principles of data analysis. Mr Van Zyl has delivered various papers at national and international conferences, and published or edited several books, chapters or client reports in areas related to demography as well as other survey-based studies. xiii

14 Preface Aims This book seeks to achieve the following objectives: provide clear definitions of migration concepts and present a literature-based foundation for the study of internal migration in South Africa; provide a comprehensive overview of internal migration (based on recent census and other secondary data, mostly those provided by Statistics South Africa); introduce techniques and approaches that can be used to analyse data on internal migration; provide guidelines for questions on internal migration in future censuses and other surveys; and help to pave the way for the questionnaire survey to be undertaken as part of the larger project on the causes of migration. As a work it is rather technical so it is aimed at analysts (e.g. academics, students and researchers) and decision makers dealing with migration issues rather than at the general public. It should appeal to those with a feel for figures and, preferably, some basic knowledge of statistics. Overview What is immediately clear from the research undertaken for this book is that despite political and economic changes, migration patterns are essentially a continuation of patterns that predate the abolition of apartheid in South Africa. A more substantive finding derived from the analyses is that relationships and patterns of migration are complex. At times the relationships demonstrate great continuity, for example the apparent continuation of migration rates set 20 years ago by segregationist policies. At others, established trends are discontinued, for example in the unexpected prominence of migration from metropoles to non-urban areas. There is discontinuity also in the role played by distance, with one pattern evident in the population of the former homelands and another in the non-homeland population. One source of misunderstanding is the divergence in absolute numbers and rates. Rates of migration appear to be contradictory. How the population is distributed by location, area type and political heritage (setting the social dimension aside) has to be borne in mind when examining the data. In the past, South Africa lacked suitable data on internal migration. This meant that historical trends could not be analysed to the extent required in a country that underwent such notable political, social and economic changes during the final decade of the 20 th xiv

15 century. The only available historical data on internal migration were those for the period , but these were flawed by the exclusion of data in respect of the former homelands of Transkei, Bophuthatswana and Venda. Census 96 provided a welcome change by making available, for the first time, data on internal migration for the entire country and population. This new source of data has invited the undertaking of appropriate analyses. However, the prior absence of suitable data has obstructed the development of experience in the analysis of migration data over time. This calls for some guidance on techniques for using national data on internal migration and for examples of the kinds of analyses that can be undertaken. This book aims to address these needs. Chapter 2 deals with the theories and models of migration. Although the underlying (root) causes of migration are predominantly economic in nature, there are some very important non-economic (mainly social) reasons why migration is often perpetuated and sometimes becomes systemic. Non-migration is seen as the result of constraints to migration caused, among other things, by the costs associated with moving (especially over longer distances) and by personality characteristics (such as aversion to risk-taking). Very often, non-migration is caused by and almost inevitably leads to an in situ adjustment to the current situation, irrespective of how unattractive that may be. In Chapter 3 the patterns of internal migration are analysed in relation to urbanisation and metropolisation trends in their historical context and with reference to the role of apartheid. The concentration of people and poverty in the former homeland areas and the resultant high population densities in these rural areas were discussed. The economic anomaly of high unemployment and low out-migration rates was highlighted. The importance of regional (spatial) planning and development for the provision of economic opportunities and services in these hitherto deprived areas should again be emphasised. The two main types of migration that have been identified are migration and labour migration. The reason for separating these two lies more in the data constraints than in any inherent conceptual or definitional differences. In fact, the general migration typology discussed in Chapter 2 treats these two migration types as very much the same thing. The data are not that tolerant though. In the analyses in Chapter 4 that deals with migration, the main question addressed is why some people migrate and others not. The analyses on labour migration in this chapter deal mainly with a similar question, namely why some people become labour migrants and others not. Various explanations for migration and nonmigration are offered, but in the end the questions remain largely unanswered, mainly owing to the absence of suitable data. Despite the data problem, an attempt is made in Chapter 4 to compare the basic migration patterns for the two five-year periods ( and ) for which relatively comparable information is available. The only analyses that are viable in view of the data constraints are those dealing with the different levels of migration among the various age and population groups. The findings indicate that despite dramatic political, social and economic changes in South Africa (including the abolition of apartheid s migration-related measures such as influx control and group area demarcations), there was an insignificant change in the overall level of migration between the late 1970s and the early 1990s. xv

16 A modified gravity model is applied in Appendix D, and it is shown that unemployment, income and racial differentials seem to play a lesser role in inter-provincial migration than crime levels. An analysis of the 1997 October Household Survey data (see Appendix D) shows that a slight majority of moves were caused by economic factors (especially work-related reasons), but that the remaining moves had been caused by noneconomic factors. The results of the modelling attempts reported in Appendix D are inconclusive and even confusing. This highlights, once again, the need for purpose-made data for the analysis of migration causes and patterns. This need should be addressed to a large extent by the HSRC-funded questionnaire survey that will be undertaken as part of the larger project. xvi

17 Overview As in many other countries, internal migration in South Africa is an under-researched topic. This lack of research attention does not necessarily stem from perceived unimportance or a lack of interest but rather from the historical absence of appropriate census data. Problems with census-based migration data are usually derived from limits to the number of questions, difficulties with the coding of origin data, and memory lapse (affecting mainly data on time of the move). The main aim of this book is to describe migration in South Africa, based on recent South African censuses and other national surveys, and to provide an analytical evaluation of the migration data that they generated. Before Census 96, South Africa lacked suitable census data on internal migration. The only partly useful data were those for the period from the 1980 census, but these were flawed by the exclusion of the former Transkei, Bophuthatswana and Venda homeland areas. Census 96 therefore provided a welcome change, with a battery of questions that covered the last move, irrespective of the time of the move, for the entire population and for the country as a whole. Using age-specific migration rates as a criterion for judging the validity of migration data, it was shown that the Census 96 data conformed, in all the important respects, to the expected pattern. Other findings pointed to the same conclusion. Our study therefore indicates that Census 96 provided, for the first time ever, an invaluable source of data on South Africa s internal migration. Census 2001 has built nicely on the baseline data provided by Census 96 by concentrating on migration since the 1996 census, and it will also improve on Census 96 by not only generating data on country but also province of birth. Although the emphasis in this book is more on census-based migration data, we also looked at other secondary national data (such as the October Household Surveys) that might help to obtain a clearer understanding of the internal migration process in South Africa. However, from an analytical point of view, these other data sources proved to be somewhat disappointing. It should nevertheless be understood that migration is a highly complex component of population dynamics. While purpose-made sample surveys can generally deal effectively with the processes and causes of migration, census-based migration data are essential for providing the context within which migration takes place in a country. Although there clearly are limits to the scope of the migration data that censuses can provide, sample surveys cannot be successful in providing the necessary insight into migration processes without reliable census data on internal migration patterns and trends. Census 96 opened up many opportunities for meaningful analyses of internal migration patterns, and Census 2001 built further upon that solid foundation. These two censuses promise to provide very useful data for comparative migration analyses. It is now up to the migration analysts to utilise this opportunity to delve deeper into the complex set of factors associated with xvii

18 South African migration patterns and trends. The HSRC has shown its commitment to help fill the existing gaps by funding empirical research on the causes of migration in this country. It is trusted that other organisations and migration scholars will do the same. If this book contributes to making it possible, we will have achieved most of our objectives. xviii

19 C H A P T E R O N E Introduction IN SOUTH AFRICA the analysis of migration trends has been hampered by the absence of comprehensive and detailed data on human movement. Until recently migration analysts relied on sample surveys that are neither detailed nor comprehensive enough for an understanding of this dynamic phenomenon. Only censuses potentially offer information with the required breadth of detail. 1 Unfortunately, before 1996 South African censuses generally failed to record migration data such as the place and timing of migratory moves within (and to) South Africa. The 1996 version, officially referred to as Census 96, was a welcome exception. Although South African censuses routinely provide data on the time and country of birth, only two previous censuses provided data on migration within the country s borders. In 1980 respondents were asked where they had lived five years prior to the census, while in 1991 a question on the duration of residence at the current address was included. Map 1.1: Location of provinces, districts and former homelands LEGEND Magisterial districts Former homelands Provincial boundaries 1

20 Post-apartheid patterns of internal migration in South Africa When used in isolation both questions have inherent limitations when it comes to analysing migration. More problematic, though, is that in both those censuses only part of the country s population was covered. The 1980 census excluded the former nominally independent states of Transkei, Venda and Bophuthatswana, and migration data for these former homelands are not available despite constituting a substantial source of migrants, in particular labour migrants (see Map 1.1). In 1991, another nominally independent territory, the former Ciskei, which also constituted a significant reservoir of migrants, was excluded from the census. Only Census 96 covered migration data for the entire country and provided useful data on movement patterns. It is hoped that the detailed analyses of some of the 1996 census data on migration presented here will contribute to a collective understanding of migration while making migration studies more accessible by showing how the information, coupled with the use of statistical techniques, can enhance understanding of the phenomenon. Migration data generated by Census 96 The box on pp. 6 7 lists all the questions included in the 1996 census that are related to migration. It shows that five fields of migration analyses were covered in Census 96. These are (a) lifetime migration (see questions 1.1 and 1.2), (b) migrant labour (see questions 2 and 5), (c) place of usual residence (questions 3.1 and 3.2), (d) duration of residence (question 4.1), and (e) origin of the most recent move (question 4.2). Taken together these questions provide, for the first time, a potentially powerful source of information on spatial mobility leading to a thorough understanding of the dynamics of migration. Despite the fact that Census 96 has been severely criticised in various quarters for not being altogether reliable, it provides an important opportunity for undertaking the sort of migration analyses required to understand its dynamics and this book explores how the data can be used in migration research. Context and scope of the study In 1996 South Africa s overall population density was approximately 33 persons per square kilometre (km 2 ), which made it the 66th least densely populated of the 196 countries listed by the Population Reference Bureau (1997). However, the population was very unevenly distributed, with densities varying significantly between and within provinces (see Map 1.2). For example, the Northern Cape is the largest province (in terms of land area) and the smallest in terms of resident population. In 1996 it had a population density of just more than 2 persons per km 2. At that time, Gauteng, the smallest province in terms of area, accommodated 432 persons per km 2. 2 Population densities are usually related to the extent of urbanisation but in South Africa some of the least developed rural districts exhibit high population densities. Particularly noticeable on any population density map are the high densities in the former homeland areas. 3 These high densities can be ascribed, among other things, to past segregationist 2

21 Introduction Map 1.2: Distribution of the population (1996): A district-based perspective 500 people per dot Kilometres policies such as the Group Areas Act and influx control. These instruments prevented the African people from migrating to urban areas (see, for example, Kok 1986), and gave rise to large-scale forced resettlements (mainly to the former homeland areas). This resulted in the formation of densely populated areas with some of the lowest levels of service delivery, infrastructure and employment in the country. Despite changes since the abolition of the segregationist legislation (including the re-incorporation of the nominally independent homelands), after 1994, the backlog is formidable. These previously neglected areas continue to be marked by high levels of poverty and vulnerability and thus continue to influence migration patterns. Migration, as usually reflected by people moving to better-serviced places with more promising job prospects, has the potential to help redress these inequities. However, if people are denied the opportunity to move (for whatever reason), past inequities can even be exacerbated. It is therefore important not to look only at the causes and consequences of migration as a social phenomenon, but also at the causes and consequences of nonmigration. As mentioned earlier, until 1996 no South African census provided the data required for comprehensive migration analyses. The project that led to these analyses being undertaken was called Causes of internal migration in South Africa. It dealt with migration patterns within the country, and with the causes of both migration and non-migration. This book was the first in a series of products arising from the project, which included an empirical survey. 3

22 Post-apartheid patterns of internal migration in South Africa Purpose of the book This book attempts to place migration (and non-migration) in perspective. For that purpose, use is made predominantly of the data contained in two census data sets. The first is the spatially highly detailed Census 96: Migration Community Profile, and the second is the 10 per cent sample of the 1996 census. The first data set offers great detail on the spatial location of the population (and thus of migrants). The second data set lacks the spatial detail but presents the full census results for 10 per cent of the population, making possible an exposition of the social profiles of migrants and non-migrants. Emphasis is placed in the analyses on obtaining a better understanding of the complex migration/non-migration differential, among other things. In the process, migration/non-migration is analysed mainly from a spatial perspective for provinces and smaller spatial entities (by utilising the data of the Migration Community Profile), and from the perspective of the individual resident (through analyses of the 10 per cent sample). The issue of appropriately defining migration is dealt with extensively in Chapter 2, but it is necessary to point out here that the term migration, as used in this book, includes all migrants, while labour migration refers to a specific subset of migrants. Outline of the book Chapter 2 provides a theoretical exposition of migration research and serves as a backdrop against which the remainder of this book should be viewed. It explores the depth of our collective misunderstanding of internal migration in South Africa. As indicated above, this misunderstanding can be attributed, at least in part, to the historical absence of suitable migration data for the country as a whole. The definition of migration receives some attention, and it is shown that in this book migration refers to those movements that involve a change in the usual place of residence from one magisterial district to another. The problems with migration intervals are discussed, and a rationale is provided for the adoption of the period as the most appropriate migration period for the majority of analyses contained in the book. Chapter 2 also covers the need for more localised surveys to capture patterns of residential mobility in the context of such area-specific studies. Chapter 3 discusses population redistribution patterns and trends. In South Africa and elsewhere, urbanisation levels increased steadily during the 20 th century and, by 1996, almost 54 per cent of the country s population lived in urban areas (defined as those places with an urban form of local government). However, great variations were observed among the nine provinces, with urbanisation levels ranging from a low of 11 per cent in Limpopo (previously known as the Northern Province) to 97 per cent in Gauteng. The latter was also by far the most preferred destination in inter-provincial migration between 1992 and 1996, followed by the Western Cape, Mpumalanga and North West. Chapter 4 provides general profiles of migrants (and, by default, non-migrants). It is shown that only about one quarter of the total population has ever migrated, and that there have been surprisingly few changes in the level of migration between and Migration in South Africa is clearly selective in terms of age, and there are some important differences in migration levels among the various employment and educational categories in 4

23 Introduction the population. A multivariate profile of migrants (versus non-migrants) shows that Mpumalanga, Gauteng and the Western Cape have the highest proportions of former migrants when the effects of other variables are eliminated. The multivariate profile also shows that people living in shacks and the homeless were generally much more migratory than those living in traditional dwellings and formal houses and flats. Chapter 5 provides guidelines for future censuses and sample surveys dealing with migration. Notes 1 Like almost all countries exhibiting a similar level of development, South Africa does not maintain a rigorous population register of where people currently reside and have moved from. 2 The three magisterial districts with the highest densities were predominantly township areas situated in each of the three main metropoles: Soweto in Gauteng ( persons per square kilometre), Umlazi in KwaZulu- Natal (7 426 per km 2 ), and Mitchells Plain in the Western Cape (6 657 per km 2 ). 3 These former homeland areas are spread over seven provinces, namely the Eastern Cape, KwaZulu-Natal, Free State, Limpopo (formerly known as the Northern Province), Mpumalanga, North West and the Northern Cape (see Map 1.1). 5

24 Post-apartheid patterns of internal migration in South Africa Box 1 Migration questions asked during Census 96 Section A: In respect of each household member: 1.1 Was (the person) born in South Africa? (Include the former TBVC states Transkei, Bophuthatswana, Venda, Ciskei) 1 = Yes 2 = No 1.2 (If No ) In what country was the person born? Write in the name of the country. 3 Is (the person) a migrant worker? (Someone who is absent from home FOR MORE THAN A MONTH each year to work or to seek work.) 1 = Yes 2 = No 3.1 Is this DWELLING (e.g. house, room, shack, flat) the place where (the person) usually lives, i.e. where (the person) spends at least four nights per week? 1 = Yes 2 = No 3.2 (If No ) Where does (this person) usually live? Name of suburb/village/settlement:... Name of city/town/farm/tribal authority:... Name of magisterial district:... If not South Africa, please state name of country:... If no usual address, circle In which year did (the person) move to the DWELLING (e.g. house, room, shack, flat) where he/she usually lives? Write in the year that he/she moved OR The person has never moved, circle 1 1 (Lived in the dwelling since birth) 4.2 (For the person who has moved) From where did (the person) move? (Before moving into the dwelling where he/she usually lives) Name of suburb/village/settlement:... Name of city/town/farm/tribal authority:... Name of magisterial district:..... If not South Africa, please state name of country:

25 Introduction Section B: In respect of the entire household: 5. Are there any persons who are usually members of this household, but who are away for a month or more because they are migrant workers? (A migrant worker is someone who is absent from home for more than a month each year to work or to seek work.) 1 = Yes 2 = No (If Yes ) Indicate the person s particulars: Age in years:... Gender:... Relationship to head of household:... Where is (the person) living: Name of suburb/village/settlement:... Name of city/town/farm/tribal authority:... Name of magisterial district:... If not South Africa, state name of country:... 7

26 C H A P T E R T W O Literature review THERE IS LITTLE that is certain about migration trends in South Africa. Until now analyses based on a comprehensive profile of the total population have not been possible. The emphasis has fallen on localised studies and/or on analyses of particular segments of the population, thus limiting the collective understanding of migration/non-migration and the ability to generalise findings to the national or even provincial level. In particular, the causes of internal migration, the processes involved and, to some extent, their consequences are poorly understood. This book aims to help fill this information gap. As censuses have their shortcomings, particularly on attitudinal data, they are unable to make up the entire information backlog. This means that sample surveys will still be needed although not to draw out the context in which migration takes place. This can be derived from the migration data generated by Census 96 and Census Current status of migration research In the past, empirical migration research tended to be restricted to distinct spatial entities, frequently the metropolitan areas (see, for example, Kok, 1984; Kok, Hofmeyr & Gelderblom, 1986; Möller, 1986; Kok, 1988; Seekings, Graaff & Joubert, 1990; Botes, Krige & Wessels, 1991; Emmett, 1992; Crankshaw & Hart, 1990; Dewar, Rosmarin & Watson, 1991; Crankshaw, Heron & Hart, 1992; Mears & Levin, 1994; Cross, Bekker & Eva, 1999). More limited research was conducted in rural areas and towns (e.g. Platzky & Walker, 1985; Graaff, 1986; Spiegel, 1987; Rule & Wills, 1989; Saayman et al. 1997; Bekker, 1999). Apart from being localised, these studies often focused on particular components of the population components generally defined by migratory behaviour. A comprehensive overview of migration therefore awaited the 1996 census results. Definitions What is migration? Standing (1984), Kok (1999) and others note the liveliness of the debate on the concept migration. Hence, defining migration is a complex issue. As no single definition can automatically be applied in all contexts, typological approaches are used to supplement the suggested definitions. When migration is used with respect to human population, reference is usually made to a range of patterns of movement. In the contemporary South African context the term is used to reflect major social changes like the movement of people from rural to urban areas. This dynamic is widely believed to drive urbanisation. In South Africa, migration is also 8

27 Literature review associated with labour migration the oscillation of workers between their homes and distant employment opportunities. From a planning perspective, it would be useful to indicate the relative importance of rural-to-urban migration and labour migration in setting the development landscape. However, comparisons of the two are complicated by the fact that very different populations and activities are involved. As this book examines both migration and labour migration, the two major components of migration in South Africa, they are defined below. In the context of development in South Africa, Kok (1999) suggested a typology of spatial mobility based on two key dimensions, namely time and space (see Table 2.1). His typology encompasses both circulatory and more permanent moves (see Gelderblom & Kok, 1994:52), but is regarded as partial because not all types of mobility are considered, i.e. the examples listed in the second column of Table 2.1 are far from exhaustive. For instance, mobility types such as those mentioned by Standing (1984:38 53) and Van de Walle (1982:92) are not considered. 1 Circulation Nomads, gatherers and wanderers Shopping trips and tourist trips Daily work trips Trips home to visit, or to return to place of employment after a period of stay (e.g. a week or weekend) at the origin of the move Long-term migrant labour absences (usually of longer than a week at a time) from home More Change of permanent residence permanent moves ( moving home ) People with no fixed place of residence Short-term circular moves involving no change of residence Short-term circular moves that do not necessarily involve a change in usual place of residence but do involve a change of residence A move taking place at the beginning or end of an extended migrant-labour period Short or long-term residence at place of destination Table 2.1: A suggested (partial) typology of spatial mobility encompassing both circulation and more permanent moves, and incorporating the more flexible approaches to defining migration Broad Example Temporal dimension Spatial dimension Classification category Description Change in Description Migrationplace of residence? No No Yes Yes Yes Short or longdistance moves Short or longdistance moves Short or longdistance moves Short or longdistance moves Short or longdistance moves defining boundary crossed? Yes/no Yes/no * Kubat (1976:11) and Standing (1984:38-39) refer to the type of moves undertaken by nomads, gatherers and wanderers, among others, as transilient mobility. No Yes No Yes Yes No Transilient mobility* Short-term mobility Daily commuting Local weekly commuting Short-term labour migration Local long-term labour mobility Long-term labour migration Permanent migration Residential mobility 9

28 Post-apartheid patterns of internal migration in South Africa The typology suggested by Kok has been informed by various sources, such as Standing (1984:38 53) and Van de Walle (1982:92). It also includes the suggestion by Pressat and Wilson (1985:144) that tourist trips (irrespective of the distance) and nomadic movements should be classified as circulation. Furthermore, Skeldon s (1990:11) suggestion that commuting and other temporary movements can be referred to as circulation was also incorporated. The suggested typology takes account of southern Africa s particular circumstances whereby long-distance commuting and labour migration are known to form a significant proportion of the spectrum of spatial movement (cf. Gelderblom & Kok, 1994:105 6). Kok s (1999) typology does not require a change in the usual place of residence of the subject. However, migration requires that a change of residence (and not necessarily a change in usual address) must accompany the crossing of the boundary of a migration-defining area. The three shaded cells in the last column of Table 2.1 depict the only migratory moves that meet both these time and distance requirements. These migration types are short-term labour migration (see the topmost shaded cell in the last column of Table 2.1), long-term labour migration (third cell from the bottom in the last column), and permanent migration (second-last cell in the last column). The non-shaded cells in the top part of the last column encompass forms of spatial mobility that can only be described as forms of circulation. The very last cell in the table deals with residential mobility, i.e. intra-area moves, which do not qualify as migration even though they do involve a change of residence. In terms of this partial typology it seems fair to suggest that one should, ideally, define migration formally as the crossing of the boundary of a predefined spatial unit by persons involved in a change of residence. This definition encompasses all the requirements referred to by Kok (1999). The spatial unit refers to the particular study s migration-defining areas, which could be political, administrative or other appropriate spatial entities. As census results on place of origin are currently restricted to magisterial districts, 2 migration-defining areas here are taken to be the magisterial districts as demarcated for Census 96. Census 96 also uses the concept of usual residence to distinguish permanent residents from visitors. Consequently, as far as this analysis is concerned, Kok s (1999) suggested (partial) typology needs to be further restricted. Although such a restriction is quite justifiable in terms of the census requirements, it complicates the analysis. In particular it prompts the adoption of at least one supplementary category of migration labour migration. Thus for the purpose of this book, migration or migratory move is defined as: a change in the magisterial district of usual residence. A residential move within the same district ( residential mobility ) is therefore excluded, as are temporary movements (e.g. visits) between districts. Consequently, migration must be distinguished from labour migration, which in terms of the typology discussed earlier represents a subset of migration. According to the census office s definition, a labour migrant is: an individual who is absent from home (or country) for more than one month of a year for the purpose of finding work or working. This could be a mineworker, a factory worker or even a gardener or domestic worker. 10

29 Literature review Thus, migration or a migratory move is restricted to the permanent or semi-permanent movement of households and individuals (sometimes also within a defined time period), and labour migration to the periodic movement of individuals (also during such a defined period, where applicable). Labour migrants are almost by definition compelled to migrate semi-permanently and unaccompanied by their households. Migration origin and destination Every residential move has an origin or source (which is the place from where the person moves) and a destination (i.e. the place where the specific move ends). For a move to be classified as migration, the origin and destination of a residential move can only be in different migration-defining areas within the same country or in different countries. Internal migration, in-migrant and out-migrant In virtually all of the analyses that follow, both the origin and destination of a specific migratory move are in the same country (South Africa), thus the moves constitute internal migration. If the origin and destination are in the same country, the person who migrates from a particular district is called an out-migrant from that area, and simultaneously he or she is an in-migrant into the area of destination. When reference is made to international migration the respective terms used are immigration and emigration (and immigrant and emigrant). Residential mobility, circulation and commuting If the origin and destination of a residential move are in the same country and in the same (migration-defining) area, the move is not regarded as migration but rather as residential mobility. The term residential mobility is reserved for those moves that involve a change of residence. If a move involves no change of residence, it is neither migration nor residential mobility, but some or other form of circulation. If such a circulatory move takes place repeatedly between the person s place of residence and his/her place of work, it is known as commuting (on a daily, weekly or monthly basis). Data adequacy, reliability and appropriateness Peter Morrison acknowledges the many practical problems experienced by migration researchers because they rarely have the luxury of dealing with large, disaggregated data sets focused on their analytical requirements. Instead they must content themselves with data that only partly satisfy their conceptual requirements (Morrison, c. 1980:8). He goes on to warn that limitations of this nature may not only inhibit the development of theory but also distort observation, and, moreover, that the relationship between concept and measurement can become perverse if analysts start manipulating concepts to fit the available data. It is necessary, instead, to adapt the available data to the conceptual requirements (Morrison, c. 1980:8). One example of this problem relates to the spatial unit selected for migration analyses (see the discussion on migration-defining areas above). According to Willis (1974), there 11

30 Post-apartheid patterns of internal migration in South Africa is no fundamental reason for accepting the boundary definition of migration, and the definition and statistical material would be closer to reality if we recorded all changes of residence (1974:5). In his view, a typology of spatial mobility can best be developed if all residential moves are considered as migration (1974:6). Rather surprisingly, though, after criticising the use of boundary definitions in migration, he actually proposes the use of migration-defining areas. This happens when he introduces his own set of requirements for appropriate spatial units in migration research. It is essential to define an area of sufficient size to include enough migrants to justify analysis of their characteristics, and a compromise must be made between small areas desirable for homogeneity and large areas desirable for the inclusion of an adequate number of migrants (1974:7). One can ascribe Willis s inconsistency to the difficulty of getting rid of the concept of migration-defining areas. The solution may therefore lie in the use of adequately small and homogeneous spatial units. If the areas are kept as small as possible, they can, as the occasion requires, be consolidated into larger spatial units. In the absence of more detailed information the magisterial district will seemingly satisfy Willis s requirements mentioned above. This raises another important consideration in favour of the use of some form of migration-defining areas, namely the need for an unambiguous specification of the origin and destination of a move. The place from which a move is made (origin) is often defined in broad terms, while the data on the area in which a move terminates (destination) are often very detailed (United Nations, 1970:1). The latter is of course only valid when dealing with the last move, i.e. in which the destination was the place where the census/survey took place. The definition of the area of origin depends on the nature of the information available to the researcher (United Nations, 1970:2). In the case of the recent South African censusbased migration data, researchers are currently able, at best, to combine enumerator-area data of the destination of the last move with the magisterial district of the area of origin. Statistics South Africa is contemplating an improvement of the data to better reflect the actual place of origin (in terms of place name ) but, to date, the data set containing these codes has not been made available. This book thus deals exclusively with magisterial districts as the smallest spatial units of analysis. Theories and models of the causes of migration Migration theory can be traced back to Ernest-George Ravenstein s The Laws of Migration, published during the latter half of the 19th century ( ), which made him the undisputed father of the modern thinking about migration (Arango, 2000:284). Another seminal work, published in the early 20th century, was that of William Thomas and Florian Znaniecki, called The Polish Peasant in Europe and America ( ) that, according to Arango (2000:284), was probably the most impressive book ever written on the subject of migration. The contributions on the application of the gravity concept in migration by Steward (1941) and Zipf (1946) were of great importance in the development of migration modelling. Another historically important publication was the book by Peter Rossi, called Why Families Move (1950), dealing with the causes of intra-urban mobility. The purpose of this section, however, is not to give an historical overview of the development of migration theory. Rather, it is important here to concentrate on the more 12

31 Literature review recent contributions to our current understanding of migration processes, and especially those dealing with the causes of these processes. In this regard the very important contributions of scholars such as Gordon De Jong (from the early 1980s up to the present) and Douglas Massey (starting in the early 1990s and still continuing) are worth mentioning. As far as the determinants of migration are concerned, it may be useful to distinguish between economic and non-economic causes. The next two sections contain a brief review of the roles of these factors and of the variables that are used to analyse them. The first section based largely on a literature review by Massey et al. (1993) speaks to the prevailing economic theories/models, while the second section addresses the non-economic models of migration. The data used in this book does not provide an opportunity to test these theories, and this review is therefore intended merely to orientate the reader to the theoretical issues involved in migration research. Economic factors that cause migration Much has been written about the economic causes of migration and it has been suggested that these are the only real root causes of migration. This is probably true in some cases but when applied to the population in general it implies that people s migratory moves are dictated almost exclusively by economic considerations, and therefore economic incentives/disincentives should be sufficient to cause/prevent migration. It is therefore more appropriate to view migration in the larger context of development (economic and otherwise). As Skeldon (1990:150) points out, patterns of mobility are intimately related to the overall process of development and any explanation of mobility becomes in a sense an explanation of development. The logic of uneven development in effect causes migration. The recent review of migration literature by Massey et al. (1993) has contributed to a better understanding of the theoretical framework within which the economic causes of migration come to the fore. This review forms the basis of the brief discussion of the economic factors to be presented here. 3 Massey et al. differentiated between the theories accounting for the initiation of migration and those explaining the persistence of migration across space and time. Although the emphasis in their review is on international migration, there is no convincing reason why some (if not all) of these theories would not apply to internal migration. Massey et al. (1993) identify four theoretically derived causes of international migration: neo-classical economics, new economics of migration, dual labour-market theory, and world systems theory. Neo-classical economics This theory, together with its extensions, suggests that migration (both internal and international) is caused by geographic differences in labour supply and demand, and by the resultant wage differentials. Key assumptions of this theory are that the elimination of wage differentials will end the movement of labour, and that migration will cease in the absence of such differentials. Individual actors are assumed to estimate the costs and benefits of moving to alternative locations, and migrate to the area where the expected discounted net returns are greatest over some time horizon (Massey et al., 1993:433 4). 13

32 Post-apartheid patterns of internal migration in South Africa The micro-economic model of individual choice (that was first formulated by Michael Todaro, 1969, 1976), combined with the concepts of human capital (developed by Larry Sjaastad) and to some extent also place utility (developed by Julian Wolpert), corresponds to the macro-economic model described above. DaVanzo (1981) gives a fairly thorough description of the micro-economic model. She explains that the utilities to be obtained in alternative localities are weighed against the utilities in the current place of residence and the cost of moving to a particular destination. A move will only be considered if the present value of a move to a particular destination is positive, and the actual move will take place to that destination where the present value is the greatest. In her view, utility cannot be directly measured in empirical research. Although this is in principle a serious drawback, DaVanzo s conclusions in this regard indicate the limited utility of the model with respect to modelling and analysing migration. However, the suggestions regarding expectancies as formulated by De Jong and Fawcett (1981) may be used to help solve this measurement problem. The variables used in the testing of these theories include: Expected income, which is defined as the probability of employment (i.e. one minus the unemployment rate) multiplied by the mean income of the economic sector a rational actor may contemplate working in (Massey et al., 1994:701). In practice this predictor can take the form of an interaction term that cross-multiplies wages and employment opportunities. A statistical test for the significance of this interaction term, compared to a regression model where real wages alone appear, constitutes a critical test comparison between the Ranis-Fei and the Todaro versions of neo-classical theory (Massey et al., 1993:455). Net gain from migration is the difference between incomes expected at origin and destination, when summed and discounted over some time horizon and added to the negative costs of movement (Massey et al., 1994:701). This involves a relatively complex calculation. For a detailed discussion of the operationalisation of neo-classical economics in empirical migration research, see Massey and Espinosa (1997:947 51). New economics of migration Oded Stark and others (quoted by Massey et al., 1993) suggested that migration decisions are seldom taken by isolated individuals (an assumption central to the micro-economic perspective), but rather by families or households. Under the new-economics regime, decision-making is seen to take place in the context of household risk minimisation (instead of the notion of individual income maximisation as proposed by neo-classical economics). Unlike individuals, households can control economic risks by diversifying the allocation of household resources like family labour. Some family members may be assigned local jobs and other economic activities, while others, perhaps with more appropriate skills, may be sent to work in more distant labour markets. In the event that local economic conditions deteriorate and activities fail to bring in sufficient income, the household can rely on migrant remittances for support (Massey et al., 1993:436). As such, the theory recognizes that in many settings, particularly in the developing world, markets for capital, futures and insurance may be absent, imperfect or inaccessible. In order to self-insure against risks to income, production and poverty, or to gain access to scarce investment capital, 14

33 Literature review households send one or more workers to foreign labor markets (Massey et al., 1994:711). Given the relatively higher wages in the industrial economy, migration offers a particularly attractive and effective strategy for minimising risks and overcoming capital constraints (Massey et al., 1994). However, the model does not preclude migration to areas with a minimal (or even a negative) wage differential. A key assumption of the new economics of migration is the notion that wage differentials are not necessary conditions for migration to occur, because households may have incentives to diversify risks through migration even in the absence of wage differentials. Another important component of this assumption is that of relative income, which implies that migration will occur not only to improve absolute income, but also to increase the household s income relative to other households in the community (Massey et al., 1993). In other words, households attempt to ameliorate their sense of relative deprivation through migration. A study in Mexico by Oded Stark and Edward Taylor showed, however, that internal migration might ameliorate relative deprivation much less than international migration does. This is because internal migrants merely end up substituting one (mainly the rural) reference group for another (the urban one), since both are within a similar social, cultural and economic setting (Massey et al., 1994). This emphasis in the new economics (or household economics ) on the household as the decision-making unit has been criticised by Spiegel (1987) and others. Some of the key points of criticism are that the concept does not deal sufficiently with the important matter of inter-household transfers especially in the context of fluidity in household composition (Spiegel, 1986). The concept is seen to assume rather simplistically that all individuals within the household have the same interests (see Gelderblom & Kok, 1994:47 52). Debates regarding the household as the migration decision-making unit are therefore nowhere near a resolution in the African context, as elsewhere. The variable central to analysis within the new economics paradigm (apart from absolute income ) is relative income, which has also been expressed as relative deprivation. Given the problems surrounding income data at the individual level, a multilevel statistical model should be constructed that not only contains the usual individual and household-level predictor variables, but also incorporates the community characteristic of income inequality, or an operational measure of relative income (Massey et al., 1993:458). Massey et al. (1994:714) refer to the operationalisation of the concept of relative deprivation in a study by Stark and Taylor. There the concept was defined (for a given household) as the proportion of households with incomes greater than the income of the specific household, multiplied by the average amount by which these incomes exceed that household s. Massey and Espinosa (1997:953 4) discuss the empirical application of this model in the context of Mexico-United States migration. Dual (segmented) labour-market theory While the two models above are essentially micro-level rational choice (decision) models, dual labour market theory argues that migration stems from the intrinsic labour demands of industrial economies (Massey et al., 1993). The theory takes cognisance of structural inflation, job motivation, economic dualism and the demography of labour supply. 15

34 Post-apartheid patterns of internal migration in South Africa Structural inflation is said to occur when employers raise wages at the bottom of the occupational hierarchy in order to attract workers for unskilled jobs. Such actions upset socially defined relationships between status and remuneration, creating strong pressures for corresponding wage increases at other levels of the hierarchy (Massey et al., 1993). Motivational problems arise at the bottom of the job hierarchy because of the general absence of avenues for upward mobility. This structural problem is presented as inescapable because there will always be a bottom level in the labour market. What employers need are workers who view bottom-level jobs simply as a means to the end of earning money, and for whom employment is reduced solely to income, with no implications for status or prestige (Massey et al., 1993:441 2). Migrants (and labour migrants in particular) often satisfy this need. Most migrant workers start off as target earners who seek to earn money for specific goals mainly to improve their status or security at home (i.e. place of origin). Their aspirations are therefore relatively limited compared to those of native workers. Economic dualism is part and parcel of bifurcated labour markets in industrial economies. Workers occupying capital-intensive jobs are generally skilled and normally work with costly capital investments. Their jobs are relatively stable compared to those in the labour-intensive sector where workers hold unstable, unskilled jobs. This leads to a segmented labour market structure, where low wages, unstable conditions, and the lack of reasonable prospects for mobility make it difficult to attract native workers who are instead drawn into the capital-intensive sector, where wages are higher, jobs are more secure, and there is a possibility of occupational improvement (Massey et al., 1993:443). Employers therefore turn to migrants to fill the shortfall in labour-intensive jobs. Empirical research in North America has confirmed that urban labour markets are in fact segmented. This research also shows that cities with large migrant populations have, apart from the capital-intensive sector and the labour-intensive sector, a third sector, namely the migrant enclave (Massey et al., 1994). It may be better, therefore, to refer in more generic terms to segmented rather than dual labour markets. The demography of labour supply perspective focuses on a decline in participation by women in low-wage labour and on a decline in fertility in industrialised countries. Collectively and individually these declines reduce the supply of those willing to work for low wages like teenagers. The imbalance between the structural demand for entry-level workers and the limited supply of such workers has increased the underlying, long-run demand for immigrants (Massey et al., 1993:443). The segmented labour market theory therefore states that labour migration is demanddriven, and since the demand for migrant workers results from the structural needs of the industrial economy, wage differentials are neither necessary nor sufficient to cause labour migration to occur (Massey et al., 1993:444). An integral part of labour market segmentation is occupational specialisation, which has been traditionally associated with migration and, in effect, is one of the principal causes of human movement. The extension of regional specialisms, the sallying forth of labour recruiters for particular occupations, and the transfer of personnel by corporations signify an 16

35 Literature review intimate and long-standing relationship between mobility and occupation that has often been ignored in the analysis of the causes of migration (Skeldon, 1990:142). Massey et al. (1993:458) state that it is difficult to verify the segmented labour market structure empirically. Since the model argues that migration is driven by conditions of labour demand rather than supply, they suggest that one should obtain a higher degree of explanatory power among the indicators relating to conditions in receiving regions than those in sending regions. If real wages and employment conditions are entered into an equation predicting movement between Turkey and Germany, for example, German indicators should dominate in terms of predictive power (Massey et al., 1993:459). Massey and Espinosa (1997:954 5) describe the preparation of data for an empirical application of this model. World systems theory In this theory, the penetration of capitalist economic relations into peripheral non-capitalist economies is seen to create a mobile population that is prone to migrate. Migration is a natural result of the disruptions and dislocations that accompany capitalist expansion. As land, raw materials and labour within the peripheral regions come under the influence and control of markets, people inevitably migrate (Massey et al., 1993:443, 445). Capitalist economic relations result in the commercialisation and mechanisation of agriculture, and in land consolidation that destroys traditional systems of land tenure based on inheritance and common rights of usufruct. This displaces people from the land (Massey et al., 1993:443, 445). Capitalists in core regions also ensure that improvements are made to expand transportation and communication links to and from the peripheral regions where they have invested. These links facilitate not only the movement of goods, information and capital, but also the movement of people by reducing transport and communication costs (Massey et al., 1993:446). The offer of wages to former peasants and the creation of incipient labour markets (fostered by individualism, private gain and social change) also promote labour migration in developing regions. The demand for factory workers strengthens the local labour market. Because female workers are in greater demand than male workers, the resulting feminisation of the workforce limits the local employment opportunities for men. The insertion of foreign-owned factories into peripheral regions thus undermines the peasant economy by producing goods that compete with those made locally; by feminising the workforce without providing factory-based employment opportunities for men; and by socialising women for industrial work and modern consumption, albeit without providing a lifetime income capable of meeting these needs. The result is the creation of a population that is socially and economically uprooted and prone to migration (Massey et al., 1993:446). Although this process echoes the situation in developing countries such as South Africa, it also takes place in the context of economic globalisation, where global cities (especially those of the United States, Europe and the Pacific) show structural characteristics that create a strong demand for migrant labour (Massey et al., 1993:446 7). Massey and Espinosa (1997:955) point out that world systems theory (like the new economics of migration ) is linked to community-level indicators. The model describes outmigration as originating in communities that are in the throes of economic development 17

36 Post-apartheid patterns of internal migration in South Africa rather than in backward, stagnant areas disconnected from national and international markets. Massey and Espinosa expected, for example, that the probability of emigration to the United States would be greater in those Mexican communities where wage rates, levels of self-employment and the proportion of women employed in manufacturing were higher, and where roads, schools and banks had already been established. They also measured capitalist penetration of Mexico by the rate of growth in direct foreign investment. However, Skeldon (1990:134) warns against a situation in which facts prove ineffectual against the armour of an a priori assumption. He is particularly concerned about the ideology-based determinism implied by the concept penetration of capitalism instead of it being viewed as a process with particular and variable consequences. The penetration of capitalism may also convey the idea that the society being affected by the expansionary power was static and unchanging, just waiting to be galvanized through contact with the West. One of the greatest myths about Africa is that it contained myriad isolated self-sufficient groups awaiting transformation by the advent of colonialists. It is important to stress that these societies were not simply passive acceptors of a new way of life but that they resisted, modified and manipulated what was being offered them or imposed upon them (Skeldon, 1990:135). Economic factors that perpetuate migration The following theoretical contributions, according to Massey et al. (1993), explain why international migration, once started, seems to be perpetuated indefinitely: institutional theory, cumulative causation and migration systems theory. 4 These are, obviously, derived predominantly from an economic perspective of international migration. Institutional theory Once international migration has begun private institutions and voluntary organisations arise to satisfy the demand created by the imbalance between the large number of people who seek entry into capital-rich countries and the limited number of immigrant visas these countries typically offer. This imbalance, and the barriers that core countries erect to keep people out, create a lucrative economic niche for entrepreneurs and institutions dedicated to promoting international movement for profit, leading to a black market in migration (Massey et al., 1993:450). The institutional developments described by Massey and his co-authors are important in perpetuating international migration, and the institutionalisation becomes more and more independent of the factors that originally caused the migration. This theory is clearly relevant in the case of international migration, but it may also play a role in some internal migration (especially where closed-city measures apply, such as South Africa s influx control policy under apartheid). Cumulative causation Not much needs to be said here except to note that causation is cumulative in the context of migration when every migratory move alters the social context within which subsequent 18

37 Literature review migration decisions are made, typically in ways that make additional movement more likely (Massey et al., 1993:451). A persistent gap in theories to explain migration is the relative lack of behavioural studies which provide a dynamic vs. a static comparison of migration move-stay decision alternatives. One example of a dynamic research focus is cumulative-cause processes such as chain migration (De Jong, 2000:307). Migration systems theory According to Massey et al. (1993), migration systems are relatively intense and stable exchanges of goods, capital and people between certain areas (e.g. countries) and less intense/stable exchanges between others. An international migration system typically includes a core receiving country or group of countries, and a set of specific sending countries linked to the receiving country/countries by unusually large flows of immigrants (Fawcett & Zlotnik, quoted in Massey et al., 1993:454). A danger in applying these theories is that migration can become an independent variable which not only regulates other societal change but ultimately controls itself (Skeldon, 1990:132). The macro-level economic factors highlighted above tend to lead to a conclusion that migration is inevitable, and that its causes do not require scholarly debate. This is a very narrow and ill-conceived conceptualisation of a dynamic process. It is important to understand that migration is also dependent on other factors. But what are these factors, and how do they influence population movement? A partial answer to these questions is suggested in the next section, which deals with the non-economic factors causing migration. The above perspectives have more in common than their focus on economic forces suggests, and they predispose analysts to seeing migration in particular ways. Internationally, migration is expected to be from the South to the North. Within countries it is expected to be from low-wage, rural hinterlands to urban heartlands. In South Africa labour migrants are also seen to be primarily black and male. Economistic approaches like those above have had an enduring effect on the understanding of migration in South Africa. The conceptual dominance of these approaches has resulted in a strong association of migrants with poverty. Migrant-sending areas are identified with the former homelands and receiving areas with the metropoles. Subsequent chapters will show the relative importance of migration between urban areas vis-à-vis ruralurban migration. The empirical evidence as to who is most inclined to migrate challenges and undermines the analytical utility of economistic approaches. Conceptual frameworks that better reflect the empirical evidence to hand are called for and may be offered by the incorporation of non-economic factors. Non-economic factors that cause migration The migration literature has identified a large number of salient non-economic factors causing internal migration. Some of these are covered in this section. It must be stated from the outset that the relevance of such additional factors cannot be determined from the data 19

38 Post-apartheid patterns of internal migration in South Africa currently at hand. This calls for further empirical research. Although economic motives appear to be a major causative factor in migration, they are unable to explain migration decision-making on the whole (see Shaw, 1975). Similarly, the emphasis on structural determinants/correlates (which are mainly economic) also undermines the search for a more balanced explanation of migration processes. The disadvantage of macro-level structural approaches is that the explanations developed tend to be deterministic The broad perspective can flatten historical specificities in favour of a naïve globalism (Skeldon, 1990:133). Value-expectancy model of migration From the work by De Jong and Fawcett (1981), Gardner (1981) and many others who applied the value-expectancy model empirically (in the Philippines, Romania, Thailand and South Africa) 5, it is clear that the most important micro-, meso- and macro-level causes of migration operate indirectly via people s values and expectations. This takes place within a causal framework. Little evidence of a strong direct link between micro-, meso- and macrolevel factors (including economic factors) and the decision to migrate has been found. As a result, the point of departure here is the micro-level causal framework suggested by De Jong and Fawcett (1981). 6 This requires a clear understanding of the causal framework of the extended value-expectancy model. De Jong and Fawcett (1981:54) developed the causal model depicted in Figure 2.1. They state that the following should be noted: The family/household should be treated as a unit, with separate analyses for the moves of individual members (e.g. single-adult siblings) and family units. The expected strength of the explanatory path is indicated in the graphic on page 23. The thick, solid lines indicate strong causal relationships, the thin, solid lines show moderate direct linkages, and the thin, dotted lines show weak causal linkages. The basic components of the value-expectancy model are goals (values or objectives) and expectancies (subjective probabilities): Values can be determined empirically by getting respondents to rate them in terms of importance (to obtain one part of the value component), and the dimensions for the values and expectancies can be obtained through multivariate statistical techniques such as factor analysis. The centrality of particular (hypothesised or empirically derived) values in a personal value system can be determined, as well as their pervasiveness across different behavioural domains and their salience in particular contexts. Expectancies can be measured by asking respondents to assess people s chances of attaining various goals in their current place of residence and in alternative destinations. Expectancies might be measured with additional dimensions also, e.g. whether outcomes pertain to self or others, whether results are expected immediately or in the longer term. The significance of these dimensions in migration behavior is apparent, as in the case of a household head who decides on a move primarily for the long-term educational and other opportunities that will be available to his [or her] children (De Jong & Fawcett, 1981:52). 20

39 Literature review From the available literature on the reasons for migration, De Jong and Fawcett (1981: 49) identified conceptual categories that seem to represent psychologically meaningful clusters. These are wealth, status, comfort, stimulation, autonomy, affiliation and morality. The factor scores of both the value and expectancy components can then be used to calculate the value-expectancy scores and the strength of the intention to migrate in respect of the place of origin and one or more possible destinations. Pairs of the value-expectancy components have a multiplicative relationship for a specific factor (dimension) or item, and the products are summed over all the factors/items being considered in order to obtain the strength of the migration behavioural intentions, which is given by the formula: MI = V i E i where: MI = strength of the migration intentions; V i = value of the outcome of item i to the person concerned; and E i = the expectation that migration will lead to the desired outcome in respect of item i. Expectancy should be measured for the present place of residence as well as alternative locations. Behaviour is not governed by motivational factors only and since migration is also facilitated or constrained by environmental and cultural factors (see Gardner, 1981), it is necessary to integrate multi-level determinants (i.e. socio-cultural, demographic, personal and economic factors) with the value-expectancy model of migration decision-making. Migration behaviour is hypothesised to be the result of (1) the strength of the valueexpectancy-derived intentions to move, (2) the indirect influences of background individual and area factors, and (3) the modifying effects of constraints and facilitators that become salient during the process of migration decision-making (De Jong & Fawcett, 1981:56). From the causal pattern depicted in Figure 2.1 it should be clear that a large number of noneconomic (and, of course, economic) variables could contribute to explaining the model. These typically include individual and household demographic characteristics, societal and cultural norms, personal traits, opportunity structure differentials between areas, information about areas, unanticipated constraints and in situ adjustments. Individual and household demographic characteristics. These are the most frequently described differentiating factors in migration behaviour. This category may include such migration correlates as life-cycle variables, family and household characteristics, factors associated with socio-economic status, the various employmentunemployment differentials, home-ownership, the extent of household crowding, ethnic differences, years in the community and past migration history. 21

40 Post-apartheid patterns of internal migration in South Africa These factors can be seen to summarise the compositional thesis of populations, some with higher and some with lower propensities to move, which help explain why certain areas have higher migration rates than other areas (De Jong & Fawcett, 1981:53). Viewed from a micro-level perspective, these characteristics are often treated as motivations for migration. For example, age differences in migration have been interpreted as indices of employment-related incentives, the migration-homeownership correlation as an index of community and family ties, and the relationship between life cycle and migration reflecting the motive for more comfortable housing (De Jong & Fawcett, 1981:53). Instead of treating them in this way, one should regard these and other similar migration-differential variables as predictors of the values and expectancies that form the main intervening variables for the analysis of migration intentions and behaviour. Societal and cultural norms These form a second broad category hypothesised to affect migration values and expectancies. Community norms and gender roles have been shown to affect migration patterns of demographic sub-groups. For example, norms concerning marriage-related migration often dictate a move by one party, usually the woman, while in other societies the new family forms an independent household, which necessitates a move by both parties. Young people may be required to migrate in pursuit of an education or to earn money for remittances to help meet family needs. The existence of such norms, we argue, would be reflected in personal values and expectancies; that is, the norms will be internalised to some extent (De Jong & Fawcett, 1981:55). Migration expectations operate in the context of social norms and gender roles. Social norms are critical elements in translating expectations into behavioural intentions and subsequent action [G]ender has a core influence on the statuses of males and females, their roles and stages in the life cycle. These help determine people s position in society and therefore the opportunities women and men have to consider in moving to the premigration stage (De Jong, 2000:307). De Jong and Fawcett (1981:55) admit that it is difficult to obtain satisfactory measures of norms, although cross-cultural comparative studies can be used to predict different value hierarchies. Personal traits Personality traits such as risk-taking ability, adaptability to change and the ability to produce the desired result (efficacy) are another set of predictors for values, expectancies and migration. A general problem with this category is that much of the research on risk taking and similar personal characteristics is flawed by the use of education or some other social or economic characteristic as a proxy measure for personal traits (De Jong & Fawcett, 1981:55). Kok (1988) developed and used a scale of risk-taking propensity and found that 22

41 Literature review this factor had a statistically significant, direct causal relationship with the value component (and therefore with value-expectancy ) in six out of the eight dimensions that he studied. 7 This category of variables can therefore be a very important indirect set of causes of migration. Figure 2.1: (Extended) Value-expectancy-based model of migration decision-making Individual and household demographic characteristics: e.g., life-cycle and family cycle, SES, employment, own/rent home, land availability, ethnicity, household density Societal and cultural norms: e.g., sex roles, political climate and policies, community norms Personal traits: e.g., risk taking, efficiency, adaptability to change Opportunity structure differentials between areas: e.g., economic opportunities, marriage opportunity, status advancement, amenities, activities Values (goals) of migration: categories of values and disvalues, their strength, salience and centrality Expectancy of attaining values: certainty of outcome to self or others in short term and long term Information: extent, relevance, perceived validity Migration behavioural intentions In situ adjustment Unanticipated constraints and facilitators Move-Stay Source: De Jong & Fawcett (1981:54) Notes Family as unit with separate analyses for members: e.g. single-adult siblings, family-unit moves Expected strength of explanatory path: strong moderate weak MI (behavioural intentions) = V i E i Expectancy measured for present residence and alternative locations. 23

42 Post-apartheid patterns of internal migration in South Africa Opportunity structure differentials between areas These differentials are a major factor in the formation of expectancies for attaining goals in the area of origin or in alternative destinations. This conclusion is borne out by an extensive literature on the economic causes of migration (see Shaw, 1975; DaVanzo, 1981). It is important, however, not to limit indicators of opportunity structure to economic goals (such as employment and income), but to include indicators of the other migration-related motive categories such as marriage-opportunity differentials, differences in education opportunities, entertainment differentials and differential amenities. It is here that the linkages between macro-level indicators and micro-level value-expectancy dimensions can readily be tested (De Jong & Fawcett, 1981:55). Information Information about areas is seen as a factor that moderates the effect of opportunity structure differentials (De Jong & Fawcett, 1981:56). Goodman (1981) illustrates the importance of the extent, relevance and perceived validity of the information about opportunities in other destinations, compared to the current location. It should, of course, be remembered that erroneous information could have as much an impact on behaviour as valid information. Consequently, it is important to assess the knowledge and beliefs and not only the actual differentials in opportunity structure. (The construction of mental maps of alternative destinations by respondents may in fact help to understand the level of knowledge and perceptions of these different places, including the current location.) Gardner relates macro-level factors to micro-level decision-making by asking how aerial characteristics come to be interpreted and evaluated as important by individuals, for their own purposes. A macrofactor can cause or influence a desire to live in a certain place only if the individual understands and perceives that the factor is of some relevance to that individual and to his or her own values and goals If the factor is seen as relevant, then it enters into the resulting desire structure along with many others. Put briefly, what does a macro-level factor mean to the individual? (1981:72). Along with the availability of information of whatever quality, however, and along with the ability of the individual to utilise this information, comes the question of the inclination of the individual to take advantage of the information (1981:78). In the contemporary South African context this question arises with respect to the impact of spatial development initiatives (SDIs) on migratory choices. For example, what impact does declaring an area an SDI have on individual expectancies? What the individual ends up with is a mental map, a cognitive map or a set of subjective expected utilities, but it should be remembered that mental maps refer only to a graphic depiction of expressed preferences for alternative residential locations (Fuller & Chapman, cited in Gardner, 1981:79). Unanticipated constraints and facilitators The constraints and facilitators (i.e. negative constraints) may include family-structure changes (such as marriage, divorce, separation, death of a spouse or a change in family size), or changes in other factors such as health, the financial costs of moving or the anticipated support from relatives and friends. These changes are often unanticipated, and may not have 24

43 Literature review been salient considerations in the original migration intentions, as caused by valueexpectancy considerations (De Jong & Fawcett, 1981:56). Constraints and facilitators (especially the unexpected ones) intervene between intentions and the decision to move or stay. They should consequently be treated as Schmeidl s (1997) proximate conditions, while value-expectancy considerations are examples of her (1997) intervening factors. In situ adjustment The decision to move is not the only outcome of a migration decision-making process. Two other outcomes have been identified in the literature. One option may be to adjust the needs of the individual or household, while the other may be to restructure the environment relative to the household so that it better satisfies needs. Either of these alternatives would result in a decision not to migrate (De Jong & Fawcett, 1981:56). (This corresponds to the conclusion by Speare, Kobrin and Kingkade 1982 that dissatisfaction should be seen as a necessary but not sufficient condition for moving.) These adjustments may include a change of occupation, alterations/extensions to the physical structure of the house or a change in life style. Some of the above sets of factors may be seen as causes of migration, while others (e.g. in situ adjustment) may cause migration not to occur. It seems that by incorporating micro-level, meso-level and macro-level causal factors into the same model, a useful explanation of migration is likely to be found. This comprehensive model of migration decision-making seems to cater well for the needs of the wider research project, but there are also other models dealing with the non-economic causes of migration that need to be considered. Residential-satisfaction model of migration The decision to move out of a community is assumed to depend on the general level of satisfaction with the place of residence, the job and the community in general (Speare, Kobrin & Kingkade, 1982:553). Social mobility aspirations may lead to residential mobility when the current residential location is seen as inconsistent with a new social status. In his (earlier) work on the role of residential satisfaction in mobility, Speare (1974) found that his residential-satisfaction index was the most important causal factor for mobility. The decision to move out is seen primarily as a function of the changes in family composition changes that occur as a family goes through its life cycle. This approach allows the identification of appropriate measures of social and community bonds for different stages of the family life cycle. Speare, Kobrin and Kingkade (1982) found that duration of stay had the strongest effect of all the variables in their causal model, and that the hypothesised intervening factors of social bonds, community bonds and wish to move also have significant direct causal effects on actual migration. 8 This approach views the combination of demographic factors (age, sex, marital status and conditions such as family size and home-ownership) as the circumstances influencing an individual s propensity to migrate. In short, one would say that whereas the rate of migration may fluctuate with the economic situation, the selective factors, relative to the migrant s position in his or her life cycle and in the family, determine who move and at what point in their lives they move. 25

44 Post-apartheid patterns of internal migration in South Africa Non-economic factors that perpetuate migration De Jong and Fawcett state that the motivation to maintain ties with relatives and friends in the area of origin is an important determinant of the decision not to move. Equally important, however, is the presence of friends and relatives at potential destinations as they exert a significant influence on the decision to move, and particularly on the decision where to move. The presence of family and friends at a distant location therefore encourages and directs migration through increasing the potential migrant s awareness of conditions and opportunities there (1981: 31 2). Similarly, the facilitating hypothesis (also known as migration auspices ) argues that having relatives and friends at a distant location encourages and directs migration by increasing the migrant s potential for adjustment through the availability of aid to relocate there (De Jong & Fawcett, 1981:32). An empirical study by SyCip and Fawcett showed that when migration has become a community tradition, background factors (such as age, gender, marital status and education) lose much of their significance. That is, it is no longer the migrant s personal characteristics that are of primary importance, but rather membership in a social network that facilitates the migration process (1988:17). This confirms the conclusion by Massey et al. (1993) that social networks do not only cause migration but also serve to perpetuate existing migration systems. Social network and social capital models allow for the fact that the expectation (or the level of residential satisfaction or the strength of a social network) of a prospective migrant may be very different from the factors that emerge from a more structural analysis of the situation. Models that have concentrated purely on the structural regularities of migration fail to describe adequately the chain of events (rather than the single factors) that accounts for decisions to move. Massey et al. (1993) suggest that network theory contributes significantly to a better understanding of why (international) migration persists, despite changes in the factors that are believed to have caused inter-country movements. According to Massey et al. (1993:448 9): [m]igrant networks are sets of interpersonal ties that connect migrants, former migrants, and non-migrants in origin and destination areas through ties of kinship, friendship, and shared community origin. They increase the likelihood of international movement because they lower the costs and risks of movement and increase the expected net returns to migration. Network connections constitute a form of social capital 9 that people can draw upon to gain access to foreign employment. Once the number of migrants reaches a critical threshold, the expansion of networks reduces the costs and risks of movement, which causes the probability of migration to rise, which causes additional movement, which further expands the networks, and so on. A crucial implication of this theory is that migration becomes institutionalised through the formation and elaboration of networks, and therefore becomes progressively independent of the factors that originally caused it, be they structural or individual (Massey et al., 1993:450; emphasis added). This conclusion by Massey and his co-authors may also largely hold true for internal migration, and should be tested in the wider project. Massey and Espinosa (1997:952 3) show some interesting ways of operationalising the model in empirical research. Social networks indeed represent an important intermediate level between the micro- 26

45 Literature review level of individual decision making and the macro-level of structural determinants, thereby helping to bridge a gap that is one of the major limitations in migration thinking (Arango, 2000:292). Evaluation De Jong (2000) incorporated many of these recent insights into a new general model of migration decision-making (depicted in Figure 2.2). This model was applied in a migration study in Thailand where the results supported the theoretical propositions of the model and provided consistent evidence on the process of migration decision-making (De Jong, 2000:317). The theories highlighted above indicate a clear need to view migration as a function of multiple motives. Even where economic motives are dominant, they do not reflect the total context of the decision to move; they may also be quite inadequate for distinguishing movers from stayers (De Jong & Fawcett, 1981:43). Abad (1981:301) has the following to add: In the value-expectancy microlevel perspective, economic goals are viewed as only one of several sets of goals that migration may fulfil. When seen against other goals, economic considerations may exert a minimal effect on some migration decisions. Alternatively, if economic goals are deemed important, they may be intimately linked to family, household or community expectations. In these respects, policies that seek to influence migration decisions via some form of economic assistance will meet with little success unless they are consistent with kinship obligations and community norms. Figure 2.2: De Jong s general model of migration decision-making Migration networks: Family/friend ties Individual human capital attributes Family migration norms Gender roles Household characteristics and resources Values/expectancies Residual satisfaction Migration intentions Migration behaviour Community characteristics Behavioural constraints/facilitators: Prior migration experience Money to move Immigration policy Labour contracts/job transfers Discrimination Source: De Jong (2000:310) 27

46 Post-apartheid patterns of internal migration in South Africa Chang (1981) is of the opinion that in order to delve behind the meaningful relationships of statistical aggregates we must rely on the contributions of micro studies (p. 309). He points out that important questions about core issues in migration are not included in census questionnaires. Consequently, such studies cannot help planners determine the actual interaction between macro variables and potential migrants behavior. The actual causes of migration are not known but only inferred from the causal relationships of aggregate variables Planners should not try to read too much meaning into such poor causal structures, as the meaning may not be there (Chang, 1981:308 9). The conclusion to be drawn from all the debates and viewpoints cited above is that while it would be wrong to exclude macro-level analyses on aggregate data altogether, these may not provide the answers required. The best (compromise) approach may be to start off from a micro-level perspective and work one s way up the higher levels of data aggregation through multi-level analyses. The value-expectancy model of migration appears to be the most appropriate one for this purpose. Unfortunately census data do not lend themselves to the application of this model, and, in this context, consideration should instead be given to applying one of the macrolevel, spatial interaction models. The (modified) gravity model is a good option and is discussed in the next section. Modified gravity model of migration An attempt is made here to apply some of the principles of the (macro-level) modified gravity model of migration to help explain inter-provincial migration in South Africa during the period Starting with Steward (1941), aggregate models of migration were specified frequently in the context of a modified gravity model. The models are of a gravity type because migration is hypothetically related directly to the size of the relevant origin and destination populations, and related inversely to distance. The models are modified in the sense that the variables of the basic gravity model are given a behavioural content, and variables expected to significantly influence the decision to migrate are included in the estimated relationship. Modified gravity models are frequently estimated in double logarithm form because this functional form yields reasonably good fits, and the coefficients obtained from it can be directly interpreted as elasticities of migrants response to changes in the various independent variables of the estimated models (Goss & Chang, 1983). However, common use of the double-logarithmic functional form to estimate modified gravity models has led to criticism by Schultz (1982). In part, his argument hinges on the geographic size of the regions for which migration is measured. If all regions had the same population and land area, migration and non-migration probabilities would (merely) reflect the costs and benefits of the various location choices. However, the regions of any country differ greatly in population and land area. A larger share of all moves will tend to occur within the boundaries of larger regions. Consequently, more non-migration will appear to exist for such regions. The result is that non-migration is spuriously correlated with origin population size and land area. Schultz (1982) sees the 28

47 Literature review standard modified gravity model as inefficient because it fails to incorporate information on the relative frequency of non-migration. He argues, though, that in the limit as the unit of time diminishes over which migration is measured, differences between the two specifications might be expected to diminish (p. 576). The reason is that the population at risk of migrating becomes a better measure of the non-migrating population when the migration interval is very short. However, to take into account Schultz s criticism, one can express the dependent variable as the logarithm of the number of people moving from place to place instead of its probability, and remove from the set of independent variables the two population sizes. An important reason for using the number of people moving from one place to another is that policy makers are interested in the volume of realised migration instead of its probability. Additionally, in dropping the two population sizes from the set of independent variables, one avoids an important correlation problem. However, if the dependent variable denotes the number of migrants moving from one area to another, an appropriate measure of the size of the population at risk of migration may be necessary. In that case the population size at the place of origin should be included in the equation. The introduction of the distance component in the gravity-based models sprang from an observation that place-to-place migration decreases with an increase in distance from origin. This has been attributed to several factors, among which the following are the most important: Distance is a proxy for the out-of-pocket costs of moving, such as those associated with gasoline and the moving van. Opportunity costs rise with distance in the sense that longer moves sometimes require more time, which in turns means more foregone earnings if the individual is not involved in a job transfer. Opportunity costs also rise with distance in that the greater the distance of the contemplated move, the more are forgone alternatives within that distance likely to enter the equation (Wadycki, 1974). Information costs rise with distance, which in turn requires greater search costs to offset the greater uncertainty associated with more distant locations. Distance serves as a proxy for the psychological costs of moving, which can be offset by making more frequent trips or trips of longer duration back to the origin, where each type of return trip raises the cost of moving as a positive function of distance (Schwartz, 1973). If past migrants tended to move to nearby places, and if current migrants tend to follow past migrants, then current migrants tend to move to nearby places (Nelson, 1959; Greenwood, 1969). For a number of reasons the deterring effects of distance are, however, declining over time. One set of reasons is related to the fact that transportation and communication systems have improved and expanded. The reduced transport costs and improvements in the transportation system therefore encourage long-distance migration. Furthermore, the absolute values of distance elasticities decline over time (Denslow & Eaton, 1984), i.e. the response of migration to the change in distance weakens over time. Because of these reasons, one may also drop the distance component from the set of independent variables. 29

48 Post-apartheid patterns of internal migration in South Africa The modified gravity model that is applied in this study (see Appendix D) incorporates these suggested modifications. The data used are obtained mainly from Census 96 and the economic data provided by Statistics South Africa. However, crime data collected from the records of the South African Police Service are also used to make provision for the incorporation into the model of some potential non-economic components that are often perceived to be particularly relevant in present-day South Africa. Problems with migration intervals An important component of the temporal dimension of migration is the reference period. The United Nations manual on internal migration states that data should be compiled with reference to specified periods of time, with a view to studying the incidence of migration. According to the manual (1970:2), the interval may be definite (e.g. one year, five years, ten years or the intercensal period, to measure fixed-term or period migration). It may also be indefinite, as in the lifetime of the population alive at a given date, to study lifetime migration 10 or when the available data lack any definite time reference. Shryock, Siegel and Associates (1976:374) point out that if the mobility period coincides with the last intercensal period, the resulting migration statistics may be more useful in measuring the components of population change or in studying the consistency of the population and migration statistics. Among the suggestions by the United Nations (1992:6) on how to prepare migration data for sub-national population projections, it is pointed out that the interval over which migration is measured should preferably coincide with the proposed projection interval. However, this practical approach does not have much to offer in terms of underlying logic for identifying migrants as those who have moved within some specified period. A time period restriction effectively excludes a number of groups, the first being those who migrated just before the beginning of the migration-defining period. Then there are those who had moved away but returned within the specified interval to the area in which they had been living at the beginning of the period (Standing, 1984:36). Also excluded are those born and those who died during the period concerned (Shryock, Siegel & Associates, 1976:374). These arbitrary reference periods are usually unnecessary in migration surveys where the full process of mobility is adequately reflected in the data. In cases where only a few migration questions are asked, such as in censuses, the time-period approach may be acceptable. The preferred approach, however, would clearly be to avoid the use of a reference period altogether. Unfortunately this is not always possible. In fact, the use of migration intervals is often unavoidable. When comparisons between areas or population components are made, it is usually necessary to make use of a standard migration interval to ensure comparability. In this book the migration interval used most frequently is the period 1 January 1992 to 10 October This period is the closest to a five-year interval with the available census data (which recorded only the year of migration). It may, however, be necessary to look at other intervals (such as ten years) as well, because these may produce a different perspective on migration levels and patterns. 30

49 Literature review Local/area-specific data: guidelines for research Area-specific research on spatial mobility usually requires an analysis not only of migration to or from places outside the area but also of the residential mobility within the boundaries of the area concerned. The generation of the type of data needed for such analyses cannot be the responsibility of the central statistics office of a country, but should be the domain of the local government(s) and/or the private sector in the area concerned. Local, area-specific survey research is therefore needed to complement the broader spatial orientation of the census or national household surveys. Such local surveys can have relatively small sample sizes (unless very detailed spatial analyses are required) and may therefore deal relatively cost-effectively with the need for data on intra-area moves. Depending on the nature of the information required, it may be possible to include a whole battery of questions on the past mobility histories of the area s residents. However, in cases where the local, area-specific survey is expected to cover a broad range of social and economic topics, the constraints imposed by the need to restrict the duration of the interview often lead to a watering down of the questions on people s mobility histories. This is usually to the detriment of the utility of the mobility data so generated. Although the need for compromises in the interest of cost-effectiveness is acknowledged, it is suggested that the mobility questions are given priority in those cases where a greater understanding of the movements of the area s residents is crucial for appropriate planning and better decision making. Conclusions Little is known about migration in South Africa at an aggregated national level and much of what is known may be in question. The emphasis of migration research has fallen on local studies or research on particular segments of the population. The dependence on census data that do not cover the country as a whole has also inhibited the development of a thorough understanding of migration in South Africa. The inadequate understanding is particularly keenly felt in the area of processes and causes of internal migration, and to some extent, its consequences. Census 96 has been the first in the country s history to provide the required data, and this book is aimed at helping to utilise this new opportunity. The theories highlighted in this chapter indicate that there is a need to view migration as a function of multiple motives. Even where economic motives dominate, they do not reflect the total context of the decision to move. They may also be quite inadequate for distinguishing migrants from non-migrants. In fact, economic considerations may exert only a minimal effect on some migration decisions, but even if economic goals are important they may be intimately linked to family, household or community expectations. Hence, policies that seek to influence migration decisions via economic incentives are not likely to meet with much success unless consistent with people s values, norms and expectations. The conclusion to be drawn from all the debates and viewpoints cited is that, while it would be incorrect to exclude macro-level analyses on aggregate data, these may not provide the answers required. The best approach may be to start off from a micro-level perspective and work one s way up the higher levels of data aggregation. The value-expectancy model of 31

50 Post-apartheid patterns of internal migration in South Africa migration appears to be the most appropriate for this purpose. Unfortunately census data do not lend themselves to the application of this model, and for the purposes of this book, consideration is instead given to applying one of the macro-level, spatial interaction models. The (modified) gravity model has been selected as the most viable option and has been described in some detail. Although there are some serious shortcomings associated with the use of migration intervals, these cannot be avoided if comparisons between areas or segments of the population are needed. In this book, the most commonly used migration interval is the period , being the period closest to a five-year interval. The interval is used in the remainder of this book for most of the comparisons and analyses, except for the analysis of lifetime migration. 11 Notes 1 Nevertheless, special-purpose typologies geared to particular analytical concerns are surely valuable as a means of imposing a sense of discipline on analysis (Standing, 1984:57). 2 Statistics South Africa plans to code this information in terms of place as well, but these data have not yet been made available. 3 Other recent reviews that are not discussed in detail here are those on the causes of internal migration by Greenwood (1997) for industrialised countries, and by Lucas (1997) and Skeldon (1990) for less developed countries. Massey et al. (1994) and Massey and Espinosa (1997) have extended the important work by Massey et al. (1993) in respect of international migration, by evaluating the economic theories and providing a framework for empirical research. 4 Another theory that attempts to explain the perpetuation of international migration is network theory. This theory will, however, be discussed in the section that deals with non-economic factors causing migration. 5 The application of the model in the Philippines was reported by, among others, Arnold (1987), De Jong (1985), De Jong et al. (1983, 1986), Gardner et al. (1986) and SyCip and Fawcett (1988). See Sandu and De Jong (1996) for a discussion of the model s application in Romania. De Jong, Johnson and Richter (1996) and De Jong (2000) report on the application of the model in Thailand. The application of the model among the white population in South Africa is reported in Kok (1988, 1990) and by Kok and Badenhorst (1990), and among black squatters by Kok and Motloch (1992). 6 Of course any other suitable micro-level framework can be used, but it seems that De Jong and Fawcett s framework offers the best opportunity for factoring in meso-level and macro-level variables together with the standard micro-level features (see also Gardner, 1981). 7 To put this even further into perspective, one should probably add that in four out of these six cases valueexpectancy had a direct causal effect on the dependent variable migration intention (see Kok, 1988:117-22). 8 The variables included in these dimensions were as follows: (a) Social bonds : (i) the proportion of relatives in the state (Rhode Island), (ii) the proportion of parents and children in the state, and (iii) the proportion of friends in the state; (b) Community bonds : (i) home-ownership, (ii) self-employment, and (iii) community participation, based on the response to the question about whether the respondent or spouse ever attended local committee meetings (specified) in the past year or so (Speare, Kobrin & Kingkade, 1982:562, 566). 9 Social capital is the sum of the resources, actual or virtual, that accrue to an individual or a group by virtue of possessing a durable network of more or less institutionalised relationships of mutual acquaintance and recognition (Bourdieu & Wacquant, quoted in Massey & Espinosa, 1997:951). 10 The term lifetime migration should not be confused with lifecourse migration (which generally covers all the migratory moves that have been undertaken since birth). 11 Lifetime migration indicates whether a person has ever moved away from his/her place of birth and, if so, where such a lifetime migrant was found at the time of observation (e.g. during the census). 32

51 C H A P T E R T H R E E Population redistribution CHAPTERS 3 AND 4 explore the wider dimensions of South African internal migration through an examination of the 1996 census results. In this chapter the extent, prevalence and general direction of migration is presented while Chapter 4 explores the social and economic underpinnings of the exhibited patterns. The focus of Chapter 3 is on population redistribution patterns and trends, 1 with a primary objective of describing the magnitude and direction of migration. The analyses consequently relate to urbanisation trends, migration from non-metropolitan areas to the metropoles (and between metropoles), inter-provincial and inter-district patterns of migration, concentrating on influences that can be observed. The analysis that follows largely reflects the movement of individuals (as opposed to households) between migrationdefining areas. These areas range from provinces and magisterial districts to areas defined by their characteristics. The latter (which are often simple aggregations of magisterial districts) allow for an exploration of movement to and from urban, metropolitan and former homeland areas. Urbanisation trends Urbanisation can be defined as the process through which the population of urban areas increases, and is usually expressed relative to the total population. 2 The process of urbanisation is widely believed to be the most significant dimension of migration. Even when these moves are not numerically dominant they are thought to be of inordinate importance in terms of social and economic development. Popular conceptions of the importance of migration from rural to urban areas are seemingly confirmed by the rapid growth of urban areas like Johannesburg, Cape Town, Durban and smaller towns. Whether appearances are borne out by the census data depends to some extent on what is taken to constitute an urban area. Unfortunately, defining an urban area turns out to be rather problematic, especially in the South African context. In most South African censuses (i.e. between 1904 and 1946 and again since 1980) the definition of an urban area was based exclusively on a de jure requirement, i.e. that some form of local government must manage the area concerned. 3 The adoption of such dichotomous approaches (urban versus rural) in defining urbanisation has been severely criticised in the literature. The contributions of Graaff (1986), Hindson (1986), Kok and Gelderblom (1988, 1989), Gelderblom and Kok (1994) and the Urban Foundation (1990) are prominent in this regard. These authors argued for a more nuanced 33

52 Proportion urban (%) Post-apartheid patterns of internal migration in South Africa approach to the definition of an urban area, in particular for a definition that acknowledged an intermediate stage between rural and urban. In Census 96 an attempt was made to introduce the concept of semi-urban areas, but this has not produced acceptable results. Consequently, one is compelled to revert to the urban-rural dichotomy for the purposes of this analysis. In 1996 almost 54 per cent of the South African population lived in urban areas defined as those places with an urban form of local government. However, great variations in urbanisation rates were observed among the nine provinces. The proportion of the population living in urban areas (i.e. the level of urbanisation) ranged from a low of 11 per cent in Limpopo to 97 per cent in Gauteng. The urbanisation levels of the population groups differ, at least in part, as a result of the policies of segregation (see, for example, Mears, 1997). Influx control (abolished in July 1986) prevented the African population from settling permanently in areas outside the former homelands, notably the urban and metropolitan areas under white control. Coupled with this was the implementation of the group areas legislation of the time (repealed in June 1990) that enforced the resettlement of millions of South Africans. The exclusion of Indians/Asians from the former Orange Free State and parts of northern Natal (as they were known then) also affected settlement patterns in the past. Although the restrictions were often effectively circumvented, apartheid inevitably curtailed the freedom with which most black citizens could settle. Despite the fact that the last of these restrictions were lifted during Graph 3.1: Urbanisation levels ( ) Year TOTAL AFRICAN/BLACK COLOURED INDIAN/ASIAN WHITE 34

53 Population redistribution the late 1980s and early 1990s, their legacy is visible in the settlement profiles of the provinces and, particularly, of the urban areas. In the 20 th century South Africa s urbanisation profile substantially reflected the differences in settlement patterns. For example, as indicated in Graph 3.1, the proportion of the African population in urban areas was well below those of the minority (non-african) groups. 4 Extrapolation of the urbanisation levels suggests that the level of urbanisation among Africans at the beginning of the 21 st century approximately matched that of South African whites at the beginning of the 20 th century. To what extent influx control and forced resettlements contributed to the relatively low urbanisation levels of Africans is difficult to estimate, but it would be fair to suggest that government actions slowed the urbanisation of Africans, especially during the 1950s and 1960s. The impact of government action on urbanisation inevitably fashioned the quality of life and social well being of citizens. Mears (1997:602) points out that South Africa s urbanisation had a disequilibrating effect on the spatial distribution of the population and their income, with the result that rural-urban migration did not close the income gap that triggered the migration in the first place. In Chapter 4 this conclusion is analysed in more detail. Metropolisation and inter-metropolitan migration The dominant role of metropolitan areas in urbanisation is highlighted by the fact that, by some measures, three quarters of migration is to metropolitan areas. It would thus be informative to look further than urbanisation in general and take a closer look at metropolisation 5 and the patterns of migration between non-metropolitan and metropolitan areas in South Africa. Metropolitan areas represent the various chambers of the economic heart of any country, further justifying a study of the movements between, to and from them. Table 3.1 indicates the estimated number of movements during the period (based on the 10 per cent sample of the 1996 census). Only the four main metropolitan conurbations (Gauteng, Durban Functional Region, Greater Cape Town and Port Elizabeth-Uitenhage) have been used for the purposes of the table. 6 The table shows that, of all the metropolitan regions, Gauteng was simultaneously the source and destination of the greatest number of migrants and respectively. This is not unexpected perhaps, given the size of its population and its central location. What the table does reveal is the extent of its domination at this scale of analysis. When considering migration between the four metropolitan regions and the residual category that incorporates all other areas it can be seen that almost half (42 per cent) of migrants moved to Gauteng (see Table 3.1). Gauteng was also the single greatest source of migration with 19 per cent of migrants originating from there. It was also the destination of most (58 per cent, i.e ) of those who migrated from non-metropolitan areas to the metropoles. Less predictably, this province was also the origin of the majority (60 per cent, i.e ) of all moves from metropolitan to non-metropolitan areas, a large proportion of which was likely to have been caused by return (labour) migrants. It is also notable that about three quarters of all the moves from Gauteng (77 per cent) and Cape Town (72 per cent) had non-metropolitan areas as their destination, notably higher than the corresponding proportions for Durban (59 per cent) and Port Elizabeth-Uitenhage (55 per cent). 35

54 Post-apartheid patterns of internal migration in South Africa Table 3.1: Migration to, between and from the four main metropoles and non-metropolitan areas in South Africa ( ) Origin Destination Non-metro* Gauteng Durban Cape Town PE-Uitenhage Total Non-metro* (38%) (12%) (12%) (3%) (65%) Gauteng (15%) (1%) (2%) (0%) (19%) Durban (5%) (2%) (1%) (0%) (8%) Cape Town (4%) (1%) (0%) (0%) (6%) PE-Uitenhage (1%) (0%) (0%) (1%) (3%) Total (25%) (42%) (14%) (15%) (4%) (100%) * Non-metro = non-metropolitan areas Source: Statistics South Africa: 10 per cent sample of Census 96. Multivariate statistical techniques used in this study A brief orientation In some of the following sections and in the next chapter three statistical methods are of particular interest in analysing migration trends. They are logistic regression, multiple classification analysis (MCA) 7 and multivariate nominal-scale analysis (MNA). 8 All three are essentially regression methods that vary, primarily, in the characteristics of the dependent variable being analysed or described and in the presentation of the results. A lay description of the differences is presented below. In most instances the variable being described in the analysis is whether an individual belongs to a group or category for example, whether he/she is a migrant or not. In MNA any individual is classified as a member of the migrant group or not. MNA thus classifies each individual as a member of the category he/she is most likely to belong to. For logistic regression the variable being described similarly ranges from 0 to 1 where 1 indicates that the individual is a full member of the group and a 0 indicates that he/she is definitely not a member of that group. Logistic regression calculates the probability or likelihood that the individual being described is a member of the group in question this 36

55 Population redistribution value calculated represents the probability that the individual belongs to the group described by 1, as all values between 0 and 1 are possible. In fact, individuals are rarely classified as a 0 or 1 but are usually partial members of both groups. Inter-provincial migration When the frame (or scale) of reference changes a different picture emerges. A shift in frame from metropolitan areas to the provinces introduces a new dimension to the picture. First, a range of moves falls from the earlier picture all those moves from a province s hinterland to its local metropole. Then a new category of moves across provincial boundaries is introduced. In the previous analysis such moves would have been captured only if they involved migrating to a metropolitan area. Inter-provincial migration is marked by a small increase in the total number of moves from (Table 3.1) to (Table 3.2). While the increase is not particularly notable, the moves discussed in the section below probably represent migration prompted by a somewhat different set of rationales and motives. In the next two sub-sections the patterns of inter-provincial migration and some of the characteristics of inter-provincial migrants are discussed. Table 3.2: Inter-provincial migration in South Africa ( ): number of people involved in every migration direction Province of origin Province of destination Total WC EC NC FS KZN NW GP MP LP (SA) Western Cape (WC) Eastern Cape (EC) Northern Cape (NC) Free State (FS) KwaZulu-Natal (KZN) North West (NW Gauteng (GP) Mpumalanga (MP) Limpopo (LP) Total (SA) Source: Statistics South Africa: Migration Community Profile data (Census 96) 37

56 Post-apartheid patterns of internal migration in South Africa Patterns of inter-provincial migration Inter-provincial migration patterns for the period in Table 3.2. show (bottom row) that Gauteng was by far the main destination in that period ( ). Then followed the Western Cape ( ), Mpumalanga ( ) and the North West ( ). The last column in Table 3.2 shows that the Eastern Cape generated the highest number (almost ) of all inter-provincial migratory moves, followed by Gauteng ( ) and Limpopo ( ). The largest set of movements was between provinces with large former homeland populations and adjacent provinces dominated by metropolitan economies. This hints at the dominance of metropolisation in the urbanisation process. However the flows in both directions, i.e. in and out of the migration-defining areas have to be considered. For example, as was indicated above, metropolitan areas contributed substantially to migration out-flows. Gauteng, along with two predominantly rural provinces with large populations in former homeland areas, the Eastern Cape and Limpopo, lost a large number of people through outmigration between 1992 and By contrast, Gauteng was by far the most popular migration destination in the country, followed distantly by the Western Cape. The effect of these exchanges of people between provinces is that Gauteng experienced a net gain of more than a quarter of a million people ( ), while the Eastern Cape lost a roughly similar number of people ( ). An analysis of the patterns of inter-provincial net migration (i.e. the number of inmigrants minus that of out-migrants) between 1992 and 1996 (Graph 3.2) indicates that Gauteng and the Western Cape experienced the greatest net gains. The Eastern Cape and Limpopo had the greatest losses during the same period. This section is completed, in the text below, by an examination of the characteristics of inter-provincial migrants, which looks at how a multiplicity of variables interact to inform the migration profile. Essentially the section examines the distinct contribution social features like age and race have on inter-provincial migration. Multivariate profiles of inter-provincial migrants Multivariate nominal-scale analysis (MNA) 9 has been applied to the 10 per cent sample to determine the profile of inter-provincial migrants for the period. In this case the dependent variable is the respondent s province of usual residence at the time of the census. Only a few explanatory variables relevant for predicting destinations of inter-provincial migration are covered in the text because the description of this analysis is rather technical and its details and results have been confined to Appendix A. But the main findings are summarised below. The MNA analysis largely confirms what was expected on the basis of earlier analyses. For example, almost half (47 per cent) of the respondents who migrated during the five-year period moved to Gauteng. By contrast, only 8 per cent of those who migrated before this period made Gauteng their destination. Gauteng s attraction as a migration destination seems to have become increasingly visible during the early 1990s. Conversely Mpumalanga, North West and Limpopo lost some of their relative attractiveness during the 38

57 Migration gains or losses (numbers) Gauteng Western Cape Mpumalanga Free State Northern Cape North West KwaZulu-Natal Limpopo Eastern Cape Population redistribution Graph 3.2: Net inter-provincial migration ( ) Province first half of the 1990s when compared to earlier trends. Only a small proportion of all interprovincial migrants moved to rural destinations in Gauteng (8 per cent), whereas a relatively and unexpectedly large proportion of inter-provincial migrants moved to rural destinations in Mpumalanga (27 per cent). The Western Cape received most (51 per cent) of its inter-provincial migrants from the adjacent Eastern Cape at the time of the census. Recent research reported by, among others, Bekker (1999), indicated that this should have been expected. Similarly, almost three quarters (72 per cent) of inter-provincial migrants from North West ended up in adjoining Gauteng. The Western Cape was the main destination for coloured migrants from other provinces (37 per cent), and KwaZulu-Natal was the main destination of Indian/Asian interprovincial migrants. The comparatively low proportion of coloured inter-provincial migrants who ended up in Gauteng (27 per cent) ran counter to the expected dominance of Gauteng as a destination for this group. North West drew only 1 per cent of coloured interprovincial migrants. The analysis also illustrates a social dimension of migration, namely that the Western Cape is more attractive to older migrants than to younger migrants, while Gauteng is a less popular destination for older migrants. 39

58 Post-apartheid patterns of internal migration in South Africa Inter-district migration To date, the place of origin (former residence) of migrants has only been systematically coded to magisterial district level. This means that it is known, for all moves, which district rather than town, village or farm migrants formerly resided in. Despite this limitation the analysis of district-to-district migration yields a wealth of data. This is primarily due to the number of observations available for statistical analysis, which increases from five (four metropolitan areas and a catchall non-metropolitan area ) or nine (each of the provinces) to 353. Correlations between migration and the social and demographic features can then be pursued with a degree of subtlety. In this section the analysis of migration flows has been restricted to the economically active population that section of the population that is employed or seeking employment while being of a suitable age. The analysis allows for the highlighting of several different aspects of migration, not the least of which is the important distinction between migration (as defined above) and labour migration. While labour migration can be considered a subset of migration it exhibits many features that set it apart from other forms of migration. The 10 per cent sample of the census reveals that at the time there were approximately 14 million economically active residents of the country. Of these slightly fewer than a million Map 3.1 Usual districts of residence of labour migrants (1996) 40

59 Population redistribution considered themselves to be migrant workers. Almost one third of the balance of the population (32 per cent) reported having moved residence at least once between 1992 and However, the vast majority of these constituted moves within the same magisterial district. 10 In effect about 15 per cent of the population had moved residence at least once during For the purposes of this section it is only this 1,6 million that reflect migrants. The relative sizes of these sub-populations are shown in the pie chart (Graph 3.3). The spatial distribution of migrants is discussed below, first with reference to labour migration and then migration. Graph 3.3: Proportions of the various migrant categories 27% Labour migration 11% 28% 28% 6% Never moved Migrant workers Moved prior to 1992 Moved within the same district ( ) Moved between districts ( ) With some restructuring of the data, the 10 per cent sample can be used to show in which district migrant labourers usually reside when not working, reflecting where their homes are. These home districts are most often located in the former TBVC and self-governing areas, or somewhat surprisingly in the metropolitan areas. (Map 1.1 indicates where these areas are primarily located.) Map 3.1 illustrates, for example, that metropolitan Gauteng is a significant source of migrant labour. This is in no way anomalous as the eight largest sources of migrant workers are in metropolitan areas bringing into question the stereotyping of labour migrants as predominantly rural residents. The large number of migrant workers originating from former homelands and the metropolitan areas is, at least in part, a reflection of the concentration of people in these areas. To identify which areas contribute high proportions of their economically active population to other areas, it is necessary to examine the ratio of migrant workers to the total economically active population. The results of the comparison are given by Map

60 Post-apartheid patterns of internal migration in South Africa Map 3.2 Ratio of migrant workers to economically active population While Map 3.2 clearly shows the dominance of the former homeland areas as a reservoir of migrant workers it also points to some unexpected deviations. For example, in Limpopo areas surrounding the former homeland districts tend to contribute a greater proportion of their economically active population to migrant labour than the adjacent homelands. The area that contributes the highest proportion of its economically active population to labour migration is Potgietersrus. This district abuts various parts of what was previously Lebowa. The spatial profile of migrant labourers origin affirms the expected role of the former homelands as a labour reservoir and metropolitan areas as destinations for those temporary migrants (not portrayed). However it also raises questions regarding anomalies, like the relatively low contribution made by the former Lebowa. A similar exercise can be used to show where permanent migrants come from and where they move to and thus to test whether the former homelands perform a similar role for proper migrants. Migration proper As the above analysis of labour migration was focused on where the workers normally resided, rather than where they were currently employed, it allows analysts to see to which areas these migrants are most likely to return. To examine the redistribution of permanent migrants it is necessary to examine both their places of origin and destination. 42

61 Population redistribution Map 3.3 District destinations of migrants ( ) Map 3.3 illustrates where migrants have chosen to settle. It clearly shows that there is a strong tendency for migrants to relocate to metropolitan areas each of the metropolitan areas marked in that figure has a large concentration of migrants. Map 3.3 also reveals that the hinterland of the country is not a popular destination for migrants with people preferring to move north of the country or to the seaboard. The 10 per cent sample also allows us to identify in which district the migrants lived prior to their move. These results are graphically represented in Map 3.4. Once again the contribution to migration of the former homelands is readily evident, as is the contribution of the metropolitan areas. The strong contributions made by these areas to migration do not necessarily indicate that residents of these areas have a higher propensity to migrate their contribution may simply reflect greater concentrations of population. It is thus helpful to examine the outmigration rates. These are expressed, once all those who are not economically active have been removed from the analysis, as the ratio between migrants and total population. These rates reaffirm the importance of the homelands contribution to migration and the contribution of the former homeland areas in the Eastern Cape in particular. In several districts of the former Transkei and in one district of what was part of KwaZulu, the number of ex-migrants was at least as great as the number of residents who did not migrate (at least when the analysis is limited to the economically active population). This points to substantial out-migration from these areas assuming, questionably, that the economically active constitute a significant proportion of their populations. 43

62 Post-apartheid patterns of internal migration in South Africa Map 3.4 Origin districts of migrants ( ) By contrast the migration rates out of the former regions of Lebowa and Gazankulu appear to be somewhat lower than both those of adjacent districts and of the former Transkei. This is somewhat surprising given both the universal stereotyping of homelands as labour reservoirs and their proximity to Gauteng. Their differential rates of contributing migrants begs the question as to how the social, economic and political heritage of areas interact with aspirations, opportunities and disincentives (such as distance) to fashion the migration profile. The impact of geographic distance on the inclination to migrate can be gleaned by examining the migration patterns between regions. The simplest way of doing this is to examine the rate at which all districts contribute migrants to a single region. A useful example is to examine the rate at which all other districts contribute migrants to Gauteng, which is shown in Map 3.5. This demonstrates how the rate at which each district contributes to Gauteng in-migration tapers off as distance increases suggesting that distance is an important disincentive to relocating. However, this trend is not borne out when observing the significant labour-migrant contributions made by former homelands and the metropolitan areas despite the great distances (see Map 3.6). Economic perspectives presuppose that migrants will be primarily attracted by the economic opportunities offered by an area. However, the potential benefits have to be offset against the risks and the economic and social costs of moving. Financial costs include those of moving people and possessions as well as the cost of returning to renew associations, etc. 44